How to Calculate BIAS on Transistor amp ?

[R.G.]

I haven't read the book, but with cheap and powerful processors such as the $5 postage-stamp sized 133Mhz RP2040 Seeed Studio it may be possible to attain near perfection (or at least as good as it gets) by closely monitoring and constantly adjusting an amp's parameters.

I'm a big proponent of using tiny processors to do things that are otherwise very difficult, so that's always in the back of my mind.
It's worth the read, though. Self does a lot of detailed instrumentation as well as in-depth simulation looking for how to do things. One particularly illustrative chart I remember is that he did a graph of gain of the output stage, both above and below crossover, at various settings of bias current/voltage.
You would think that at some setting of bias, this would settle down to a smooth line at crossover, but it never does. Every possible bias setting has a wobble right at crossover. Self's conclusion is that there is some setting that is as good as it can get for a given pair of output devices, but never perfect.
The same line of thinking led him to model both Darlington/Emitter Follower output stages as well as complementary feedback PNP/NPN pairs. These have different and different sized crossover wobbles even when optimally biased.
There is a chapter on thermal compensation as well. Pretty much every new solid state amp using push-pull output stages uses a Vbe multiplier thermal compensation instead of discrete diodes like they used to. Self goes over the circuit tricks to make this stage be effectively an analog computer for thermal compensation. Apparently you can somewhat "model" the thermal response of the transistors on the heat sink, including the thermal lags with the thermal R-C issues. As an interesting point, the quickest way to get the "thermal information" of how hot the transistor chip inside the package might be is to put the thermal sensor over the chip on the outside of the transistor package, not on the heat sink.
Something that Self points out is that Sanken has actually made power transistors with integrated diodes on the transistor chip itself and brought this out to leads so the heat of the chip itself could at least potentially be sensed for temperature right on the chip.

Fascinating stuff.

[Pixel]

I have a PA amp that uses germanium power transistors. These transistors are paired with thermistors to counteract the temperature sensitivity of the germanium.

[R.G.]

I worked out a mod for the Thomas Organ USA Vox amps that use germanium outputs so PNP silicons could be substituted. Massively more reliable. Germanium is both massively more leaky than silicon (~1000x) and has a much lower maximum temperature limit.

[Chuck H]

I'm a big proponent of using tiny processors to do things that are otherwise very difficult, so that's always in the back of my mind.
It's worth the read, though. Self does a lot of detailed instrumentation as well as in-depth simulation looking for how to do things. One particularly illustrative chart I remember is that he did a graph of gain of the output stage, both above and below crossover, at various settings of bias current/voltage.
You would think that at some setting of bias, this would settle down to a smooth line at crossover, but it never does. Every possible bias setting has a wobble right at crossover. Self's conclusion is that there is some setting that is as good as it can get for a given pair of output devices, but never perfect.
The same line of thinking led him to model both Darlington/Emitter Follower output stages as well as complementary feedback PNP/NPN pairs. These have different and different sized crossover wobbles even when optimally biased.
There is a chapter on thermal compensation as well. Pretty much every new solid state amp using push-pull output stages uses a Vbe multiplier thermal compensation instead of discrete diodes like they used to. Self goes over the circuit tricks to make this stage be effectively an analog computer for thermal compensation. Apparently you can somewhat "model" the thermal response of the transistors on the heat sink, including the thermal lags with the thermal R-C issues. As an interesting point, the quickest way to get the "thermal information" of how hot the transistor chip inside the package might be is to put the thermal sensor over the chip on the outside of the transistor package, not on the heat sink.
Something that Self points out is that Sanken has actually made power transistors with integrated diodes on the transistor chip itself and brought this out to leads so the heat of the chip itself could at least potentially be sensed for temperature right on the chip.

Fascinating stuff.

KABLOOEY!!!

That was my head exploding. Mick is also looking into current/voltage feedback relationships in another thread. Is it possible that the demise of tubes will inspire the best of what can be done with SS amps now in the digital age?!? One can hope. Nearly all of this goes over my head but it's great to read that the higher minds are developing what they can with what's still available and viable.

I might have thought most of these considerations had been worked out by the hi fi market and engineers by now, but I guess not? Or is all this as it relates to instrument amplification instead of reproduction?

[R.G.]

KABLOOEY!!!

That was my head exploding. Mick is also looking into current/voltage feedback relationships in another thread. Is it possible that the demise of tubes will inspire the best of what can be done with SS amps now in the digital age?!? One can hope. Nearly all of this goes over my head but it's great to read that the higher minds are developing what they can with what's still available and viable.

I might have thought most of these considerations had been worked out by the hi fi market and engineers by now, but I guess not? Or is all this as it relates to instrument amplification instead of reproduction?

