Subbing NPN output transistors for PNP?

[JohnRoberts]

The NPN input long tail pair, will have a small base current flowing into the base. This current drawn through the local resistor to ground will create a very small - DCV at that base..... it is just sloppy design. 

So, the negative value input bias is due to the NPN pair. It would be positive for PNP.

Do we know what the offset is in that circuit?

Yes, current flows into the base of NPN, out of PNP. The bias current should be predictable from dividing the nominal operating current of the input transistors by their beta (hfe or current gain). I just looked up that transistor and it's data sheet claims a range of beta from 110x to 800x so anywhere from  .5mA/110 to .5ma/800. Hopefully devices from the same batch will be much closer than the min/max specification.

JR

 

[JohnRoberts]

> His SIM said it was all hunky dory.... 

You forgot the  ;D ;D ;D

> can full SPICE do Safe Operating Area + thermal stuff?

If we can calculate it, SIM can too, but that was my initial resistance to SIM, if we have to do 100% of the work, or more, to load the models with data what's the point?

My sim "liked" running a 12AX7 (330V 1.5W max) at 1,500V 250 Watts.


Thermal bias stability seems to be a lifelong study. If you ask SPICE the question at two (or a sweep) temperatures you can know the static trend. This must be multiplied by cooling to know if it will run-away. If you put the bias diode ON an output device (not shown in that plan) and your sink is not woefully small, IME it probably won't runaway. (Undersized sinks are necessary for profit but are "unprofitable" in DIY; be generous!) But the best fix for runaway tends to be bad for small sounds after large sounds (the new on-die devices help).

Yup, the Vbe multiplier (tr5) should be thermally bonded to the output devices or at least the hot part of the heat sink. If the drivers are not also bonded to the same thermal system, we could get over compensation, with class A current dropping as output devices heat up.  A difficulty for applying SIM to thermal stability is difficulty to determine all the input variables like actual thermal resistance of the heat sinks to ambient, etc (we ASSume there are heatsinks  ).

Current meter in-line with the amp power is a brave way to bias. Little slip on the trim, 50mA meter smokes (most DMMs do not have huge overload rating on current). And the higher goal is >30mV on the little emitter resistors..... volt-ranges overload gracefully. The "34-38mA" target seems low for 0.22r resistors, although I know my ear would hardly notice.

Second-Breakdown is a more complicated question. I suppose one of SPICE's boxes can accept an equation which will burp when SOA is violated.

For modest power just throw more robust than needed devices at it... for bigger amps the secondary breakdown can be managed with diode break manipulations within active current limiting.

Back when a big transistor was two tanks of gasoline, such questions mattered. But at today's jellybean prices, forget the question, the answer is LOTs of output devices. Two TO3 per 50W was my old safety rule, and that's cheap. If not about profit/loss, eight per 200W (25W/pair) isn't excessive; also beats heavy-metal heatsinks.

Another evolution is plastic vs metal packaged power devices... 75W from a TO3 (metal) pair was not unusual, now throw 1.5x-2x as many plastic devices and all are happy, a little more silicon and worse thermal resistance junction to sink, but more devices so less power per device.

A serious internal debate when plastic power devices became the new standard was max operating temperatures. It looks like many amplifiers use plastic devices with inadequate thermal headroom for worst case, but survive in the real world, so worst case must not happen very often.

> Intentional beta mismatch of the input LTP to get nominal DC balance is not a common design practice

There's 1K up top and 620 below. Both have 0.7V drop. For a BALanced pair, the bottom should pass twice the current of the top resistor. This suggests, not 1K:620, but 1K2:620 or 1K:470. Basic cocktail-napkin proportioning. With a low-pay helper, or on breadboard, the values won't be exact 2:1 due to TR4 base current, Early effect (and what does does the 22K top of TR2 do? Anything good?) and 1N4148 not being same drop as ZTX753.

With this change, input and output DC offset should be 5mV without selection.

Yup one of several head scratchers in there,  surely easier to just use the right resistor values.

I've never seen a "brass cased Tantalum". Is that some extra-good part which won't short-out in a decade? (Shorted C6 is not a disaster, just 100mV output offset which any 80W speaker will survive.)

The (anonymous?) designer knows something of amp design. I'd do it different.... I'm not sure how much of that is differences in taste and experience.

I do think it has more parts than ABsolutely needed; and if I'm adding more parts MY experience leads to Protection before some of this other elaboration. The short-path through 220 TR4 560 TR7 TR9 annoys me too. (A flat tire should not ruin the steering-wheel.)

