Okay, so what are you saying exactly atmasphere? Let's be clear so it can be discussed without misunderstandings.
Yeah I thought Shannon-Nyquist and Harold S Black's feedback equations and theories were pretty universally applicable to amplifiers of all sorts.
And yes, I agree that high levels of feedback in tube amplifiers with transformers can be problematic, with low levels of loop feedback you are extending the harmonic spectra in the output of the amplifier. (This is what Baxandall was talking about)
I will also confess that I like the sound of asymmetrical behaviors of class A tube amplifier stages, and don't use any feedback in my SE amplifiers, but I am not fooling myself either.
And yes, I agree that high levels of feedback in tube amplifiers with transformers can be problematic, with low levels of loop feedback you are extending the harmonic spectra in the output of the amplifier. (This is what Baxandall was talking about)
I will also confess that I like the sound of asymmetrical behaviors of class A tube amplifier stages, and don't use any feedback in my SE amplifiers, but I am not fooling myself either.
I'm saying, as Crowhurst did, that distorting the feedback signal prior to mixing with the audio signal is not a good thing. I'm a big fan of differential amplifiers BTW- all of our OTLs and two of our preamps are fully balanced and differential. We've produced fully differential circuits longer than any other manufacturer in the world at this point. They are indeed lower distortion. But that does not mean they are zero distortion.
Crowhurst wrote exactly that.
An OTL, FWIW, can have a lot more feedback since it has better phase margins. Futterman claimed to run 60dB in his amp. He was careful about his feedback loops and IME didn't transmit all that knowledge in his schematics. So many Futterman style OTLs (including his) did have oscillation problems due to the load interacting with the feedback loop. He was careful to also use ferrite beads. Some of his amps held up quite well.
They are. But they don't expect you're going to distort the feedback signal! You might process it to keep out unwanted frequencies (due to phase margin, that sort of thing). If you've not read this, its a treat. I'm sure you have though. At any rate, pay attention to page 2.
https://linearaudio.net/sites/linearaudio.net/files/volume1bp.pdf
Hi atmasphere,
When I do upgrades, I take great care to use components and take-off points so as not to distort the feedback signal. Running it into a carefully matched diff pair does not distort the feedback signal. If you want to argue that, then the exact same thing happens to the input signal. Since we apply the signals to very (very) close to the identical parts, the distortion is complimentary and drops out of that equation.
I approach these circuits from a test and measurement viewpoint. A lot of hand matching that would never be viable in mass production, and components that would not survive wave soldering. But you know, the basic solid state circuits used an well designed equipment generally do not distort any signals to the extent you may be suggesting. Remember too, we have vastly higher open loop gain to work with, using parts that respond well beyond 50 MHz in signal stages, and generally over 5 MHz in power stages. So very wide bandwidth that we control with compensation and stage current levels.
In tube amplifiers, transformer coupled audio amplifiers have a much restricted frequency range and do have phase shift issues that are not avoidable. So that does limit how much feedback you can apply. OTL amplifiers do not suffer from this obviously, so you can employ much greater levels of feedback (good). The fact remains that tubes, even high current types, are generally unhappy with low impedances and delivering current is very hard on cathodes. A solid state device suits this role very naturally, which is why the industry moved to solid state. The only place you see tubes these days are very high voltage and high current levels (radio station transmitting tubes and industrial welders). But these tubes do not lend themselves to home audio very well - and are horribly expensive!!
Crowhurst wrote considering the technology of the time. So he was dealing with low transconductance devices in circuits with minimal feedback. The framework he lived with does not exist today, so what he wrote has to be put into perspective with the technology we have today. Even the resistors and capacitors he had at his disposal were vastly inferior to what we use today in just about every aspect. Those carbon composition resistors had terrible voltage and temperature characteristics! Never mind the capacitors of the day. So what he wrote back then was generally true, only the concepts apply today. His findings would have to be re-interpreted, so new experiments made. A great deal has changed since then.
