Zen vs. Douglas Self Trimodal Power Amplifier?

Hello everybody! My first post on this forum. Fairly technical mood around here but I see a lot of common sense posts also that I can undestand. I see a lot of Nelson Pass fans here. I was looking through available PCB's and came across this little wonder.

http://www.dself.demon.co.uk/trimo.htm
and
http://www.signaltransfer.freeuk.com/trimodal.htm

It lets you select mode of operation between A, B and A/B. I guess for critical quiet listening class A would be prefferable, but when you want some power you can switch over to class B and give up a little quality. Anybody heard/built this amp? Any thoughts? I think Monarchy Audio sells something along those lines and it seems to be respected in certain cicles. But then again some swear by Bose!
 
I havent built the trimodal but i have the book from Self and have read about it. It is a good concept but it is too complicated to be good quality. It al depends on what you want. The Zen amp is not good for most cases. Like I mean if you want very high sound levels or to use it with with unsensitive speakers. It is really a good amp for low levels or for active speakers.
 
Have you read all Self articles on their original form (EW&WW magazine) with all the follow ups? And then the book?

I do agree that the Trimodal is far too complicated and that simpler can be better (not always).

What I intend to do is modify a Pioneer receiver's amp section, which resembles Self's design a lot, though lacking his "tricks" to improve distortion figures of course. Only a few pcb tracks have to be cut to add the improvements.

Self does not approve on what I want to do, though probably because he may be afraid that I may sue...

In any case, the circuit I will be using is his "simpler" one. The output will be regular emitter-followers (EF) arrangement instead of complementary feedback pair (CFP), mostly because it would be tricky to modify that on the Pionner's pcb.

All parts will be improved too, and the first stages will get a regulator.

What do you think?
 
Something that Douglas Self never quite makes clear (nor does G. Randy Slone) is whether they have capacitance downstream of the regulators. I gather they do not. Any circuit I've ever heard which directly dumps the output of the regulator into the active circuitry runs out of steam when asked to produce any sizeable dynamics. Some equipment (and I'm thinking in particular of Jeff Rowland's stuff, here, as I believe he uses that strategy) sounds really, really sweet. Your first impression is that this must be the finest sounding stuff in the universe, then someone does something rude, like hitting a drum, and...nothing happens. It's as though there's a limiter unit inside his gear.
If that is, indeed, what Self and Slone think regulation is all about, I can understand their disdain for regulated rails. Only...it's not that difficult or expensive to put a couple of caps after the regulator.
The short answer...go for it...but be careful. I've never had much luck modifying existing PCBs as the traces never seem to be anything remotely like what I want, and it would take an entire spool of wire to do the little point-to-point patches necessary to bring things to a conclusion.

Grey
 
Nobody would say that a regulated power supply is a bad thing. The reason that no one uses one in an amp (maybe only in very expensive, complex units) is because a regulated power supply is as hard to make as the amp.

Say that the amp is rated max 3 amps output with 4 transistors in the output/channel. To make a regulated PSU for 3 amps that is capable of going up to high frequencies around 20-100 KHz is not easy. It´s not like making a power supply for general use. You need at least 4 transistors and double the heatsinks again for the power supply in the worsed case. I say you would need more so the PSU is fast and doesnt reach it´s max capabilities making it hang over. The design has to be very good or else the improvement isn´t a lot better. Maybe you could have the same improvement in sound with just putting double capacity after the rectifiers. We are talking of a cost here of plus 60-80 %.
 
I, myself, haven't found things to be as bleak as promitheus does, particularly as regards to price. The voltage regulator doesn't have to drop as much power/heat as the output because, although clearly the current must be accounted for, the voltage drop isn't nearly as bad. A decent regulator generally has a dropout of, say, 1 to 3V, so in theory you could get by with, let's say, 5V of headroom. In the real world, I'd provide more (on the order of 10%, work backwards from a reasonable voltage sag in the mains line), but certainly not the 20-30-40% that Self/Slone seem to think is necessary. Since we're not dropping as much voltage, the power dissipation is correspondingly less.
Self & Slone seem to be *very* stingy with their capacitors, acting as though it's profligate waste to put anything more than 10,000 uF caps into the main bank. No wonder they feel they need such headroom on their rails! What they regard as normal, acceptable ripple makes me cringe. In their view, it's all taken care of by the CMRR and NFB, but fer cryin' out loud, guys, we're DIY--what's a few more dollars on bulk caps? We're not responsible to anyone but ourselves for the amount of money we put into our projects. Even when I'm running unregulated rails--hell, *especially* when I'm running unregulated rails, I go with large filter caps in the main bank.
(Actually, I suspect that a lot of their intransigence on the audibility of things stems from their attitudes towards their power supplies, and the fact that they don't seem to want to bother matching their gain devices. [Yes, they will--grudgingly--match the devices for the input differential.])
Regarding the frequency response of the regulator...I quite agree with promitheus that a wide-bandwidth regulator is a good thing, but here again I feel that a bank of good caps after the regulator (I'm prone to using lots of film caps at this point in the circuit) helps take care of the high frequencies. Caps here also take care of any residual noise generated by the regulator itself.
In any event, as I understood carlmart's post, he's only planning to regulate the front end, which is a considerably less daunting thing, current-wise. Perhaps I misunderstood.
In regulators, as in other places in audio circuitry, I find that less is more.

