Tube NFB

[boji]

add some NFB to reduce it to a reasonable level.

Before my book arrives... Just curious: Reading Bluegrass Dan's thread, can it be said that in the tube world, NFB is deployed only to reduce gain unlike opa's or do they serve the same purpose?

Thanks

 

[Bo Deadly]

Sure does. NFB can improve linearity dramatically.

 

[boji]

Thanks square!  I think I'm imparting more magic to a tube than just acting like a number of transistors.

 

[ruffrecords]

Thanks square!  I think I'm imparting more magic to a tube than just acting like a number of transistors.

NFB provides a number of advantages and it is completely independent of the active componets of the system it applies to.

NFB can:

Reduce distortion
Flatten frequency response
Lower output impedance
Raise input impedance
Improve power supply rejection
Reduce the effect of component value changes with time.

The only downside is it can be unstable if not applied properly.

Cheers

Ian

 

[boji]

Excellent excellent...  Thank you Ian!

 

[JohnRoberts]

NF was invented when tubes were the only gain stages available. (Black over a century ago?)

It was primarily used to improve frequency response, but it also reduces distortion.  Imagine the frequency response after running through hundreds (thousands) of telephone repeater amplifiers.
===
I've told this story before but when I was advising wrt feature set of the well respected peavey tube mic preamp, I suggested that we consider a variant with lower NF to have more tube distortion (we provided an internal jumper but shipped it set for clean).

After no customers requested more tube distortion, :  we permanently connected the full NF for low distortion behavior.

JR

 

[PRR]

> primarily used to improve frequency response

Actually gain stability (though freq resp is part of that issue). Tube amp stages vary more than 1dB each. In a 100-repeater chain that gives +/-100dB error with age, temperature, replacement. With repeaters spaced-out across miles of countryside, going to each one to trim was not a happy chore.

 

[Matador]

I was thinking: the fundamental feedback equations for closed-loop gain is hand wavy about assuming that closed-loop gain is only dependent on the feedback network, provided open-loop gain is 'sufficiently' high.

Would this be the main rub with practically applying feedback in tube gain stages?  A modern IC op-amp might have an open-loop gain of around 106, however one doesn't typically see this in a standard tube gain stage topology.  Could it not be argued that most of the complexities with NFB in tube circuits arise from this assumption not holding (unlike with transistor topologies)?

 

[JohnRoberts]

I was thinking: the fundamental feedback equations for closed-loop gain is hand wavy about assuming that closed-loop gain is only dependent on the feedback network, provided open-loop gain is 'sufficiently' high.

Would this be the main rub with practically applying feedback in tube gain stages?  A modern IC op-amp might have an open-loop gain of around 106, however one doesn't typically see this in a standard tube gain stage topology.  Could it not be argued that most of the complexities with NFB in tube circuits arise from this assumption not holding (unlike with transistor topologies)?

Op amps only have that much gain at very low frequency (because of dominant pole stability compensation)...  Loop gain margins*** of as little as 20-30 dB at HF are still useful to linearize transfer function.

Yes Paul,  gain and response were both managed (along with everything else). 

JR

***Loop gain margin is the difference between open loop gain and closed loop gain... I wrote about this in the 80s

 

[boji]

I was thinking: the fundamental feedback equations for closed-loop gain is hand wavy about assuming that closed-loop gain is only dependent on the feedback network, provided open-loop gain is 'sufficiently' high.

Would this be the main rub with practically applying feedback in tube gain stages?

This is how you properly ask a q on tewb NFB.  ;D

/heart for bothering (and expanding) on it everyone.

 

[ruffrecords]

I was thinking: the fundamental feedback equations for closed-loop gain is hand wavy about assuming that closed-loop gain is only dependent on the feedback network, provided open-loop gain is 'sufficiently' high.

Would this be the main rub with practically applying feedback in tube gain stages?  A modern IC op-amp might have an open-loop gain of around 106, however one doesn't typically see this in a standard tube gain stage topology.  Could it not be argued that most of the complexities with NFB in tube circuits arise from this assumption not holding (unlike with transistor topologies)?

I would say no.  The basic NFB equations make no assumptions about the size of the open loop gain.

NFB in tubes is certainly a little different because the open loop gain tends to be a lot smaller than in op amps. But the main problem for tubes is stability. With op amps, you can close the loop at dc and the open loop gain continues also down to dc. This means there are no low frequency stability problems with op amps. They only need to manage HF stability which they do by taking a sledgehammer to the open loop response with a dominant pole at around 10Hz or so. You can afford to do this if you have 120dB ope open loop dc gain.

With tubes there are big dc voltage difference between stages and between output and input. This makes it nigh on impossible to close the loop at dc or have an open loop response down to dc. This means when you close the loop there are always at least two pole/zero pairs at low frequencies which means it will only be stable for a limited range of open and closed loop gains. That's why the REDD 47 tube mic pre has such a limited range of gain settings for example.

Stability is critically dependent on the amount of feedback which is just the open loop gain in dB mius the closed loop gain in dB. So if you have a tube pre with 60dB open loop gain and you arrange the NFB so the closed loop gain is 40dB then you have 20dB of NFB and you cann arrange for the amp to be stable at both high and low frequencies under these conditions. Now suppose you want to vary the closed loop gain from 40dB down to 20dB say. The normal way to do this is to vary the NFB. The problem is  at the lower gain setting there will be 40dB of NFB and your carefully set up NFB stability no longer works.

One solution is to simultaneously vary the open loop and closed loop gains so their difference (= the amount of NFB) remains nearly constant and the circuit remains stable. The classic example of this is the V76 amplifier.

