Eric's Aleph-X Amplifier Construction Notes

A DIY version of the excellent Pass Labs XA series amplifier

Eric's DIY Home Theater Projects

Misc DIY Audio Projects

Other Useful Links

 

Aleph-X Amplifier Construction - Table of Contents

 

Introduction and Acknowledgements:

This project was made possible through the generous assistance of Nelson Pass, Grey Rollins, Hifizen, Carpenter, William, and the rest of the community at DIYAudio.com. The notes on this page are a combination of my own experiences and a synthesis of those reported in the threads listed below. These notes are very detailed (but are well organized by topic with an index above) and contain general information on amplifier building as well as the specifics of this amp. This amplifier is a DIY effort to duplicate the exceptional ~$10,000 XA line of commercial amplifier offered by Pass Labs. The XA series amplifier combines the best features and performance characteristics of the legendary Aleph series and the highly acclaimed X series amplifiers (see the passlabs.com web site for more information on these). A number of people at DIYAudio contributed to the circuit design, the design and production of printed circuit boards, as well as construction and testing advice (special thanks to Magura and William!). This particular design is not so much a "project" that comes with a schematic, parts list, and step-by-step building instructions but rather is an open design that can be relatively easily be scaled to meet any set of output requirements dictated by your particular set of speakers and listening preferences. For example, the a40 amp that I built is available from the PassDIY site as a downloadable PDF file containing everything that you need to know to build one 40wpc stereo amplifier. No such single document exists for the Aleph-X amplifier (except for this web page). There are several versions of it in existence that range from 40w to 100w versions of the amp, each with different power supply rail voltages, different bias points, different numbers of output mosfets, etc. As each of these designs works, there is no single "right" way to build the amp except for the way that best suits your needs. The original schematic (see below) runs on 15 volt rails, 4.5A bias, dissipates about 130w of heat per channel, and will deliver approximately 38 watts into both a 4 ohm and an 8 ohm load. The following references are good (and lengthy!) reading about this amplifier and its evolution.

Relevant Threads on DIYAudio.com:

The Aleph-X -GRollins (it all started here with Grey on May 24, 2002)
Aleph-X Builders Thread - protos
Aleph-X Wiki- hifiZen (the original Wiki disappeared some time ago - this is a PDF copy I made of it in April of 2004)
Aleph-X Official PCB rev Beta & 1.0 - hifiZen
Industrial AlephX high powered version - Netlist
Aleph-X: High-Power Version - Blitz
Another Aleph-X coming up! - Edwin Dorre
Help needed: Firing up the Aleph-X100 - Blitz
One Aleph-X working, One to go - Wuffwaff
Aleph-X Offset Problem - Andy Pairo
Aleph-X bias current - SteveG
Another Aleph-X up and running - Xavier1000
AX100 100w Aleph-X Monoblocks - GL
Chassis Construction
- BrianGT


Amplifier Description:

The Aleph-X is a high current, pure Class A (excellent performance, simple design, lots of heat), dual gain stage amplifier (a Mosfet input stage adds voltage and a Mosfet output stage adds the current). The circuit is best described (by Nelson) as balanced single-ended Class A, consisting of two balanced Aleph amplifiers sharing a single differential input pair of transistors. This innovative design allows distortion from both halves of the output to cancel at the speaker.

The particular version that I am building delivers approximately 100wpc into an 8ohm load and about 150wpc into a 4ohm load (thanks to William on DIYAudio for sharing his hard work - I'm just a copycat...). Each channel uses 12 output mosfets arranged in 4 banks of 3 matched devices each and dissipates approximately 400 watts all of the time. Total draw from the wall measures about 3.8A for the completed mono amp when fully warm (one hour warm-up time). Power consumption is somewhat higher when the amp is cold and diminishes (just a little) as it warms up. Each channel is powered by a 1500VA transformer (though a 1000VA transformer would do just fine) with dual 18v secondaries, has over 500,000uF of capacitance, runs on 22v power supply rails (+22v, 0v, -22v), and is biased at about 9A, or 1.5A per output transistor. At 1.5A bias, each of the transistors in the output stage will dissipate approximately 34 watts (well within the safe operating area). Total heatsinking per mono amp needs to be a minimum of about 0.06c/w to limit thermal rise to the neighborhood of 30c. My amps have a total of about 0.072c/w of heatsinking and top out with a 32c temperature rise above ambient. This makes for a very large, heavy, and hot amp! Final physical dimensions measure just over 14" wide, 20" deep, and 12" tall, including 2" bolts used as legs. An earlier version of this amp used only 4 heatsinks instead of 6 for a total of about 0.10c/w which resulted in a 50c temperature rise! This web page describes the building process for my "second generation" (read: larger) amps. I built three identical amps, one for each of the Left, Center, and Right channels of my basement home theater. These amps handily drive my Avro Part open baffle speakers and provide pure sonic nirvana!

What does it Sound Like?

To me, these amps sound ^%&*#$! amazing! MUCH better than my Marantz MA500 monoblocks, much nicer than my Adcom GFA 545II, and better than the DIY Pass Labs a40 that I built a few years ago and still listen to every day. I defer to Nelson's impressions on the commercial version of this amp:

"The sound of this amplifier is a quantum leap over the parent’s. They retain the sweet warmth and lushness of the Aleph series without the fluffy colorations. The dynamic contrast is even better than the X series. The bass has as much control over the speaker as the original X amps, but is a bit more neutral and carries more subtle nuance. The midrange is a little deeper and the sound stage wider than the Alephs. Are these amps better than Alephs and X’s in every respect? Yes, except for the higher power/current ratings of the X amps."

Perhaps more eloquently, Nelson says that it sounds "like chocolate and peanut butter."

Which Version of the Amp to Build (the parts list)?

A Note Concerning Intellectual Property:

This amplifier is a combination of the Aleph series and X series of commercial amplifiers produced by Pass Labs and are protected by various copyrights and patents. These designs are strictly intended for individual, amateur, Do-It-Yourself purposes and are not to be duplicated for commercial gain.

There are at least two versions of this amp (low- and high-power) that have been built and tested by a number of people. The parts list for each can be found by following the appropriate link below: Choose the one that best suits your power needs. The general "rule of thumb" for determining your configuration is to identify the power needs of your speaker and listening environment. Though rail voltage and bias current are related to one another and interact to large extent, power into an 8-ohm load is primarily a function of power supply rail voltage (continuing to increase the bias beyond a specific point will not yield addtional output power into an 8ohm load), whereas power into a 4-ohm speaker load is primarily a function of bias current (though at some point, you cannot increase the bias any further without also increasing the rail voltage - here is where things get REALLY warm!). As a general rule, Nelson indicates that the Aleph-type amps generally sound better at higher bias settings. To make things a bit more real, you can easily deliver 100w into 8ohms with +/-22v rails and 5A of bias current (though you only get 50w into a 4ohm load). Holding the rails steady at +/-22v and increasing the bias to 9A keeps output into 8ohms the same, but enables the amp to deliver closer to 150w into 4ohm speakers. There is a big penalty to pay for (nearly) doubling the bias current - you need to get rid of significantly more heat! (not to mention the additional strain on your power supply...)

Grey's Original Version that runs on +/-15v rails at 4.5A bias, uses 4 fets per channel, produces about 40w into both 8 and 4 ohm speakers (at 50% AC Current Gain), dissipates 34w per mosfet, or 135w per channel. Each Q is biased by 0.5v/0.22ohms or 2.27A (I=V/R, ohm's law), or about 34w. This brings the total bias current per channel to 4.55A, total dissipation would be 15V x 2 sides of the circuit x 4.55A = 136.5w. If you bump the AC Current Gain to 66%, it will boost the power output for 4 ohm speakers to about 72w without increasing the heat dissipation, but this will degrade sound quality to some extent. Heat sinking for this amp needs to be about 0.15c/w or more per mono. The transformer needed for this level of power output is a minimum of 300VA with dual 13v secondaries per channel.

High Powered Version that runs on +/-22v rails at about 7 or 8A bias (the bias point is easily adjustable, provided your heat sinks can handle the heat), uses 12 fets per channel, produces about 100w into 8 ohm speakers, 150w into 4 ohm speakers (at 50% AC Current Gain), dissipates ~25-35w per mosfet, or somewhere near 320-400w per channel. Changing AC Current Gain to 66% will increase power output to near 180w into a 4 ohm load, though comes at the cost of somewhat degraded audio performance. Heat sinking for this amp needs to be 0.06c/w or more per mono. The transformer needed for this level of power output is a minimum of 750VA with dual 18v secondaries per channel. If you build the high-powered version (or any version requiring you to use multiple output transistors in parallel to share the bias load), you will need to duplicate everything that is enclosed in the turquoise boxes in the schematic (below). For example, using 12 mosfets per channel means having three of each output transistor in the schematic (Q1, Q2, Q10, Q11) as well as three of each gate resistor (R7, R9, R36, R38) and three of each source resistor (R5, R6, R40, R41). Not to mention a more stout power supply and copious amounts of heat sinking. Plan accordingly...

If you desire something different in terms of power output than what is listed above, you are venturing off into strict Do It Yourself territory. William on DIYAudio developed an Excel spreadsheet that is useful for designing your own amp. You can specify the power supply rail voltage, the bias current, the number of Mosfets you wish to use as well as the size of your heatsinks. The spreadsheet will calculate output power into 4/8 ohm loads, heat dissipation, and transformer ratings. I have also added a new worksheet to this file that can help you calculate and set the AC Current Gain for the completed amp (see below for this discussion). Click here for William's AXE1-2 spreadsheet. Without a fairly good understanding of how the circuit works and a fair amount of test equipment, you are not likely to get very far. Its much easier just to pick one of the linked versions and copy it (as I have done).

For those of you that want a little more help in designing an amp that meets certain power output requirements, here is a quick tutorial to get you going. William's spreadsheet is already filled in with "typical" power supply voltage, bias current, and heat sink size data that will result in an amp that can deliver about 100w into either an 8ohm or 4ohm load while keeping final running temperature at fairly reasonable levels. Start with cells B12 and B13 - these determine your power supply. Play around with these first and note the changes in power output indicated in cells B17 and B18. There are interesting relationships between Voltage and Bias Current to note. B15 is where you specify the number of output devices for your amp. This needs to be a multiple of 4 (by design) and use the value in B23 as a guide for changing B15. If B23 goes much below 20, decrease B15 by 4. If B23 goes much above 30, increase B15 by 4. Play with these cells until you arrive at a power specification that suits your needs. Then have a look at B66. This is where you specify the total heat sink capacity (per mono) for your amp. Then have a look at B80 and B82. You want to increase your heat sink capacity (decrease the value in B66) until B80 is kept sufficiently below 100c and B88 is kept under 45-50c or so. Play around for a while until you get a feel for the relationships. Then you can start shopping for heat sinks.

Grey provides a good summary of the design: "The Aleph-X circuit is capable of driving an arbitrarily low impedance as long as the bias is set high enough. As I've noted once or twice, the voltage sets the wattage capability and the bias tells you how low an impedance you can work into. Obviously, there are tradeoffs. Power dissipation gets ugly, fast. MOSFET reliability decreases. Cumulative Gate capacitance begins to be a burden, although different people will give you different answers as to when this happens. I tried to make the Aleph-X scalable. How large or small you build it is up to you, but please do it intelligently. "

Explanation of the Circuit & Schematic:

Grey explains: Taking it from the front end: (parts for each "side" of the balanced amp are group in parentheses)

Input Differential--(R23, Q5) and (R25, Q7)
Current Source--D1, R17, R20, Q6, R24, R26, V2
Protection & anti-ground loop--D2-5, R21
Input Network--(R18, R19) and ( R28, R29)
Feedback--(R16, C2) and ( R30, C4)
Current limiting/fault protection--(R10, R13, Q4) and ( Q9, R35, R39)
Output--(Q2, R6, R9) and (Q11, R38, R41)
Current Source--(R2, R3, Q1, R5, R7, R8, V1, R11, R12, C1, R14, R15, Q3, C3) and ( C5, R31, R32, Q8, V3, R33, R34, C6, R36, R37, Q10, R40, R42, R43)
Output grounding/DC offset control--(R1, R4) and (R44, R45)

Yes, you can quibble a bit about some of the groupings, like whether R6 and R41 ought to be included in the limiting/protection unit for instance, but it'll do for a start. If you want to, circle each group on the schematic with a pencil. Some units, like the Current Source, are completely, absolutely up for grabs. Wanna put in a different current source? Knock yourself out. Pull out the parts I listed above and insert the current source of your choice. Your one and only concern is that your new current source deliver something on the order of 20mA and be a bit variable. Unless you want to start changing other things...

The Aleph Current Source is really the only unusual thing, and the patent is actually pretty readable. The X part is nothing more than the way the feedback runs across from one side to the other. That patent is a little more obscure to read, though. The Outputs are nothing special. That's a normal way to hook up a gain device--grounded Source. Wanna put in bipolars? Okay. Tubes? Okay. Just be sure that you make some adjustments so that they bias up properly and you'll be fine. Having a differential is pretty much necessary, it's inherent in the X scheme. Besides, it's a nice, elegant way to do the feedback, the phase splitting (if needed), and get some gain going, all in one compact package. The Current limiting/fault protection unit can be removed entirely if you want. The Protection and anti-ground loop stuff, too. Just be careful with static at the front end if you take it out. Incidentally, you can do voltage feedback to the Aleph current source instead of current feedback and it'll still be Aleph. But that's another story.

R10/R13 and R39/R35 form voltage dividers that set the 'on' point for Q4 and Q9. Think of them as setting .65V for the Vbe. If the voltage drop across R6 or R41 gets too high, the voltage across R13 and R35 will approach .65V and the NPN will switch on, limiting the output MOSFET it is attached to. You can set this wherever you like or remove it entirely if it bothers you. To disable the current limiters, remove (or never install) R10, R13, Q4, R39, R35, and Q9. The circuit will hum along quite happily assuming that you don't have a fault condition at the outputs.

The sum of the resistance of V1/R11 and V3/R33 will alter the bias somewhat. In principle, you can remove them entirely (infinite resistance) and the Vbe of Q3/Q8 will take over entirely. Putting in resistance lowers that voltage, hence lowering the current in the MOSFET. By all means, experiment with different values there. This is going to be one of those issues that is influenced by the temperature of the MOSFETs, which in turn will be set by the bias current, rail voltage, and heatsinking.

I had the very devil of a time figuring out why my Alephs weren't biasing up properly. Eventually, I figured out that it was because they were stone cold (they're water cooled) compared to Nelson's production models which ran at supernova temperatures. Once I got that riddle solved, I was much happier. Everyone's thermal environment will be different so experimentation will be the order of the day. The only way around this that I can see would be to specify heatsinks for specific variations on the circuit, but since DIY people are notorious for taking matters into their own hands, that solution wouldn't work well in the real world. This is not a good beginner project...but then again, there are those budding DIYers who will scoff and try it anyway. Baptism by fire, you might say. A learning experience. The value of the resistors from the output to the current source will be interrelated with the value of the resistors from the outputs to ground. The lower one is, the higher the other (there are limits to this, obviously--you don't want a dead short in either location).

From the Aleph-X Wiki:

The hybrid of Aleph and supersymmetry topologies result in a balanced single-ended class A power amplifier, where a single differential pair feeds a balanced pair of Aleph-style output stages. Thus, it is a two gain stage design, where Q5 and Q7 form the first gain stage (diff. pair), while Q2 and Q11 form the second gain stage on each side. Q1 and Q10 are the primary power transistors for the dynamic current sources with controlling elements Q3 and Q8.

The input impedance is set with R18, R19, R28, and R29 at about 47k. D2-D5 provide input protection (overvoltage, electro-static discharge), while R21 can be used to help break ground loops. The front end diff pair Q5/Q7 is fed by a constant current source consisting of D1, R17, R20, Q6, R24, R26, V2. The voltages developed across R23 and R25 drive the primary output power transistors Q2 and Q11. Grey has included overcurrent protection using Q4 and Q9. These transistors are inactive (off) during normal operation, and the max current set point is determined by the R6, R10, R13, and R41, R39, R35 networks.

R2, R3, R5 and R40, R42, R43 are current sensing elements for the dynamic current sources (they must all be the same value). Looking at the left side, Q3 senses quiescent current (Iq) through Q1 as a voltage across R5. This voltage arrives at the base of Q3 via R8. R12 forms a voltage divider with R8, influencing Q3 when AC current flows to the load (speaker) via R2/R3 (C1 decouples R12 from DC voltages present on R2/R3). Thus, as Q2 draws more current during a signal peak, Vce at Q3 increases. This in turn draws more current through
R15, lowering Vgs at Q1, and Q1 will carry less current. Similarly, when Q2 is carrying less current, Q1 will carry more. The ratio of current variance in Q1 will be set by R8 and R12. At a 50% ratio, Q1 will theoretically carry no current when Q2 sinks 2xIq, and Q1 will source 2xIq when Q2 sinks zero current. In practice, the ratio is set slightly lower. C3 is a bootstrap capacitor to keep the current source operating properly as the whole sub-circuit rides up and down on the output voltage. R11 and V1 provide some biasing current to Q3, and allow some adjustment of the output stage bias / Iq. Once set (and the amp is nicely broken in), both V1 and V3 can be replaced with fixed resistors, if you desire. A good reference for more info about the Aleph current source can be found in the Zen Version2 writeup on the PassDIY web site.

R1, R4, R44 and R45 help control absolute DC offset at the outputs by sinking current to ground when the outputs sit either above or below 0V potential. The fairly low value of around 30 ohms means these resistors must be rated to dissipate substantial power when the amplifier is delivering maximum power to the speaker.

Successful Aleph-X builders appreciate that such a simple and elegant circuit requires careful construction, component selection and matching in order to deal with factors normally solved with added complexity in standard amplifier designs. While early prototypes of this circuit showed great promise, some difficulty was encountered with managing the common-mode (absolute) and differential DC offsets of the circuit.

How Much Heat Sink?

How much heat sinking do you need for this (or any other Class A) amplifier? LOTS! and then MORE! If your sinks are not HUGE and HEAVY, they're probably too small! A completed amp (high-powered mono, or low-powered stereo) should weigh in near 50-70 lbs. Any less and you've probably skimped somewhere where you shouldn't have. As an aside, Nelson recommends getting at least 15lbs of amplifier for each $1000 you spend on a retail amp... Looks like I've built monoblocks worth about $10k...

More practically, here is a rule of thumb that seems to work pretty well. In general, you want to limit the thermal rise of the heat sink to around 25c above ambient (thus providing a final temperature of about 45-50c). You typically want to be able to comfortably rest you hand on the heat sink of the operating amplifier for 5 to 10 seconds. Anything shorter than this and you are probably running too hot, which threatens the longevity of your output transistors, and possibly also the life of your speakers (once your output transistors fail). Nelson provides a good description of different thermal operating points:

Blimey Hot
10 seconds hands on
45c
Crikey Hot
5 seconds hands on
50c
Bloody Hot
2 seconds hands on
55c
X*?@!
60c

So how do you figure all of this out BEFORE you build your amp and discover that it is too hot (because your heat sinks are too small)? Glad you asked! Here is what I recommend. Thermal ratings of heat sinks are specified in units of c/w, or thermal rise (measured in degrees centigrade) divided by the wattage that they need to dissipate (w). This measure assumes a reasonable ambient air temperature (typically 20-25c), that the heat sinks are black anodized, have free air circulation (not using a fan and given plenty of breathing room-usually 8 inches on all sides and don't sit directly on the floor), and that the fins of the heat sink are oriented vertically. Be aware, though, that different vendors will have different assumptions when measuring and quoting the thermal dissipation of their heat sinks - sometimes you cannot reliably trust the figures. The bottom line is that you need BIG and HEAVY heat sinks - and LOTS of them. To get an idea of what you need, follow these steps:

  1. Divide 25c (max thermal rise that you want) by the quiescent power consumption of your amp (per channel for building monoblocks or for 2 channels together if building a stereo chassis, keep in mind this is the wattage dissipation, NOT output wattage - they are very different specifications).
  2. Take this number and de-rate it by 25%. This de-rating takes into account all of the inefficiencies of mounting your output transistors to the heat sink and the less than favorable conditions that you are likely to create in building your amp even if you follow the advice above (trust me, no matter what you do, it won't be ideal).
  3. Take this number and round it down to the nearest 100th decimal place.
  4. This is the size of a heat sink that you really need. Then think of a way to get more...

So here are a few examples:

The a40 amp (see link at the top of the page) is specified to dissipate 200w total for two channels (its a stereo chassis). Dividing 25c by 200w yields 0.125c/w. Next, derate this by 25% by dividing your c/w rating by 1.25. This yields a more realistic figure of 0.10 c/w. Then round it down (resulting in a larger heat sink specification). Thus, the final heat sink rating should be about 0.09c/w in order to dissipate 200w and keeping thermal rise limited to 25c. How close does this come to reality? My completed a40 amp actually draws 190 watts from the wall socket (slightly less than specified), has a total heat sinking capacity of 0.084c/w (slightly more than calculated), and has final thermal rise is 24c (almost exactly to target specs).

As another example,the Pass Labs Aleph2 monoblock delivers 100wpc into an 8 ohm load and dissipates 300 watts of heat all of the time. Nelson Pass states that the Aleph2 requires heatsinking of 0.06c/w per monoblock. Using the theoretical calculation dissipating 300 watts of power while limiting thermal rise to 25c above ambient (25c/300w) should require 0.083 c/w worth of heatsinking. Derating the theoretical result of 0.083c/w by 25% (0.083/1.25) provides us with a more realistic figure of 0.0656c/w and rounding down to the nearest 100th yields exactly the target value of 0.06c/w indicated by Nelson.