Grey
 
regulate yes, regulate no

First of all: can someone explain to me what those funny message icon globes mean? I must be making a face very much alike those when I look at them, trying to guess which should I pick. Today I went for the toothed smile...

GRollins is right when he says that I am intending to regulate just the low current side of the amp, that is all but the output transistors. Full regulation is something I am still not too sure about.

I certainly do not agree with Self (or Slone if he says the same) over the regulation thing. The regulator should be good though, as well as the parts you use. First stages voltage, after regulation, should also be higher than the output stage unregulated voltage. And that is not easy to implement. Many people, like N.Pass simply go for a voltage doubler and then regulate that, so as to be able to use the same transformer.

Some weeks ago there was a 3-part piece on <http://www.tnt-audio.com/> on "Solid State Power Amplifier Supply", by Dejan Veselinovic, that I think should be read by many people. Myself always learns something new on this question.

What type this regulator should be, totally discrete or using 3-pin regulators, may not be the issue. On my case I have had very good luck using 3X7 types. Those are the ones I intend to use on the Pioneer's regulator, even if left "floating", that is unprotected from an eventual short circuit. That is because those types have a maximum working voltage of around 40v: within that they are fully protected. But you can use them on higher voltages as long as the difference between input-output is never greater than 35v or so. No protection though.

There are new regulator types, like the LT1085 and similar, that supposedly improve on the 3X7, but the latter are the minimum you should go for. Much better than 78XX or 79XX families.

To go for a full regulation, that I may try some time to see how it sounds, you need plenty of space and current to spare. In this case I wonder if you don't limit dynamics somehow.

But larger (and $$$) models from Krell or Mark Levinson are wholy regulated, if I am not wrong. And they get good reviews

When I make up my mind and go for a try I would like to start by using James Boak's and Kit Ryan's designs published in The Audio Amateur some years ago. Builders reported great results from using them.

This year there was un upgrade article on Audio Electronics (former TAA and present Audioxpress), by Walt Jung, on his top grade supplies published some years ago. Pity these supplies are designed for lower voltages, up to +/-24v.

In any case, over this regulators question, a friend of mine says they are not the right way to go because regulators use feedback to lower impedance and that feedback gets imprinted on the sound. A valid question that I can't really say much about, except that I have never listened to unregulated (amps or preamps) that sounded better than regulated (somewhere in the circuit) ones.

Until proven different, I think regulation IS the way to go. For me, an audio circuit using a regulated supply sounds better than an unregulated one.

Going for a high bandwidth on the regulator should be good too, but I don't know how to work on them for being stable and have a large BW at the same time.

This supply question deserves a quite thorough discussion, that hopefully we will carry on. Though maybe it exceeds the original issue of "ZEN VS DOUGLAS SELF TRIMODAL POWER AMPLIFIER" that started it.


Carlos
 
One big issue about full power supply regulation is
that the full supply voltage has to be dropped across
the regulator, the >transformer< being a weak link here.
Because of line and load regulation variances, one has
to drop ten or fifteen, maybe more, volts per rail to
stay within regulation from no load to full load.

While I've had past experience working with switch mode
regulation, I don't feel confident to attempt designing
a regulator based on this topology, nor on a resonant
mode converter, which would have lower noise.

Best one can usually do for a practical design is to
put plenty of iron into the transformer and plenty of
capacitance in the power supply. Decouple generously
on the driver board and sub-regulate there.

According to various sources, a diff amp with a current
source tends to have good power supply rejection anyway.

It's a bit of a complex subject since power supply design should also include issues like rectifier noise and RFI
intrusion from the AC line.