But I've built, repaired, and designed many amps and my preferences may be unique.

One tell I always looks for in amp designs that use trimpots for setting class A bias is where the trim is located. In one leg of the Vbe multiplier, a failed (open circuit) trimpot will bias the output stage full on and likely release it's smoke. In the other leg, the trimpot open failure commands low class A current, and no components are sacrificed to smoke. So in this designer's favor, the trimpot is in the correct leg, but I will deduct credit for using a 10T trimpot. While it may make it harder to blow up his VOM, it is generally not that critical of an adjustment. I've seen credible designs with no trims at all (but strict control over device production tolerances).

I hope we don't scare off the OP with this trek down memory lane... Just do it, and see what happens. You will learn more if the first try doesn't work, than if it does. Then again some people don't want to learn how this stuff works. 

JR

 

[abechap024]

This is great stuff guys! I don't really want to post, just because I have so much to learn, just soaking it all in. That is very interesting that the transistor matching could've been avoided with some different resistor values. Also how tr5 should be thermal coupled with the output transistors.  I'm just hooking different stuff up, trying different things. All the discussions here are very thought provoking.

That small signal circuit design book/ and amp book by Douglas Self is now on my list after reading part of it on Google, seems like a very good read.

Throughout this power amp ordeal, I've thought more than once, about just using a chip! Very tempting...but at the same time I am learning a lot about transistors. I think I'm going to stick it through with these....I've gotten both (L/R) making sound - but they both still need a little tweaking.
cheers!

 

[sahib]

The NPN input long tail pair, will have a small base current flowing into the base. This current drawn through the local resistor to ground will create a very small - DCV at that base..... it is just sloppy design.  

So, the negative value input bias is due to the NPN pair. It would be positive for PNP.

Do we know what the offset is in that circuit?

Yes, current flows into the base of NPN, out of PNP. The bias current should be predictable from dividing the nominal operating current of the input transistors by their beta (hfe or current gain). I just looked up that transistor and it's data sheet claims a range of beta from 110x to 800x so anywhere from  .5mA/110 to .5ma/800. Hopefully devices from the same batch will be much closer than the min/max specification.

JR

Sorry John. I did not mean to be sarcastic. I am good with transistor fundamentals. I was trying to make a point to Ricardo but in the absence of sufficient info in my post it did not make any sense I suppose..

I have checked out the company's web page. The schematic seems to be identical except the caps are shown correct way round.

http://avondaleaudio.com/diy-audio/power-amplifier-module-ncc200/

They seem to be positioned to serve the audiophile side of the market. I am not sure if there is a way of asking them to come to explain the design and why it is not as dodgy as Ricardo suggests it out to be.

However, I still don't get why D.Self would have the input caps connected opposite way in two identical designs. His designs are hardly sloppy.

 

[ricardo]

So, the negative value input bias is due to the NPN pair. It would be positive for PNP.

Do we know what the offset is in that circuit?

I was stung by PRR to do a full offset analysis to prove I ken kont en reed en rite .

Assuming Vbe about 0V6 rather than 0V7, hfe TR1 220, TR2 200, TR4 100.  dVbe = 25mV ln(Ic1/Ic2)

I get -24.3mV if working (new) and -1.658V if C6 is SC (after some on/off cycles).

Nasty but no Holy Smoke so I grovel at Guru PRR's feet in shame.


I suspect old fogeys like PRR , JR, Guru Baxandall & I have mass graves of Power Devices in their gardens.  This urge to pontificate on Power Amps is catharsis for our acts of ethnic cleansing.  But rather than re-design Abe's circuit component by component, why not point him in the direction of a simple circuit which will work or at least won't blow up.

My suggestion is Douglas Self.  I have criticisms but I'm carping.  You can get PCBs for some of his circuits.  Some of his Power Amp book is at http://douglas-self.com/ampins/dipa/dipa.htm You DO want to buy the book.  Much stuff which I'd always wanted to test but never found time to do.

PRR, there's a great chapter on thermal stability which discusses the pros & cons of where to place the sensor, time constants, idle, bias after burst of power etc.  But after pretending to think long & hard about this , I'm still undecided of the best thing to do.  Instead, my approach is to have a feedback structure which make the choice of bias less critical. eg No visible artifacts at 25mA/device, and then bias to 50mA.  Sensor on the H/S is then as good as any.  I don't like his use of insulating pads to twiddle the thermal paths.

