diyAudio (
-   Pass Labs (
-   -   Pass Son of Zen w/current sources (etc.) (

GRollins 21st September 2001 08:11 PM

This is the first post of an intended series relating to variations on Nelson Pass's Son of Zen amplifier circuit.
First off, credit where credit is due: I'd like to thank Nelson for sharing the circuit with the DIY community. For that matter, I'd like to thank him for sharing so many of his other ideas with us, as well. These circuits provide an interesting playground for the imagination, as well as good sound for the ears. They are (relatively speaking) kind to the wallet, and are, if approached correctly, educational.
The original circuit:


The SOZ circuit is an amazingly simple circuit. A simple differential pair functions as a single gain stage. That's it. Finis.
(Those who wish to argue whether a differential is a single gain stage can exit through the door at the rear of the room. Bye.)
By definition, a differential is class A circuit, and therein lies the rub. The SOZ is a red-blooded, fire-breathing, carpet-singeing heat machine. It is the amplifier of choice for Eskimos, penguins, and polar bears who need a space heater that also happens to give off music. Others tend to look askance at amplifiers that require heatsinks the size of Volkswagens. It helps if you have a relative at the local power company who can get you kilowatts of AC for free.
Devising ways to get rid of the heat has caused many a sleepless night for would-be SOZ builders. The first thing most people think of is to buy standard passive heatsinks. Shortly thereafter, they come to the realization that life isn't simple; you can't just walk right down to the corner electronics shop and pick up a couple of large heatsinks for $1.98 each. First off, big heatsinks are hard to come by. Worse yet, they cost serious money.
What to do?
Well, one option is to rob a bank and go buy some fancy heatsinks.
Not recommended.
The local constabulary will be after you in no time, and I doubt that the judge will allow you to cart your shiny new amplifiers off to the jailhouse.
After careful consideration, I discarded this option. I chose instead to try water cooling the pair of Aleph 2s that I built. (Alephs suffer from the same tendency to melt inquisitive cats that the SOZ does. Depending on how you feel about your resident cat, this may be an attractive concept.) Not only did it work, it worked well. Better than I expected, in fact. See the Water Cooled Audio Circuitry thread elsewhere on this site for details.
But some people, mysteriously, just don't want 5/8" hoses snaking across their listening room floors. There's no accounting for taste, I suppose.
So that leaves us hunting for other options. Forced air cooling works well, of course, but still assumes that you can find medium-sized (as opposed to huge) heatsinks, and that you are willing to tolerate the noise of the fans.
Some have suggested Peltier modules, but to my knowledge no one has actually built a SOZ using them. Whether it's a practical option, I will leave to others to decide.
And, rather abruptly, we're scraping the bottom of the barrel. There just aren't that many ways to get rid of unwanted heat.
Back to first principles. Is there any way to reduce the heat produced by the SOZ amp? Well, yes. You could scale down the amp, but that's an unsatisfactory solution for those who were rather hoping to increase, not decrease, the power. A quick perusal of the schematic, however, suggests the possibility of substituting current sources for the lower 8 ohm resistors.
That brings us to our current discussion.


