• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

High-End Tube preamp with ECC88

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I can not find this phenomenom, but I do not say that somebody else can not hear the difference.
What you think is the reason to this ?

I understood that you think local NFB does not spoil the sound.
What you think makes the difference ?

I have not really considered the underlying reasons for why more gNFB sounds worse, I am simply reporting my personal experience.

One of the oft quoted reasons is that as low order harmonics go down, high order harmonics go up, and gNFB causes this. The human ear is exquisitely fine tuned to these high order harmonics. Phase shift issues seem to play a large part also.
Fundamentally though it is important to remember that THD measures are fairly inadequate at characterizing what is really going on in any given amp as there are multiple other influences on the sound.

This is an interesting article which expresses what seem like reasonable opinions. It highlights the advent of the Williamson design as the cause of the popularity of gNFB which absolutely needed copious gNFB to work;

http://www.dolphin-hsl.com/articles/distortion.html

Shoog
 
Last edited:
...One of the oft quoted reasons is that as low order harmonics go down, high order harmonics go up, and gNFB causes this...

Shoog

This is not the case with tubes. GNFB will reduce all harmonic components almost linerly. I have once studied this, but unfortunately I could not find my files at the moment.

I think that those people who can hear the difference with local and global NFB should make more deeper analysis about the possible reasons.
We, who do not hear it are not right persons to study this.
 
Ultimately DIY audio is art as much as it is science. We do it for the fun and stress relief.
Pulling together the equipment and skills to analyze what has been done already, and still left undone, by trained acoustic engineers is not what some are here for.

Shoog
 
I have not really considered the underlying reasons for why more gNFB sounds worse, I am simply reporting my personal experience.

One of the oft quoted reasons is that as low order harmonics go down, high order harmonics go up, and gNFB causes this.

This is true for both local or global NFB- it's not particular to global.

Mathematically there is no difference between the two- precisely the same trasfer function can be obtain with local, global, or any combination of feedback (loops within loops...).
The only significant difference between local/global NFB in practice, is that a global loop necessarily encloses a more complicated circuit, with more poles/zeros and possibly a more ugly open-loop transfer function. In theory you could still make this behave in precisely the same way as a series of local NFB stages, but actually doing it is tedious / difficult.

In practice then, global NFB is more likely to result in a 'peculiar' transfer function than local NFB. Not because there's something inherently bad about global NFB, but because its so bloody difficult to compensate for all the stuff inside the loop.
 
Last edited:
Shoog said:
Overall it shows that you are listening to gNFB rather than a valve amp.
That is precisely what NFB is supposed to do! Remove the effect of any active devices, so all you hear is the feedback network distortion which will always be much smaller. Even a poor resistor will be more linear than any valve or transistor.

Ultimately the proof of the pudding is how many really outstanding valve preamplifier's employ gNFB ??
Much smaller signals so much less distortion so much less need for any linearising technique. You are comparing apples with oranges. Also, remember that some people deliberately choose to use preamps with significant levels of distortion because they prefer the 'sound'.
 
This is true for both local or global NFB- it's not particular to global.

Mathematically there is no difference between the two- precisely the same trasfer function can be obtain with local, global, or any combination of feedback (loops within loops...).
The only significant difference between local/global NFB in practice, is that a global loop necessarily encloses a more complicated circuit, with more poles/zeros and possibly a more ugly open-loop transfer function. In theory you could still make this behave in precisely the same way as a series of local NFB stages, but actually doing it is tedious / difficult.

In practice then, global NFB is more likely to result in a 'peculiar' transfer function than local NFB. Not because there's something inherently bad about global NFB, but because its so bloody difficult to compensate for all the stuff inside the loop.

I would agree entirely with this.
Even compensating for a 4 element (valves, caps & transformers) supply chain is difficult and failure is a dead/poor/wrong sounding amp.

Shoog
 
Last edited:
That is precisely what NFB is supposed to do! Remove the effect of any active devices, so all you hear is the feedback network distortion which will always be much smaller. Even a poor resistor will be more linear than any valve or transistor.


Much smaller signals so much less distortion so much less need for any linearising technique. You are comparing apples with oranges. Also, remember that some people deliberately choose to use preamps with significant levels of distortion because they prefer the 'sound'.

There is far more going on here than a simple substitution of one set of distortions for a smaller set.