While toroidal power supply transformers are popular, one
weakness is that they do have good HF coupling between
the primary and secondary, unless electrostatic shielding
is incorporated. This is why I prefer split-bobbin transformers for preamp and such power supplies. But try
and get that with big iron!

There are various tradeoffs you have to live with in the
engineering world; those are some of the few I know about
and had encounters with when I had a real job as an electronics technician ten years ago. I got to look over the
engineer's shoulder as the new supplies were designed, debug
'em and fix 'em in production. I really miss having that
kind of a job. :(
 
Damon,
For me, the key to bulk power supply design was clearly expressed when you wrote "plenty of iron [in] the transformer and plenty of capacitance in the power supply." Self/Slone, for instance, believe in puny little power supplies, counting on CMRR to take out the nasties. Then they stick a class B output section in. This is a good recipe for collapsing rails, as at high power a class B output section will present a widely varying load to the power supply. If the power supply isn't stiff enough to handle it...who cares? Or so they (and many others) seem to say.
With a stout power supply, you can minimize the variation of the rail voltage, simply by having enough electrons flowing in from the transformer to keep the output happy without having to draw down the 'savings account' in the bulk supply caps.
Of course, one could always go for a class A output, which doesn't vary at all...
Yes, heavy iron costs money, but I used to have this little thing I'd say when I was faced with a customer who couldn't decide between two different pieces of equipment because of price. First, I'd warn them up front that I was about to play with their heads. (Honest Abe, that's me.) Then I'd say--Okay, I'm a magician. I have a magic wand. When I wave my magic wand the calendar will advance one year. (I'd wave my imaginary wand.) I'd then say--You're listening to your stereo. Now, are you saying to yourself I'm glad I bought XXX, or I *wish* I'd bought XXX? Then I'd remind them that pain (even financial) is temporary. A year later the amount they spent would only be a memory, but the stereo would still be with them. The moral of the story being: Save up, beg, borrow, or steal (okay, maybe not steal) enough money and build the bigger power supply. If you need to, rationalize it by telling yourself that you can reuse the transformer and caps (if they're big enough) in your *next* project. Think of the money you've saved!

Grey
 

blmn

Member
2001-02-01 2:43 am
.
Carlmart and Damon,

Actually, the big problem with regulated PSUs for the output stages is the cost and, as Promitheus said, the design itself must be done carefully. Since the design allows good responses to transient current demands and no problems with HF instability, no problem except the price and additional heat generation.

The standard textbooks use three factors to define how good is the power supply: input regulation factor, output resistance and temperature factor.

To be good the first one needs good mains connection, the second needs good transformers, wiring and high quality components and the last one needs stable components under many temperature conditions. All these things will change if you consider a regulated PSU, but the main idea is the same.

A good transformer is not always a bigger one. Good transformers are always done with good iron (low losses, good flux density), wire (good insulation and the correct area) and construction (as Damon said in his statement about toroids). If you want a really good transformer, forget about the cheaper ones (little or big), or design your own.

In regulated power supplies, the amount of voltage under the series regulator is (not only) a function of the current drawning by the load. So, in many cases you really need much more than a few volts to maintain a good regulation factor. I tested it myself, in some cases (40 Amps peak capacities from car audio SMPS amps, I needed 60% of over voltage).

Self just says the following about reservoir capacitances is his book (page 220):

"For amplifiers of moderate power the total reservoir capacitance usually ranges from 4,700uF to 20,000uF, though some designs have much more. Ripple current ratings must be taken seriously. It is often claimed that large amounts of reservoir capacitance give "firmer" bass; this is untrue for all normal amplifier designs below clipping."

He says very interesting things about decoupling caps considering all the amp stages. He didn't recommend underated transformers, cheap components or even less capacitance than necessary.

The calculation of reservoir capacitor is not a mysterious or a not determined thing, you can find it in many textbooks. Considering the ripple voltage needed, the peak current and using good quality caps, wiring and diodes these calculations work very well. As far as I know, the rule of thumb proposed by Slone is his book (1000uF per 10 watts), works well too and didn't disagree with the equations about the matter. Of course if you use larger amounts of reservoir caps you will have less ripple and a better capacity of response on high current demands (there is a practical limit for it too), but you must be aware of surge currents and your bank account. A good transformer is much more important, in this case.

