Negative Feedback

[Wavebourn]

The particular shape of the transfer curve affects how the amplifier sounds. The position on that transfer curve you choose for signal zero also affects it. So if you want a particular result, you achieve it by shaping the transfer curve(s).

This assumes that the distortion is audible to begin with. In order to achieve any particular result in terms of sound, the difference must be audible. But since in the big picture, perfection in the recording chain does not exist, there is not much harm from subtle distortions if they serve some purpose. I accept this even though I always focus on eliminating distortion.

Distortions are not necessary audible as distortions. With a transfer function that resembles distortions of a mechanical media distortions of a solo instrument or a voice sound as if more alive, more dynamic. However, when they intermodulate with other notes that do not cause musical by-product of intermodulation, it sounds as a dirt. It is an Achilles hill of no feedback audiophile triode amps.

 

[Sorento]

In 1960s RC20 manual the rca50 already had 3 nested fb loops:
plate to driver plate
plate to driver cathode
global secondary to input cathode

https://www.engineeringradio.us/blog/wp-content/uploads/2010/09/tube-audio-amp.jpg

 

[Wavebourn]

Yes, exceptional amp. 3 carefully calculated nested feedback loops and regulated screen grid voltage. High end ahead of time. But due to unusual tubes specified copy-paste did not happen. Engineers did not understand the concept, it's value. Too bad.

 

[StevenOH]

One of the many articles in the "Wireless World" about NFB and non-linearity:

 

[triode_al]

Can't take so much negative feedback without a grain of positive one.

Now that's a positive contribution. I have dreamed up the following:
Yes, negative and positive loops can be combined. It is not fancy at all.

1. In fact, there are very nice preamplifiers using this, such as from the famous Jean Hiraga f.e. in his (1982?) ECC83 RIAA.
2. In the Dynaco PAS-X phono stage the 47kΩ gives positive feedback of 5 dB open loop. That helps the RIAA correction.
3. The positive FB improves the dynamics a bit and gives +6dB open loop;
4. it also stabilizes the cathode voltage somewhere with an out-of phase signal so the noise floor drops some 30 to 50 dB - much better than a capacitor decoupling.
5. and it can give less distortion.
6. When you are the point of 'nulling' of the positive feedback (at one frequency) and going beyond you can see the phase starts to shift of the distortion that is left.

 

[triode_al]

Yes. The problem (for understanding) is that a sine phase shifted still looks like a sine and thus gives the impression it is delayed.
That is why this sort of things is much better demonstrated with squares or triagle waves. Then you clearly see a phase shift but no delay.

jan

Once I made a copy of a Audio Research ECC82 line stage. It has a negative voltage, to bias the input tube cathode; the feedback to the input cathode was on DC level. (How to explain - normally the cathode is positive, requiring an extra capacitor in that loop; and so in the AR design it was referencing FB to a negative Vb while the summing point can be at 0V)).
And then I tested the single coupling cap between the two stages. (ECC82 + Ecc82 followed by a ECC82 cathode follower). There were very big differences to be seen from the choice of capacitor with a >200kHz square wave: overshoot inside the loop.
My reasoning was - the cap "delayed" the burst sometimes significantly; but now you say so = maybe this was phase shift; but like I am very unmethodical, I did not make big notes of which caps were best (too bad I am a fuzzy head guy 🤯).

 

[MarcelvdG]

I recreated Baxandall's distortion tests and there is a giant wrinky pink elephant in the room which Jan Didden mentioned in his article. It mainly applies to transistors that are driven to the point of almost turning off, IE producing 10% distortion before the addition of feedback.

View attachment 1388471

What has simulating a junction diode got to do with reproducing measurements of a device with a quadratic characteristic? The logarithmic current-to-voltage relation of a junction diode has nonzero Taylor coefficients everywhere, see for example https://socratic.org/questions/how-do-you-find-the-taylor-series-for-ln-x-about-the-value-x-1

With a device that has only zeroth-, first- and second-order terms, you don't have any higher-than-second-order distortion without feedback. With feedback, you get a square-root-like error signal that does have higher-order terms. That's true independent of the signal level.

With a device that has nonzero Taylor coefficients everywhere, the effect can very well be masked by the higher-order distortion that is already there anyway.

In fact, you see that clearly in Feldtkeller's calculations from 1936:

For the quadratic device (gamma = 2), the third-order distortion coefficient goes from zero to non-zero, peaks at a loop gain around -0.9 and then gradually drops again. For the power-of-1.5 device (gamma = 1.5), there is only a minor increase when the loop gain goes from 0 to -0.8 and then it gradually drops again.