If you do anything screwy with your heat sinks like use ones that do not have black anodize, spray paint them (shame on you!), mount them horizontally (don't do this), block the air flow above or below them (duh), or place them in a closed cabinet (NEVER do this!), you need to increase your de-rating fudge factor to something more like 1.50 or even 1.75 depending on how bad things are. Also remember that your completed amp needs a little breathing room under it as well. This helps improve the vertical airflow through the heatsinks so they operate at maximum efficiency. Make sure that the lowest point on your heatsinks are at least 2-3" above the nearest surface to allow enough air through them. If you are looking at using a smaller number of larger heatsinks, be sure that the back plate of your extrusion is at least 0.25" thick. There are specific formulas for determining the most efficient thickness for any given heatsink, but if you stay close to 0.25", you're in good shape. If you end up with a sink that is much thinner (anything below 0.20"), it won't be able to spread the heat as evenly and as efficiently across the entire radiating area of the sink. The end result is that the outer regions of the sink will run much cooler than the middle portion of this sink where your output transistors are mounted. Thus, you will have paid a great deal of money for heatsinks that aren't able to operate at their maximum efficiency. [Returning to the topic of painting heatsinks, a professionally applied thin layer of paint can work wonders for a raw aluminum heat sink to increase its emissivity, but most people cannot apply an consistently even, thin coat of paint. The result will most likely be uneven layering of paint that is much too thick and instead function as an insulative layer surrounding your heatsink. This clearly will not help matters.]

Black anodize enhances the emissivity of aluminum heat sinks and helps keep your output transistors cooler. Typical raw aluminum has an emissivity coef of 0.07 whereas black anodized aluminum has a coef that is closer to 0.77 (values closer to 1.00 are better). Have a look here for more details on heatsink surfaces and lengths. Unfortunately, there is NOT a linear relationship between heatsink length and its ability to dissipate heat. A 100% increase in the length of the heatsink translates to something like a 40% increase in heat dissipation (square root of 2, to be precise). Like I said before: BIG and HEAVY! And then some...

In conclusion: Perform your calculation for the power that you need to dissipate, determine the appropriate sized heatsink, then derate its dissipation by 25% of the calculated dissipation rating and round this answer down. This is the actual size of a heatsink that you need - unfortunately, you always need more heatsinking that you think! Now, while it is next to impossible to find a single heat sink that is rated at 0.06c/w for a chassis that needs to dissipate 300w, it is much easier to find several heat sinks each rated somewhere in the range of 0.25c/w to 0.45c/w and make sure that your output transistors are evenly distributed across your sinks. When using multiple sinks, just add their c/w ratings to one another to obtain the value that you need. Have a look at Rod Elliott's Heatsink Thermal Rating Calculator Spreadsheet if you've found an unknown heatsink and need to guesstimate its thermal dissipation rating.

Heatsink sources: You can purchase heatsinks new from places like Alexandria, rTheta, Wakefield, Conrad, and probably a few others that I am forgetting right now. Buying new often costs a fair bit of money and your order is often subject to a minimum dollar amount that may be difficult to reach for just a single amp. Of these vendors, M&M had the lowest minimum order of $200, but they've been purchased by Alexandria and that might have changed. A few years ago, I did business with a local distributor for Seifert (who makes VERY nice heat sinks), but their minimum order is $500. If you are working on a smaller budget, you will need some patience as you spend time surveying surplus vendors (see the links at the bottom of this page) and e-bay for a large enough quantity of large enough sinks (large ones are relatively difficult to come by). For my Aleph-X amps, I purchased 20 pieces of extrusion #11645 (0.8c/w per 3" length and 75.5sq in of radiating surface per linear inch) that were each 10" long (yielding approximately 0.43c/w for each piece- remember, doubling the heat sink length only increases dissipation only by a factor of 1.41) and had them apply black anodize. I paid close to $30 each for an order of this size, plus a flat fee to anodize the lot. Remember, wider is better than taller for dissipating heat and look for a thick mounting plate (ideally, 1/4" or more) for the transistors.

Be sure to read through Rod Elliott's page on heatsinks and transistor mounting.

Power Supply:

Construction Notes and The DIY Disclaimer:

I am not an expert and, I suspect, neither are you (or you wouldn't be reading this). This project involves electrical wiring that connects to the 120v AC house wiring. These voltages are lethal and you should take all appropriate precautions. If you don't know what these precautions are, be sure to take the time to learn before going any further!

You can arrange your power supply in a number of ways. Some people just use the standard transformer-bridge-caps approach (which I did for my a40 amp), while others recommend CRC or even CLC for the higher powered configurations in order to reduce power supply ripple due to the higher current draw from the caps. Below are two power supply configurations, the first (left) is a CRCRC from Hugo of DIYAudio (including a photo), and the second (right) is a CLC from Kristijan. The first power supply provides 30v rails, while the second one will provide 20v rails. After reading through the power supply information below, you might want to download a copy of Duncan Amps Power Supply Designer software. Its a very nice piece of software that will let you simulate the size of your transformer, secondary voltages, various configurations such as CLC or CRC, and bias load placed on the power supply so you can explore power supply ripple and other factors before you build your own.

Constructing a Star Grounding Point:

No matter which implementation you choose for your power supply - just caps, a CRC, CRCRC, LC, or CLC - do yourself a favor and create a high-quality star grounding point using 12ga wire. This is the single biggest step you can take toward eliminating hum and buzz in your speakers. Much of the source of hum comes from a poor grounding configuration or ground loops, meaning that you have more than a single grounding point in your amplifier (easy to address), or across pieces of equipment within your system (sometimes harder to address).

A schematic of a star ground is pictured above along with photos from my own power supply grounding scheme. Note how all ground wires connect to a single point within the power supply - this is the star ground. The picture on the left shows the overall power supply and the picture (the small black and white wires attached to the terminal screws on the cap are for the power on LED) on the right shows a close up of the star grounding point. If you are experiencing hum that originates within your amplifier, a star ground is probably the best bet for eliminating it. If your are experiencing sum that originates due to the interaction of components within your system, there is a different approach you can try. Hum sometimes results from a ground loop across two or more pieces of equipment in your audio chain and is one of the reasons that optical TOSLINK cables became popular - they removed one electrical pathway for ground loops across boxes. Another way to eliminate this type of ground loop is to isolate the signal ground in your chassis from AC Mains ground. I've done this with the black thermistor that you see in the photos above. A standard CL-60 has a resistance of about 5 ohms when cold, thus it provides some measure of isolation, but will still allow current to flow to ground should this be necessary (as a result of a wiring fault or component failure). Another method is to connect signal ground to AC Mains ground through a diode. Each of these methods works and behaves somewhat differently. In my a40 amp, I have used a diode to join signal ground with AC Mains ground - using a direct connection or even a thermistor here causes the amp to hum at the output. In my Aleph-X amps, there doesn't seem to be much difference between a direct connection, using a thermistor, or using a diode. Thus, I used a thermistor because I happened to have one laying around in my parts box when I was building the power supply.

Finally, do yourself a favor and solder ALL of the crimp connections you made on the wires for your power supply. This will eliminate whatever little bit of resistance that remains in your crimp connector and help to keep things more secure in the long run. Remember, there is a GREAT DEAL of power circulating here!

Choosing a Transformer:

With Class A amps, the calculation of how large a transformer to purchase is pretty straight forward: the VA rating of your transformer should be a minimum of 7.5x the target output power for an 8ohm load. For example, if your Aleph-X amp will provide 100wpc into 8 ohms, you need a MINIMUM of a 750VA transformer. Many people opt for a 1000-1500VA transformer in this case to provide an increased margin of flexibility for increasing the bias or trying different configurations. The a40 amp that I built provides 40w into 8 ohms and is a stereo chassis fed by a single transformer. Multiplying the output power (40w) times 2 channels and then multiplying by the 7.5 factor results in a minimum rating of 600VA. This is exactly the transformer that I used. While I could have gone for a higher VA rating (this transformer has a slight buzz to it), the one I have now works just fine without any problems. In this spirit, I would recommend a 1000VA transformer in the Aleph-X power supply rather than a 600VA transformer that it specified in the above power supply schematic (on the right). In my experience, at 1500VA transformer runs more quietly (less mechanical buzz) than does a 750VA. In addition, the larger transformer will have better power regulation and your power supply rails will droop less under load.

Another element to plan for is various types of losses in the power supply. Using "traditional" bridge rectifiers will likely cost you about one full volt from your power supply rails. Adding a resistor to the power supply for a CRC filter will drop the rails further. Finally, placing a load on the power supply will take things down even further. Here is how things work out for my 1500VA transformer. Without load, the transformer secondaries provide 20VAC (the transformer was designed to provide 19VAC under full load). So we have 20v * sqrt(2) = 28v. But the rectifier costs you about one full volt and you'll drop a little bit of voltage due to the resistance inherent to the power supply caps, so my measured DC voltage is effectively 26.6v. Next, I added a 0.2ohm power resistor (for the CRC) and an additional 220,000uF cap (thanks, Peter!). At just over 9A of current draw, the 0.2ohm resistor drops 1.8v from the rails, taking things down to about 24.5v. The load itself of the amp pulls the rail voltage down another 2.5 to 3.0v, so my final design results in rails that are about 22.0v while providing 9A of current. If you add a thermistor to the primary of your transformer to help suppress the inrush current, you'll lose another 0.3v or so from the rails. Everything you add costs you rail voltage. This could be good or bad, depending on your situation (perhaps lowering the rails keeps your output transistors and heat sinks a bit cooler). Thus, my advice for you is to start with an idea of your targeted rail voltage and work backwards. For an interesting comparison point, a smaller 750VA transformer displayed at 13-15% voltage sag under full load, the 1500VA transformer features a voltage sag of 7-10%. See the size comparison below. The 1500VA is on the left, measures just over 8" in diameter and weighs in at 22 lbs. The smaller 750VA is measures just about 6" in diameter and weighs about 14lbs. The CD in the middle is for size reference.

OK, so now we know how large the transformer needs to be (its VA rating), the next thing to figure out is the required voltage to be supplied on the secondaries of the transformer. The (theoretical) calculation for specifying the proper secondary voltage for your transformer depends on the type of power supply you want to create: Cap-input or inductor-input.

Capacitor Input Power Supply:

The most typical configuration is a capacitor input power supply (three are shown above). For a cap-input power supply, take the rail voltage that you need and divide it by the square root of 2 (1.414). Thus, if you need 32v rails, you should get a transformer with 22v secondaries (32/1.414 = 22.6, in theory). In reality, you lose somewhere between 0.7v and 1.0v due to rectification (the specs for the diodes or bridge will typically indicate for the forward voltage loss), sometimes a little more at the caps, and then the rails will droop once under load from the amp (this is especially true for Class A amps which have a constant and high current draw). For my a40 amp, 24v secondaries provide about 32.5v after rectification and at full load. Thus, dividing the needed rail voltage by 1.35 is more accurate in this case. If you are planning a CRC or a CLC filter (see below for details), you will probably lose another little bit (on the order of 1 to 2v) depending on the resistance you use, so plan accordingly when choosing a secondary voltage level. Each of the power supplies shown above is for a single channel. If you are looking to build a stereo chassis, plan on doubling things up. At a minimum, you want to have a dedicated bank of both positive and negative power supply caps for each channel. This provides good channel isolation and good economy since the transformer is shared by both channels. For the true audio purist, nothing short of one transformer for each channel (total isolation and therefore highest cost) will do.

Choke Input Power Supply:

For a choke input power supply, take the rail voltage that you need and divide it by 0.9. Thus if you need 32v rails, you should get a transformer with 36v or 37v secondaries You will still lose voltage over the rectifier and caps, plus a little additional voltage with the resistance of your inductor. When choosing the inductor that you will use, keep an eye on the resistance (measured in DCR) of the coil as higher resistance will cause the inductor to heat up considerably. Consider using 12g or 14g inductors. This is also likely to impact the necessary venting for your chassis to deal with the extra heat that will surely impact bias and DC offset settings (discussed below).

Some sources for transformers include Plitron, SumR, Toriod, Avel, Antek, and Signal. Many people had custom transformers wound by Victoria Magnetics, but they don't appear to be around any longer. More recent reports on various forums indicate that there may be quality control issues with recent Avel transformers that are made in China...

Dealing With Capacitance:

Due to the high current draw of this amp, an extreme amount of capacitance is called for. The typical "bigger is better" clearly applies here, but some care is needed (which is typical for Class A amps due to the higher-than-typical current draw when compared to Class AB and B amps). Hugo's power supply (left) features nearly 150,000uF per rail at 30v; Kristijan's (right) has 100,000uF per rail at 21v; my power supply has 290,000uF per rail at 22v. Do be sure that your power supply caps are rated about 20% higher than your planned rail voltage so that the caps can comfortably handle the voltage and to provide a safety margin. 25v caps are just fine for an amp with 21v - 22v rails. Also, when you get into capacitance this large, it is often useful to have an inrush limiter (typically a thermistor) on the primary (line side) of the transformer to reduce the stress on your power switch and rectifiers as you power up the amp (it will also keep the lights in your house from dimming quite so much). One recommended thermistor is a CL-60 (Mouser #527-CL60) that is rated at 10ohms (at no load) and 5A maximum steady state current. For a larger 1500VA transformer that is most likely wired with two sets of primaries, put one in series with each primary so you don't burn them up by drawing too much current through just one. Under load, the resistance drops to about 0.18 ohms and you will also see a corresponding drop in your power supply rail voltage - likely about 0.2 or 0.3 volts. And they run HOT! My CL-60 runs somewhere near 140-150c - well above the boiling point of water, so keep some distance between wires so you don't melt the insulation away.

Also, while we are talking about high currents in the power supply, the best power supply switch is a double pole switch that has 4 or 6 terminals that is rated for 20A at 250v. Essentially, this is two switches that operate in parallel, so the current is shared across two sets of contacts, thus reducing the stress and potential for arcing and pitting the switch contacts. To reduce the potential for arcing at the switch, you can bypass the switch with a 0.1pF 600v ceramic disk capacitor (see images of the chassis power entry below).

A Word of Warning: This much capacitance is HUGE! It can be LETHAL! Don't be DUMB! Take off any dangling necklace or bracelet or metal watch band that can accidentally touch the terminals of your power supply caps. Before working on an amp that has been powered up, make sure that your caps are discharged! The best way to do this is to install bleeder resistors across the caps in your power supply. These are usually 4-8k ohm and rated to 5w-10w. Alternatively, you can connect a power resistor across the terminals with alligator clips once the amp is powered down. DO NOT use your screwdriver to short the terminals! This will destroy your screwdriver (arcing literally melts metal away), possibly your caps, any maybe you in the process. Please be careful! The best way I have found to discharge caps is using a small automotive light bulb (like a blinker light) that can handle 12v or 18v. If you have more voltage, just wire two bulbs in series. They glow brightly at first and then fade as the caps discharge after a few seconds. Best of all, a light bulb won't cause arcing!

Using a CRC or CLC Filter (a Pi Filter):

If you really want to design a great power supply (one that minimizes ripple [cyclic highs and lows in the rail voltage] and provides cleaner power to your amp) the designs of choice are CLC (capacitor, inductor, capacitor) or CRC (capacitor, resistor, capacitor) power supplies. So which do you choose, CLC or CRC? Nelson had some comments on the DIYAudio forum about this topic. His comments indicate that either a CRC or CLC will do the job. However, he tends to prefer CRC networks because 1) they are cheaper 2) they have no resonance like a CLC does 3) no mechanical noise like a CLC and 4) no magnetic field to be picked up as you might with a CLC. Additionally a CRC takes up less room in your chassis. On the other hand, CLCs reduce ripple further than a typical CRC can. I used a CRC in my set of amps... Inductor values should be in the neighborhood of 1.5mH to 2.2mH (don't fret over whether you should use 1.4mH or 1.8mH - each will have largely the same effect) whereas resistor values in the range of 0.10 to 0.20 ohms will work well. Larger resistors reduce ripple further, but burn off more rail voltage and heat in the process. In both cases, be sure to watch the power handling capability of your inductor or resistor. These amps draw large amounts of current, so inductors should be made with 12ga or 14ga wire and able to handle 5-10A , whereas the resistor will need to be an aluminum housed type that has metal flanges for bolting it to the chassis for heatsinking and able to handle 25-50w at 50c operating temperature. To calculate the power dissipation of the power resistor, measure the voltage drop across the resistor, square it, and divide this by the resistance. For instance, in my completed amp, the voltage drop across the 0.2 ohm resistor in my CRC configuration measures 1.8v. Thus, power dissipation is (1.8^2)/0.2 = 16.2 watts. I am using a 50w resistor that is mounted to the metal chassis with an electrically conductive sil-pad to dissipate heat, so I'm in good shape. Additionally, measuring the voltage drop across the resistor will provide the bias current for that half of the amp. Using that same measured 1.8v drop across the resistor and dividing by the resistance (0.2 ohms) indicates the amp is drawing 1.8v/0.2 ohms = 9A. Ohm's law is a wonderful tool! A quick side note for using an electrically conductive sil-pad (like Bergquist Q-Pad 3 that are meant for rectifiers) for mounting the resistor to the chassis in the event of a failure of some component of your power supply: it is desirable to have the power supply short to ground and blow the main power supply fuse before it causes down-stream damage to the rest of your amp (output devices) or your speakers.

Rectifier:

There are a number of options here. Some have used individual STPS 80 H100 TV Schottky's, while others have just used a standard 600V 35A bridge that is readily available from Mouser, Digikey, or and number of surplus vendors. In either case, be sure that your rectifiers can handle the current surge at turn on (I would think a 10A steady-state diode is plenty, check the specs of the device you are considering) and are mounted directly to the metal chassis (with a thermally conductive electrical insulator, of course) or on their own heat sink to help them stay cool. For reference, a typical IRF bridge rectifier rated 600v 35A can handle temperatures up to about 80c while carrying a load of about 10 amps. In practice, mounting your rectifier to the bottom or back of a metal chassis should keep its temperature somewhere near 40-50c. If you don't use some sort of thermal transfer interface (sil-pad, grease, etc) between the rectifier and chassis or heatsink, the heat may cause your rectifier to fail prematurely (as mine did). Both of the rectifiers in my amp recently failed after running continuously for 24 hours to break in a new pair of speakers. As the diodes began to fail, the transformer began to groan more and more loudly until the fuse popped. Playing around with new power supply components for a little while revealed that it was indeed the rectifiers that died. Just for kicks, I temporarily rebuilt the power supply exactly the way it was to study the thermal behavior. With my non-contact infrared thermometer, I measured the base plate of the amp at 35c, while the rectifiers clocked in at 82c - YIKES! No wonder they failed, they were living life on the edge of the power dissipation/case temperature curve. The story wasn't much better for the 0.2ohm 50w power resistor that was in my CRC power supply - it was measuring close to 65c. Both of these devices need a thermal interface material under them to improve heat dissipation and increase their life span!

Matching Components:

Since this current gain stage in this amplifier design is essentially two Aleph circuits operating in parallel, it is important to make sure that certain components are carefully matched to one another. Of critical importance is matching the input differential pair, followed by the output mosfets. Many people will stop here and then use 1% tolerance resistors for the rest of the circuit. While ending your matching efforts here is probably sufficient, one of the points of DIY is doing things because you can and because you can make something better. Following that thought, I also matched the resistors across the halves of the circuit, especially the mosfet source resistors to help insure proper current sharing. Finally, and perhaps not important at all, I also matched the capacitors across the two halves of each amplifier. One useful technique for organizing the matched sets of resistors was to use a piece of foam (left over from insulating the concrete block walls in my basement before I finished it) with a strip of masking tape on it. I could write the part number and its measurement on the tape and then poke the resistor lead into the foam to hold it in place. This way, I was able to store all of the matched resistors for each amp in a single piece of foam as shown below. Since the lead is long enough, there is no danger of the parts falling out should the whole thing tumble to the floor as mine did several times.