Regards
 
DIY means Do It Yourself. A simple enough statement. It's a wide umbrella which covers everything from someone putting a $4.95 Radio Shack driver into a cigar box to keep the back wave from cancelling the front, to giga systems that would cost more, if bought as manufactured items, than any random three of us could afford if we pooled our resources.
Once upon a time, there was a man named David Hafler. He made his reputation with tube circuits (amongst them the famous Dynaco ST-70, which sold about ninety-nine million units way back when--some people still like 'em), then went on to start the Hafler company. Probably the best known Hafler product was the DH-200, a 100W/ch solid state amp, available as a kit or factory-made unit. A good, reliable, decent sounding unit that went on to sell another ninety-nine million units. I hope David got rich off of them. He deserved it. Good product at a great price.
But for all that it was a separate amp, it was mid-fi. I don't recall a single person who ever felt otherwise. The build quality was excellent (for the price), but the amp tended towards being somewhat harsh on top (a later revision, the DH-220, cured some of the sonic problems) and got positively rough when pushed hard.
The power supply was simple. A standard EI transformer (big toroids were rare in those days) through a bridge rectifier, into a single pair of caps. The value of the caps? 10,000 uF, one for each rail. Now, condsidering that the one power supply took care of both channels, that would work out as being right in line with Slone's rule of thumb for capacitance/watt. Only, the thing was, you could hook a meter to the rails and watch the voltage fluctuate as the music came and went. The first thing most audiophiles did was to tag more capacitance into the amp. Around here they used the Mallory CG series, but in other places they used Sprague or whatever was available locally. It made a huge difference in the ability of the amp to handle peaks...and firmed up the bass. Something even a teenaged kid could hear over a pair of old Advents. He borrowed the caps from his older brother just to see what all this cap stuff was all about. No ego, no preconceived notions, just a couple of borrowed caps and an open mind. The way I heard about it was that the kid brother didn't want to give the caps back when he found out what a big difference it made and the two of them had a bit of an argument about it.
Now, if a curious kid with a ragged-out pair of Advents and a DIY mid-fi kit can hear the difference caps make, it makes you wonder whether Slone ever listens to his stuff, or simply does it as textbook exercises.
Ripple calculations are pretty basic. Within reach of where I'm sitting as I write this I've got ten or fifteen books, all of which will have at least one chapter (or, in the case of Gottlieb, the entire book) on power supplies. In the book shelf on the other side of the room I have another dozen books which will also tell you how to deal with ripple. Kind of them. (Incidentally, for those who might wish for a basic education in electronics [as opposed to Horowitz & Hill, fer instance, which can be pretty opaque to the uninitiated], allow me to recommend the ARRL Handbook. That's the book the Ham radio guys use. No, I'm not a ham operator. Never have been. And yes, there's an awful lot of radio stuff in there which may or may not be of interest to you. But the treatment of basic electronics is second to none in terms of being accessible and practical. Those guys don't have time for a lot of fancy-pants theory, they want a radio that works. There are times when peoples' lives have depended on Ham operators when all other communications have failed due to natural disaster or war or whatever. You can get your theory later from H&H or Diefenderfer or whoever is popular this year. If I'd had the ARRL book when I was a kid, I'd have learned electronics ten times faster.) Anyway, I regard ripple calculations as a good starting point. A bare minimum. To get any performance out of an amp, you're going to need more. Mysterious? No. There's no mystery. When the rails sag, the operating point for the gain devices change, whether bipolar, FET, or tube. When the operating point changes, that causes distortion. (CMRR cannot guard against a sagging rail. That's not what it's for.)
Actually, I guess there is a mystery: Why would anyone defend a circuit that allows dynamic distortions of this nature?
Granted, short of a well-regulated power supply, there will always be some ripple; some collapse of the rails. It's also subject to diminishing returns, as are most things. No problem. But why sit at the bottom of the returns curve with a minimal power supply, when you can slide at least part-way up? (Subject perhaps to having food to eat, spousal-approval-factor, etc.)
Fortunately, there's room under the DIY umbrella for all of us, from speakers in cigar boxes on up. The choice is up to the individual as to how much performance they might want, but to deny that more performance is possible, whether the individual chooses to attain that performance or not, is foolish.
The owner of a Camaro knows that Ferraris are out there, whether he will ever own one or not. His Camaro may even have been hot-rodded to the point where it will accelerate as well as the Ferrari, but will it do everything else as well as the Ferrari? Nope. It's still just a Camaro. But the wise Camaro owner doesn't deny the existence or validity of a Ferrari just because he doesn't own one.
Incidentally, if I had a circuit that required a 60% overvoltage on the rail, I'd start looking back upline to find out where the bottleneck was. Given that it's a car system, there aren't too many places to have to look, although given the limitations of a car's electrical system, some of them may well prove to be intractable.

