What is it about Negative Feedback?

[Sub Sonic]

As far as I am aware (I’m no electronics guru), negative feedback can be part of an amplifier circuit and is used for error correction/distortion reduction, reducing output impedance and possibly other things.

 

Various parties are anti Negative Feedback, or anti Global Negative Feedback.

 

If Negative Feedback is used for error correction, why are some amplifiers said to sound better without it, or with less of it? Are there objective reasons why it is better or worse, or does it come down to subjective listening? Is it dependent on the characteristics of the speaker attached to the output?

 

Any thoughts appreciated.

 

SS

 

Edited January 19, 2019 by Sub Sonic

[andyr]

I believe what Hugh ('Mr AKSA') Dean tells me. 

 

Which is:

• amps with high levels of global NFB sound less pleasing to the ear than ones with no - or low levels of - global NFB.
• local NFB is not such a big problem.
• high levels of distortion are not, per se, bad - tube amps can sound beautiful ... even though they have much higher levels of distortion than typical ss amps.
• what is important for a good-sounding amp is that it has a decreasing harmonic distortion profile.  (So H2 is higher than H3 ... is higher than H4 ... etc.)

Andy

 

[Wimbo]

In solid state, I always linked negative feedback with a slow undynamic amp.From memory, its used to correct the output to give lower (on Paper)distortion.

[Guest Eggcup The Daft]

51 minutes ago, andyr said:

I believe what Hugh ('Mr AKSA') Dean tells me. 

 

Which is:

• amps with high levels of global NFB sound less pleasing to the ear than ones with no - or low levels of - global NFB.
• local NFB is not such a big problem.
• high levels of distortion are not, per se, bad - tube amps can sound beautiful ... even though they have much higher levels of distortion than typical ss amps.
• what is important for a good-sounding amp is that it has a decreasing harmonic distortion profile.  (So H2 is higher than H3 ... is higher than H4 ... etc.)

Andy

 

These statements are, of course, generalisations, and other aspects of any particular design may be more important.

 

Many amps use global NFB and most people will have heard an amp with global NFB that they like, and will own or have owned one. It's not the only choice or compromise that an amp designer has to make. A lot of designers, maybe the majority, will defend their use of global NFB in creating a stable, low distortion amplifier

 

As I understand it the case against global NFB is based on two things, the distortion pattern and time smearing. Placing NFB around a complicated amp design will result in the feedback signal being far enough behind the input signal to compromise it, especially at high frequencies. It's also noted in discussions of this subject that it may be necessary to have additional circuitry to prevent the feedback signal from sometimes turning positive and that that additional circuitry adds problems of its own.

 

As far as the harmonic profile goes, I've seen other claims - for example, that higher odd harmonic profiles are the ones to avoid, and that very high harmonics are less worrying because they can't be heard

 

The principles seem good, but it's all about implementation, like much else in audio.

[andyr]

1 hour ago, Eggcup The Daft said:

These statements are, of course, generalisations, and other aspects of any particular design may be more important.

 

Sure.

 

1 hour ago, Eggcup The Daft said:

Many amps use global NFB and most people will have heard an amp with global NFB that they like, and will own or have owned one. It's not the only choice or compromise that an amp designer has to make. A lot of designers, maybe the majority, will defend their use of global NFB in creating a stable, low distortion amplifier

 

I didn't say global NFB was bad, per se.  I said an amp with high levels of GNFB tends to sound worse than an amp with low levels of global NFB.

 

Andy

 

[Ittaku]

There's a fascinating amplifier in the current edition of Stereophile called the CH Precision I1 that allows you to adjust the degree of global NFB from 0 to 100%. Never has there been a clearer way to assess electrically and audibly what effects GNFB do - in this design at least - compared to just local feedback. (Note default setting is 0% global, 100% local.)