-Chris
When I do upgrades, I take great care to use components and take-off points so as not to distort the feedback signal. Running it into a carefully matched diff pair does not distort the feedback signal. If you want to argue that, then the exact same thing happens to the input signal. Since we apply the signals to very (very) close to the identical parts, the distortion is complimentary and drops out of that equation.
I approach these circuits from a test and measurement viewpoint. A lot of hand matching that would never be viable in mass production, and components that would not survive wave soldering. But you know, the basic solid state circuits used an well designed equipment generally do not distort any signals to the extent you may be suggesting. Remember too, we have vastly higher open loop gain to work with, using parts that respond well beyond 50 MHz in signal stages, and generally over 5 MHz in power stages. So very wide bandwidth that we control with compensation and stage current levels.
In tube amplifiers, transformer coupled audio amplifiers have a much restricted frequency range and do have phase shift issues that are not avoidable. So that does limit how much feedback you can apply. OTL amplifiers do not suffer from this obviously, so you can employ much greater levels of feedback (good). The fact remains that tubes, even high current types, are generally unhappy with low impedances and delivering current is very hard on cathodes. A solid state device suits this role very naturally, which is why the industry moved to solid state. The only place you see tubes these days are very high voltage and high current levels (radio station transmitting tubes and industrial welders). But these tubes do not lend themselves to home audio very well - and are horribly expensive!!
Crowhurst wrote considering the technology of the time. So he was dealing with low transconductance devices in circuits with minimal feedback. The framework he lived with does not exist today, so what he wrote has to be put into perspective with the technology we have today. Even the resistors and capacitors he had at his disposal were vastly inferior to what we use today in just about every aspect. Those carbon composition resistors had terrible voltage and temperature characteristics! Never mind the capacitors of the day. So what he wrote back then was generally true, only the concepts apply today. His findings would have to be re-interpreted, so new experiments made. A great deal has changed since then.
-Chris
emphasis added.
Yes, a differential pair distorts the input signal as well as the feedback. The distortion isn't complimentary; your conclusion above is incorrect. Its the same for both sides, except for phase (ideally). In a differential amplifier even orders are cancelled but odd orders are not (cubic non-linearity). The feedback signal is getting distorted by having odd orders added.
Of course no differential amplifier is perfect. Plate, collector or drain loads have to be matched as well as the active devices themselves. You've done this work, so you know that the matches aren't perfect; the differential effect isn't perfect either. But even if it was you'd still be dealing with the inherent non-linearity if the active devices. The only way there would be no distortion would be if the devices themselves were perfectly linear and such things do not exist.
I wonder maybe is better to use single input device as is in all Futterman OTL amps and also on many other conventional tube amps with OPT ? , where input signal is applied to grid and negative feedback signal is applied to cathode ?
, btw, so many transistors power amps from 70` uses such simple one transistor input stage (singleton) and sounded pretty good .
, btw, so many transistors power amps from 70` uses such simple one transistor input stage (singleton) and sounded pretty good .
The advantage of this approach is you get even orders and well as odd orders. The disadvantage is there is overall more distortion. I think a better way to do it is to mix the feedback with the audio before it gets into the amplifier- the way opamps do it. Resistors are more linear than tubes or semiconductors.
I forgot to mention the Kronhite tube amplifier of the late 1950s early 1960s. This amp used up to 80ddB of feedback and produced IM distortion of less than 0.005%. If there's a will there's a way department.
Last edited:
The Kronhite manual. Amplifier designed by Richard Burwen.
Last edited:
I used to own a set. My regular amp was an ARC D51 and the Kronhite bested it easily. This was back in the early 1980s...
So now you have conflicting testimony on that one. I think the set of amps I had might have been in danger of power supply filter cap failure, something I'd not considered at all at the time. With any of this vintage stuff you really want to make sure the equipment is actually running correctly prior to proclamation of its sound...
I think this is one possible way to mix audio signal with negative feedback signal using only resistors in practical OTL amp ? ,
where at input is one very high gain AF pentode as only voltage gain stage driving bootstrapped P.I. stage which again drives SEPP-OPS in
Inverted Futterman mode , known and called also as Technics OTL variation .