Grey
 

djk

R.I.P
2001-02-04 4:23 am
USA
It is a waste of time to regulate the current gain stage supplies.It is worthwhile to regulate the voltage gain stage supplies.The Hafler XL280 and XL600 did this.As long as the amp does not clip you cannot hear the difference between 1,000µF and 100,000µF.If you pre-clip the input signal with a soft-clip/limiter circuit your amp will always be under loop control and not make horrible sounds when the feedback loop overloads and the outputs stick on and go into common mode conduction.I sold a hopped up Altec 9440 to a local club.400W/ch with a single pair of 10,000µF caps.A 1.5Kva transformer.It replaced a Crest 6000 professional series.800W/ch with a total of 120,000µF in caps.A 3.6Kva transformer.It cost me $4 for the diodes and caps I put in the Altec.
 
djk,
I certainly agree that regulating the front end is a good idea. To me, given the relatively low currents in question, it seems to be a no brainer. Regulating the high current to the back...lotsa work. Whether I'm just being lazy or not, I've never gotten around to regulating the back end of my tube amp, although I've regulated the front end to a fare-thee-well.
For club use, which falls more under the category of pro gear (basically mid-fi stuff), I think I'd agree with you. Money and effort invested for little or no return. The speakers are *far* from hi-fi and the intended audience doesn't give a rip. Sell them another beer and crank up the volume another dB or two. Limiting the input is probably not a bad idea as DJs and/or the sound men for live groups are notorious for not caring. (Yes, the better ones do, but you can't count on having a good 'un at the controls.) Clubs are a great application for Crest, QSC, or Hafler gear. Lots o' watts in a compact rack mount chassis.
For home use? As Jan and I were discussing elsewhere--there are better things--although a case can be made for the struggling musician who uses the amp for double duty. But Crest and Altec just don't show up in hi-fi systems inless the wallet is small and the desire for music is large. (Those whose primary use for their home system is for party music are in roughly the same boat as clubs...just a smaller room and the neighbors are more likely to complain when it gets loud.)

Grey
 

blmn

Member
2001-02-01 2:43 am
.
djk,

Yes, I think you are right. Working in normal conditions there is no difference between 1000u and 100000u, and I think there is no problem to raise value of the capacitors until the limit of voltage drop. I changed the values some times and I achieved more excursion before clipping. In your case you did you do any test or measurement before the change?

Regards
 
Regulating the input circuits is a very good idea but not allways a must. the advantage of this is that the input can be set at a higher voltage then the output. In this way the output can work at a lower voltage with less dissipation with a lot of drive voltage from the input.
In high power pro equipment a switched mode power supply is a good idea because of space. I have seen a few very good amps with 1U rack height rated at 1000 Watts rms. Not bad.
One thing I have seen in power supplies is that not only a big capacitance is needed (and makes a difference in bass) but low series resistance and inductance is needed too.
I made the same amp twice ones using 4 caps rated 10 mF each/63v of good quality (Philips high current long life type). The second time I used 28 caps rated 2mF2 for a total of 61mF thats only 50% more but the difference was very very noticable and it cost allmost the same.
 

blmn

Member
2001-02-01 2:43 am
.
GRollins,

Incidentally, and not considering your bitter words (as I told you before, they are not important here), I was talking about a bench power supply for testing high power car audio amps. There are huge amounts of current running there. It has nothing to do with car battery systems. The SMPS power supply itself has usually near 90% efficient.

In the real life the efficiency of a linear power supply depends heavilly on the output current. Many factors that are not under the designer control in this case or are not possible to solve without very expensive solutions.

Besides, think about .05ohm total circuit wiring resistance over 40 ampéres peak demand plus the VCE of transistor regulator plus the VBE of the rectification diodes and you will have near 60% of overvoltage (considering 12VDC output) needed, right?

In this case, batteries are much better than linear power supplies if properly charged. I'm thinking in change my testing gear on this way.



Regards
 
blmn,
Now the curious thing is...some high end folks are running *home* gear off of batteries (super quiet). This is not something I've got a hankering to try at the moment, but there are those who swear by it. (And I'm not sure I want to go to the trouble to series 50 batteries to get enough V+ for tube gear...whew!)
Maybe some day when I've got no other projects on the board (fat chance), I'll get around to it.
Actually, I once owned a tiny Mark Levinson phono head amp that was powered off a 9V battery. I'd forgotten all about that. Cute little booger.

Grey