[Sub Sonic]

39 minutes ago, Ittaku said:

There's a fascinating amplifier in the current edition of Stereophile called the CH Precision I1 that allows you to adjust the degree of global NFB from 0 to 100%. Never has there been a clearer way to assess electrically and audibly what effects GNFB do - in this design at least - compared to just local feedback. (Note default setting is 0% global, 100% local.)

Mmm thanks, but at $38k US I might settle for anecdotal answers

[Guest Muon N']

47 minutes ago, Ittaku said:

There's a fascinating amplifier in the current edition of Stereophile called the CH Precision I1 that allows you to adjust the degree of global NFB from 0 to 100%. Never has there been a clearer way to assess electrically and audibly what effects GNFB do - in this design at least - compared to just local feedback. (Note default setting is 0% global, 100% local.)

I'd imagine it has been done before more than once, but then  It's just an adjustable resistance so easy to implement so I'd be surprised if not.

 

[michaelw]

52 minutes ago, Ittaku said:

There's a fascinating amplifier in the current edition of Stereophile called the CH Precision I1 that allows you to adjust the degree of global NFB from 0 to 100%. Never has there been a clearer way to assess electrically and audibly what effects GNFB do - in this design at least - compared to just local feedback. (Note default setting is 0% global, 100% local.)

 

the 1995 cary cad-805 also had user adjustable global feedback, albeit a smaller range from 0-10db.

[Guest Muon N']

Manley also https://www.manley.com/tech-pages/negative-feedback/

[Guest Eggcup The Daft]

3 hours ago, michaelw said:

 

the 1995 cary cad-805 also had user adjustable global feedback, albeit a smaller range from 0-10db.

Hard to tell if it was a smaller range, you quote 0-10dB but the figure for the CH is "0-100%" which carries, er, less meaning.

[Buffle]

It’s difficult to imagine how an adjustment to set the amount of global NFB would be particularly meaningful.  Reducing NFB increases gain, or more accurately NFB reduces gain.  There’d have to be a proportionate reduction in gain.  But designing with enough NFB in mind to surpass the “makes things worse” point means designing a circuit with sufficient gain in the first place, which is likely to be different than would be designed if no global NFB was the target.  So dialing in no NFB might mean going into the range in which the circuit might not be operating optimally.

[Guest Eggcup The Daft]

7 hours ago, andyr said:

amps with high levels of global NFB sound less pleasing to the ear than ones with no - or low levels of - global NFB.

 

4 hours ago, andyr said:

I said an amp with high levels of GNFB tends to sound worse than an amp with low levels of global NFB.

These two phrases are subtly different. I didn't actually question either version though. I said that amplifiers with global NFB are common and liked. I was attempting to say that the situation is more nuanced than you made it sound, that's all.

 

For the record, I own a Simaudio Moon amp that is stated to have no global NFB and chose it over others that have global NFB (low levels as I understand it). I'm not aware that amps with high levels of global NFB necessarily sound worse and would be interested to know if anyone has solid evidence to back such a claim.

 

Similarly, would high levels of local negative feedback be regarded as an issue? Don't some of the better regarded opamps use high levels of negative feedback, for example?

 

[walker1000]

All the amplifiers I have owned used negative feedback and I have nothing negative to say about any of them.

Edited January 19, 2019 by walker1000

[Guest Muon N']

14 minutes ago, Buffle said:

It’s difficult to imagine how an adjustment to set the amount of global NFB would be particularly meaningful.  Reducing NFB increases gain, or more accurately NFB reduces gain.  There’d have to be a proportionate reduction in gain.  But designing with enough NFB in mind to surpass the “makes things worse” point means designing a circuit with sufficient gain in the first place, which is likely to be different than would be designed if no global NFB was the target.  So dialing in no NFB might mean going into the range in which the circuit might not be operating optimally.

As I understand it many designs that employ GNFB (especially gobs of it) do so to add stability to a circuit that would be unstable without it?