Which Components to Match

A single channel of an Aleph-X amplifier is essentially two (traditional) Aleph amplifier circuits running in parallel: one amplifier for the positive signal input and a second amplifier for the negative signal input. Thus, the Aleph-X circuit requires careful matching of components across the two sides (each amplifier) of the completed circuit in order to perform at its best. Also, in order to avoid problems with DC offset at the speaker output (having your amp put out a constant +5v to your speakers), it is necessary to match several sets of components. I matched the following sets of items, highest priority and most critical matches are indicated first:

  1. Q5 and Q7 that form input differential. Ideally, these should be matched to within 1mV from Gate to Source at the rail voltage and bias current that they will see in the completed amplifier. See matching notes below for additional details.
  2. R23 and R25, which bias Q5 and Q7. Since these two resistors set the bias level for the input differential, it is essential that these match at the 0.1% or greater level.
  3. Within each bank of power mosfets for the current source (Q1 and Q10) and the output stage (Q2 and Q11). If you follow Grey's original design (~38w into 8 ohms), the completed amp uses only 4 transistors, one each for Q1, Q2, Q10, and Q11 and this type of matching does not apply. However, if you are looking for more output power (I'm looking for 100w into 8 ohms and 150w into 4 ohms), you will need to parallel your output mosfets so that they share equally the current and do not burn themselves up in the process. My completed amp will have 3 mosfets in parallel for each of Q1, Q2, Q10, and Q11 (12 in total for each channel). It is essential that each mosfet within the set of 3 closely match one another. That means each of the three parallel mosfets for Q1 must match one another, same with the three parallel mosfets for Q2, and so on. Nelson recommends matching to within 0.1v, while others advocate matching to 0.01v or even 0.001v at your intended rail voltage and bias current setting.. Be aware, though, that there are diminishing returns for matching beyond the 0.001v level as the devices will experiences different temperatures when mounted to the heat sinks of the completed amp. Different temperatures will impact the gain of the devices, and vice versa. Since you can't completely control this, it seems that the conventional wisdom indicates matching at the 0.01v level (groups within 10mV of one another). I was able to match mine to the 0.001v level (groups matching within 1mV of one another).
  4. Across banks of power mosfets. For the output stage, it is also critical to match all mosfets for Q2 with all mosfets for Q11. So for my amp (that uses a total of 12 power mosfets), I need 6 carefully matched mosfets (spread across Q2 and Q11). It is also desirable, but less critical, to match mosfets for Q1 to those for Q10. Thus, you would really need to have two sets of 6 mosfets that are carefully matched. If you are so inclined and have a large enough pool from which to choose, you can match all 12 for the entire amp. All you need is plenty of time and a big supply (if you need to find 12 that match, you might need to purchase 24 to 36 at the same time) of mosfets to test.
  5. Source resistors for the power transistors. Each of the mosfets in Q1, Q2, Q10, and Q11 will also need source resistors (R5, R6, R40, R41) that can dissipate a fair amount of power over a sustained period of time. Since these resistors will ultimately affect the bias level for each mosfet, they need to be carefully matched as well. For my design, I paralleled three 1ohm 5watt resistors to result in a single resistor that measures 0.333 ohms and can handle approximately 15w of power dissipation. I first measured each 1ohm resistor and then grouped them together so that each set of three matched as closely as possible. The result is a set of 12 resistor bundles that measure 0.333 ohms and a set of 24 resistor bundles that measure 0.332 ohms. A bit anal? Perhaps, but they match well and I did it because I could.
  6. All remaining resistor values across the two halves of the circuit. Since this amp is essentially two amplifiers (each resembling an original Aleph Amp) running in parallel, both "halves" of the circuit need to match as precisely as possible. It seems that tolerances of 0.1% here are sufficient. If your meter has enough resolution (a regular $20 3.5-digit DMM is fine here), you may be able to match at the 0.05% to 0.01% level- better matching will never hurt. To be able to build 3 amplifiers, I ordered enough of each part to build 4 amps. This has provided enough components of each value to perform careful matching, plus will allow some extras should I inadvertently allow the magic smoke to escape from any of the parts when I fire it up.
  7. Resistors to Match to One Another
    R1/4 and R44/45
    R10 and R39
    R18 and R28 0.1%
    R2/3 and R42/43
    R11 and R33
    R19 and R29 0.1%
    R5 and R40 (all) 0.1%
    R12 and R34
    R22 and R27
    R6 and R41 (all) 0.1%
    R13 and R35
    R23 and R25 0.1%
    R7 and R36 (all)
    R14 and R31
    R46 and R47
    R8 and R37
    R15 and R32
    R9 and R38 (all)
    R16 and R30 0.1%
  8. Capacitor pairs. This is most likely not necessary at all, but I figured since I had matched everything else, I would try to match these as well. Since the tolerance on caps is not as tight as that of resistors, matching at the 0.1% is not possible (and may not even yield any benefit), but I matched them as closely as I could because I could.

Matching Mosfets

Matching mosfets seems to cause a great deal of distress among amplifier builders. It really is not terribly difficult to accomplish, but it does require access to equipment that many people do not have lying around. Also, purchasing the necessary equipment just to build a few amps is cost prohibitive. You need to have access to a regulated bench power supply and a precision volt meter, preferably one that reads to the mV level such as a 4 1/2 digit or greater DMM (the 6.5 digit multi meter that I used retails for $1200 - yikes! Glad I had access to the University's EE lab on campus. Alternatively, make friends with someone who is an electrical engineer...). Your typical $10-$30 DMM is not well suited for this task as its precision may not be great enough. Nelson posted a set of articles about matching and testing mosfets on the Pass DIY web site that are helpful reading. Essentially, you want to set up the circuit below for making your measurements. The figures in the diagram have been optimized to test the IRFP240's and IRF9610's at the actual voltage (22v) and currents (10mA for 9610s and 1.125A for the 240s) that they will see in the completed amp. Note the changes in the test jig: the 9610 is a P-channel mosfet, whereas the 240 is an N-channel mosfet.

Matching IRF9610s (Q5 and Q7): My design calls for 22v rails and the circuit biases the pair of them at 20mA. Thus, matching should be performed with 10mV applied to each one. For matching your mosfets using this technique, you can start by either marking each device with a piece of tape or sticker with a number on it, or just use the number that is imprinted on the casing itself. Incidentally, if you have a series of mosfets with sequential numbers imprinted on them, they were made at the same time and from the same silicon wafer. Having 10 or more from the same lot greatly increases your chances of finding well-matching pairs. Make a piece of paper that has a place to record the device number and its voltage measurement at several points in time - I measured mine at specific time intervals of 1, 2, 3, 4, and 5 minutes to see how they behave over time. Next, build yourself a test jig like the one pictured above. Use half of a dip8 socket (available from Radio Shack for about $0.70) and a small piece of perf board (also from RS). In order to make the test jig more stable during use, I screwed it down to a small block of wood and then labeled the connection points so I wouldn't forget which was which... To determine the appropriate value of resistors for your test jig, just use Ohm's Law with your target voltage and current settings. I targeted 22v rails and assume that the mosfets will drop 4v of power, so 22-4 = 18v. The 9610's need to be matched at about 10mA each for this circuit, so that's all you need to start. Use Ohm's Law for the rest: I=V/R, or R=V/I, R=(22-4)v/0.01A, or 1800 ohms. Paralleling a 2k2 with a 10k resistor will achieve 1k8 as specified in the circuit and normal 1/4 or 1/2 watt resistors are fine since we are only dealing with milliamps of current. The value of the test jig is that you don't have to disconnect/reconnect all of your probe wires each time. I simply socketed each device, then connected the positive power supply clip. When the test period was over, I disconnected the positive supply clip and then removed the transistor from the socket.

Before you start measuring, lay all of the pieces out on the table in close proximity to one another and let them sit for 20-30 minutes so they are all at the same starting temperature. Find a comfy chair - you'll be here a while because you need to measure all of your devices at the same time. If you measure half of them now and half of them on another day, you won't be able to compare the results because the environment will be different. I found that it took about 1 hour for every 10 devices I tested. You also have to be very careful how you handle these devices while you measure them as they are extremely sensitive to their temperature environment. Use pliers instead of your fingers to place them into your test measuring jig. Even brief contact (1-2 seconds) with your fingers will heat up the part prior to testing and will adversely impact their final measurements (hotter devices result in a lower voltage reading). Also, simply walking past them or breathing on them while testing them will affect your results as well!! Be sure any room air conditioners or heaters are turned off, all fans are off, windows and doors are closed, etc. You even have to be careful where you place them in relation to your regulated bench supply as many of these have built in fans that create a breeze. If you don't believe me, just try gently blowing on one once the voltage reading has stabilized - the voltage reading will change quickly as the device changes temperature! I placed my test jig down into a box that would help isolate the devices from vagrant breezes while I measured them. Using your pliers, gently insert the device into the socket and connect your power supply. Use a stopwatch that clearly indicates seconds to keep careful track of time.

At first, the voltage reading will decrease fairly rapidly as the device begins to heat up, but after about 4-5 mins the vgs is getting stable and the voltage drop is getting very slow - the forth decimal place was changing by one digit with each 5 to 8 seconds that passed after 4 minutes. To measure these, I used a 6.5 digit multi meter with a regulated bench power supply at the voltage and current settings indicated above. This particular meter provided 4 decimal place resolution (0.0001V) and within my set of 40 devices (that included 3 groups of sequentially numbered fets) yielded nine sets of 3 closely matched devices. In three cases, I found pairs of devices whose voltage change tracked identically across the entire 5 minute test period at the 0.0001v level (they also had sequential numbers stamped into them)! Looks like I've found my input differentials for my front left, center, and right amps... Nelson indicates that he matches these devices to within 10mV for the Aleph series of amps, but those who have built the DIY version of the Aleph-X indicate that the amp is extremely sensitive to the matching of these devices to minimize DC offset problems (see below) and recommend matching them to within 1mV or better over a period of up to 30 minutes! My pairs are matched to the 0.1mV level after 5 minutes.

Rematching these same devices a few days later revealed different final measurements (I suppose due to different ambient temperatures), but the original groupings were still within 0.001v of one another.

Matching IRF240's (Q1, Q2, Q10, and Q11): The main point in matching these devices is for building higher powered versions of the amp that require the use of more than one mosfet for each of Q1, Q2, Q10, and Q11. Typically, you want to keep the power dissipation of each mosfet in the 20-30w region. If your output goal indicates each mosfet would be dissipating much more than 30w each, its time to think about paralleling two or more mosfets so that they share the current dissipation. Equal sharing is what requires careful matching.

Measuring these devices is essentially the same as indicated above for the 9610's, but you cannot test these for as long of a period of time without attaching them to a heat sink as they really heat up quickly! All of the same precautions about temperature variation for the 9610's apply here as well. Also, if you are testing the 240's anywhere near 0.5A or higher, regular 1/2w resistors do not have sufficient power dissipation ratings, it is recommended that you use resistors rated at 30w or more. Since the pin spacing on the 240's is different from the 9610's, they won't fit into the same DIP-8 socket that you made for the 9610's. The 240's do, however, fit into every other hole in an 18- or 20-pin socket, but they are a little tighter... I purchased several sets of IRF240's from Matthew Olson of DIYAudio that he matched in groups within 10mV using 15v and 0.5A power supply for 30 seconds. Since I have access to good test equipment, I re-tested them at 22v and 0.75A for 45 seconds. Using Ohm's Law, I calculated the value of the resistor R=V/I, R=18v/0.75A, so R= 24 ohms. From what I have read on the subject of matching, it seems that many people prefer to match the devices at the actual voltage and current levels that they will see in the completed amp. Unfortunately, the regulated power supply I am using can deliver a maximum of 1.0A at 25v, so I cannot test them at 1.125 as I had intended to.

As expected, when measured at different voltage and current levels for longer periods of time, my final measurements differed from Matthew's by as much as 0.15v, but his original groupings held up very well! In contrast to measuring the 9610's, the voltage reading for the 240's changes very quickly as they heat up, making precision voltage readings difficult to perform at 0.001v levels. I had measured the 9610's with the meter set to 4 decimal places, but this level of accuracy is not feasible for the 240's, so I made measurements with the meter set for 3 decimal places. Even at this setting, the last digit changed rapidly, making it challenging to record precise readings at specific time intervals (15s, 30s, and 45s) for each mosfet.

After 45 seconds at 22v and 0.75A, the transistors got quite hot and could still be handled without burning my fingers, but I don't think I'd test them at much higher current or for any longer without using some form of heatsinking... Also, I used three 10w power resistors in parallel in order to achieve 25 ohms (100 + 100 + 50 ohm resistors are close enough to the target of 24 ohms). Together, these resistors should be able to handle about 30w of power, but they got quite hot to the touch and after an hour of testing, are capable of burning your fingers! When you arrange your test jig, make sure your power resistors have some breathing room to cool off. While testing a higher voltage and current levels for longer periods of time allowed me to be more selective, but did not fundamentally alter the groupings that Matthew had made. Overall, it looks like testing at 15v and 0.5A is just fine for this application. In the end, I was able to match six sets of 6 devices at a time to within 0.003v to 0.005v of one another (that's within 0.03% accuracy for both the 9610s and the 240s!). Nelson recommends matching at a level of 0.1 to 0.01Vgs, so it looks like my groupings will be just fine!

Matching at Different Bias Points: If you need to measure your IRF240's (or 9610's) at a different bias point, just use the formula Current = (Voltage - 4 ) / R, where Voltage is your actual rail voltage (and the setting for your regulated power supply), the "-4" accounts for the approximate voltage dropped by the mosfet in the circuit, and R is the value of the resistor. Adjust the output of your test equipment and the resistor in your test jig accordingly and measure at whatever bias point you need.

IRFP044 vs IRFP240 vs IRFP244 vs IRFP250 : A number of amps have been built using each of these transistors. Nelson has characterized the differences among the three latter transistors as follows: The 250's have greater current capacity, but twice the capacitance of the 240's and 244's. They will give a better bottom end, and will sound a bit different, for better or worse, depending on your situation and taste. There is very little practical difference between the 240's and 244's. Some have reported that the IRFP044 sounds a little "darker" (more narrow frequency bandwidth and greater high-frequency roll off) than the 240, though this is not really surprising as the 044 has greater capacitance than the 240 series. Essentially, greater capacitance provides slightly better bass, but at the cost of slightly reduced treble. The practical difference, though, is their current capacity. The maximum current and operating temperature ratings have lead many to use the IRFP240/244 for rail voltages of 20v or higher, while using the IRFP044 for 15-20v rails. The lower voltage versions of the amp tend to feature higher current, and the 044's can handle significantly more heat.

Device
Max Voltage
Max Current
Max Temp
IRFP044
60v
57A
175c
IRFP240
200v
20A
150c
IRFP244
250v
15A
150c

Ian posted some of his impressions of using IRFP044 vs IRFP244s in his Aleph-X amps:

After using the IRFP044s, I then changed over one channel to the IRFP240's with the same power supply conditions (Grey's original schematic). Nelson's comments about gain devices effecting the sound were borne out here. In comparisons, the IRFP240 appears to have a somewhat thinner presentation and more brilliant high frequency response. The IRFP044 is richer sounding with more body in the midrange but otherwise less linear .....The jury is still out but I tend to think that while using only one pair per side of each, neither is better than the other and the IRF240 perhaps offers more latitude to use in various numbers for the right tonal balance. I doubt if I would use the IFRP044 for a tweeter amp in a bi amp setup and more likely the midrange. The entire thread can be found here.

Matching Source Resistors

Source resistors in this circuit are R5, R6, R40, and R41 and typically have values of less than 1 ohm - the ones I'm using are 0.333 ohms each (achieved by using three 1 ohm resistors in parallel). Most readily available $10-30 Digital Multi Meters (DMM) are really not well suited to accurately measuring low value resistors (typically, under 5 to 10 ohms or so). There are two ways to accomplish this task: 1) Use a regulated bench power supply to pass a small voltage and current through the resistors and then measuring the voltage drop across the resistors. This requires fairly precise (read: expensive) equipment. Since voltage drop is directly related to its resistance (through Ohm's law), matching the resistors according to voltage drop is essentially the same as matching them by resistance - but is much more precise. 2) An easier method is to put a larger (and known value) resistor in series with the one you are trying to measure. Choose one that is 8-10 ohms and measure it. This way, you are measuring the resistance of the entire apparatus (meter, known resistor, resistance of the wire probes), then just add in the new resistor and make a note of the reading. Either way, what is most important here is that they match in value since they will be placed in parallel and directly control the bias level for the output mosfets.

To match mine, I used a regulated bench power supply set to 0.1V. With this setup, each set of three paralleled resistors drew approximately 0.232A. Just use the alligator clips from your power supply and connect them across the bundle of resistors. Then, connect the leads from your volt meter across the resistors too. Depending upon your equipment, you may need to wait several seconds for the voltage reading to stop fluctuating before taking the reading. Be sure to "wiggle" the alligator clips back and forth on the resistors leads a little to make sure they are making good contact before taking your reading. On the meter I was using, the forth decimal place was in constant fluctuation even after a few seconds, so I rounded and only used three decimal accuracy in determining the resistance values of each resistor bundle (3x1ohm). To determine the precise resistance, just divide the voltage by the current. This method easily provides resistance accuracy to 3 decimal places and allowed matching of the resistor bundles to 0.3% accuracy. I now have several sets of resistors that measure 0.333 ohm and several sets that measure 0.332 ohm.

Matching Other Resistors

Here is the place where your standard $10-20 DMM will come in handy. Matching resistors with values that exceed about 10 ohms is a task that your typical DMM is well suited for. Measure them one by one, lay them down where they won't be disturbed and make a list of the readings. When you're done, just go through your list and pick the resistors that match. It may be helpful to re-measure to make sure you picked the right ones. After confirming their measurements, just poke them into your foam block, and write their value on the tape underneath the resistor. With my three digit LCR meter, I was routinely able to match all of my resistors to tolerances of 0.1% to 0.01%. Most of them matched in the range of 0.05% to 0.01%.. Having all of the matched resistors readily available in the foam blocks helped enormously while I was stuffing the boards. I could pull the resistors I needed and double check their values against the parts list and schematic before I put them in the circuit.

Stuffing the Circuit Boards:

As an interesting point of trivia, I think approximately 750 of these boards were produced as the result of the original group buy that took place in December of 2002. Over the years, I'm assuming the Peter Daniel's group buys have produced another several hundred Aleph JX boards. I wonder how many of these boards have actually been populated and turned into working amps...

Some time ago, I began purchasing the parts necessary to populate the circuit boards I have for the Left, Center, and Right amplifiers for my new basement home theater. When your parts arrive, be sure to measure each and every resistor and capacitor that you have received to check for errors! Several times in the past I have received resistors of different value than what I had ordered. For example, I had ordered 220 ohm resistors and received 220k ohm resistors instead!

In addition to matching the components as described above, I also took one extra step before soldering anything in place: I used a small piece of 0000 steel wool to polish the leads. Now, you may think its not worth the bother, but have a look at the left most picture below. The resistor on bottom has had its lead cleaned while the resistor on top has not. The difference does not show up in the picture as much as it does holding it in your hand, but the difference can be both seen and felt as you work with the part. I'm not sure it will make a realistic difference, but intuition seems to indicate that a cleaner lead will result in a better physical and electrical solder joint. The center photo below is the Panasonic 220uF FC series cap bypassed by a Wima 10nF FKP2 polypropylene film capacitor. The piggybacked assembly was then inserted into the circuit board. Finally, the right-most picture below is the 100K ohm multi turn potentiometer (V1 and V3). These are great little pots, they vary continuously from about 1 ohm to 100K ohms. Turned fully clockwise while looking at the printing on the front sets it to about 1 ohm, while each full 360 degree turn counter clockwise increases the resistance by 4K ohms. For the 200 ohm pot (V2), each 360 turn changes the resistance by 10 ohms. While they allow for fairly high-precision adjustment, several people have indicated that multi-turn pots may be less robust than single turn pots as they are more susceptible to changing their value over time or when bumped or dropped.

I installed every component directly on the PCB, with the exception of Q1, Q2, Q10, Q11 (all of the output mosfets), R5, R6, R40, R41 (source resistors for each mosfet), and R7, R9, R36, R38 (gate resistors for each mosfet). These resistors were directly attached to each mosfet and tied together as shown below in the section on chassis building.

The pictures below are of the actual PCB from the 2003 group buy. The associated schematic for this PCB is in the middle. After spending several evenings measuring and matching my parts, it took me about 3-4 hours to stuff each board. Looking at the completed board, you'll also notice that not all of the spaces are filled. If you build the stock ~38w amp, all of your components will fit neatly onto the board and there will be no empty spaces. However, because I'm looking for higher power output and using paralleled output devices, they will be mounted directly to the heatsink. All of the source and gate resistors will be attached to the mosfets in a point-to-point fashion, and then tied back to the PCB. The same is also true for the output resistors that attach to the source of the mosfets.

Mounting the Input Differential (Q5 & Q7):

With my first prototype of this amp, I took extra steps to mount the differential pair (Q5 & Q7) on the main heatsink by the output mosfets. After playing with the behavior of this amp for several weeks, I was not convinced that this step was necessary at all. The operating temperature of these transistors doesn't seem to matter very much so long as Q5 and Q7 track at the same temperature. So, I just mounted them back-to-back on the PCB with a sil-pad between them and used plastic shoulders to keep them electrically isolated from one another. Alternatively, you can use a sil-pad and a plastic screw to mount them back-to-back while keeping them electrically isolated from one another. It's easiest to attach them to one another prior to inserting them into the circuit board. After you attach them, be sure to check for electrical isolation with your ohm meter - the mounting tabs should be isolated from one another. In the prototype, I mounted Q6 externally (it just floated on some wires). With increased ventilation through the new chassis, I think it will perform just as well inside the chassis just mounted to the PCB, so that's where it lives now. PCB mounting of the differential pair reduces the amount of additional wiring that is necessary and does not change the behavior characteristics of the completed amp at all.

Installing Jumpers:

This particular circuit board has a great deal of flexibility built into it in order to accommodate different needs and philosophies of building. I am simply copying someone else's parts layout, so I installed a wire jumper at the following locations: J1a, Q12a (jumper the outer pins together), R48, and R49.

Making PCB External Connections with the "Ext" Pads:

Grey's Original Version If you have opted to build Grey's original version of this amp (15v rails, 4.5A bias, ~40w output), then the only external connections that need to be made to the circuit board include the power supply rails, signal input, and speaker output. For this design, the circuit board will hold all of the necessary components (except the power supply). No other external connections are necessary.

High Powered Version If you have opted for increased power output (using additional output MOSFETs connected in parallel), you will need to use the EXT pads on the PCB (each EXT pad is individually labeled in the left-most PCB image above). The discussion below uses these labels. Click here for a PDF document that illustrates how to wire the MOSFETs in parallel and connect them to the EXT pads. If you need to parallel more output devices, just extend the wiring for each Q shown in the document.

Ext 1a/1b This is the positive rail (+V) and connects to Drain of both Q1 and Q10. Use heavy gauge wire here (12ga or 14ga) because it carries a great deal of current.
Ext 2a/2b This connection is for the Mosfet Gate (input signal). Connect Q1 to Ext 2a and Q10 to Ext 2b.
Ext 3a/3b

Use Ext 3a for the Source of Q1 and the Drain of Q2. Use Ext 3b for the Source of Q10 and the Drain for Q11. Use heavy gauge wire here also. This eventually becomes the speaker output.