One Holy Smoke issue.  http://douglas-self.com/ampins/dipa/dpafig33.gif isn't a full working circuit.  If you use his "Beta Enhance" TR12, you MUST have a resistor (eg 5k1) in its collector.  Otherwise, when the amp overloads, TR12 will die & kill TR4, 8, 9 and your speakers. Then TR6, 7 will probably die in sympathy.

Don't forget some protection for your speakers too.  At least a fuse or PTC resistor though the protection is marginal.  Loadsa amps have DC sensing on the O/P to open a speaker relay if the amp dies.

Abe, I second Guru's PRR & JR's advice on loadsa Power Transistors & big H/S.  Self has good recommendations on devices. These are MUCH better and more robust than the old 2n3055/2955 stuff which I started killing many years ago.  The UN team will be coming to dig up your garden.

 

[JohnRoberts]

PRR, there's a great chapter on thermal stability which discusses the pros & cons of where to place the sensor, time constants, idle, bias after burst of power etc.  But after pretending to think long & hard about this , I'm still undecided of the best thing to do.  Instead, my approach is to have a feedback structure which make the choice of bias less critical. eg No visible artifacts at 25mA/device, and then bias to 50mA.  Sensor on the H/S is then as good as any.  I don't like his use of insulating pads to twiddle the thermal paths.

I came up with an interesting biasing scheme back in the late '80s. I put a differential pair across the two .22 ohm emitter resistors, and dialed in the differential to be at equilibrium with X mV across it, I then fed the error current from the differential pair into the Vbe multiplier to set the class A bias current directly by measuring the voltage drop across the emitter resistors. I did the whole thing with one 5 npn transistor array, back when they still made those.

The beauty of this approach is it doesn't care a lick, about what the temperature of anything is, it just regulates the class A current, hot or cold to be the same. I was ready to put this into production on a small 35W amp to prove it with large numbers. Nothing reveals oversights like production volume, but the engineering manager I was working for at the time, had no balls, and made me use a conventional biasing scheme. 

I recall prototyping this temperature independent bias circuit into a guitar amp that used darlington output devices and was notorious for sounding like crap until it warmed up. With my bias circuit it came up cold sounding as good (or bad) as when hot... 

Abe, I second Guru's PRR & JR's advice on loadsa Power Transistors & big H/S.  Self has good recommendations on devices. These are MUCH better and more robust than the old 2n3055/2955 stuff which I started killing many years ago.  The UN team will be coming to dig up your garden.

I'm no guru... more a jack of many trades, who has already made most of the common mistakes already.

Abe, needs to do the math, on how much current he needs, and whether the power devices he wants to use will put out the amps he needs. If not, then maybe double up...

JR

 

[ricardo]

I came up with an interesting biasing scheme back in the late '80s. I put a differential pair across the two .22 ohm emitter resistors, and dialed in the differential to be at equilibrium with X mV across it, I then fed the error current from the differential pair into the Vbe multiplier to set the class A bias current directly by measuring the voltage drop across the emitter resistors. I did the whole thing with one 5 npn transistor array, back when they still made those.

Now THAT I want to see!  Any schematics / scribbles on a napkin, John?

 

[abechap024]

I came up with an interesting biasing scheme back in the late '80s. I put a differential pair across the two .22 ohm emitter resistors, and dialed in the differential to be at equilibrium with X mV across it, I then fed the error current from the differential pair into the Vbe multiplier to set the class A bias current directly by measuring the voltage drop across the emitter resistors. I did the whole thing with one 5 npn transistor array, back when they still made those.

Now THAT I want to see!  Any schematics / scribbles on a napkin, John?

Yes that sounds very clever!

 

[JohnRoberts]

I came up with an interesting biasing scheme back in the late '80s. I put a differential pair across the two .22 ohm emitter resistors, and dialed in the differential to be at equilibrium with X mV across it, I then fed the error current from the differential pair into the Vbe multiplier to set the class A bias current directly by measuring the voltage drop across the emitter resistors. I did the whole thing with one 5 npn transistor array, back when they still made those.

Now THAT I want to see!  Any schematics / scribbles on a napkin, John?

OK, this was many years ago,,,and never put into production. Here is quick pass as what I would do today, but I suspect bench testing would refine and change this.

The values are all just off the top of my head (disregard the PN on the transistors, just what was easy in my schematic capture program). The circuit shown is designed for equilibrium with 2 mA total nominal in the differential pair devices controlling the class A current. The nominal class A current is established by 1 mA x 15 ohms divided by the two .22 ohm emitter resistors, for more or less class A current just scale the 15 ohm value, and/or the emitter resistors.