 

I was going to mention earlier that if a variable control was included in a design it would need to be stable at all settings including '0'

[andyr]

9 hours ago, walker1000 said:

All the amplifiers I have owned used negative feedback and I have nothing negative to say about any of them.

 

Your mission (if you choose to accept it) is to write 100 times on the blackboard "I now understand global NFB is different to local NFB!". 

 

Andy

 

[sir sanders zingmore]

4 minutes ago, andyr said:

 

Your mission (if you choose to accept it) is to write 100 times on the blackboard "I now understand global NFB is different to local NFB!". 

 

Andy

 

I understand that they are different but I don’t know what the difference is. 

[Ittaku]

9 hours ago, Eggcup The Daft said:

Hard to tell if it was a smaller range, you quote 0-10dB but the figure for the CH is "0-100%" which carries, er, less meaning.

It's a shift from 100% global to 100% local feedback, which is also subtly different to just adding and removing GNFB. Either way, I still thought it would be interesting to dial it and see just how it sounds different between those two.

[andyr]

37 minutes ago, Sir Sanders Zingmore said:

I understand that they are different but I don’t know what the difference is. 

 

The difference - AIUI, Trevor - is that:

• global NFB connects the final output signal of the circuit back to the input, whereas
• local NFB connects the output from a gain stage within the overall circuit, back to its input.

 

How exackly you do this ... I have noooo idea! 

 

Andy

 

[blybo]

just like @Eggcup The Daft I run a Sim Moon amplifier (Neo ACE), at least partially due to their Zero GNFB design across their range. I'm not electronically minded but in discussions on here around 18 months ago, it was concluded that zero GNFB designs were more likely to sound more engaging at lower volumes/outputs, something I was particularly interested in finding because my wife is constantly turning the volume down at home, and because my speakers are reasonably efficient.

[davewantsmoore]

19 hours ago, Sub Sonic said:

If Negative Feedback is used for error correction, why are some amplifiers said to sound better without it, or with less of it?

If you do negative feedback "badly", it sounds very bad.

 

Designing such a control loop with high performance, isn't trivial.     At the risk of over generalising, how it works out is that if you have a high performance control loop, you want it to use lots of feedback (more is better).     Using more makes it sound worse, if you control loop isn't 'perfect'.     Using a "weak" not well performing control loop, will improve some headline measurements like THD/IMD .... but otherwise sounds bad.

[davewantsmoore]

19 hours ago, Sub Sonic said:

Is it dependent on the characteristics of the speaker attached to the output?

Yes.... but only when people aren't making a sensible apples to apples comparison.

 

Negative feedback lower the output impedance of the amplifier.    Meaning we can have a situation where no feedback amp has "high output impedance" and feedback amp has "low output impedance".

 

If the speaker does not ave a flat impedance curve .... then the high output impedance amplifier will cause a different balance of frequencies (Eg. more bass, more treble, peaks and dips in places) to come out of the speaker.     And THIS will sound very different.

 

It isn't the sound of negative feedback per se.... it is just the sound of different frequency response  (just like if you grabbed an EQ and started messing with the frequency balance).

 

If you correct the two situations (the high output impedance and lowoutput imedance amps) to have the same FR, then this will be apples vs apples.

[Ittaku]

24 minutes ago, davewantsmoore said:

If you do * "badly", it sounds very bad.

Fixed that for you. As always, it's never about technology, but implementation.

[sir sanders zingmore]

29 minutes ago, davewantsmoore said:

Using a "weak" not well performing control loop, will improve some headline measurements like THD/IMD .... but otherwise sounds bad.

What are the measurements that would tell you it will sound bad?

[Zaphod Beeblebrox]

15 minutes ago, Sir Sanders Zingmore said:

What are the measurements that would tell you it will sound bad?

There's quite a few, but an easy one would be to drive the amp into a capacitive load with a square wave. Overshoot, or ringing suggests that some instability is present. Instability can sound unpleasant.