Ext 4a/4b Use Ext 4a for the Gate connection of Q2 and Ext 4b for the Gate connection of Q11.
Ext 5a/5b This is the negative rail (-v) and connects to the Source of both Q2 and Q11. Use heavy gauge wire here.
Ext 6a/6b These are for the output resistors for the speaker terminals. Depending upon the number of resistors you used here to achieve the correct value, you can either put your resistors directly on the board (and thus you don't need to use these Ext pads), or use these connectors and put your resistors somewhere else and connect them across Ext3a/3b and Ext 6a/6b.
Ext 7a/7b These are for connecting the output to ground. Depending upon the number of resistors you used here to achieve the correct value, you can either put your resistors directly on the board (and thus you don't need to use these Ext pads), or use these connectors and put your resistors somewhere else.

 

Chassis Construction:

You can build a chassis any number of ways. There are a few places online that have a variety of ready made sizes. Par-Metal is one, but they run a little on the high side for pricing. Your best bet is simply to find a local sheet metal, steel, or aluminum provider and see if you can scrounge some aluminum scrap or left overs at the end of the day. Any cutting of scraps can be done on a table saw with a blade intended for non-ferrous metal and some cutting oil (WD-40 works well). Depending on how much volume they do, this may or may not work if you are looking for really cheap scraps (most places charge a nominal fee per pound). The items in the scrap bin at my local metal shop were all too small to be useful, so I ordered enough aluminum to build three chassis for under $100 - the benefit here is that they did all of the cutting for me (using a hydraulic shear). The top, bottom, and rear panels are all 1/8" thick aluminum sheet and the front plate is 1/4". I used the diagram on the left below as a rough assembly guide for my chassis.

Another great chassis approach is documented here: Another Aleph-X up and Running by Xavier1000.

The first thing I did when I received my aluminum order was to spend some time filing all of the edges (metal shears can leave very sharp edges and corners) and washing off all of the grease since they would be handled frequently. Since I am building multiple chassis, I sourced all of my screws, washers, nuts, taps, etc from McMaster-Carr - they are significantly less expensive than the local hardware store that sells screws individually. For those of you who are more industrious and have access to a machine shop, the image on the right was created byAR2 on DIYAudio.com and is just exceptional work by any standard! This is one of the nicest DIY chassis I've seen in a long time and is the result of a great amount of time, effort, and care. For construction details as depicted below on the left, be sure to see Brian's thread on Chassis Construction.

Preparation of Mosfet Mounting Locations:

Before you begin joining multiple heatsinks side by side (if you need to do this step at all), I would recommend preparing the sites on the sink where you plan to mount your power mosfets. Mosfets should be centered on your sinks as much as possible, both top to bottom and side to side. If you are mounting more than one mosfet on any single sink, its probably best to leave a bit of space between them so there is roughly even space on either side of each mosfet and space between the mosfets and the edges of your sink. This will help insure even distribution of heat across the sink which helps keep your carefully matched mosfets at approximately the same temperature.

I recommend drilling these holes with a drill press so you get holes that are exactly perpendicular to the surface of the heatsink. A crooked hole will result in uneven mounting pressure or a gap on one side of the mosfet which will surely lead to premature part failure. If the back plate of your heatsink is sufficiently thick (approaching 1/2") you can drill and tap a hole so you only need a screw, or you can drill the hole all of the way through and use a nut and washer on the other end (this is what I did). If drilling clear-thru holes, make sure you carefully align your drill so your resulting hole is evenly spaced between the fins of the heat sink. When you have drilled your holes, be sure to use a larger drill bit to remove the raised bur from your new hole. This step has already been performed in the image on the left below. Rod Elliott's web page concerning heatsinks is required reading on this topic. The holes have already been counter-sunk with a larger drill bit and very gentle pressure on the drill press. The two holes in the bottom right were originally planned for mounting of the input differential pair, but it turned out that it was not necessary for them to be mounted there.

After drilling your holes, I would recommend a careful inspection of the site where you plan to mount your mosfets. If the back of the heatsink is ABSOLUTELY flat AND smooth (rather rare and only accompanies heatsinks that were fairly expensive to source...), then leave it alone. If, however, you can feel "ridges" on the back of the sink with your finger nail (these are remnants of a lesser-expensive aluminum extrusion process), you need to fix this before your mount your mosfets. The easiest "fix" is to use a small, flat piece of aluminum (I used a 2" length of the 1/2" bar stock) as a sanding block wrapped with 300 or 400 grit sandpaper (the black wet-dry type) and a drop of oil. You can finish off by repeating this with 600 grit sandpaper. In just 3 or 4 minutes, you can have a VERY smooth mounting area for your mosfets. When I sanded my spots, the black anodize was removed from the sink so I was left with bare aluminum. See the right most image above. Once you have prepared your mounting spots, just leave them alone for a little while. Do not mount your output devices until you have joined your heatsinks together (if you need to do this at all) because they'll just get in the way and you risk damaging them while you work on the rest of the chassis.

Joining Heatsinks Side-By-Side:

Since my heat sinks were only about 6" wide, I needed to join three of them side-by-side in order to form one side of the completed chassis. To do this, you'll need your heatsinks, a few scraps of aluminum for clamping and spacers, as well as at least three clamps: two medium size C-clamps (capable of opening 3-4") and a larger clamp with capacity to open wide enough to span the width of your heatsinks. Things also tend to go more smoothly if you have a drill press. Even a small table-mounted drill press will work just fine - this is exactly what I used. Use one of the clamps across the two heatsinks in order to hold them together while you position the 1/2" square aluminum bar stock across the top edge of the sinks. If you are joining more than two sinks together for the side of your chassis, just work on two at a time. So, if you have 3 sinks to join, join the first two, then go back and join the third. My original prototype was configured with 2 sinks per side (total of 4 per chassis), but it got hotter than I wanted with a thermal rise of 52c. Thus, my remaining amps were built with 3 sinks per side (total of 6 per chassis - which keeps thermal rise at about 32c). This is why the last image on the right has a "Frankenstein's Monster" look to it for adding the third heat sink - I didn't want to undo the work I did to join the first two sinks together, to the third one was just "tacked on." So much for planning ahead. This also illustrates the value of building a prototype...

I used a small scrap of 1/8" aluminum as a spacer along the top edge of the sink so that when the top plate was mounted to the chassis, it would sit flush with the top of the heatsink (see the third image below). Use two C-clamps to clamp the bar stock in place, taking care to use another scrap of aluminum on the fin-side of the heatsink so you don't bend up the fins with the clamp. Then just drill your holes through the bar stock and the back of the heatsink, taking care to position your hole so the drill bit emerges centered between the fins of your heatsink. If you end up a little off-center, its OK, you can just file a flat spot on the head of your screw to help it fit closer to the heatsink fin (I used my Dremel tool with a grinding wheel on the screw head - see the right-most image below). I made sure that I had at least two screws in each heatsink to keep things from shifting around while I worked. For assembly, I used 25mm M4 size metric screws with socket heads, spring-style lock washers, and finished things off with an M4 sized nut. Doing this again, I think I might choose to use 3/4" square bar stock and something a little heavier than M4 screws, but what I used worked out fine. I certainly would not use anything smaller than M4, though. I don't know that smaller screws would have the strength that you will need. Some people advocate using thermal paste between the heatsinks and the bar stock to help spread the head more evenly across the chassis. I didn't bother with this step and temperature measurements of the completed amp don't reveal any "cold" spots on the chassis at all. One completed side of the amp is shown in the image on the right. Three sinks are joined together and the output mosfets are installed, labelled, and the copper wiring (12 ga) is for the positive and negative power supply.

Mounting Mosfets:

In case you missed it above, Rod Elliott's web page concerning heatsinks is required reading on this topic. Be sure to read the entire page! Yes, its long - read it anyhow! Once your heatsinks are joined, mounting holes are drilled, and you've smoothed the mounting area - you are ready to mount your mosfets. There are a number of ways to achieve this. Your first choice is to use mica insulators and heatsink compound "goop" or Sil-Pads. Many people claim that mica and goop is a better (thermally conductive) approach. I used Bergquist 1500ST which is extremely thin (0.2mm) and very thermally conductive (0.22 c/w) and electrically insulative. It is critical that the metal backing on the Mosfet be electrically isolated from the mounting surface of the heatsink - otherwise, your mosfets will short when mounted to the electrically grounded chassis. In the image below, you can see the blue sil-pad, the mosfet, a large washer to spread the mounting force across the entire mosfet, a lock washer hidden under the nut, and the tip of the screw. Don't over-tighten this screw/nut! I held the nut with a pair of needle nose pliers and using the short end of my Allen wrench, tightened the screw by hand only. This provided enough pressure to get good thermal contact, holds everything in place, but doesn't risk cracking the transistor. As a final check after mounting each mosfet is to get your ohm meter and check each of the three pins for electrical continuity with the heatsink. You want to make sure that each of the three pins are isolated from the heatsink. If you find any connectivity at all, you've done something incorrectly and need to dismount the mosfet to investigate what went wrong. At this point, you may not be able to reuse your Sil-Pad. The Sil-Pads are available in specific sizes for different transistor package types, or as 4"x4" sheets. I got a 4"x4" sheet that I cut into sizes appropriate for my needs. The only thing missing was the mounting hole, but this was easy enough to add myself with a leather punch tool (see image on the right).

Chassis Bottom Plate:

I started by taping a piece of paper to my bottom plate and just moving things around until they fit well. Items mounted to my bottom plate include the transformer, the power supply caps (with mounting clamps), and the power resistor for the CRC power supply (rectifiers were mounted to the heatsinks with electrically conductive Sil-Pads). Once you have a physical configuration that you like, just trace the outline of each item with a pen. Be sure to leave room on the sides of the bottom plate where your metal bar stock (that attaches to the heatsinks) will overlap. I then covered this template with wide packing tape so I could use it for each of my amps without having it get destroyed part way through the job. I drilled three different size holes - I used a 5/32" for clearance for the M4 size screws that would hold all of the caps to the bottom plate. I used 3/8" holes around the perimeter of the transformer and in a few other places to provide some air flow, and I used 1.5" circular hole saw for major ventilation holes.

I then drilled out the bottom plate of my chassis using the paper template. The smaller holes were drilled on my drill press (its much easier to put a hole in metal where you actually want it with a drill press than drilling freehand as the freehand drill bit tends to walk around a little before digging in) and the larger 1.5" diameter holes were drilled with a hole saw in my handheld drill (using some WD-40 as a lubricant). For the smaller holes, I used a 1/4" drill bit on the backside of the aluminum to remove the raised up lip created by drilling. Go gently here, you only want to remove the lip, not really drill a larger hole! For the larger 3/8" holes, a 2" wide putty knife worked well to just hack off the raised lip around the holes.

Next, I cut a small scrap of aluminum to use as a shield between the transformer and the PCB (see the second image from the left). I just placed it where I wanted, held it in place with my foot, and drilled through both the shield and the bottom plate at the same time. I then used some threaded rod and a few nuts from the hardware store as the standoff for the shield. The threaded rod comes in 36" lengths and I cut it down with a hacksaw. Exact alignment of the holes is not critical since you are working with 4"-6" rods that will have some degree of flex to them. The legs for the amp are nothing more than some 3" long 3/8" bolts. This provides enough ground clearance to let the heatsinks get some actual airflow under the fins and to allow air under the chassis to help with cooling. In the next few images above, you can see the power supply caps and transformer already in place with the threaded rod standoffs. Next, the PCB mounting plate is installed. You can see the short screws that will hold the PCB above the aluminum shield. Be sure to test-fit the PCB standoffs before you spend time tightening down the shield. Finally, you can see the entire bottom plate of the amp assembled in the right-most image above.

One final note - don't mount your bridge rectifiers to the bottom plate of the chassis as you see in the right-most image above. This does not provide adequate cooling for them and they will fail - mine did when I left the amps on for 24 hours to break in my new speakers... Instead, mount them to the heatsinks as indicated below.

Joining Chassis Bottom to Heatsinks:

To begin, I turned the heatsinks upside down and put the entire bottom plate assembly into place. I found that doing this with the transformer attached to the bottom plate added a degree of stability due to its weight. This kept things from shifting around too much while I worked. The key to the next step is to use a piece of masking tape to mark where you want to drill your holes to mount the bottom plate to the heatsinks. Since my bar stock is 1/2" wide, I drew a line 1/4" in from from the edge along the entire length of the bottom plate. I then separated the parts a bit so I could see where I already had screws joining the bar stock to the heatsinks - don't drill through the screws you've already installed. I tried to concentrate the new mounting holes around the transformer, since this alone weighs in near 24 lbs and needs to have a sturdy connection. In hindsight, I probably should have gone for 1/4" aluminum for the top and bottom plate and 3/8" for the face plate for sturdier construction. These holes were drilled free-hand being sure to hold the drill vertical so you end up with holes that aren't crooked. To join the bottom of the chassis to the heatsinks, I found it useful to use a set of clamps, as the joined set of heatsinks is not perfectly flat across the back (though it was very close). The holes toward the rear of the chassis were tapped so I could just insert my screws, whereas the holes toward the front of the chassis were drilled clear through so I could put a washer and nut on the other end. Each is really equally easy (the time it takes to tap a hole is about the same as the time it takes to squeeze your fingers or needle nose pliers down the side of your chassis and get both a washer and a nut onto the screw you just inserted - with all of the time I spent dropping them along the way), so its your choice. My choice was dictated by the addition of a third heatsink to each side (my first prototype amp ran rather hot!) so the screws I had on hand were not long enough to reach through two thicknesses of bar stock

Tapping Holes in Aluminum:

Tapping holes is not nearly as difficult or as onerous a task as some people might have you believe. The entire process only takes 2-3 minutes per hole start-to-finish and is really quite easy. It is essential to use some form of oil to lubricate the hole, otherwise you risk breaking off your tap while its stuck in the hole you drilled - then you are faced with the near-impossible task of extracting the broken tap). Fortunately, this hasn't happened to me. Just put a drop of oil on the tip of your tap and start turning the tap by hand. Go slowly and evenly. The tap will "bite" right away and should produce a little corkscrew of metal coming out of your hole after a few turns. When it starts to bind up and getting tight (you'll know it when it happens), carefully back the tap out and clean it off with an old toothbrush. Clean the hole of any remaining debris with a pipe cleaner (I folded mine making a double thickness for more effective cleaning), re-lubricate the tap, and start threading the tap back into the hold. You should be able to tap more deeply than you did the first time. When it binds up again, remove the tap, clean the hole, and you should be done. Its really just a pretty quick process. Just work gently so you don't break your tap! Speaking of the taps themselves, carbon steel taps are a no-no as they generally snap off easily no matter what... use high speed steel taps only! Use either spiral fluted or two flute taps makes the chore ALOT easier. See the Tapmatic website for more detailed information on types of taps.

You can see the result in the right-most image above: the power supply and heatsinks are joined together and ready for the face plate and the rear plate.

The rectifiers for each side of the amp are mounted to the lower portion of the heatsink, just below the output stage. In my original prototype, I mounted the bridge rectifiiers to the bottom plate of the chassis and after about 24hours of use, the rectifiers failed due to extreme temperatures. Live and learn...

Adding the Face Plate and Rear Plate:

Add the face plate to the amp before attaching the rear plate. The rear plate will have all kinds of things sticking out of it (power switch, fuse mount, speaker binding posts, etc) that prevent you from standing the amp on its end to attached the face plate if you put the rear plate on first. To add the faceplate, I found it useful to at least fit the top plate of the amp between the heatsinks where it belongs and then to mark the location of the mounting holes with masking tape and a pen. Take care to properly measure the distances in from the edges of your plate so you hit the middle of the bar stock you attached to the front of the heatsink. If you're off, you may have a hard time fitting your nut on the backside, or risk tear-out if you are too close to the edge. Once you mark the position of the holes, I used a punch and a hammer to put a dimple in the aluminum sheet. This keeps the drill bit from wandering while you start the hole. Use a small drill bit first, as it will clear your hole more quickly. Then use a larger diameter drill bit next. Because positioning these holes is a little more critical, you may want to only drill one hole at a time and put your screw in place before drilling the next hole. This keeps things from shifting while you work and eliminates unpleasant surprises, too. I used heavier screws here, probably M8 or so with large heads. Once you have 2 screws in place, the last two holes are easy.

Next, do the same thing with your rear plate. Stand the amp on its face and mark the position of the holes with masking tape. The position of these holes is a bit more critical since you are now trying to hit the 1/2" by 1/2" square end of your aluminum bar stock. Just as before, drill one hole at a time, tap it, clean it, and insert the screw. Then do your second hole. Thread it, clean it. Repeat for the 3rd and 4th screws.

For the top plate of your amp, its exactly the same thing. Mark your holes with masking tape, be sure not to hit any of the screws that are already in place, drill, tap, clean and you're done (almost).

Finishing the Face Plate/Top Plate:

One relatively easy way to clean up the appearance of your face plate and/or top plate of your amplifier is to create a brushed aluminum look. You can do this with some wet-dry sandpaper, a sanding block, a straight edge, and a little bit of WD-40. My straight edge was a scrap of 2x4 lumber left over from a previous project. Be sure to put something down under your work area as the sandpaper and oil will make lots of black oily aluminum slag that will dribble and splatter as you work (and its VERY difficult to remove from the carpet!). Put your aluminum plate down on a sturdy work surface, align your straight edge with the direction of your "brush strokes," add a few drops of oil, and begin sanding away. Be careful to keep your sanding block pressed tightly against your straight edge as the grooves it cuts will be obvious if they are not straight. If your aluminum has somewhat deep gouges or scratches, start off with 220 grit paper to remove the heavy scratches. Then work your way up through 300, 400, and finally 600 grit paper. Sand for a little while, move your straight edge over a few inches and sand some more. After some effort (depending on how badly scratched your aluminum is), you'll have sanded the entire surface. Carry it over to the sink for some soap and water clean up and see what it looks like. Scratches that didn't come out will be obvious and need more work. After a while, you'll have a nicely finished sheet of aluminum.

After drilling the mounting holes, I had someone at the projects lab do some custom CNC engraving for me. I provided an image of what I wanted things to look like and he worked up in the CAD software and let me preview it. We made only a minor tweak or two - a tribute to his talent, and then he engraved them. The image on the right is after the LED hole was drilled and a bit more sanding (600 grit paper) was done. Before the final sanding, I filled the engraved areas of the plate with black enamel paint (left over stuff from my Avro speaker project). The final result looks great!

The top of the chassis received similar treatment to the front and rear panels. I had a large rectangular hole cut in the top for ventilation. It was then sanded in a manner similar to the front place and then painted. Finally, the ventilation hole in the top was covered by a length of gutter topper (mounted from the inside) to keep objects and little fingers out. The best part is that the gutter topper was just a few $$ from the local hardware store! Alternative finishes include sand blasting or powder coating. While sand blasting can be done at home with the proper tools, powder coating usually isn't so easy to do on your own. Any of these methods will yield a chassis with a nicer appearance than a randomly scratched sheet of aluminum on top of it.

Rear Panel & Power Entry Wiring:

Pay particular attention to the power entry for your amp and the grounding of your power supply. Make sure your EIC power inlet (image below) is grounded immediately to the chassis where the AC power enters. Depending on the type of EIC socket you have (mine is metal, but some are plastic), you want the AC Ground to immediately connect to the metal chassis with a screw and a round eyelet wire terminator that is crimped onto the wire. This grounds the chassis to AC Mains ground and prevents the metal chassis from ever carrying current in the event of component failure.

The AC Neutral gets wired directly to the primary of your transformer via the white and orange leads (neutral on all transformers should be white, transformers with dual primaries have white and orange for neutral). The AC Hot goes first to a 7A slow-blow fuse, then to a heavy-duty switch (10A 240VAC) that is bypassed with a 0.01uF 600v cap to prevent arcing and pitting of the switch contacts, through a pair of CL-60 thermistors and to the black lead of your transformer. If you have a large transformer with two primary windings, run another thermistor to the brown lead (one thermistor per primary of the transformer). The thermistors will limit the inrush surge as the caps begin to charge when you turn on the amp (each monoblock has 580,000uF of capacitance - over half a Farad!). The reason for using two thermistors is to keep the current through each one relatively low as the CL-60s are limited to 5A steady state. My completed amp draws almost 3.9A from the wall socket, so were getting pretty close to their design limits which limits their lifespan. The thermistors will reduce your power supply rails by about 0.2vDC. Be sure that the primary wires on your transformer are twisted tightly together. The same goes for the secondary wires. This helps prevent AC noise from bleeding into every other circuit in the chassis due to the high currents and likelihood for stray magnetic fields. On the Primary side of the transformer, the white & black wires should be twisted together, as should the orange and brown wires. These colors are industry standards for transformer primaries.