This is a low gain circuit and I probably limited it to even less pull in range. As drawn it has +/- roughly one diode drop, which is a lot. This proforma circuit is just to demonstrate the concept (IIRC I used a NPN for the Vbe multiplier mainly because I was using a 5 NPN xstor array). There are many errors in that circuit but the vast majority only result in a small linear change to class A current, so a huge temperature change will only result in a modest bias current change.  This is a low gain circuit inside the overall high gain global feedback loop, so it runs out of trim range harmlessly when the output exceeds 2x the class A current in either direction. The differential pair devices don't actually hard saturate so they will recover quickly between cycles. 

I didn't target this for audiophile applications,  but rather cheap, robust biasing, without any factory trims. My recollection is that the distortion wasn't significantly worse than typical trim-less schemes but I didn't focus on seeing how low that could go.

Note: By now I wouldn't be surprised if somebody is using something similar inside a power IC. AFAIK this was a novel design at the time, but that was a couple decades ago. Since I was working for a boss at the time with diminished cohones I never tested this in production (he didn't want to risk follow up engineering changes or perhaps failure of this unproved design). When I was an engineering manager I was a little more adventurous, but that's just me. I can see his POV, even if I didn't agree.

JR

 

[ricardo]

> OK, this was many years ago,,,and never put into production. Here is quick pass as what I would do today, but I suspect bench testing would refine and change this.

I grovel at your feet Guru JR 

This has been a Holy Grail from Jurassic times when I did dis stuff for real .. to get rid of a time consuming AOT on production Final Test.  Peter Walker told me that was the real reason for Quad Current Dumping.

http://avondaleaudio.com/diy-audio/power-amplifier-module-ncc200/
: Ha! A Golden Pinnae outfit. This explains a lot. Naim were famous for oscillating and releasing Holy Smoke.  You had to use ONLY Golden Pinnae speakers from their Auric List.

I stick to my guess that the designer cobbled together stuff from the comics w/o understanding.  Some comics OK so some of it is OK.
But his bias/stability stuff tells me he's "Lets try this.  Dunno what it's supposed to do but we got loads more transistors if the Holy Smoke escapes."

The RCA App. Note
Dem we de days. I was a Motorola man. They had an application note which explained how to design & select stuff based on hfe etc & even how to do SOA protection simply, clearly but accurately.  If anyone has a copy, I'm sure Abe would appreciate this.  Circa 60s/70s.  Amps from 10W to 75W.  Introduced the MJ802/4502 which were IMHO, the finest Power devices of their era.

> However, I still don't get why D.Self would have the input caps connected opposite way in two identical designs. His designs are hardly sloppy.

Even gurus make mistakes. His web page has corrections for his books including 2 replacement schematics.  Where did you see these 2 circuits?  I'm a beach bum so dun hef buks.

 

[JohnRoberts]

I repeat I am not a guru, but I have probably met and worked with some.    8)

I think Jack Sondermeyer who was chief analog dude at Peavey when I worked there, probably wrote the RCA solid state manual when he was working there, before Hartley got him to move to MS. I suspect Hartley hired him to cut back on his phone bills. . Back in the early days it was hard to make truly reliable solid state amps that weren't slower than dirt.

We used truck loads of motorola TO-3s before they sold off their Mexican factory several years back and got out of the metal power business. We had several house numbers that were Motorola parts with guaranteed SOA and predictable Vbe @ current, set up for our quasi standard power level amplifiers. We even had a special (Motorola too IIRC) dual diode in the system that was used for biasing up those motorola output pairs. 

We don't need no stinkin trimpots....

JR

 

[abechap024]

Very cool John! Amazing what can be accomplished with some good know how and experience. Getting rid of trim-pots and things is a step in the right direction if you ask me, making the design more of a self calibrating machine.
Cheers,
Abe

 

[ricardo]

> We used truck loads of motorola TO-3s before they sold off their Mexican factory several years back and got out of the metal power business. We had several house numbers that were Motorola parts with guaranteed SOA and predictable Vbe @ current, set up for our quasi standard power level amplifiers. We even had a special (Motorola too IIRC) dual diode in the system that was used for biasing up those motorola output pairs. 

> We don't need no stinkin trimpots....

John, can you remember what sort of Quiescent Current range you got with these special parts in production?  2 to 1?

And what you'd expect from your special circuit?