Don't get your hopes up. I'm not posting a complete, paint-by-numbers schematic. Yet. I will do so as soon as possible. In the meantime, I thought I'd pass along a few points from the notes I've been taking as I've poked and prodded at the original design.
--As folks who've watched other threads here may have noticed, I'm not a big fan of current sources. Don't get me *ideal* current source is a wonder to behold. They work well in simulations. They accomplish marvelous things.
In theory.
In practice, they're afflicted by capacitance and other annoying properties that keep them from realizing their potential. That said, they have their uses. In this case, we're going to try grafting one onto the SOZ in order to reduce the power dissipation. Those who feel inclined to debate the virtues (or lack thereof) of current sources are invited to do so elsewhere. There are already several threads on this site for this topic. Go resurrect one of them.
--The amplifier is not going to be a dial-a-wattage circuit such as Nelson suggested in his article. In fact, I'm not sure that the stock version would work well in that configuration. Why? Device matching. The output devices used in the SOZ need to be matched. To match a device, you pick a target idle current and match the output devices at that current. Unfortunately, in the real world, devices aren't linear enough that a pair matched at, say, 1 amp will stay matched when you ramp up to 2 amps. Yes, with a large enough box of MOSFETs, you could probably come up with a pair of devices that would track well at any given current. You could also grow old trying to find that magic pair. It's your money. It's your time. If you feel that it's worth the trouble and expense then go ahead. However, most people really only need an amplifier that has a certain fixed power output. This implies a fixed idle current, which in turn means that we can find a decent pair of semiconductors that will do the job. As a related point, if you're driving the circuit with current sources, you'll need a way to adjust the output of the current sources as you turn the power up or down. It wouldn't be that difficult to arrange, I suppose, but to me it seems like a lot of effort for little return.
--The current sources either need to be carefully matched, or you will need some way to adjust them. This will be your DC offset adjustment.
--The current sources will be dissipating heat. From what I'm seeing so far, it's somewhat less than three times the dissipation of the output devices. Plan on more heatsinks.
--I'm using parts I have on hand, so far. I have not (yet) gone to water-cooling for the SOZ. It's air-cooled; no fans. I'm using Motorola MTP7N20E MOSFETs for both the outputs and the current sources. Resistors vary. I've got two of the big Dale 8 ohm 1% 250W load resistors that I typically use for testing purposes. Here, they've been pressed into service as part of the circuit itself. The 1 ohm resistors are either NTE 5W ceramics or some 1 ohm Dale wirewounds. The power supply is primitive at present. Just a simple transformer/rectifier/cap job that I threw together. Nothing fancy. Yes, there's some ripple. I intend to toss some more caps (and whatever else I can find in the junk box) in there to smooth out the ripple. There are only so many hours in a day. I'll get it done eventually.
--The gain of the SOZ circuit is low. The stock version I whipped together had about 14dB gain (single-ended). The current source version is even lower, a little over 6 dB (also single-ended). If you intend to drive a SOZ directly, you'll need some voltage swing from your preamp.
--The amp is small...on the order of 2 or 3 watts output. It has only one channel. This keeps it from taking over my workbench. It also reduces finance problems. I'll consider building a big one if I like the results.
--I have several things planned after getting the kinks out of the amp as it stands now. The first one is to mate Nelson's Balanced Line Stage (aka Bride of Son of Zen/BOSOZ) to the SOZ in order to get some gain into the poor thing. It will also allow for better matching with a single-ended source, functioning as a phase-splitter. I also plan to parallel two SOZ stages (cutting back on current in each in order to lighten the load on the devices). This will require more matched devices, but assuming that you've got enough on hand, there's no reason you couldn't keep paralleling them to reach any arbitrary power level. What this does to your electric bill is your problem. I've got a few other ideas in mind, too, but this will be enough to use up more free time than I can reasonably expect to have over the next month or two.
--No, this is not the X/Aleph hybrid circuit I proposed in Petter's X thread. That's a separate project that I will approach later.


[Edited by Jason on 09-27-2001 at 01:26 AM]

Petter 22nd September 2001 09:24 PM


This would be just about the simplest circuit you can make + you get the benefit of the current sources over the Zen (efficiency). I might try this one myself once finished with my other projects. The plain Zen does not cut it for me.

Let us know how you get on!

Re Peltiers: Remember these things also dissipate power ...

It would be quite interesting to operate this amplifier in source follower mode. Since the output power is so low, you could probably do that without much trouble. Alternatively it would be interesting to operate as you probably have done with one current source on bottom and dual ones on top. A third alternative is to decouple the stages and run them with a voltage gain of (Rdrain/Rsource) per half.


GRollins 22nd September 2001 10:30 PM

The idea of a single current source is attractive, but will limit the amplifier to low power output, since the dissipation in the current source(s) is(are) so high. As it is, even with two current sources (decoupled via the two vertical 1 ohm resistors and the horizontal one bridging the two halves), you'll outstrip the thermal capabilities of the transistors in the current sources before you get near those of the output devices.
Not to worry...I've got a trick up my sleeve.
I unearthed a pair of Dale 2.5 ohm 1% 25W resistors (ah, the joys of a well-stocked junk box...) this afternoon. If I stack them on top of the current sources instead of--or perhaps in addition to--the vertical 1 ohm resistors, then part of the voltage drop will occur across the resistors, instead of within the current sources. The more the resistors do, the less the current sources have to...yes, I'll be getting back into more heat dissipation (boo!), but presumably regaining gain (yea!). I was begrudging the drop in gain, as I have plans to burn off some of that gain later.
(Hint: Think X. Yes, Sam, you were right...)
Petter, you've spent more time on the X circuit than I have. We may need to put our heads together on this. I have a few ideas, but haven't gotten to the point where I can try them yet.
I hope to spend some time on the critter this evening--time to warm up the 'scope. (Yes, literally. I've got an old tube [they sound better, you know...] Tektronix that takes a few minutes to settle in.)