Signal Input Wiring and Final Chassis Assembly:

Using RCA / XLR Inputs:

The signal pin out from the RCA jack is: Center pin = Positive Signal, Outer collar = Shield/Ground.
The signal pin out from the XLR jack is: Pin 1 = Shield/Ground, Pin 2 = Positive Input Signal, Pin 3 = Negative Input Signal. The pins should be labelled on the actual XLR jack itself. My thanks to William for providing the following diagram:

If you want to use both an RCA and XLR jack on your amp for flexibility, wire the center pin of the RCA jack to Pin 2 (positive input signal) of the XLR, then through a 4.7uF cap, then to the +IN on the circuit board. Wire Pin 3 (negative input signal) of the XLR jack through another 4.7uF cap, then to the -IN on the circuit board. Wire the RCA outer collar to Pin 1 (shield/ground) of the XLR, then to the Shield on the circuit board. Finally, install a switch that shorts Pin 1 (shield/ground) and Pin 3 (negative input signal) of the XLR jack before the input caps. When you use a balanced XLR input, keep the switch open. When you use an unbalanced RCA input, close the switch (this will provide 6dB of gain for the amp). These two capacitors will prevent any DC signal from going into your amp and then being amplified and sent to your speakers. Speakers + DC currents usually = burned voice coils. Depending on the value of R46 and R47 that you chose, you are likely to see DC voltage across your speaker terminals as high as 10v when you first power up your amp if you don't include these caps. You may want to choose a high-quality capacitor here since the input signal must pass directly through this cap. Screwing up your input signal before it gets to your amp is a sure fire path to disappointing results. An interesting read is Tony Gee's Capacitor Comparison web page.

My original construction used Dayton Metalized Polyprolyene capacitor because they are cheap and I had them on hand (though some claim that metalized film will result in restricted bass extension). For better sonic performance, you can step up to the big stuff like Intertechnik Audyn Cap True Copper MKP, Mundorf Silver/Gold/Oil or Clarity Cap MR MKP. Essentially, what you are doing with the cap is creating a first order (6dB per octave) RC high-pass filter at about 2-3Hz (a high-pass filter lets frequencies above the established threshold pass through the filter). The corner frequency (-3dB point) of this filter is given as f = 1 / (2 *Pi * R * C). For example, consider what happens if you use an 1.5uF input cap with an amp that has a 38k ohm input impedance (such as my Aleph-X amp):

f = 1 / (2 * 3.14 * 38,000 * 0.0000015)
f = 1 / 0.358
f = 2.8 Hz

The most important effect of the cap is that it prevents any DC from making its way into your amp, though this is not the only effect. An RC high-pass filter also creates a phase-shift in the input signal that is a function of the corner frequency derived above. The general rule is that the greatest magnitude of phase shift occurs within a factor of 0.1 to 10 times the corner frequency. Signals less than 0.1 times the cutoff frequency are shifted close to 90 degrees while signals more than 10 times the cutoff frequency show little or no appreciable phase shift. This is what dictates the necessity of a somewhat larger value capacitor for the input cap than you would expect. For example, if we choose a 1.5uF input cap for an amp with an input impedance of 38,700 Ohms, our corner frequency works out to 2.8Hz as shown above. The trouble is that there will be appreciable phase shift up to about 28Hz with this value for an input cap. Thus, for this example, we are better off choosing a larger cap - something in the range of 2.2uF to 4.7uF. A 2.2uF cap will provide a corner frequency of 1.87Hz and a phase shift that is all but gone by about 18Hz. With high quailty film and foil caps, though, this small change in value translates into significant price differences. As a final note, when using input caps with balanced inputs, it is important to use the same cap value on each leg of your balanced input and across multiple amps. Consistency is the key. A more detailed read about corner frequencies and phase shift can be found here.

If you are using the PCB that I have pictured here on this page, don't worry about the input resistor to ground or grounding the collar of the RCA jack, as this has already been built into the circuit board. Do provide electrical isolation between the RCA collar and the metal chassis, though (this maintains only a single grounding point which prevents ground loops and other potential weird stuff).

 

Question: I can adjust the bias nicely and can also adjust the relative DC offset to zero. BUT as soon as I connect the negative XLR input to ground (For RCA input) I get a relative DC input of 4V! Absolute DC offset is 6V with XLR and 8V!!! negative XLR input grounded (for RCA input). What did I wrong?

Answer: I had the same problems myself using single ended inputs. Whatever I tried the dc offset at the output would be always very high and always drifting- very difficult to adjust. With balanced input there was never a real problem after adjustment and warm up things were quite stable despite initial high dc. I don´t know how Pass labs manage this in their amps.
When you connect one input to ground the absolute dc adjustment is feeding back to the relative adjustment and is almost impossible to adjust as it will slowly drift away. It is as frustrating as keeping water in your cupped hands. It will never stay .I think Nelson suggested lowering the output resistors to ground.
- from Protos on 10/30/2007

Final Assembly:

I can't believe the quantity of screws, washers, and nuts that I have gone through already! I placed a $100 order with McMaster and I think I will need more before I am finished! Keep in mind, though, I'm building three identical large chasses at the same time. Where possible, I drilled holes all of the way through and put lock washers and nuts on the other end of the screws to minimize the amount of hole tapping that I needed to do. The tapped holes I made hold the back panel to the ends of the bars the run the length of the heat sinks, the top plate, and part of the bottom panel (where the aluminum bar stock was doubled to extend the heat sinks).

Above are images of the front, top, and side of the amp. I made three identical amps in all, one to power each of the left, center, and right speakers in the front array. Together, they make very effective room heaters for the winter months!

Now that your amp is together and complete, you are only half of the way there! its time to open things up again and begin the tweaking phase after making some measurements! For some of this, you can just remove the top plate so you have easy access to measurement points. But when it comes time for final bias and offset adjustments, you need to perform these steps with the amp fully warmed up and the top plate firmly in place. For me, this meant carefully attaching several sets of alligator clips and running the wires out the bottom of the amp through my ventilation holes for measurements.

The First Power-up & Measurements:

I was simultaneously excited and hesitant to flip that switch for the first time. I went over everything at least three times looking for errors. And then I checked it all again... Then I let it sit for 3 or 4 more days working up the nerve to flip the power switch for the first time... Before I turned it on, I re-checked all of the wiring again... then again...

You might also want to do yourself a favor at this point. Find a new notebook, put the date at the top of the page, and keep a clear record of what you measure, what you change, etc. I didn't do this when I built my amp. Many of my notes are here on this page, but I've made so many tweaks and changes over time, I'm starting to lose track of things. Looking back on the past few years of construction changes, I'm wishing I had a more complete and accurate record of what I did, when, and what the results were...

Procedure:

Power-up for the first time should be done (slowly) with a Variac (basically a large variable transformer that plugs into the wall outlet) to help prevent burning sets of matched output devices or causing other unexpected problems that result from wiring errors. Slowly doesn't mean to turn up the voltage into your amp until you hit 120v and something fries and smokes. This defeats the purpose. By slowly, I mean that you need to turn up the input voltage until you get something like 10v on your rails, then stop and do some measurements. If you don't have a variac, wire a 60-100w light bulb in series with the house AC outlet that powers the transformer primaries. Since I don't have a variac, I used the light bulb approach which provided 7-8v rails, not quite enough to get everything up and running, but close enough to check things out. After I flipped the switch for the first time (I double and triple checked ALL of my wiring several times - comparing chassis wiring to the schematic several times to be sure something wasn't screwed up), the bulb glowed brightly for a few seconds and then dimmed a bit once the power supply caps were charged. Think of the light bulb's intensity level as an current meter. Glowing brightly means more current is being drawn, glowing more dimly means less current is being drawn.

What to look for:

Before you power up, start by setting V1, V2, and V3 to their minimum settings (turn fully clockwise). V1 and V3 set the bias at its minimum and you get a good starting point with V2 (which controls DC offset across the speaker terminals - see further notes on these below). Using the light bulb method, my power supply rails were only running at 8v (instead of the targeted 22v). So far, so good. No crackling noises, no smoke to see, no burning smell, nothing overheating- all good signs!

Check relative DC offset (DMM across speaker terminals) - this should be zero volts if you used input caps as described above. Check absolute DC offset (one speaker terminal to chassis ground). You will probably find a few volts here (up to 10v or so) and it will likely be increasing or decreasing at some somewhat steady rate as the amp starts to warm up a little. If necessary, adjust VR2 until you get the DC voltage across the speaker terminals as close to zero as possible. Don't worry if it changes right away (any change should be small - on the order of a few tenths of a volt or so over time), its likely to continue changing until your amp reaches full operating temperature. Look for 11mA bias for your differential pair (measure approx 4v across R23 and R25, one will likely be a volt of so lower than the other, this is OK at this point), you should be able to measure about 0.1v or so across source resistors for Q1 and Q2. Q10 and Q11 may not have any voltage drop across their source resistors depending on your rail voltage. Turn the V1 and V3 pots to make sure that your bias and offset change a little bit (it may not change very much, though). After checking all of the above items, give it the acid test: Hook up an old test speaker (use an old one incase you fry it, this way, you wont care...) and give it an input signal - I used my daughters MP3 player. You should be able to get something that resembles music out of the amp, though with an underpowered amp, it will sound pretty rough. Once you have verified that things are working and your amp can indeed amplify an input signal, you're ready for a full power test.

After this confidence builder, I removed the light bulb and went directly for 120VAC from the wall outlet. After powering up again, I measured each bank of source resistors and verified that bias changed by adjusting the V1 and V3 trim pots. The heat sinks began to get mildly warm after a few minutes. Even better signs!At this point, adjust the bias point for something closer to 0.4VDC across each of your source resistors. Go back and re-adjust VR2 for zero relative DC offset. Keep an eye on the temperature of R44/45 and R1/4. If these are heating up (warm/hot to the touch), its because the relative DC offset is out of whack (high positive or negative voltage relative to your power supply rails), so go back and adjust VR2 until the voltage across the speaker terminals is close to zero again. Keep an eye on bias current and DC offset as it warms up.

DC offset looked pretty good at this point. I biased the amp so I can read a very consistent 0.500v across all source resistors. With the RCA input shorted (for testing purposes only), I'm getting 0.039v across the speaker terminals (relative DC offset). Absolute DC offset runs 0.812 on the positive speaker terminal (to ground) and 0.775v on the negative speaker terminal (to ground). Resistors R44/45 drop 0.785v and R1/4 drop 0.778v. I cannot measure any voltage drop at all across R2/3 or R42/43 (because there is no input signal). Heatsinks began heating up and topped out near 50c at the top (room temp is currently 17c). The output transistors inside the chassis seem to be running at 60c, while measuring the outside of the heatsink between the transistors yields 56c - looks like I have a pretty good thermal bond between the transistors and the sink!

Success! One Aleph-X up and running!


I'm not able to comment on sound quality right now as I'm running it from an MP3 player with an old crappy test speaker that has $20 worth of 25 year old drivers in it (and they're suffering from lots of foam rot). The amp does, however, play music! More testing and adjusting is necessary before I connect the amp to decent quality speakers. I need to gain some trust in the operation of the amp first.


Adjusting Bias and Offset (Trim Pots V1, V2, and V3):

Be sure to do all your final adjustments (see below) on a fully warmed-up amp. Let the amp warm up for an hour before making any critical measurements or adjustments. If in doubt, watch for drift in the DC parameters, and wait for them to settle. Be aware that merely opening the chassis can affect the bias points and offset measurements due to cool air getting in. So if possible, you should try and keep the chassis fully assembled while you do any final adjustments. One way to allow for this is to attach thin wires and run them out through a vent hole or something, then use them as remote probe points, or a place to put the temporary trim pots for the adjustment phase. Just be careful you don't let any of these dangling wire-ends short to each other! Keep them organized, separated, and secured in place with some tape.

Some notes from Peter Daniel about adjusting the trim pots: If you didn't make the enclosure yet and the amp is not in a permanent chassis, I wouldn't rush with replacing trim pots with fixed resistors. The chassis will change the thermal heat distribution inside the amp and I can almost guarantee that you will have to readjust everything again once the chassis is complete. This is also one reason that the amp wasn't very stable with DC offsets. When I lift the cover of my AlephX, DC can run 30mV one way or another in a first minute. That's why also, in my A75 amp I made holes in a chassis to adjust DC and bias with a thin screwdriver without having to remove the top panel.

Grataku also provided some useful advice. One thing I have learned from the prototype is that I am going to mount the board horizontally and on top to have ready access to the ccs trim pot. However, V2 is really the only thing that needs tweaking in situ to trim the absolute DC when everything is at the steady state. V1, 3 and the AC Current Gain seem to stay the same.

Trimpot V1: Used to adjust the bias level of "left side" of amp, directly biases Q3, which then affects current gain through Q1 and Q2 in the output stage. See "Setting Bias" and "Adjusting Relative DC Offset" below.

Trimpot V2: Used to adjust absolute DC Offset. Adjust only after setting V1 and V3. See below.

Trimpot V3: Used to adjust bias level of "right side" of amp, directly biases Q8, which then affects current gain through Q10 and Q11 in the output stage. See "Setting Bias" and "Adjusting Relative DC Offset" below.

Setting Bias with V1 and V3

V1 and V3 are used to set and/or fine tune the bias current. Set V1 and V3 before adjusting V2. In keeping with the traditional Aleph values, a good starting value is to set these so that you see a nominal 0.5V across the Source resistors (R5, R6, R40, and R41) of the output MOSFETs. Regardless of what you set it to, make sure both sides of the circuit match. The pots give you a fair amount of latitude in setting the bias. As a general rule, more bias current sounds better, but remember to keep an eye on the power being dissipated by your output devices; don't want to cook them. The first time you power up, begin with both V1 and V3 set to 0 ohms (full clockwise when looking at them from the front). This results in the lowest possible bias setting for your amp (approximately 4A for the 100w design). Then, slowly increase the bias by increasing the value of both pots (turn counterclockwise). Keep turning them up until you get approximately 0.5V across each of R5, R6, R40, and R41. It is best to set the bias and adjust the offset (discussed below) with the input shorted to ground (just use and RCA plug and solder a wire across the two terminals) and without a speaker or any other output load connected to the amp.

Keeping an Eye on Voltage, Power Dissipation, and Temperature:

Remember - Keep an eye on power dissipation of the Mosfets and the corresponding heat sink temperature! The one reality with Class-A amplifiers is that you can never have enough heat sinking! While these output mosfets are pretty robust devices and can handle a great deal of power, you need to balance power dissipation with their temperature. In the graph below (from the IRFP240 data sheet), you can see that we are operating these devices well within their comfort zone, and certainly well within their Safe Operating Area. My design has each mosfet running at a case temperature of about 55-60c while dissipating about 1.5A. Also, from the data sheet, we learn that the maximum power dissipation (Pd) with a case temperature (Tc) of 25c is 150w. Further specs indicate to derate the maximum power dissipation by 1.2w for every 1c above 25c. Using my own amp as a reference point, the case temperature for each transistor is below 60c. Using the derating factor, each transistor can dissipate a maximum of 108w at 60c. With 9.4A of bias on 22v rails, each of my transistors is dissipating about 35w, approximately 1/3 of the maximum power dissipation at this temperature. The overall conclusion is that my power and heat situation is well within acceptable limits for long-term reliability.

Calculating the Actual Bias Setting:

OK, so you've made the above adjustments by turning the trim pots and you've measured the voltage drop across the source resistors. So what is the actual bias point of your amplifier and how much power can it produce into an 8ohm load or into a 4 ohm load? Here is where William's spreadsheet is very useful in modeling the output behavior your amp (it automates many of the calculations). My experience indicates that the spreadsheet is "fairly" accurate, but following up on the test bench in the EE lab on campus with my speaker dummy load resistor and an oscilloscope reveals the 4ohm figures to be a pinch on the optimistic side (though not my any really meaningful amount). Using the spreadsheet, plug in your target voltage and bias levels into the spreadsheet, along with the number of mosfets you are using and it will project output power at 8ohms, 4 ohms, etc.

For those of you that want a better understanding of bias setting, read on: The source resistors in my output stage are dropping 0.500v (adjusted as indicated above) and they measure 0.333 ohms of resistance. Using ohms law (I=V/R), bias current = 0.500/0.333 = 1.5A per mosfet. With 3 mosfets per bank (each output Q in the schematic is one bank) this is 4.5A per each of Q1, Q2, Q10, and Q11. Thus, I'm drawing 9A per "side" of the balanced amplifier. This is the bias setting for the amp. My target was 22v rails with 9A of bias. I can verify this 9A figure by looking at the CRC filter on each rail: the 0.2 ohm resistor is dropping 1.8v. Again using I=V/R, I=1.8v/0.2 ohms = 9A, so each rail is pulling 9A from the PSU.

9A * 22v rails means each side of the amp is burning just shy of ~200w, or ~400w for all 12 mosfets in the amp. Using my VOM, the entire chassis is pulling 3.88A out of the wall socket at 121v or 465 watts in total! I guess this verifies the relatively high bias point... The remaining power (465w - 400w) is easily accounted for in the power dissipation (heat) that comes from the 0.2 ohm resistors in the CRC power supply (approx 16w each), the rectifiers (probably 10-12w each), and the thermistors (3-4w each) on the transformer primaries.

Plugging this data into the spreadsheet, it indicates that the amp will produce 100w into 8 ohms, 160w into 4 ohms, and 80w into 2 ohms. Thus, the peak current delivery lies just about 4ohms (higher bias allows you to drive more difficult low impedance speakers). Given this configuration, increasing the bias does not effect power output into 8 ohms at all, only power into more difficult (4 ohm and 2ohm) speaker loads - thus the output power is voltage limited in this instance. In order to increase the output into 8ohms above the 100w target, it is necessary to increase the voltage on the power supply rails. There are three options for accomplishing this: 1) replace the rectifier bridge with individual high-speed, soft recovery diodes for less voltage drop [this will likely restore 0.3v at best to the rails], 2) remove the thermistors from the transformer primary [this will also restore about 0.3v to the rails], and 3) reduce the value of the resistor in the CRC filter [this has the greatest potential impact since the 0.2ohm resistor is already is dropping 1.8v from the rails]. If I replace the resistor with one that is 0.1ohms, I will increase the rail voltage by 0.9v. If you are looking for higher rail voltage, every little bit helps... Its easy enough to bypass the thermistors and parallel the resistors and see how much voltage I get back. One final option if you are looking for higher power into 4ohm speakers (like I am), you can explore setting the AC current gain slightly higher. See the discussion below on this topic.

Adjusting Relative DC Offset with V1 and V3

Relative DC offset is measured from the positive speaker output to the negative speaker output on the same channel. The trim pots (V1 and V3) on the outputs will allow you a bit of control over the relative offset (by tweaking bias settings). If you used matched devices across the amp [that is the Aleph constant current source devices (Q1 and Q10) match on each side and that the gain devices match on each side (Q2 and Q11) and matched the differential pair (Q5 to Q7)] then relative offset should be pretty low (less than, say, 0.2v) and should be very stable over time and across temperature as the amp warms up. Set the pots such that you have the same voltage drop across the Source resistors on the output MOSFETs comparing one side to the other. This fine tuning is performed with small adjustments once the initial bias level has been set and the amp is fully warmed up. Make your (small) adjustment, let it sit for 5 minutes, and then re-measure voltage drop across the source resistors. Things will move around a little while a new heat sink temperature settles in for the adjusted bias point. It'll change again when you put the lid on...

When you're done adjusting the relative DC offset, its time to check the absolute offset (see below), as the two interact. Again, make sure things are fully warmed up (an hour should be enough to get the amp nice and warm). The output current adjustments can help with small relative DC adjustments at the output. - Grey

One particular thing I have noticed is that all adjustments affect each other, including temperature. Bias settings influence heat dissipation, Q6 influences Absolute DC Offset and Q6 is highly temperature dependent. You can adjust Absolute DC Offset with V2. Adjusting V1/V3 also has influence on DC in some way. Increase or decrease the bias by as little as 0.5A overall and a new adjustment has to be made to the offset. I tend to believe there’s a sweet spot somewhere. - NetList on 04-08-2005

Adjusting Absolute DC Offset with V2 and the 'McMillan Resistors' (R46 and R47) :

V2 sets the absolute DC offset (voltage as measured from speaker positive output relative to power supply ground) after the bias has been set with V1 and V3. Absolute DC offset is measured by attaching a test lead to one output, the other to ground. Set the front end current source adjustment (V2) so that the absolute DC offset is as close to zero as possible (again, after allowing time for warm up)- Q6 is very sensitive to temperature (as are Q5 and Q7 since they are the same type of part). In my experience, if you cannot get the absolute offset down close to zero (under half a volt or so) you may have your bias point set too high. Let the amp warm up for at least 30 minutes first. An hour is better. Readjust. It'll drift a little. As long as you're close, you're okay. The speaker does not see the absolute DC offset. The only reason you fiddle with it at all is that it will cut into your potential output voltage swing to either positive or negative, causing premature clipping. It will drift a bit, particularly with heat variations. Don't worry about it. - GRollins on 03-15-2004

There are numerous ways to stabilize the absolute DC offset, but the easiest is to load each output to ground with a resistor (R1/R4 and R44/R45). In any case, the issue is usually reduced maximum output when the absolute DC is more than a volt off. 200 mV is better than you generally need. - NelsonPass on 04-06-2005

Changing R46/47 (the McMillan Resistors) from 4k7 to 3k9 to 3k3 (perhaps even 2k7 or 2k2) decreases the initial level of DC Offset, but at the cost of degraded sound quality. From what I have observed, it appears that the McMillan resistors control both the initial DC offset (high resistance leads to higher initial values of absolute DC offset) and the rate at which the absolute DC offset declines. Using 3k3, the DC offset starts at about 3.0v fell pretty quickly - within 20 mins, it was at zero. Using 10k, the DC offset starts out much higher (14.0v in my case) and falls much more slowly and thus, it takes the amp longer to reach steady state.. An interesting other effect is that Absolute DC Offset will cause the amp to behave differently while it is warming up. Using values of 3k3, I had 3.0v of Absolute DC Offset and all of my heat sinks warmed up at the same rate - or at least not perceptably different by putting your hand on the sinks to compare their temperature. Replacing the 3k3 resistor with 10k offsets the power delivered to the fets. If you´ve got a positive Absolute DC Offset then the positive side will dissipate Ibias x (Vpower supply -V abs dc) and the other side will dissipate I bias x (V supply + V abs dc offset) which is a very big difference at 14V startup voltage. In my case this works out to the output transistors (Q1 and Q10 that are powered by the positive rail) dissipating 20v - 14v = 6v, so 6v x 7.5A = 60 watts, while the current source transistors (Q2 and Q11) are dissipating 20v + 14v = 34v, so 34v x 7.5A = 255 watts. Quite an imbalance that you will immediately feel in the temperature of your heat sinks within the first few minutes of power-up. As the amp warms up, the Absolute DC Offset will fall, and power balance will equalize eventually (typically within 30 mins). For my prototype amp, settled on using 6k6 resistors here (two 3k3s in series) and having Q6 mounted on the main heat sink just about the power mosfets (use a sil-pad and plastic screws/washers/shoulders to electrically isolate Q6). When I first turn the amp on (cold), I get 8.0v Absolute DC Offset. Within 1 minute, this has fallen to 4.8v, within 5 mins is down to 2.8v and by 15-20 minutes is very close to 0v. It drifts around a little (remember, Q6 is VERY temperature sensitive), but this is fine. If you are adjusting your amp without the top cover in place, you will need to re-adjust everything after installing the cover. My top plate is 1/8" thick aluminum sheet and is perforated with a series of 1/4" diameter holes. Several of these holes are strategically placed directly above the trim pots VR1, VR2, and VR3 so I can access them in place with the cover on. A set of wires was temporarily attached to the source resistors, to the output to ground resistors. These allow me to set the bias, balance it across both sides of the amp, and measure AC Current Gain while the cover is in place. Just be very careful! Keep your wires organized, label them, tape them down so they don't touch each other or the grounded chassis, this allows you to measure and adjust a fully warmed up amp without removing the top cover. When you are happy with the settings, open the case and remove the wires and put the lid back on.