 

[sahib]

However, I still don't get why D.Self would have the input caps connected opposite way in two identical designs. His designs are hardly sloppy.

Even gurus make mistakes. His web page has corrections for his books including 2 replacement schematics.  Where did you see these 2 circuits?  I'm a beach bum so dun hef buks.

Audio Power Amplifier Design Handbook, page 243.

Might well be an error.

Back to the Golden Pinnae Outfit's amp. PRR thinks the designer knows something but you are betting on a different horse. Could you expand on his bias stability and why it is fire hazard?

 

[JohnRoberts]

> We used truck loads of motorola TO-3s before they sold off their Mexican factory several years back and got out of the metal power business. We had several house numbers that were Motorola parts with guaranteed SOA and predictable Vbe @ current, set up for our quasi standard power level amplifiers. We even had a special (Motorola too IIRC) dual diode in the system that was used for biasing up those motorola output pairs. 

> We don't need no stinkin trimpots....

John, can you remember what sort of Quiescent Current range you got with these special parts in production?  2 to 1?

I don't know... mainly from "just barely works to doesn't blow up". That stuff was all refined years before i started working there, and I didn't work in the power amp area. It was dialed in for the metal T0-3 motorola devices, to just plug and play with mostly proven circuits. Occasionally there might be the batch of transistors that weren't to spec, but units never released smoke, just ran hotter or colder than normal.  .

I do recall hearing about one trim-less guitar amp using plastic darlingtons, that could be seriously starved for bias current on winter mornings in MN, until it warmed up.  :-(.

And what you'd expect from your special circuit?

I guess nobody wins the secret prize for finding my mistake.   R18 should be more like 470-510 ohms, for equilibrium at nominal 1 mA per device in the differential pair.

The nominal class A current at balance is defined by 1mA in Q14 x the 15 ohms in series with it's emitter. That means the diff pair is in balance when 15 mV offset is across the two .22 ohm emitter resistors, making around 35 mA of class A current. As the load starts drawing current from only one output, the voltage drop across the other emitter resistor falls to nothing, so the full 15mV becomes effectively across only one of the .22 ohm resistors (up to 70 mA). Beyond that, the class A regulator just runs out of pull in range, but since this is all going on inside the global feedback loop, the amp just keeps on trucking in class B.

Note: It's roughly 2mA class A current error for every 1 mV of DC error in the differential pair, the sensitivity to error in the rest of the circuit is more complicated but mostly low gain so thermally stable. I used a transistor array, but suspect this circuit would work, albeit with more slop in the nominal class A current, using discrete transistors for the diff pair.

It has literally been a couple decades since I had this on the bench so my memory is fuzzy, but I think I limited the adjustment pull in range to +/- 300mV or so. I suspect in production I could tighten it up to even less than that, but I never got that far.

JR

 

[ricardo]

> Back to the Golden Pinnae Outfit's amp. PRR thinks the designer knows something but you are betting on a different horse. Could you expand on his bias stability and why it is fire hazard?

Duu.uuh!  I meant offset and HF stability as 2 separate issues.  He dunno wat he do cos ..

His offset trim is REALLY yucky and he probably don't have much experience in production if he recommends Tants.  But as Guru PRR points out, the offset when they die, which I make to be volts, won't give smoke.

His 690R+470p, 390R+470p and 47p+2k2 networks tell me he's had HF instability & oscillation but doesn't know how to cure them.  He's just fudging. There are simpler & better ways to ensure unconditional stability.  Professor EM Cherry's articles are worth studying even if you don't adopt his complex NDFL methods.  Self (& Baxandall toward the end) are slowly coming round to his way of thinking.  His analysis can be used for lesser methods like the ones in Self's book.

HF oscillation in da old days was a guarantee of Holy Smoke.  Modern transistors, like Self's recommendations, are much more robust. But you're a brave man or a fool to bet on your amp surviving full blown HF oscillation.  But see my post of 20may about the sound of Golden Pinnae amps.  Brave men or ...  ???

I didn't mean to comment on bias stability cos I dunno where he has his sensor.  Actually the pics show it isn't on the H/S which is another strike against him especially for a kit.

 

[sahib]

His 690R+470p, 390R+470p and 47p+2k2 networks tell me he's had HF instability & oscillation but doesn't know how to cure them.  He's just fudging. There are simpler & better ways to ensure unconditional stability.  Professor EM Cherry's articles are worth studying even if you don't adopt his complex NDFL methods.  Self (& Baxandall toward the end) are slowly coming round to his way of thinking.  His analysis can be used for lesser methods like the ones in Self's book.