GRollins 26th September 2001 06:08 AM

Before I go any further, I'd like to acknowledge Jason's kind assistance. He did the conversions from the original BMP files (the best I could do on short notice) to the PNG format used here, and offered to host the resulting files.
This is the schematic of the circuit as it stands now:

<img src="">
The upper portion of the circuit is still recognizable as the SOZ. I have left the part numbers unchanged to aid comparison. The lower 8 ohm resistors have been replaced with current sources.
Incidentally, I said above that the Dale 2.5 ohm resistors I had unearthed were 25W--they're actually 50W. Either they grew up, or I need new glasses.
I'd like to caution anyone who might wish to use this circuit as-is to try to build a higher powered version of the amp that the current sources will not react properly to higher rail voltages. To reset the bias currents, you will have to change R18 and R19. Failure to do so will starve the output devices for current, leading to poor performance and the downfall of civilization.
So how much current do you want?
Assuming that you want something similar to the current in a stock SOZ amplifier:

rail voltage/10.8 = desired current per leg

Note that rail voltage means *one* rail, not the combined voltage of the positive and negative rails. For example, suppose you wanted to use +-30V rails:

30V/10.8 = 2.78A

so you would set the current in each current source for 2.8A, or so.
By changing the value of R18 and R19 according to:

.65V/A = R

There are three things to note if you should decide to forge ahead and make a more powerful version of the circuit.
One is that the above formula will only be approximate, as the collector resistors for the MPSA18s, R12 and R13, also effect the current. As the rail voltage rises, you might want to put in proportionately higher values for those resistors.
The second thing is (and this is true for this circuit at any size), due to variations in the parts, you will probably need to trim the values of R18 and R19 by placing resistors in parallel with the resistor of the weaker side. Use small values--probably 47 ohms or less. For example, I have a 3 ohm resistor in parallel with one of mine. Matched devices and tight tolerance parts in the current sources will keep this kind of fiddling to a minimum, but it isn't really neccessary as long as you have the time and patience to adjust the circuit. You can measure the current in each leg at any number of convenient points. The 8 ohm resistors at the top will probably be easy to get to.
The third thing is to substitute something heftier for the MOSFETs. The Motorola parts are simply what I had on hand. You are welcome to use anything that suits you, but pay attention to the thermal demands of the circuit.
The actual balance between the two halves of the circuit will be your DC offset, so you can approach the problem by attaching the leads to your meter across the speaker outputs. I would not recommend hooking a speaker to this circuit until you have checked the DC offset. Unpleasant things happen to speakers that are subjected to large amounts of DC--the cone lurches full transit, followed shortly thereafter by a thin whisp of acrid smoke as the voice coil sidesteps into the fifth dimension. Use a load resistor to simulate a speaker. Any small value (say, 2 to 10 ohm) 5W ceramic resistor will do just fine. Just whatever you have in your junkbox.
This circuit is very much a work in progress, but is functional as it is. If you build it, music will come out.
I'll be working on a few refinements as time permits.


[Edited by Jason on 09-27-2001 at 01:26 AM]

grataku 26th September 2001 04:35 PM

Grey, it's very nice of you to do all of that!
High on my wish-list is a small balanced system with balanced source, the balanced linestage, and MAYBE a 10 W SOZ.
Could you tell me again how much power does the CS SOZ version save? From what I remember about 20% right?
Did you notice that part of the text is hidden below the pics?

tschrama 26th September 2001 06:55 PM

Hi... I think you could save some dissipation if the minus voltage rails are raise to -12 or -9 or even minus 6 volts.. this would not effect the output power of the amp.. at least that's what I understand of the circuit..... the dissipated power saved can be easely calculated if you konw the Iq: 2 amperes with a minus voltage supply of -9 instead of -18Volt saves 18Watts... some 25% (?) .....


promitheus 26th September 2001 09:05 PM

I think the best is to reduce the negative voltage rail. It has nothing to do with the output power or the quality, itīs just there for simplicity of the power supply.