From William on 02/01/2007: The McMillan resistors (R46 R47) are connected between the outputs (+ and -) and the sources of the input differential pair. Their value ranges from 2k2 to about 10k and they form feedback loop. Suppose the output at the speaker terminal absolute output becomes positive. Now the current through the diff pair will become higher and the voltage over the drain resistors (R23 / R25) becomes higher. Now the output fets will open up a bit (Rds gets lower) and the dc voltage at the output becomes lower. It works the same way the other way round. I tried different R´s starting from 2k7 and ending up with 10k. Higher resistor values lead to better sound quality, but will also result in higher levels of absolute dc offset at startup (though offset declines as the amp warms up).

From William on 06/23/2007: For the McMillan resistors you should use the highest value you can get away with for a decent absolute offset behavior. I noticed big improvements up to about 10-12k with the standard input and up to 20-22k with the Jfet input. I didn't install a pot but tried different values, measured and listened. As for absolute DC Offset the startup value normally doubles when the Mcmillan resistors are doubled. So 4V dc offset with 5k will change to 8V dc offset with 10k resistors. This is the cold startup value for a 0V value when warm.

Another alternative for lowering absolute DC offset is to reduce R24 from 470 to 320 ohms.

The final configuration of my newer, larger amps, is to use 10k ohm resistors for R46 and R47. This provides about 14.6v of absolute DC offset. Offset falls to about 9v within the first minute and is down to near zero within the first 30 minutes.

Below is a table of Absolute DC Offset measurements and corresponding values of R46 and R47. The greater the value of R46 & R47, the greater the initial level of absolute DC offset and the longer it takes to fall to zero volts as the amp warms up. But, higher values of R46 and R47 (up to about 10k) lead to improved sound quality (particularly in the bass that the amp can reproduce). These measurements were made with a incomplete chassis (no top cover in place). I suspect the behavior will change (offset levels will decline more quickly) with a completed and more closed chassis.

It is important to note that these readings were made on an amp with no input capacitor. Input caps act as high-pass filters to keep DC signals out of your amp. A high amplitude DC input signal is a sure tweeter killer, thus most people install them on their DIY amps. Being somewhat atypical, my amp has no such input cap (yet). I'll have to perform another round of measurements when my next set of caps arrives.

Time
3k3
6k6
10k
Start up
3.0v
12.4v
13.4v
1 min
6.7v
6.7v
5 mins
3.3v
3.8v
10 mins
2.1v
2.5v
15 mins
1.7v
1.8v
20 mins
0.0v
1.2v
1.3v
25 mins
1.0v
1.1v
30 mins
0.8v
0.8v
1 hr
0.3v

Adjusting the AC Current Gain

Something in this part of the discussion is "off". My description below yields opposite results than those posted by others. Somewhere there is a mistake and I haven't had the time to track it down just yet.

Before adjusting the current gain, get all of the DC bias parameters set where you want them... adjust the bias current on each side, and get the absolute and differential DC offsets trimmed. Once you've completed the bias adjustments (using VR1 and VR3), you can replace the trim pots with a fixed resistor, and should never have to worry about it again (though, you should wait until you are completely done tweaking before replacing the pots - I'm still playing around 4 weeks after initial power up...), even if you have to come back for minor adjustments to the DC parameters. It should only need adjusting if you make major changes to the bias current.

Do I Need to Bother With This? If you are simply copying a known design (as I am), you probably won't need to bother with AC current gain settings or how to make them. Where this is important is if you are building your own custom version of this amp.

Some Background on Current Gain: As used here, 50% AC current gain means that at maximum output half the current comes from the Aleph Current Source's AC current gain (Q1 and Q10 at the top of the schematic and PCB), half is from the static bias setting (that you adjusted using VR1 and VR3 above). Grey's original design is biased to approximately 4.5A. 4.5A = 3A + 1.5A. The 3A is DC bias current and 1.5A (50% of 3A) is maximum peak AC current gained by the active current source. 67% gain would mean that at max output, 67% of the current comes from the AC gain, so a bias of 1 amp could produce an output current of 3 amps (1 amp bias, 2 from AC gain)

As shown in the Aleph 30 service manual, the Ratios of R114 and R115 and the parallel combinations of R120,121 and R124-7 set the current gain. Since the output resistors add up to half the value of the emitter resistors in parallel, the AC gain would be 50% if R114=R115. Reducing R114 increases the AC gain. Since the Aleph 30 has R114 at 750R and R115 at 1K, the gain is greater than 50% - Bob Ellis

Nelson Responds to Bob's Post: This calculation ignores the inverse transconductance of the Mosfets and limited gain of the npn circuit, so it is a first order approximation. From experience, the actual gain will be less than this calculation, which is why I always prefer to measure the current ratios through the resistors. More details are contained here: Aleph 2 AC Current Gain Discussion

Essentially, it all boils down to a few relatively simple measurements - you just need a DMM with a 2v range setting so that it can measure AC voltage down to two or three decimal places (three decimal places allows greater accuracy). If your meter's lowest AC voltage range is 200v, it will only allow you to measure 1 decimal point which is insufficient for this task. Also, this should not be adjusted until both the relative and absolute DC offset has been adjusted to as close to 0v as possible and the amp is fully warmed up (60mins or more).

What you need: a signal source that provides a nice sine wave - just one frequency somewhere in the range of 60Hz-1000Hz (a CD with discrete test tones, function/signal generator, etc), your amp, two DMMs (one will do in a pinch), and a resistor bank (to serve as a dummy load) that can handle some output power without burning up. A picture of my dummy load is below:

For my dummy load, I found some surplus 16ohm 120w power resistors at MPJA (see links below) that I connected in parallel two at a time for an 8ohm load (that can dissipate 240w of power) and four at a time for a 4ohm load (that can handle 480w of power). I just lined up all four side-by-side, screwed them down to a block of wood, wired one end of each resistor together, and wired the other end two at a time with a regular wall-mounted light switch in the middle. This gives me an 8ohm or 4ohm resistor with the flick of a switch and provides plenty of power handling capacity. Some wire and a cheap set of speaker binding posts finishes off the rig so regular speaker wire can be used for testing. Others have used five 40ohm 50w dale resistors in parallel. Whatever you can find and arrange so that you end up with an 8ohm load that can handle approximately 200w of power or so should be fine. PartsExpress usually has some reasonable dummy load resistors. A word of caution - if you are sending your amp a high-level input signal (max output from this amp without clipping requires an input signal of about 2v peak-to-peak) your resistors will likely get VERY hot (on the order of boiling water hot - 100c) while being driven by the your amp so don't touch them while while you are working (or right afterward, either)! These giant resistors turn from green to yellow/brown when they begin to heat up and they produce an interesting odor as well the first time I used them. Being more creative, another possibility for a dummy load that would handle even more power is one of the electric oil-filled radiator heaters from Walmart (or your local hardware store). On high, they are typically a 9 ohm resistor floating in oil with a great heatsink. Unplug yours, flip the switches to high and measure the resistance on the power cord with your DMM. You should be able to safely dissipate 1500 watts with no problem...

How to Measure and Adjust AC Current Gain:

Though there has been some confusion, setting the AC Current Gain is pretty straight forward. Since the Aleph-X design is really two traditional Aleph amps creatively strung together, you need to measure/adjust each "side" of the Aleph-X amp individually. For the first time through, focus on R12 (you will adjust the value of this resistor) and R2/3 and R5 (you will need to measure both of these to determine which way to adjust R12). The procedure below comes from Edwin Dorre from DIYAudio and is also contained in the PDF of the Wiki that is linked at the top of the page.

1) Put a 2.0k pot in place of R12/R34
2) Connect an 8-ohm or 4-ohm ‘dummy’ load to the speaker output (a real speaker will do in a pinch but is not a great option due to sound levels)
3) Feed the amplifier input with a 60Hz sine wave at line-level - typically a volt or so (or perhaps 1kHz if you’re using an oscilloscope - 60Hz is suitable for most multimeters). If you have a CD with discrete test tones, just run the RCA out from your CD player directly into the RCA in for the amp. I used an old DVD player with a test-tone CD. At 1000Hz, the player puts out a 0.8v peak-to-peak signal which worked very well.
4) Measure the AC voltage across the entire set of paralleled R2/R3 output resistors, and across R5 from any mosfet from Q1 (a big advantage here is two voltmeters).
5) Calculate your AC Current Gain by following the procedure below (the formulas for doing this are contained in the spreadsheet linked above for calculating bias settings)

My completed amp had the following resistor values: R12=1k3, 3 x 0.22 ohm as paralleled output resistors (R2/R3 bank) and 3 x 1.0 ohm as current source resistors (R5 for a single mosfet for Q1).
With my DMM set on the 2vAC range, I measured 0.161vAC across R2/R3 and 0.121vAC across R5. Now just follow the calculations:
0.161v measured / ( 0.22 ohms / 3 resistors) = 2.195A
0.121v measured / ( 1.0 ohms / 3 resistors) = 1.10A
(1 – ( 1.10 / 2.195 )) * 100 = 49.9% AC Current Gain. In the general case, this is an ideal setting (50%) and is exactly how the production Aleph amps are configured.

You can optimize power into a 4-ohm load by adjusting the value of R12 (and R34 on the other "side" of the circuit board). Start by playing around in the spreadsheet for setting the bias to identify what AC current gain setting maximizes power output into 4 ohms. If you set it too high, the amp becomes current limited (can no longer deliver current, essentially running out of power), so be careful of the setting you choose. I originally chose to target a setting of 60% based on the spreadsheet. After some experimenting, I found a value of 1k6 for R12 provided exactly what I was looking for (I just put the original 1k3 in series with a 332 ohm resistor). With my DMM set on the 2vAC range, I now measure 0.175vAC across R2/R3 and 0.105vAC across R5 and ran the formulas again:
0.175v / ( 0.22 / 3 ) = 2.386A
0.105v / ( 1.0 / 3 ) = 0.954A
(1 – ( 0.954 / 2.386 )) * 100 = 60.0% AC Current Gain - exactly on target. Nelson indicates that you can safely run AC Current Gain as high as 66% without any problems, provided that your bias set high enough. With it set to 60%, power into an 8-ohm load does not change, power into a 4-ohm load increases, and there is no increase in total current draw or heat dissipation. Essentially, the higher the AC current gain, the more the current source can multiply the current. This does, however, also lead to degraded sound quality, so - like everything else- there is a cost to this tweak.

With a larger resistive value for R12 the current source will deliver a higher percentage of the work. A lower resistance value will lower the percentage. If you want a switchable variable/constant current source, install a switch to break the R12 circuit.

Now go back and re-run the entire set of measurements using R34, R40, and R42/R43. Ideally, you can just copy the value of R34 from R12, but its always good to double check things first. If the two sides of the amp are not balanced, you will get symmetric clipping at maximum output.

When setting the amp up for more than 50% ac-current-gain the upper half (current source) will shut off during negative current peaks causing the amp to get in some sort of AB mode, however, it's worth noting that if you bias the amplifier at the expected peak output current, you won't get this effect, and you can set the current gain arbitrarily. - Nelson Pass

Some relevant reading can be found here. This is the explanation by Mr. Pass himself in the Zen v2 article. Clipping should be symmetrical if the ac gain is set to be the same on both sides. If clipping is asymmetrical the current gain is probably off on one side.

Output Power and Waveform Testing:

Easy Stuff to Measure and Verify- Start Here:

I say that this stuff is easy to measure because you don't need any fancy equipment just yet - just a good old digital multi meter. Further on is a section dedicated to measuring with an oscilloscope that provides a different set of insights. The problem is that not everyone has an oscilloscope at their disposal and they tend to be somewhat expensive to purchase, even used. While I was waiting for some time to hook my amp to an oscilloscope, I set out to measure, watch, and adjust the amp. I let the amp warm up for a full hour (without the top cover installed). During this time, absolute DC offset started off at 9.0v upon power up. This dropped to 5.39v at the end of 1 minute and continued dropping so that it measured 3.9v by 5 mins, was at 1.86v by 30mins, and at 0.5v after 60mins. At this point, I readjusted bias to 8.0A (I was trying to be more kind to my poor suffering transformer that was audibly complaining about the strain it was under) and reset Absolute Offset to as close to zero as I could (-0.08v). Relative DC offset remained varied from 0.00v to 0.02v, and rail voltage measured a steady 21.1v. I was then able to measure AC Current Gain at 49.7% on one side and 49.4% on the other side. Just about perfect (50% is the ideal target for optimal sound quality)! Measuring the voltage drop across R23 and R25 (both 392 ohms) revealed about 4.63v. Using ohm's law, this indicates that the front end differential pair (Q5 and Q7) are each running about 11.8mA of bias - also right on target (11mA each is the desired bias level)!

I spent the next little while measuring the Source resistor that feeds each of my output mosfets. In the past, I just clipped a test lead to whichever single resistor was most easily accessible, so this is the first time I measured the voltage drop across each of the 12 source resistors. This first chassis (my prototype) is constructed using 4 heatsinks and I have 3 mosfets mounted to each sink. With the RCA input shorted and no load connected to the speaker terminals, I made the following measurements:

 

Left Side of Amp
Heat Sink : Final Temp
Transistor
Vsource
Rsource
Amps/Fet
Watts/Fet
Avg Dissipation
Sink 1: 70c
Q10 left
0.440v
0.333ohms
1.32A
27.88w
28.20w avg Q10 bank
Q10 middle
0.450v
0.333ohms
1.35A
28.51w
Q10 right
0.445v
0.333ohms
1.34A
28.20w
Sink 2: 68c
Q1 left
0.448v
0.333ohms
1.35A
28.39w
28.32w avg Q1 bank
Q1 middle
0.448v
0.333ohms
1.35A
28.39w
Q1 right
0.445v
0.333ohms
1.34A
28.20w

 

Right Side of Amp
Heat Sink - Temp
Transistor
Vsource
Rsource
Amps/Fet
Watts/Fet
Avg Dissipation
Sink 3: 71c
Q11 left
0.442v
0.333ohms
1.33A
28.01w
28.11w avg Q11 bank
Q11 middle
0.448v
0.333ohms
1.35A
28.39w
Q11 right
0.441v
0.333ohms
1.32A
27.94w
Sink 4: 72c
Q2 left
0.446v
0.333ohms
1.34A
28.26w
28.28w avg Q2 bank
Q2 middle
0.449v
0.333ohms
1.35A
28.45w
Q2 right
0.444v
0.333ohms
1.33A
28.13w

It looks like all of that matching of mosfets and source resistors had paid off - the power dissipated by the mosfets match very closely with one another, both within and across banks. The worst mismatch is with the Q10 bank at 2.2% and the best matching is within Q1 bank at 0.6%. All four banks themselves are matched to within 0.74%. I think I would be hard-pressed to do much better.

Total power dissipation of the mosfets equals just a hair over 8.04A on one side and 8.02A on the other side, or a total of 339 watts for the entire amplifier. Measuring total draw from the wall socket reveals consumption of ~378 watts, so my guess is that the remaining 40 watts are being consumed by the components on the PCB, and burned off as heat with the thermistor, the transformer, and the two bridge rectifiers (which seems reasonable given their temperature rises). At this point, I'm not going to change ANY settings until I can hook it to a scope again to look at the output waveforms - I suspect I'll see much cleaner output this time now that the two sides are better balanced and Absolute DC Offset has been trimmed. Playing musing sounds nice, but again, I only have one working channel at this point so its difficult to really judge.

I measured the heatsink temperatures with my non-contact infrared gun and I searched each sink for the highest temperature I could find. Predictably, this ended up being just above the center mosfet on each sink. I was surprised, though, to find a maximum temperature of 72c. Ambient temperature on the floor where I was working measured 19c, so we're looking at a thermal rise of 53c! Yikes! With an eye toward longevity and more quiet operation, my plan is to replace the 750VA transformer with one rated at 1500VA and to re-build the chassis with six heatsinks instead of the four I am using now. This will provide 50% more heatsinking capacity (total chassis of ~ 0.07c/w as opposed to the ~0.105 c/w) as each sink will hold two mosfets instead of the three that currently live there. Hopefully, this will reduce chassis temperature by about 12c - a much more reasonable result.

The table below provides a quick snapshot of the output power of the completed amp for several different bias points. Constant for each measurement are the power supply rails that were fixed at about 20.5v. After having my theater up and running for about 6 months, I began exploring different bias settings due to the massive amounts of heat thrown off by three of these completed amplifiers sitting side by side. Reducing the bias point for all three amps from 9.0A to 7.2A saves a total of about 240w of power consumption and reduces maximum sound output from the speakers by less than 3dB.

dV Source Resistor
Amplifier Bias Point
Total Power Consumption
Output Power 8 ohms
Output Power 4 ohms
Temperature top of sink 22c ambient
0.50v
9.0A
3.45A - 410w
85w
~140w
48c
0.45v
8.1A
3.05A - 360w
85w
~110w
46c
0.40v
7.2A
2.75A - 330w
85w
~85w
44c

 

More Complex Things to Measure:

OK - here is where the rubber meets the road and we see what the reality of the situation brings! For this task, I headed to the EE lab to get some actual measurements of output capability. You will need a signal generator (or a CD with test tones), your dummy load, and a scope to watch the output waveform. There are a few things you may want to measure: ripple voltage on the power supply caps, maximum output capability before clipping, and overall waveform quality (both for sine wave and square waves). Below are my measurements including a sample of what you see when you hook things up incorrectly.

Measuring Power Supply Ripple Voltage:

This is perhaps the easiest to measure. Set you scope for "AC Coupling" and connect the leads of one test probe to your final power supply cap and see what kind of trace you get on the display. Be careful to connect positive with positive and negative with negative (the positive lead is typically the little hook with a retractable plastic slider over it and the negative lead is typically the alligator clip with a little rubber boot over it). If you mix these up, you'll end up grounding your power supply because the negative lead on the probe (the little alligator clip) is a direct connect to the 3-prong ground that goes into the wall socket - not a good thing to mix these up. With a 21v power supply, a bank of 3 caps (CCRC totaling 290,000uF), and an 8A draw, I get about 200mv of ripple (left image below) on the caps before the resistor. The important characteristic of this graph is the slope of the two "sides" of the curve, immediately to the left and right of each peak.. The left side represents the voltage charge in the cap growing as a result of the DC pulse that comes from the rectifier. This should be somewhat vertical (it will never be truly vertical because the cap takes time to absorb the voltage from the rectifier). The right side shows the voltage "draw down" on the caps by the amp during the time between the DC pulses from the rectifier. The slope of this side should be more gentle and elongated as seen below in the left image. This reflects the voltage decline of the caps while the amp is operating before it gets the next DC pulse from the power supply. With the 0.2 ohm power resistor between my two 35,000uF caps and the final 220,000uF cap (CCRC), ripple is reduced to about 5mv (middle image below) on the final cap that actually drives the amp. These images are just thumbnails, I didn't post a larger version of them as there is really not much else to see here. Nelson indicates that ripple below 600mV shouldn't pose any real problems, but part of the DIY philosophy is to make things better because you can, so CCRC it is, but I need a bigger transformer with 19v secondaries instead of 18v secondaries to give me some room to play without sacrificing rail voltage. One benefit of inserting the resistor is that it reduced some of the PSU-induced hum at the speaker. Without the resistor, I have is a very small hum when you press your ear to the speaker. Putting the resistor in the power supply reduced the hum to the point where I am not worried about it at all. A final image (on the right) is what a bad capacitor looks like on the scope. Notice that the voltage trace drops just as fast on the right side of the spike as it rose on the left side of the peak and then basically "flat-lines" while it awaits the next DC pulse from the rectifier. This is an old cap that has dried out and can longer hold a charge - thus it dissipates its voltage just as quickly as it received it. This cap will induce a larger than normal ripple voltage into you amp (sometimes on the order of whole volts) and results in sub-par performance. If you find a cap that behaves like this, replace it with a newer cap and make sure it behaves properly in your amp (each individual cap should exhibit a trace like the one on the left).