It can also be argued that he knows how to cure hf instability/ oscillation as the networks he introduced points in that direction. Perhaps he chose to stick to this particular plan and solved his problems accordingly. I realise that my post is not contributing anything technical but I wouldn't like others to think that the use of 47pf +2K2 (on the feedback path) is  heresy  as it is as common as asprin.

Abe, since you have this circuit on your operating table you can perhaps give us a feedback on its performance.

 

[JohnRoberts]

His 690R+470p, 390R+470p and 47p+2k2 networks tell me he's had HF instability & oscillation but doesn't know how to cure them.  He's just fudging. There are simpler & better ways to ensure unconditional stability.  Professor EM Cherry's articles are worth studying even if you don't adopt his complex NDFL methods.  Self (& Baxandall toward the end) are slowly coming round to his way of thinking.  His analysis can be used for lesser methods like the ones in Self's book.

It can also be argued that he knows how to cure hf instability/ oscillation as the networks he introduced points in that direction. Perhaps he chose to stick to this particular plan and solved his problems accordingly. I realise that my post is not contributing anything technical but I wouldn't like others to think that the use of 47pf +2K2 (on the feedback path) is  heresy  as it is as common as asprin.

Abe, since you have this circuit on your operating table you can perhaps give us a feedback on its performance.

I can see both sides of this argument.. The schematic has all the conventional compensation techniques (degenerated input LTP, dominant pole compensation, output trap, and output load isolation coil...  that said some of the extra RxC shunts inside and to the global feedback parts look a little questionable.

For a quasi-comp topology one could argue for adding some lag to the one output stage without the extra inversion to make the zig speed similar to the zag speed, while in general you just need to make the dominant pole (c across TR4) large enough to swamp out the worst of the two. Note: He has added some more lead to the slower inverting output stage so perhaps, kind of sort of balancing the zig/zag speeds.

I still have my old resistor and capacitor diddle boxes which were used in the bad old days, to design by trial and error. I am not smart enough to pass judgement on this design by simple inspection. Build it and see what happens. If using different output devices, i'd be tempted to revisit all the apparent diddles (just about everything with an *).

JR

PS: I remember reading that old Cherry paper in the AES Journal but don't recall it well enough to repeat any useful points. IIRC he made the controversial claim that common collector and common emitter output stages could both be stabilized just as easily (not in my experience).

 

[abechap024]

Well this circuit is a basically more or less a clone of a Naim 140. (Though avalone audio did take out/change stuff)

I have some china pcbs of more or less the same circuit.
Performance wise - its distorting quite a bit - haven't pinned down the exact reason yet...It dawned on me this morning that somewhere I don't think my output transistors are getting biased correctly...the unit doesn't pull very much current at all when Idle and what I'm hearing seems to be miss/biased/wrongly/biased transistors...just kinda jumpy in/out distortion.
At lower volumes the distortion doesn't seem to be quite as bad.

Interesting points on all fronts though, It has me thinking that I should toss these PCBs and just start from scratch trying to implement a new improved circuit with the various ideas brought up.

 

[JohnRoberts]

Well this circuit is a basically more or less a clone of a Naim 140. (Though avalone audio did take out/change stuff)

I have some china pcbs of more or less the same circuit.
Performance wise - its distorting quite a bit - haven't pinned down the exact reason yet...It dawned on me this morning that somewhere I don't think my output transistors are getting biased correctly...the unit doesn't pull very much current at all when Idle and what I'm hearing seems to be miss/biased/wrongly/biased transistors...just kinda jumpy in/out distortion.
At lower volumes the distortion doesn't seem to be quite as bad.

Interesting points on all fronts though, It has me thinking that I should toss these PCBs and just start from scratch trying to implement a new improved circuit with the various ideas brought up.

#1.. confirm no spurious oscillation (with a scope).

#2 check class A current.. with no audio playing measure DC voltage across each .22 emitter resistor. Using ohms law (i=mV/.22ohm) this will give you current in each power device. Should be same but dc offset and speaker load may cause some error at 0V. I would expect zero crossing distortion to clean up at 10-30 mA of class A.

Note: I don't expect this to be uber hifi, but if it works it shouldn't totally suck.

I assume you got the driver situation worked out for the quasi comp outputs. The last schematic you posted has the pull down all tangled up. If it works at all it wouldn't be cool to the touch or otherwise.

JR