Neat stuff, all right!
The hifi guys are chasing the last 0.001% of distortion, and lot of it lurks inside crossover issues. This is clearly not so much an issue for instrument amplifiers, but it does feed our tech dreams, all right. And in many cases lets us know where to dig for some specific thing we're chasing.
Another thing that Self was recommending was doing a first-approximation biasing by voltage/current setup, then fine tuning an amplifier by watching a distortion analyzer and tuning for lowest distortion reading. It turns out that for push-pull transistors, distortion decreases as the idle current increases from zero (that is, out of Class B toward Class AB), reaches some minimum, then increases again as it's biased further towards Class A. It's not the case that more bias toward Class A is better until you actually get to Class A. The start and end of the overlap region where both output devices are on contributes two "anti-crossover" notches. Weird.

Adding current feedback to raise the output impedance of a solid state feedback amp has been around for a while. It does loosen the grip of the power stage on the speaker and let speaker funnies shine through more. I've used this before, and it does change the tone of a solid state amp. I believe it's used in some commercial amps these days.

[x-pro]

They are chasing 0.001%. At the same time, forgetting that even the best speaker system cannot have distortion less than 4%.

A good engineer always has common sense.

[R.G.]

They are chasing 0.001%. At the same time, forgetting that even the best speaker system cannot have distortion less than 4%.

A good engineer always has common sense.

Yeah. But, as with all things, there is the question of objectives. If the objective is to get lowest reasonably audible distortion, The chase for vanishingly small distortion would have ended long ago at about 0,01% distortion.

But if the objective is to sell more high priced amplifiers to so-called "Golden Ears" who swear that they can hear differences in the amount of oxygen dissolved into speaker cables, then "common sense" does not come into play, only the ability to claim 0.0000000001% distortion for Lah-Di-Dah (r) High End amplifiers sells more of them.

The bosses and accountants that hire engineers will make the engineers care very, very much about getting distortion amounts down to vanishingly small. If the engineer says in the interview to be hired "You know, it's silly to chase small distortion numbers." they will simply hire another engineer that can be made to care.

The vast majority of engineers have common sense enough to keep getting pay checks.

[Mick Bailey]

Fender sold amps with a degree of current feedback in the early 70s - maybe before then. It's one aspect to that goes unnoticed unless you search through schematics and then find it's faintly common. So much so that manufacturers don't even mention it as a feature. Even the small Marshall MG practice amps use mixed-mode feedback. Perhaps Randall went further than most with their amps that used current-only feedback to raise output impedance.

Rather than chasing down distortion in power amps, my view would be to go the other way and replicate tube distortion. When you look at the SS attempts at tube simulation and amp modelling, it tends to be all based around a preamp feeding a low-distortion power amp. I'm not aware of anyone who is making SS power amps that better represent what happens with a tube amp. I guess that in the end a waveform is a waveform and it doesn't matter how or where it's created. It would be interesting though to build a power amp that can be driven into distortion just like its tube counterpart.

[The Dude]

Of course personal preference, but I prefer preamp distortion feeding a low distortion power amp. The reason being that you don't have to run an amp full tilt to get a similar result. Of course it doesn't sound exactly the same, but close enough, IMO. Lots of players need to be able to do an outdoor gig with volume one weekend, quietly do a small club the next weekend, and maybe practice in the bedroom the next. It's nice to be able to just turn the master down instead of patching a power soak, or cover the amp in plexiglass, or some other solution.

[Mick Bailey]

My thought is that the power amp could be made intrinsically scaleable to eliminate the need for power soaks etc. To give the type of ourput you'd get from a cranked non-MV amp.

[x-pro]

They've been doing this for a long time.
Rack ones have a sensitivity switch.

[R.G.]

Somewhat related - Thomas Organ Vox power amplifiers. In the early 1960s, bigger amps, 20+ watts or so, often used a driver transformer to run two stacked output transistors. This setup was prone to failure with the slow, leaky output devices. Driving them too hard often killed the output devices. The biasing was odd by today's standards, done by a voltage divider network that was not adjusted much in production - no trimmers.

Thomas attacked the failures with their "Watchdog Limiter", a soft-ish clipping circuit at the input to the power amp. This clipping circuit was adjustable, and the instructions for setting it involved running the amp at full NON-clipping output voltage so the output was not naturally clipping. Then the Watchdog Limiter was adjusted to limit a volt or two lower. That way the amp was not clipped, but the input was limited to just before the amp would clip.
The limiter had a soft-ish clipping nature, much softer than the power amp itself would do by itself - and the power amp was never overdriven by its input signal.

This was clever solution with a hidden advantage for tone. The amp had whatever amount of feedback it had, but much less than the high feedback types that came later, so it was a "softer" speaker driver than it might otherwise have been. The distortion was that of the limiter, which could be adjusted. Its tone was better than most feedback-fitted power amps. On the whole, the Thomas Vox amps with the limiter were better tonally than their competitors. I put them midway between pure solid state amps and tube amps, purely to my ears. Your ears may, of course, hear it differently.

The input limiter trick adds one more tool to adjusting the sound of the power amp. A clean, higher-feedback power amp can have its output impedance adjusted to change its "grip" on the speakers, as well as having its input limited in a custom manner to allow you to design the clipping nature. If you think about it, miking a tube amp, then feeding the mike into a high-feedback PA amp is quite similar.