 

Normal PSU cap:
100-300mV ripple - this is typical ripple performance

Second cap bank in CCRC:
5-10mV ripple - this is excellent ripple performance
Extreme ripple of a bad PSU capacitor - this cap needs to be replaced

 

Determining Power Bandwidth (Frequency Response):

Ideally, you want your amp to faithfully reproduce a clean sine wave from 10Hz all of the way up to about 200kHz. Connect the signal generator to the amplifier input and connect your dummy load across the speaker output. In order to properly display the output waveform, you need a 2-channel oscilloscope with the ability to display the sum of the two channels in a single trace. Here is where things a little bit non-intuitive if you don't keep in mind that a single channel of an Aleph-X is really two amps working together (that's what makes the "X" part): Connect the positive leg of each channel's probe to each side of the dummy load and the negative leg of each channel's probe to the amplifier power supply ground. This hook-up procedure is critical for the Aleph-X amp because of the balanced design (one side of the amp drives the positive speaker terminal, the other side of the amp drives the negative speaker terminal). If you just use one channel to measure across your dummy load, you are grounding the negative side of the amp which will provide a heavily distorted waveform as pictured below. With a "regular" and more typical store-bought amplifier, the input signal ground (barrel of the RCA plug) and output signal ground (black speaker terminal) are just that - they are grounded to one another (and is often implemented as a straight pass-through with no alteration) and often are also grounded to the chassis and the 3-prong wall outlet ground (when present) - it is only the positive input signal that gets amplified, so just using a single probe works on your more typical amps.

Once you have properly connected your 2-channel scope across the dummy load resistor, drive the amp with a sufficient input signal to result in an output signal somewhere in the 1v to 10v (peak to peak) range. If you have an auto-ranging scope, it will automatically measure and display the peak-to-peak voltage (the difference in voltage between the top-most and bottom-most part of the trace). If your scope does not have auto ranging capability, you can also use a typical DMM set to AC range. A DMM will read Root Mean Square (RMS) voltage, which is not the same as peak-to-peak voltage. Fortunately, you can simply convert RMS to peak-to-peak by multiplying the RMS value by the square root of 2 (1.414).

OK, now we've got all of the hook-up and and measurements procedures ready to go. So how do you determine the power bandwidth? Start with a 1kHz input sine wave and make a note of the output voltage. Then, just increase the frequency of the input voltage until the output voltage drops to approximately 0.7 times (70% of) the output voltage you recorded at 1kHz. That´s your upper -3dB point (and it should be in excess of 100kHz). Start again at 1kHz and then decrease the frequency until you see the same 0.7 drop point. This is your lower -3dB point (which should be close to 7-10Hz).

pF value for C2 and C4
Bandwidth
0pF (no cap installed)
240kHz
1pF
210kHz
3pF
140kHz
5pF
120kHz
10pF
80kHz
Pass DIY a40 amp
240kHz
Marantz MA500 monoblock
47kHz

Clearly, the 80kHz bandwidth that results from a 10pF cap in the feedback loop is too narrow (bandwidth should be above 100kHz). If you look below in the section dealing with square wave output testing, you can see the impact of different capacitance values on the resulting output waves. The square waves correspond to bandwidth limits - lower bandwidth results in more rounded square waves.

Power Bandwidth and use of C9 & C10 to Solve Output Oscillation:

As a result of some discussion on DIYAudio concerning the role of C9 and C10 in the circuit, I decided to make some comparison measurements. These optional caps are used to solve the specific problem of high-frequency oscillation, typically referred to as "motoboating" due to the sound it produces in your speakers. These caps are not shown in the schematic at the very top of the page (originally developed by Grey), but they are present on the PCB from the group buy. I made some measurements of input and output voltage levels at the visible onset of clipping in the scope. I let the amp warm up for an hour first. Here is what I found with my amps configured with 9610 for the input differential:

Input
Frequency
C9/C10 Present, 4 ohm load
C9/C10 Absent, 4 ohm load

Input Voltage Peak-Peak

Output Power at Clip Onset

Input Voltage Peak-Peak

Output Power at Clip Onset

1 kHz
3.66v
150w
3.76v
153w
5kHz
3.64v
139w
3.94v
163w
10 kHz
3.05v
96w
3.63v
138w
15 kHz
2.80v
80w
3.67v
140w
20 kHz
2.56v
65w
3.47v
125w

These measurements were made with 22.4v rails and 9.4A bias current. Measured voltage drop across source resistors averages 0.515v for each of Q1, Q2, Q10, and Q11. Final heat sink temp is about 55c.

With C9 and C10 in the circuit, high frequency power output was reduced (-3dB) when compared to lower frequency output. After removing C9 and C10 from the circuit, the maximum sine wave performance before distortion becomes obvious is clearly improved. Looking at square wave output performance, there is slight improvement compared to when these caps were installed in the circuit. There is now a bit of a small bump on the leading edges of the trace (perhaps I'll increase C2/C4 just a pinch to compensate), but overall, the vertical edges of the traces are now more vertical than they were with C9/C10 in the circuit (most noticeable at 20kHz). This, however, is a very small difference. Listening to music on the newly modified pair of amps doesn't reveal any immediate difference at low volume levels (none was expected), but does show some improvement in the treble area when the speakers are driven hard. As an added benefit, the power bandwidth (-3dB output point) has increased from approx 140kHz to closer to 210kHz with C9 and C10 omitted from the circuit. The straightforward conclusion at this point is to keep C9/C10 out of the circuit unless they are needed. I think I originally included them simply because other had and I didn't fully understand that they are intended to be used to solve a specific problem.

How do you know if they are needed? The easiest way for me to determine this is after I changed my input differential from the 9610's to jFets (see section below on this). After making this change, when I power down the amp, the speakers made a very conspicuous "motorboat" sound as the power drained from the PSU caps. This is indicative of high frequency oscillation that has a detrimental effect on the power bandwidth and ultimately on the logevity of your output mosfets if not cured. Installing C9 and C10 is the cure, though you will need to experiment with different values to see its impact on your power bandwidth. Remember to let the amp fully warm up before doing any serious measuring with your scope.

Power Bandwidth and Input Sensitivity:

A short time ago, I discovered that it takes nearly 2.5v RMS input signal in order to drive my amps to full output. All of my other amps (Marantz, Adcom, Behringer) are able to achieve their full power output with input more on the order of 1.0 to 1.2v RMS. To increase the input sensitivity of the Aleph-X amp, just double double values of R16 and R30 (change them from 100kohms to 200kohms). Be sure to recheck the amp for offset performance and oscillation after making this change.

After making this change, I picked up 5dB in output compared to the unmodified amp. Using the scope, I was hitting max clean output with a 1.4vRMS input signal. However, the bandwidth (-3dB point) dropped from about 210kHz to just about 100kHz. After making this modification Sine and Square waves look clean all the way through 100kHz, above that, the square waves begin to round over and look line sine waves, but this was normal before the mod. Absolute DC Offset behaves as usual (starts at about 14v, drop to 10v with 2-3 mins, and by 30 mins is near zero) and there were no signs of oscillation that I could find. Relative DC offset remains below 2mV at all times.

Grey indicates increasing the input impedance will decrease the bandwidth due to the Gate capacitance of the MOSFETs - Whether this is important to you or not is another matter. It's not a deal killer to most people. Just thought I'd throw it in. Stuff like that bothers me, but I realize that I'm in a distinct minority in my quest for wide bandwidth (preferably accomplished with little or no NFB--any idiot can get wide bandwidth with oodles of NFB). Note for the record that I prefer higher input impedances whenever possible--a holdover from my tube days--but that bandwidth matters more to me now than high impedance. If I've got a circuit that, like most of the Aleph variants, starts rolling off ca. 100kHz and I do something that drops that to 80kHz I stop and ask myself just how badly I want that modification of the circuit. I nearly always go back to the 100kHz version. Someday I'll discover a definitive explanation as to why it takes a solid state circuit with 250kHz bandwidth to sound as open as a tube piece with 50kHz bandwidth. It's not logical that it should be so, but if audio were that cut and dried, it probably wouldn't be nearly as much fun.

Determining Output Quality and Distortion:

No, we're not going to measure the actual distortion level (THD, or Total Harmonic Distortion - this requires a fancy and expensive meter), but we can easily spot distortion when it occurs on the scope. As before, start with your dummy load connected to the amplifier output and your signal generator connected to the input. Connect both channels of your oscilloscope across the dummy load as indicated above and set the scope to display the sum of the signal across the two probes. Just for reference, I've included an image of what happens when you use only one probe with this amp - which effectively grounds one half of the amp circuit. On the left is a sine wave (left) and on the right is a square wave (right) that are both heavily distorted (the input signal is displayed in the top trace and the output signal in the bottom trace). Don't do this to your poor amp. You won't really hurt anything, but it won't be happy...


Sine Wave Testing:

Next, run an input sine wave of sufficient amplitude to get 1V at the output across the dummy load. Use your signal generator and run a variety of sine waves into the amp, ranging from about 10Hz up to around 100kHz. The output sinuses should all look smooth and undistorted throughout this frequency range. Then, increase the input signal so that you get a 10v signal across the dummy load and repeat the frequency sweep. The output waveform should be clean an undistorted for both the 1v and 10v output signal sweeps. Below is a 10kHz sine wave from the output of the amp (when properly measured with two probes). I could post more of these at different frequencies or at different amplitudes, but all of the resulting sine waves from 10Hz to about 200kHz look exactly the same as this one - nice, smooth, clean, regular waveforms like the one below.

Square Wave Testing:

Next up, repeat both the 1v and 10v output sweeps using a square wave function as input to the amp instead of the sine wave you used above. Square waves will feature somewhat different levels of distortion in the output waveform, but the important part is to see corners that are as sharp as possible and lines that are as straight as possible from about 500Hz right on through 20kHz (or beyond). Traces of square waves below about 200Hz will typically feature slanted lines on top and bottom, but this is OK. See the image below on the left of a 50Hz trace. The key point here is to experiment with different pF values of capacitance in the feedback loop (C2 and C4 in the parts list linked above) to see which value provides the nicest square wave across the audio spectrum. Below is a chart I compiled that compares resulting square wave output at 1kHz, 10kHz, and 20kHz for values of 0pF (no cap), 1pF, 3pF, and 5pF. All measurements were made into a 4ohm load at 10v peak-to-peak output levels. Notice the vertical overshoot (especially at 20kHz) in the waves that result from 0pF and 1pF values. This is indicative of not enough capacitance in the feedback loop. Next, look at the 5pF traces and see how the leading edges are rounded off - this is indicative of too much capacitance in the feedback loop - this limits the bandwidth of the amp. This leaves us with an optimal value (in my case at least) of 3pF for C2 and C4 in the feedback loop. There is a small ripple in the leading edge of the trace, but not as much as with 1pF and 0pF. There is some roundedness in the trace, but not as much as with 5pF. All in all, 3pF represents a good compromise value here.

Finally, have a look at the right-most image above. The traces on the left show the absolute output of each side of the amp (amplification of positive input signal and amplification of the negative input signal) with respect to the power supply grounding point. Each trace here shows the inherent distortion of the output mosfets. The genius of the super-symmetry design is to sum these so that the distortions cancel each other out when you look at the relative output of each side of the amp together - this is shown in the trace on the right hand side of the right-most image above. What a great design!

Square Wave Output and Power Bandwidth Comparison:

Taking a look at the three amplifiers in the bandwidth discussion above, it is easy to see the relationship between bandwidth and the ability of the amp to reproduce a square wave. Below are three square waves, each measured with a 10kHz input signal, but with three different amplifiers that feature three different power bandwidths. Clearly, as the bandwidth decreases, the ability of the amp to reproduce a 10kHz square wave also decreases. Bandwidth above 100kHz is a good thing as shown in the images below. Now this doesn't mean that we feed our amps square waves and listen to them on any sort of basis (in fact, I would advocate NOT doing this at all), but clearly the ability of an amp to take the amplified musical output signal and make it not only stop but also turn it on a dime is a good thing. You don't want the output signal to overshoot its mark, nor do you want it to not be able to achieve the mark. Any amp worth its salt should be able to cleanly reproduce a nice, square 10kHz square wave with ease. The a40 and Aleph-X amps do this handily (technically, the a40 is a pinch more technically accurate or "analytical" if you follow the industry lingo). The MA500 misses the mark to a visible extent, though I suspect this has something to do with the THX certification process and the built-in "cinema re-equalization" that helps to reduce treble harshness on movie soundtracks.

10kHz on DIY Pass a40 - 240kHz bandwidth
10kHz on DIY Aleph-X - 140kHz bandwidth
10kHz on Marantz MA-500 - 47kHz bandwidth

 

Determining Peak Output Power Level:

Again, connect the positive leg of each channel's probe to each side of the dummy load and the negative leg of each channel's probe to the amplifier power supply ground. Run a pure single-frequency sine wave into the amp (something in the range of 60Hz to 1kHz is fine) like you did when you adjusted the AC current gain, slowly increase the amplitude of your input signal and watch the resulting output waveform until you see the top and bottom of the wave begin to flatten out. This is clipping and the more the amp clips (because the input signal is too great) the flatter the curve becomes. This is where it is most easy to destroy a loudspeaker because extreme clipping can drive your speakers with DC which causes trouble with voice coils that don't get the opportunity they need to dissipate heat. For reference, the image on the left shows a normal, clean, unclipped, and undistorted sine wave. The image on the right shows a heavily clipped sine wave. Note the "clipped off" peaks at the top and bottom of the waveforms. There is also some distortion in the lower portions of the waveform, and the legs of the wave are not clean and smooth. The image in the middle shows a close up of a waveform just at the onset of clipping for the amp. Looking closely at the top and bottom of the waveforms shows some distortion beginning to show up at the most curved areas of the wave. Distortion and clipping are indicated by the yellow arrows in the enlarged image when you click on it. The critical point that you want to observe when you a determining peak power output is the exactly point where clipping begins to set in. This is your maximum undistorted output into that particular load.

Undistorted Sine wave
Onset of Clipping
Heavily into Clipping

When you see clipping in the display of the scope (top and bottom of the sine waves are no longer smooth but begin to show signs of distortion or flatten out), make a note of the voltage being delivered by the amp into your dummy load (an auto ranging scope will tell you the peak-to-peak voltage, or you can measure RMS with your meter set to AC and multiply this by 1.414). For examples, while I was creating the middle image above, the scope indicated that the amp was driving 36.5v into the dummy load at the point where output begins to clip. My dummy load was set to 4 ohms at the time. Use Ohms Law (I=V/R) to calculate power output: I = 36.5v / 4ohms = 9.125A. Then multiply 9.125A by the one-way measured voltage swing of 18.25v (36.5v divided by two). This indicates that the amp is producing 166.5w into a 4 ohm load when clipping begins to occur. This is pretty significant output for a Class-A amplifier! Measuring the amplitude of the input signal I can see that it takes a 1.9v input signal to drive the amp into clipping. This is just about where I expected it would be. Most amps begin to clip just before 2.0v of input.

Repeating the above example with an 8 ohm load, I am able to measure 36.8v before I see clipping. Repeating the calculation, I get: I = 36.8v / 8ohms = 4.6A, then multiply 4.6A by the observed 18.4v and get 84.4w into 8 ohms. There you have it. This amp can drive about 85w into 8ohms and about 165w into 4 ohms before clipping at the current bias setting.

Jan Didden from DIYAudio made the following suggestion for output testing: The idea is that a real speaker load often represent a complex network not of just a resistor but also capacitors and/or inductors that varies with frequency.... Especially the capacitive part can sometimes provoke an amp into instability/oscillations. So, to check the power output of your amp in various steady state loads, the resistors are fine. If you also want to verify amp stability, try paralleling those resistors with capacitors varying from perhaps 10nF through 100nF, 1uF, 10uF. Especially the latter (10uF) is hard on the amp with high frequencies because it represents a low impedance, so you probably don't want to do that longer than just a few seconds. Always slowly turn up the signal and watch the scope for funny patterns (things that don't look like nice smooth sine waves), turn it off if you see that!

Using the Scope to set (and more precisely balancing) bias for optimal clipping behavior:

This discussion comes from Hugo in DIYAudio (the full discussion can be found here).

Interestingly, I set the bias solely with the scope. I let the amp warm up for almost an hour, trimmed the absolute DC offset very nicely to +/-10mV and then connected an 8ohm load to the output and a 1kHz sine wave to the input. I raised the input voltage to the point where I could see a clipped sine wave on the scope and I swear it did look ugly. Each half of the amp got a probe attached and I set the scope in ADD mode. (Sum of both channels). Then, I started trimming the two bias pots to make the wave form clip symmetrically.

I found that there seems to be some kind of sweet spot where a certain bias level is optimal; too little or too much bias results in earlier clipping. I would call it some kind of turnover point and carefully trimming both halves of the amp gives the highest unclipped wave form possible. It always appeared to me that the bias limit was equal to the heat sink capacity. I can’t explain why but I found out this seems not to be the case. After settings were done I checked with the DMM and the bias for both halves was perfectly equal for each half of the amp. At some point the scope picture looked quite similar to the one I saw from William. I should check again if the bias was lower or higher from the sweet spot, don’t remember.

Replacing the Input Differential with JFETs - a significant sound quality upgrade!

There have been a few discussions online about replacing the 9610 mosfets that comprise the front-end input differential pair with Toshiba 2SJ74BL JFETs that have more linear behavior characteristics. I am in the process of making this change in my amps right now and will update this section in the near future.

The changes to the circuit are relatively easy to make, but I am still working through some small problems. These include somewhat reduced power output (about 1-2dB before clipping), somewhat reduced bandwidth (which is still in excess of 100kHz), and increased relative DC offset which tops out near 100mV once the amp is warmed up. Once I get these issues straightened out, I will post more complete details. For now, I am using the following circuit diagram to replace the 9610 devices (thanks, Graeme!). If you are running a low-powered amp that does not have parallel output mosfets, you can probably just get away with directly substituting a single JFET for each 9610 and removing the gate resistors (replace them with a jumper). However, if you have parallel output mosfets for higher power output, it is likely that you will need parallel your JFETs otherwise they won't have the drive to overcome the input capacitance of the power mosfets and you will see reduced bandwidth. You may also need to experiment with different values for C9 and C10 to prevent parasitic oscillations that will rob output power and cause output power to collapse beyond clipping - try values in the 100pF to 1,000pF range.

Listening impressions indicate this is a very worthwhile upgrade to the amp! The biggest hurdle in making this change is getting your hands on well-matched 2SJ74BL parts (matched Idss in excess of 10mA is preferred). A great reference on matching JFETs can be found here: http://www.diamondstar.de/transistor_matching_jfet.html. These devices have been discontinued by Toshiba, though a number of places still have some in stock. Be aware, though, you will need to purchase a large quantity (often > 30-50) in order to find multiple sets of four devices that match one another - they tend to vary widely. Alternatively, there are several sources on DIYAudio - check with Spencer (http://www.fetaudio.com/store). In terms of sound quality, the JFET differential provides improved treble response and somewhat reduced midrange response. I would characterize both of these changes as welcome, yet minor, improvements. I have always thought the Avro speakers were just a pinch shy at the top end of the spectrum (most noticeable when listening to music in Pure Direct mode on my preamp which bypasses the Audyssey room EQ). This upgrade restores the "missing" treble I've been looking for. Turns out, it was the 9610's in my amp, not the speakers...

Score one for the JFETs.

While it was not a stand-out quality with the 9610 front end, chaning to the JFETs also makes the midrange sound a bit less "tubby" for lack of a better word. Like the increased treble, the change in midrange is subtle, but evident on direct comparison. I find this be a welcome change!

Score another one for the JFETs!

I think these changes most directly highlight the greater linearity of the JFETs across the audio spectrum. Given the relatively small magnitude of these changes to the sound quality, though, it is quite possible that you might be able to make similar impacts using some EQ in your preamp or receiver. Overall, the tonal balance of the amp is more realistic and pleasing with JFETs. I see this as a REAL benefit to audio purists who listen to music without any type of EQ in the audio chain.

The final impact, though, is by far the most substantial and the specific reason that I find this upgrade so very compelling! This impact is not subtle at all! The amp with the JFET front end makes my enormous five foot tall speaker COMPLETELY DISAPPEAR in the room! While the sound from the 9610 front end is quite nice (and handily outperforms most every other audiophile-grade amp out there), there is now an interesting contrast between my left speaker (9610 differential) and my right speaker (JFET differential). The speaker driven by 9610's seems to "call more attention" to itself - something that I didn't notice before making the change. This is somewhat hard to explain thoroughly, but when listening to music and movies, I am now more aware of the left speaker and its physical location in the room. By contrast, the speaker driven by the JFET amp just seems to be GONE! Sound is clearly being produced, but I am unable to precisely identify its physical location in the room as I can with the other speaker. At first, I thought this was perhaps related to SPL levels - somehow the modified amp was no longer balanced with the others anymore and the sound was a few decibels shy in comparison. A few minutes with my SPL meter showed this was not the case at all. Something about the JFETs has made an already top-notch amplifier EVEN MORE TRANSPARENT! Sound just seems to fill the room without having a clear point of origin. With just one amp modified, the soundstage has widened considerably on the right side of the room and imaging is more apparent. I can't wait to get the entire front array of amps upgraded to JFETs!

MAJOR score for the JFETs!!