One further step is to close the loop at dc. This removes one LF  pole/zero pair and can make the amplifier unconditionally stable at LF. Then you only need to worry about HF stability. This is what I do in the EZTubeMixer preamp design (as well as varying open loop gain with closed loop gain).

Cheers

Ian

 

[alexc]

nfb in tube circuits is a very fine thing, as Ian and others have described,  but imho ,  it does alter the harmonic structure in the higher-order freqs 

....  at least in  'single tone' style testing, which of course is quite a simplification of things,  and not at all  a pinch on the 'real world' of dynamically varying audio  [ie. music].

I use conventional types of  nfb almost all the time, no question  - except in some single-ended output tube (spud) applications, where it is interesting to try to see (and hear!)  the changes  from a 'no nfb'  transitioning to a  'some nfb'  etc.

It's a challenge to retain the obvious benefits of proper nfb arrangements, without causing over-proliferation of higher order harmonics.

...

One of the things I most like about nfb, is it doesn't cost $$  for big improvements [most of the time] in audio quality. Especially in tube builds where costs can really mount for the good components.

But for sure it's a many-faceted phenomenon.

 

[ruffrecords]

nfb in tube circuits is a very fine thing, as Ian and others have described,  but imho ,  it does alter the harmonic structure in the higher-order freqs 

Can you be more specific about exactly what you mean by  the phrase "alters the harmonic structure". I assume you are not talking about the myth that NFB adds harmonics not present in the original output.

Cheers

Ian

 

[Tubetec]

I think feedback definately  has an effect on the way an amp overloads ,and the spectrum of the distortion produced .
Of course its true to say an amp with feedback applied will produce less distortion than without , but  if you keep increasing the input voltage you will see an increase in higher order products when it eventually does distort , its something to do  with the feedbacks abillity to correct being hampered during moments of overload , Audio Note had an interesting essay on it at one point ,probably still available on their site ,negative effects of feedback or something  like that was the title.

Its not all bad , feedback in moderate amounts in guitar amps definately contributes to what we call tone , in the case of push pull much of the second harmonic cancels ,revealing more upper harmonics . The Marshall amp which was derived  from the Fender Bassmann uses a presence  control in negative feedback loop from transformer secondary . The Vox Ac30 on the other hand doesnt use feedback to linearise , instead it runs class A .

 

[bluebird]

Anyone know if there is advantages to having NFB come from an output transformer like in a Pultec to a circuit like the UA 610 where its coming from the plate of the output tube?

(theres an EQ circuit in the feedback loop of the 610. I took it out for simplification)

 

[shabtek]

feedback taken from a tertiary winding can linearize the transformer/include its nonlinearities  in the nfb loop

 

[Winston OBoogie]

Anyone know if there is advantages to having NFB come from an output transformer like in a Pultec to a circuit like the UA 610 where its coming from the plate of the output tube?

(theres an EQ circuit in the feedback loop of the 610. I took it out for simplification)

Non linearities of the transformer can certainly be reduced by including it in the FB loop, but the poles and phase shifting  inherent in the transformer may cause instability and require adjustment of the forward path and feedback loop in the form of added zeros, lead and/or lag networks to compensate.  Generally speaking though, I found this less necessary if the loop gain isn't  too large.

 

[Winston OBoogie]

Can you be more specific about exactly what you mean by  the phrase "alters the harmonic structure". I assume you are not talking about the myth that NFB adds harmonics not present in the original output.

With zero, or very small amounts of,  -tve FB in a tube (or otherwise) amplifier, the distortion profile rises monotonically with input level.  Depending on the topology and tube type, this might be a strong 2nd, followed by 3rd etc.  Or a higher 3rd  follower by 2nd...

With higher loop gains, this does tend to change somewhat.  The soft clipping we might get with a pure resistive load for instance is reduced and pushed closer to the voltage rails with the profile of the distortion products being somewhat compressed. 

I do think that the observations of Peter Baxendall and others regarding the optimum minimum amount of feedback can hold true for tube circuitry just as much as it does with solid state.

 

[Winston OBoogie]

Addendum to that last reply: One difference with tube circuitry of course is that we usually have higher rails and therefore a bigger overload safety margin.  Also that some tubes are really quite remarkably linear right out of the gate.  This means we can get away with lower amounts of feedback and keep our fairly benign distortion profile.
IMO the paper by Walter Sear and Russel Hamm written all those years ago regarding tubes Vs  transistors as used in mic amplifiers still makes a strong case for their (tube) use where there are uncontrolled loud transients or a large dynamic range.

Addendum to Addendum: I wrote those last replies in the wee hours after a full day and night of travelling so, apologies for the rambling 

 

[ruffrecords]

I think it is important to consider  function when discussing NFB in tubes. When talking about tube amps many people immediately think of power amps. In power amps you are typically operating within 10dB of clipping - with a 100W amp once you are over 10W you are in this zone. So the headroom is small and the transition from normal operation to clipping is more likely to be reached. The gentle transition between the two in a tube power amp with little or no NFB probably explains why tthis is preferred by many.

Inside a mixer, things are different. With a 250V HT supply it is not too hard to design a preamp capable of swinging over a 150V range - this is about 36dBu. So if the nominal internal operating level is 0dBu you have an enormous 36dB of headroom. This is immune to almost all overload conditions. Even an output stage driving a 600 ohm load via a 2:1 transformer can achieve +30dBu before clipping and I do  not know of any interface that would not clip before this does. So in this instance negative feedback can be put to good use.

Cheers

Ian