GRollins 27th September 2001 04:08 AM

Sorry about the text being hidden. I swear I thought I got that settled last night, then today it's creeping back up under the schematics. Help me keep an eye on it.
How much wattage savings?
Simply going from a stock SOZ to one with current sources in place of the lower 8 ohm resistors (R3 and R4 in Figure 1)...well, it depends on how you look at it. If the rails are held constant, and the current through the circuit is the same, then you're not saving any heat, really.
What happens is that the voltage drop across, say, R3 is about 15V, to round off the numbers. If you use a current source, the voltage drop is on the order of 9V or so. If the current is the same, let's say 2A, then:

Resistor--> 15V*2A= 30W dissipation

Current source--> 9V*2A= 18W dissipation

for a heat savings of about 40%. However, the output devices cheerfully expand to take up the 'missing' 6V so now *they* are generating more heat. What you get in return is more voltage swing at the outputs. Or you could drop the rail by that 6V, and save the heat...your choice. Either way, it is more efficient in the sense of the amount of heat created vs. the amount of wattage available at the outputs.
Note that the 1 ohm (or in my case 2.5 ohm--R5 & R6) and the upper 8 ohm resistors (R1 & R2) are going to dissipate the same amount of heat as long as the current remains constant, so the total increase in efficiency is going to be less than 40%.
thijs & promitheus,
I tossed the idea of asymmetrical rails at Vince (vdi_nenna) back this past winter or spring sometime. It came up again in another thread not long ago...and had surfaced at random points between those times. Actually, that's the reason I'm undertaking this part of the project. A lot of people are interested in the concept, but are stymied by the actual execution of a current source. I figured I'd give things a boost by getting a working circuit posted, just so we could get past having a thread every month or so about putting current sources in a SOZ. My own interests lie somewhat further down the track. Given my feelings about current sources, I may very well pull the fancy stuff out and plug in some resistors (thus going back to a stock SOZ) once I get the current source bugaboo laid to rest.
Personally, I'm not real worried about heat the way others are. All I have to do is tag the thing into my water-cooled system if it starts getting too hot. So far, heat hasn't been that much of a problem.
As for the actual value of the negative rail, it will need to be more than -5 to -10V, as the MOSFETs are enhancement mode devices. The sources are running about 5V below ground just to bias up the output devices. The signal will vary the level of the sources somewhat, at least with a single-ended input. And, last, but not least...the current source itself needs a little elbow room. I haven't measured the minimum voltage for the current source to get its act together, but it's probably on the order of about 5-7V. Add all that together and you're looking at about 15V or so, maybe even a bit more. Right now, I'm only running 18V rails, so I haven't bothered to float the ground down a mere 3V or so; it's just not worth the trouble.
I'm at the limit of what I can do with this transformer. It's out of my junk box, about 30V CT, probably about 100VA or so. The rails, unloaded, are +-21V, but when it's up and running, they collapse to +-18V. Clearly, the poor thing needs to be given a rest.
Not to worry. I've got some larger ones on hand. I just wanted to see what I could do with a minimal power supply. The answer: Minimal output--about 2 or 3W. No surprise.
The Things To Do list (the short term one) includes:
--Build a bigger power supply. Once I get wider rails, I can float ground and have asymmetrical rails. Another option is to use two separate transformers, a smaller one (but still high current) for the negative rail, and a larger one for the positive rail. I'll see what my junk box has to offer.
--Pull out some heftier MOSFETs. I can get a bit more out of the MTP7N20Es, but not enough to make it worthwhile.
--Match said MOSFETs at a fairly high current. The ones I'm using now were matched at .5A, but I'm running them at about 2A. They aren't tracking cleanly. Again, no surprise.
--Devise a decent adjustment system for the current source. Paralleling smallish resistors gets tedious, but you can't just run down to the store and buy a .5 ohm pot. I've made some notes on ideas, not the least of which is to use a current source that uses a separate voltage as its current reference (but not error correction) so that I can vary that more easily. Actually I may pursue that on the present setup, as it won't require higher rails and such to experiment with.
One thing at a time. I had to spend part of today checking on one of my bee hives. UPS steadfastly refuses to deliver the 48 hour days I ordered yea, these many years ago. Given time, I will get there.


Jason 27th September 2001 06:28 AM


I resized the images making sure there was as little loss of quality as possible, so that it would be easier for people with 800x600 monitor resolutions to see the schematic in one hit. Let me know if that's an issue!

promitheus 27th September 2001 06:54 AM

The pictures look ok now.

All times are GMT. The time now is 11:15 PM.

Search Engine Optimisation provided by DragonByte SEO (Pro) - vBulletin Mods & Addons Copyright © 2018 DragonByte Technologies Ltd.
Resources saved on this page: MySQL 18.75%
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2018 DragonByte Technologies Ltd.
Copyright ©1999-2018 diyAudio