Some other observations about this upgrade: It does not seem necessary to thermally/physically couple the SJ74 devices to one another. However, it DOES seem necessary to couple the ZTX550 devices to one another, otherwise relative DC offset will wander as the amp warms up. Right now, all four ZTX550 devices are "free standing" in my amp. If I put my finger on top of the 550s one the "Q5 side" of the circuit, the offset goes quickly positive as a result of body heat. If I put my finger on top of the 550s on the "Q7 side" of the circuit, the offset goes quickly negative due to body heat. If I pinch all 4 in a row together with my fingers, offset approaches zero, so I'm wondering if the outer two 550's in my lineup of 4 are dominating the relative offset swing. I am thinking of trying another arrangement with better thermal coupling of the 550's. The tradeoff is that this complicates the point to point wiring that I am currently using. If you don't keep your wires short (less than 1" or so) you'll pick up extra hum in your front end...

Another important consideration with the ZTX550 devices is to match them using the hfe function of your DMM. After measuring a batch of 30 that I purchased, I found hfe measurements varied by as much as 25% across the lot. Within this set of 30, I was able to create three groups of four devices that all matched to within 1.5% of one another. A larger group would likely lead to better matching, but a few minutes spent matching will help as well.

Also, I notice that when rel offset is low (~14mV on powerup) the hum/hiss (tweeter and midrange drivers) from the speaker is very, very quiet. After everything warms up (and offset is now ~80-100mv), there is more pronounced hum/hiss from the speaker that matches my other amps (still inaudible from more than one foot from the tweeter). I don't know if this is linked to overall thermal state, or just relative offset (or if these two are actually more identical that I'm thinking). I have some more work to do...

More to some soon!

A few relevant discussions are linked below:

AlephJ-X by William
AX100 100W Aleph-X Monoblocks by GL
Aleph J-X Group by Peter Daniel
Aleph J-X Amp Project
The next best think to 2SK389 / 2SJ109 by Patrick
One Aleph-X working, One to go by William

 

Solving Speaker Hum

Speaker hum is perhaps one of the most frustrating problems to deal with in DIY audio. Web searches will undoubtedly reveal a host of explanations and solutions. Nelson Pass posted the following advice for tracking down and solving hum:

"When you run into this sort of stuff, you need to isolate the problem down to [one of] the following categories: 1) oscillation: high frequency buzzing, none of the other approaches helps: play with frequency compensation, look at output with a scope. 2) ground loop: If the problem doesn't go away when you disconnect the source and short the inputs, it's not a ground loop, probably. 3) transformer pickup: Move the transformer / bridge / cap assembly away physically, and the problem goes away. 4) poor ripple rejection: CLC or CRC filter in the supply fixes this."

Tackling these in order, we have the following:

1) High frequency oscillation - this will be evident in a scope when looking at square wave output. For the Aleph-X amp, this is addressed through the use of C9 and C10. See the discussion above on Power Bandwidth.

2) Ground Loop. This is most directly addressed through the use of a high quality star grounding scheme. See the discussion above concerning the use of star grounding in the Power Supply section. Make sure your chassis is directly ground to AC Mains ground. Signal ground may be connected to AC Mains ground either a) directly, b) through a thermistor, or c) through a bridge rectifier. Though it may be tempting to use a small resistor instead of a thermistor or a small diode instead of a larger rectifier - if things fail you will be dealing with large currents that need to be safely directed to ground without burning out the resistor and leaving you with a dangerous situation. This is something you might want to experiment with. My completed a40 amp is dead quiet with the PSU ground connected to AC ground through the diode of a bridge rectifier, but the speakers hum when I replace the rectifier with a thermistor. The Aleph-X amps are dead quiet with just a thermistor, but produce audible buzz at the speakers when the power supply ground is directly connected to the AC ground. Go figure... The point is to have a configuration that allows current to flow should it be necessary. Thermistors have a nominal resistance built into them (about 5 ohms in the case of the CL-60), so current can flow should the need arise (power supply fault of some sort). Diodes allows current to flow out of the chassis (for safely handling a power supply fault) but block current from flowing into your amp (thus forming a ground loop).

3) Transformer Pickup. Try rotating your transformer. Move the wires. Move the transformer / bridge / cap assembly away physically to see if the problem goes away.

4) Poor Ripple Rejection: This can be from a) not having enough capacitance in your power supply or b) using old surplus caps that are dried out. Try replacing your caps, or consider using a CRC or CLC power supply design. Either way, measuring power supply ripple with your meter or a scope should reveal something interesting here. See the above section on measuring power supply ripple.

 

 

So, How Much Does it Cost to Build One?

This is highly dependent on how you go about things. Obviously, buying all new parts will cost more. Scavenging surplus/used parts can help you save a bundle. Its very clear that 4 items will consume 80-90% of your total costs: Chassis, Heatsinks, Transformer, and Power Supply Caps.

Using six sinks per mono chassis (total of ~ 0.07c/w) added up to about $200 for me. You can do better price wise if you find surplus/used, but you have to be more choosy and patient - its hard to find large and useful sinks on the surplus market! A 750va toriod will run you about $175 or so (1500va is closer to $300, including shipping for that huge and heavy ball of copper). Aluminum for your chassis can be had for $40 (or nearly free if you are thrifty and scavenge scrap at the metal shop). Figure $20 - $40 on tools, screws, nuts, washers, wire terminals, wire, etc. depending on what you already have laying around. Probably about $75 to stuff the PCB and wire all of the mosfets. Then throw in another $25 - $40 for stuff like RCA, XLR, speaker binding posts, power switch, fuses & holders, EIC power entry, EIC power cord, etc. For me, this works out to about $700 per 100w mono chassis.

Building a lower powered mono version (or a low-powered stereo chassis) is substantially cheaper for several reasons: 1) fewer heatsinks are required, 2) you can get by with a single (smaller) transformer for both channels (you may even run both channels off of the same cap bank...), 3) fewer output fets, less matching required, 4) source resistors are readily available in 0.22ohm ratings and fewer of them are needed. Overall, I suspect that I could build a lower powered stereo chassis for somewhere near $400 to $450.

How Long Does it Take to Build One (or two)?

This depends on how much time you have on your hands, how much knowledge you already possess, and how much money is in your pocket. One you have all of the parts and knowledge, assembly is pretty straight forward. Looking back at the time I've spent, it took me about 3-4 nights (after the kids were in bed) to match & sort parts and populate my 3 boards. One afternoon in the lab to match my mosfets. Another 3-4 nights to join the heatsinks to one another (3 chassis worth). 2-3 nights to drill, mount, and wire up the output stages. One or two evenings to wire up the power supplies. Another (mostly) full day to drill, tap, and assemble the chassis. Another night to wire the output stage to the board, connect the inputs, the power supply, the switches, etc. Four more days staring at things, double checking the wiring, and generally being afraid to flip the power switch for the first time and see things go up in smoke (fortunately, this did not happen!). Another night to do the initial testing of voltage, bias, offset, etc. Another few night fiddling with bias points and power dissipation. Then more nights fiddling, tweaking, changing, measuring, tweaking some more, observing with a scope, listening, etc.

So, start to finish for me it was about two to three solid weeks of working in the evenings after the kids are in bed. In reality, it took me about one month for full assembly.

The nice part is that you can gather and assemble your parts slowly over time. Use this time to read, read, read (see the links throughout this document). There is much to learn if you are not an electrical engineer by education or by trade (like me). There is still much to learn even if you are so qualified. Other projects (and life in general) will likely take priority along the way. Make lots of notes and keep them organized (you are reading most of my notes now). I started following this discussion right from the beginning (May 2002). Now the calendar indicates that it is 2010! Go Figure... The intervening years have been filled with building a new house, the birth of three children, career advancement, and other curiosities... Have fun!

 

General Q&A Concerning: Adjustments/Checking/Measuring/Troubleshooting

How Can I Increase the Power Output of Grey's Original Circuit?

To increase power output (into the same load), you'll need to increase the number of devices, the supply voltage, the bias current, the closed loop voltage gain and the trigger point of the protection circuit (if you use it).

More simply, Nelson suggests going up to 66% AC Current Gain for low impedance speakers. For example, you could set bias to 3.5A , 66% ac current gain and 26w pd with only 4 fets giving 53W into 4ohm loads.

Tweak From Grey:

If you want to change gain, you've got several options. You can:
--Change the feedback loop (R16 & R30)
--Change the ratio of the resistors in the inputs (R18/R19 & R20/R29)
--Change the loading of the front end differential (R23 & R25)

Note that all of these effect other things. The feedback ratio, for instance, will change the gain, but will also change the distortion characteristics. If you get really radical and throw in a cascode, you're going to have to do some serious thinking about how that's going to influence the DC offset on the load resistors ...which in turn biases the output MOSFETs.
The design looks complicated, but isn't really. Just take a blank piece of paper and cover half of the schematic. Suddenly, you're back to something very much like the original Mini-A from whence it came. (I have a master plan, you see ...everything dovetails together...) If you've got sufficient parts on hand, you can build a Mini-A and test any ideas on it before trying to fit them into the Aleph-X. Remember, I tried to make the Mini-A a 'cheap' project. There will be some things that don't translate exactly--in particular things that relate to output DC offset, but many things will and the Mini-A is a great test bed for possibilities.

--------------

Measurements (from Wessol 12-30-02 Aleph-X thread)

The Beta AX lives...
Using Grey's original CCS.
From the beta schematic I changed the following:
R1/R4 and R44/R45 total 62-ohm instead of 31-ohm
C2 & C4 are 5pF instead of 10pF
R19 & R29 are 10K

The Results:
PS Voltage 14.3VDC w/ 220,000uF per rail
Avg DC offset is running 1.4mV
Avg Abs DC offset is 47mV / 55mV
R23 4.24V
R25 4.31V
R5 .499V
R40 .501V
R6 .504V

------------------------------------

From Chad :

I will try to be as specific as I can be, and I will be referring to Grey schematic for nomenclature:
First of all I setup a proper rig to test match the mosfet, and matched both the output and differential mosfets. Contrary to Peter findings I got a pair of 9610 with Vgs within 2mV over a 1 minute monitoring period and that was enough to get a corresponding voltage drop on r23 and r25 when I soldered them in place back to back with a sil-pad in the middle. The Output mosfet could not be matched to better than 10 mV on my sample of 20 I could only find 4 between 3.85-3.86 V Vgs. This translates in about 40mV DC offset. That IMHO is the result of different transconductance. Maybe this large effect would average out if several mosfets would be paralleled resulting in a much lower relative DC offset.

TO FIRST ORDER things that DON'T MATTER AT ALL:

-Value of out to gnd resistance.
I am assuming that some type of value is needed to control the DC offset between + and - out this to work HOWEVER, short of eliminating that completely, playing with values between 20 and 100 ohms made no difference at all in the absolute or relative DC offset. THEREFORE, this resistor might as well be 100 ohm to save some power. It is POSSIBLE that while music is playing the value of the DC will be kept to a more constant value with a lower resistance but as I said this is all to first order.

Things that matter ALOT:

VR1 and VR3 have a large impact, since they set the bias current through the output transistors. You can measure this across R6 / R41. I was a little surprised to find that very large variances in this current had little to no effect on the DC bias at the outputs. In any case, I'm running 2.5A bias per side for a total of 5A bias per channel, and I haven't had any difficulty with the standard component values. You may want to increase the value of R10 and R39 if you are running high bias currents.

-Values of R24 to R26 and V2.
R24, 26 and V2 decide how much current is going through the output differential.
R23 and R25 control the threshold at which Q2 and Q11 will start conducting.


-Values of R15 and R32. The voltage drop across these resistor control the conduction of the Q1 and Q10 mosfets. Not as effective as r24, 26 and V2 but somewhat effective to control absolute DC value at the outputs. The reality all these resistances need to be adjusted simultaneously I have a 2k trim pot. in right now which is turned all the way up to 2k.

I used a strange CS a hybrid of HH and the standard that used an LM329 instead on the 9.1 zener and a 5k to –15V for r17. The current R24 and 26 are 205 ohms and V2 is 200 ohm. I initially used 301 ohm for r23 and 25 but reverted back to 390 ohms. It should make no difference at all. At one point I was listening to the sound of the input differential biased to 15mA just for kicks. On my test crappy speaker it sounded ok.
I went back to the 100K and 10pf loop (r16 c2) from the 200k and 5 (more XFB?).

Here are some voltages:
PS voltage +/- 13.5 VDC on my test setup. As soon as the VM guy replies to my email I will order the real transformer.
+/- DC ~50-70 mV when playing music with balanced in and 6.6 uF input caps on both hot and cold.
The absolute DC hangs around 0 and +/-35 mV.
Source R drop top 490 mV, bottom 460mV
Vgs top 4.2 V, bottom 4.07V
Differential load drop (r23,25) 4.510 and 4.512V.

At this point it would be nice to figure out how to mess with R9,10, and 13 to increase the bias on Q2 and Q11 thereby evening out the source resistor drops. Since adjusting the CS resistor affects both the lower and upper output mosfets AND r15 and r32 take care of the top bias only a third way of adjusting the bias of only the bottom mosfets could prove somewhat useful. Maybe at that point the resistor to ground would just be there for show.

At this point the only comment about the sound is that it was really audibly distorting when the absolute DC offset was 8V and now it sounds a heck of a lot better. Very promising actually.

Yes the rectifier need it's own heatsink. So far I don't hear any turn on-off noises or hums. I have a high efficiency test speaker that I want to try before I really say.

-------------------------------------

Problem: Amp seems to be oscillating - getting a motorboat sound on the speakers

Solution: 1) I used 3n3 for C9 and C10 (4n7 also worked) to get rid of oscillations. 10pF for C2 and C4 (5 would probably be enough). C7/8 are still not used and the amp works fine - William. or 2) Motorboating occurs when bias pots are turned all the way down. I suggest playing with V1, V2 and see what happens - Hugo.

--------------------------------------

R1/R4 getting hot means that too much absolute DC is present at the outputs. (That is from output to ground and should preferably be less than 1V). Assuming your amp is technically OK, you should be able to adjust that DC with VR2. - NetList 08-20-2004

Make sure your current source and diff pair works. Check the voltage across R23/R25. Something between 4V and 5V is OK - NetList 08-20-2004

Voltage over R2, R3, R42,R43 drops only when current is flowing through the speaker and that means this is only AC, so in normal quiet state there is no DC drop at those resistors. Tell me the voltages over: R5, R6, R40 , R41; they should be equal around 400..600 mV depending on biasing. But they should be as close together as possible, assuming the Q1, Q2, Q10, and Q11 were matched properly and VR1 & VR3 were setup properly. - piotrzurawski on 08-21-2004

If absolute dc offset gets too high (above Nelson's 66% ?) my A-X starts to hum - dieringe on 09-22-2004

Problem: Heat sink stays cold after 5, 10 minutes; R23 and R25 measure only about 3.3V : Solution: VR2 somehow end up at the highest 200 ohms. After I found this out and turn it all the way down, I got my Q7 up to 4+V and the heat sink starts working now....

Question: What is the general opinion on the compensation caps/ are there already values known to be used with the boards? I could start without them and ad when necessary. How can I test this, what input signal and what load should I use? - William. Answer: I usually start without them, but you'll need a scope to watch the circuit. As a rule, an ordinary resistive load is adequate to provoke oscillation. If you are looking more extensively, try hanging a .047 to 1 uF across the output terminals. If you really want to see how bulletproof it is, drive an unterminated length of Litz or Mogami or similar low inductance/high capacitance cables. - NelsonPass on 12-14-2004

----------------------------------------

Problem: (This is from a discussion of the A-30 circuit) I fussed a little and it played for a couple seconds after a power cycle then went quiet. Then I noticed that it smoked a little RC circuit I placed on the GND connector of the PS boards

Solution: (Nelson's reply) When the RC to ground at the output is cooked, it always means high power ultrasonics, almost always system oscillation.
I say system because an amplifier's internal oscillation is almost always at low power levels, but if you bleed the output of an amp back to the input, you often see full power up around 100 KHz or so. Most often this is due to bad input grounding on the cables. It could also be the MOX putting out power ultrasonics.

-------------------------------------------

Problem: Powering up I can adjust the bias just fine and have 0.44v over the 0.22ohm source resistors, but then the problem I'm having is that I cannot fix the absolute DC offset with the 3rd pot, both sides measure ~ -15.5V for the entire travel of the pot. I've replaced the differential input and current source fets thinking it is probably them, which didn't help. All parts were hand matched very well

Solution 1: -15.5 volt means the output fets are completely open. This means that the current through the current source is too high.
You can measure it by looking at the voltage over the drain resistors from the input fets. Try to raise the value of the resistor parallel to the 3th pot and turn the pot to max resistance. This should do the job.

Solution 2: I got the same, and I found the Zener Diode of the input current source was installed in wrong direction.

---------------------------------------------

Question: Has anybody tried to increase the input impedance for aleph x? I'd like to change it from 20k ohm balanced to 40-47k ohm balanced. I have built the 100W amps with 220k ohm resistors instead 100k ohm to have better input sensitivity. Lets say it is kristijan schematic, only 100k ohm changed to 220k ohm. In what direction should I go when replacing 10k ohm input resistors with 20-23k ohm?

Solution: Input impedance is roughly sum of two series resistors (+ and - side); you can up them to, say, 27 K -but you must scale up 100K resistors for same factor (2,7x) . Also you must scale up (2,7x ) 10K resistors from inputs to gnd. After that you need to scale down (2,7x) 10pF caps across (now) 270K resistors. I found using a high input impendence caused a phase variation between the channels of my aleph-mini amp. It was only a few degrees at 20kHZ, but could be seen very clearly at 100kHZ. Additionally, increasing the input impedance will decrease the bandwidth due to the Gate capacitance of the MOSFETs, though Nelson indicates, "Not by much, as it turns out."

-----------------------

From Zen_Mod about bias setting and DC offset:

when you have "proper " bias through outputs , but offset is positive - that means that Aleph CCS-es are more "open" than needed ( programmed for more current than is going through lower mosfets) .

in that case you need to close down Aleph CCS-es , decreasing value of b-c resistance , of their programming bjt's ;
so - decreasing value of R11+VR1 chain ( on one side ) and R33+VR2 chain ( on other side) ;

if happens that you knock VR's to zero , that means that you need to decrease value of their adjacent resistors

 

Some General Construction Tips:

Yes, many of these were learned to hard way, so the reason for this section is to try to prevent you from going through the same frustrations that I did! In no particular order, here are a few things I've learned by working on this project off and on for the past few years.

1) Measure each and every part prior to including it in your circuit. Mistakes happen at many different levels. In one case, I was sent resistors that measured 220k ohms when I ordered 220 ohm resistors. Also, in many cases, it is useful to match parts across a stereo amp or match components in the output stage with one another. Another benefit of doing this is that it will decrease the likelihood of making a dumb mistake like soldering the wrong part value into your circuit.

2) If building more than one channel at a time, install all of the same parts at the same time. For instance, if you are building 3 monoblocks (like me), install the R1 resistors on each of the three boards before you move on to installing R2 on each of the boards. This helps to reduce mistakes because you have just measured each resistor again before installing it (to verify that you are installing the correct part), and you can visually compare each of the three circuit boards after installing each part to make sure that you have installed the part into the correct position. Seems silly, but it helps!

3) Once the amp(s) are working, make changes to only one channel at a time when working on a multi-channel amplifier! Often times, a change that you intend for the better results in the introduction of a problem. This is especially true when moving wires inside your chassis in order to reduce or eliminate hum. Other changes, if improperly applied, have the potential to let the smoke escape from vital components of your amp - its far better to toast one channel rather all of them at once! As a parallel to this rule, power up only one channel at a time after making a change...

4) Be especially careful when soldering to avoid cold solder joints. A cold solder joint occurs when the materials that you are soldering together have not yet reached the melting point of the solder, thus you have a hot blob of solder on a cold piece of metal. The result is a solder joint that is neither electrically nor mechanically solid. This typically happens when you are soldering a small part to a much larger hunk of metal that hasn't fully heated up yet (hence the description "cold"). If you find yourself having trouble making good solder joints, consider a higher wattage soldering iron. For dedicated soldering on a PCB, a 25-40w iron should be all you need. A cold solder joint can cause a great deal of trouble because when visually inspected, it looks good. When the component itself is measured with a meter, it measures good. But, the connection with the rest of the circuit is intermittent at best. I've been soldering for years and have never had a problem with this before, but it caused some frustration in fixing a problem with this amp. A cold solder joint on one of my resistors resulted in running the full rail voltage to the speaker outputs which would have immediately fried any speaker! Ironically, I suspect the cold solder joint resulted from the use of a soldering heatsink that I had clipped to the leg of the resistor to prevent the resistor itself from overheating while I was soldering it. This problem is easily solved by carefully and methodically re-melting each of your solder joints with the tip of your soldering iron and then letting them cool again.

 

Some Useful Articles:

Electrical Grounding 1, 2

Building Power Supplies 1, 2, and 3

Power Supply Decoupling

Heat sink selection & transistor mounting

Links to other power amps

ePanorama.net

Surplus Parts Vendors:

Here are some surplus parts sites in the US: All Electronics, BG Micro , Tech DIY, Alltronics , Apex Jr , Excess Solutions, Electronic Goldmine , Halted ,Surplus Sales

MPJA, MECI, Skycraft, Fair Radio Sales, Alltronics, Jameco, HSC Electronic Supply, Brigar Electronics, Cascade Surplus,

General Retail Vendors:

Digikey, Mouser, Newark, Allied, MCM Electronics, Galco,

Boutique Parts Vendors (exotic resistors, capacitors, etc.):

Michael Percy Audio, Borbely Audio, Angela Instruments, Audio Note, The Parts Connection, Welborne Labs, Williams Hart Electronics, Sonic Craft, Madisound

 

Good luck!