Why will people still consistently prefer the zero gNFB preamp to the one using gNFB. The feedback has changed the quality of the sound in some significant way which is not revealed by looking at the THD figures.

This gets down to the who why do caps sound different when they measure the same issue. Measurements are inadequate to assess the quality nof a complete amp, and using gNFB to produce good measurements will not guarantee the amp is pleasant to live with.



Shoog
 
Depending on the circumstances, the allure of an effects box (for high distortion preamps) or pure fashion and preconception (for low distortion preamps).
I would agree with the first point - but entirely disagree with the second.

Unfortunately since you would maintain there is no difference between a plain CF and a SLCF, we really can't take this discussion further without ending up disparaging each others opinions.

Shoog
 
Nelson Pass ( a designer i respect greatly) has some relevant thoughts on the issue of harmonic distortion and feedback.
He concludes that THD measurements are no way adequate to analyze the complex nature of a real world signal passing through a distorting device (amplifier), and that Global feedback in particularly has a way of introducing a hugely complex compound distortion which bares absolutely no relationship to its THD measurement;

https://passlabs.com/articles/audio-distortion-and-feedback

Needless to say, Nelson Pass strives to use the minimum amount of Global Negative Feedback possible. Put that down to fashion if you like, but I think there are thousands who would disagree with you.

Shoog
 
I've built several amps that would not tolerate greater than 8dB of gnfb without degrading the sound quality noticeably, with a measurable increase in high order harmonics (out to the 20th). Adjusting the GNFB while watching the spectral response allowed for accurate prediction of when the amp would sound good and when it would not. The levels of the harmonics when measured in percent were so low as to be effectively unmeasurable in that they were in the noise. The harmonics were in the -100dB range(ref 1Vrms=0dB). When measured in dB, it was a different matter.

Measure it with the right tool and use the proper units of measurement and it correlates well.

It is no wonder that %thd is a useless measurement. It is based on the wrong units of measurement, but it is easy to banter around and sounds good.
 
Shoog said:
The feedback has changed the quality of the sound in some significant way which is not revealed by looking at the THD figures.
Why do people keep dragging out THD, as though they think 'we' believe in it as strongly as they ignore it? A discussion can't get very far when one side criticises views the other side has not said and does not believe.

Needless to say, Nelson Pass strives to use the minimum amount of Global Negative Feedback possible. Put that down to fashion if you like, but I think there are thousands who would disagree with you.
Since when has truth had anything to do with gurus or democracy?

TheGimp said:
I've built several amps that would not tolerate greater than 8dB of gnfb without degrading the sound quality noticeably, with a measurable increase in high order harmonics (out to the 20th).
Thank you for demonstrating a well-known problem with insufficient feedback. People often seem to think that if a little feedback is bad then more must be worse. Not so.

It is no wonder that %thd is a useless measurement. It is based on the wrong units of measurement, but it is easy to banter around and sounds good.
On here THD seems to be mainly bantered around by people who ignore it, but think others swear by it.
 
Last edited:
This is not the case with tubes. GNFB will reduce all harmonic components almost linerly. I have once studied this, but unfortunately I could not find my files at the moment.

It should make no difference. After all, Norman Crowhurst's theoretical and practical research concerning the effect of NFB on low level, high order harmonics was done with VTs.

I think that those people who can hear the difference with local and global NFB should make more deeper analysis about the possible reasons. We, who do not hear it are not right persons to study this.

Believe me, I would like to. One of my projects included variable gNFB, from none at all, to about 13db of gNFB. While playing with that twiddle knob, the difference was obvious. Running open loop produced the expected "edginess" or overly "aggressive" sound. Dialing in more first took the edge off, then by the time you hit the full 13db of gNFB, the sound was definitely developing that "solid state" sound. It might not have been a bad compromise for a production amp, and it wasn't bad with some forms of music (soft rock, classical) but left others (metal, techno) sounding "subdued". Reducing the gNFB brought metal and techno "back to life".

You may not "hear it" only because aural memory is very short lived, and highly unreliable. You need to be there, and listen as different levels of NFB are selected in real time.

Then there is the phenomenon of "getting used to". You can listen to some truly horrid sounding designs until you "forget" how appalling they really are. You can even come to prefer it, given enough exposure. This doesn't happen in other areas of design. A poor video amp never starts looking good no matter how long you look at a video display with all the wrong colors, smeared large objects, missing low level detail, broken vertical lines, and all the other artifacts of phase shift and high frequency attenuation.

On one forum for music of the 1960s, one contributor said he didn't want to listen to the good remasters of old songs on decent equipment precisely because he didn't want his memories of listening to these songs on a pocket transistor AM radio (and you know what sonic abominations those things were) messed with. Yeah, I can understand how being able to actually understand the lyrics of Walk Away Rene (Left Banke -- a song whose lyrics are notoriously difficult to hear clearly unless played through premium sound systems. Even the 4 Tops got the lyrics wrong on their cover version) just might do that to youthful memories.

This is what audio grifters mean by "burn in" after they con you into smearing their "magical" goop on the cones of your expensive speeks. (And, yes, this was a real product -- C-37 I believe it was called -- and a real claim. At least the "magic stones" just sat on top of your speeks, not destroy them.) Or why they tell you their $1000/foot "magic cables", or their lousy beeswax capacitors rolled by 20 y/o virgins on their bare thighs in a Tibetan Buddhist monastery by the light of the full moon on Beltane eve cost $Hundreds whereas coupling capacitors from Mouser cost pennies, need "burn in" time. Again, time to convince yourself you wasn't had, and that it made a real difference when it didn't really make any difference at all, or made the sound worse, not better. Or why some audiophoolz spend unnecessary BUX on silver or silver plated copper wire. It makes a big difference at 400MHz, but none at all at 400Hz. I save the silver wire for the UHF projects, not waste it on low frequency projects.

If you see a claim that a component needs "burn in", ask yourself: why didn't they burn it in at the factory before it went out the door?
 
The sound quality improved from no GNFB to around -6 to -8dB GNFB. From there sound quality continued to degrade until I ran out of available gain to continue increasing the feedback.

So two conclusions can be drawn (1) increasing feedback beyond -8dB degrades sound quality, or (2) I had insufficient gain available to allow me to increase feedback to the point where it again improved sound quality.

IIRC, I could only go to about -15dB GNFB before output dropped to the point where listening tests were useless due to the decrease in volume.
 
Data is available, Stuart. Take for example typical Williamson, 3-stage capacitive - coupled, with global feedback by voltage in series with input signal.
When OL gain goes down due to non-linearities and losses in output transformer and output tubes the driver pushes harder. It causes grid currents on peaks that charge coupling capacitors shifting bias point.
Also it causes increase of voltage swing between grid and cathode of the 1'st stage (highlighted very well in Jan Didden's presentation on the BAF!), increasing distortions of the stage that corrects errors.

I.e. global feedback causes distortions modulated by envelope. Edit: sorry, more strict would be to say that it changes this modulation, and can make it more pronounced.

In order to maintain stability we need to add compensation that means feedback on higher frequencies is lower, and this lack of feedback on highs even highlights this "dynamic modulation of distortions".

The problem is, when we measure data using steady signals we measure what we hear less rather than what we hear more but that we don't measure: dynamic change of distortions modified by feedback itself. When you hear change in spectrum the faster it happens the more it is audible. It is the main contradiction between hearing and measurements. But the data is available. You can push your amp harder than in real life, in frequency, speed of changes, levels, then extrapolate and predict distortions in the real life. And you can analyze processes that happen in each stage when they are surrounded by common feedback loop. Just pay more attention on dynamics that is more audible than static measurements reveal, and we all will be happy. :D

Global feedback is not bad. What is bad, it's understanding.
 
Last edited:
Data is available, Stuart.

We're talking about preamps here. Overload in power amps driven to clipping is a different story, thus the rather elaborate measures I used in my power amp to minimize recovery time, or your soft clipping circuits.

Global feedback is not bad. What is bad, it's understanding.

Precisely. As with any other aspect of circuit design, if you don't understand it, you're likely to get a bad result. Some attribute it to the design element rather than their own inability to understand and use the tool properly.
 
Since when has truth had anything to do with gurus or democracy?

At this point you lose all credibility. The F5 amp is probably the best amp I have ever heard - which ultimately is what matters here.

He is not a Guru from selling snake oil, he sells solidly engineered products based on sound understanding of his chosen components. I would take his word over almost anyone on these boards.

Shoog
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.