No global negative feedback

its transient response is very much determined by the source impedance driving it.
I'm not convinced. This is because the electrical 'damping' seen by the speaker is the entire loop impedance of the speaker voice coil, crossover components, cables and finally the amp output impedance. The impact of the amp's output impedance on the damping is substantially reduced by all the other bits, perhaps 6 Ohms against 0.01 or 0.1.

Paul
 
but you'll find, as with the CJ pre amp and Radford power amp mentioned in the article, that these were designed from the outset to be no/low NFB designs.
The Radford amp would have been designed to have the maximum NFB possible given the limits of phase shifts, transformers and restricted open loop gain.

Paul
 
There was some research done years ago by Harold Leak on just when distortion becomes audible. If I can find the link I'll post it but his conclusion based on subjecting listeners to the same circuit was that anything above 0.1% THD could be reliably and repeatedly detected. This lead to the naming of the Leak 'Point One' amp.
 
I'll stick with the increased output impedence making the largest difference, although I disagree with Rob about the effect of high output impedence on speaker frequency response - it's insignificant compared with the affect of the room acoustics and likely to go totally unnoticed in most cases.

I'd certainly agree with that in many cases. IIRC from what you were saying yesterday about your own speakers, 16 ohm load and >100dbw sensitivity the effects will be swamped as you say.

However, driving a complex 4 Ohm load is a different story.
 
What makes you think it is a complex load... :confused:

Actually Schallwand used to sell them with a SONY playstation 1002 plus a Sonic-T amp, both modified, but only damping, supports and PC, for people on a tight budget, I ran them with a all in one TEAC, 20 Tripath watts, with no problems, on the contrary...
 
I'm not convinced.Paul

I wouldn't expect you to be;) As I understand it ,the output impedance of the amplifier must be added to the dc resistance of the speaker cable and speaker voice coil resistance in the damping equation. So it's theoretically possible to have a negative output impedence with an even higher damping factor? I think Nelson Pass has experimented with that.

But it's an interesting proposition and Dave Berning claims to have measured the results.

I made a gadget for observing the optimum damping of a speaker cone in response to a step transient driven from various impedances. The device uses a lightweight conductive foil that can be taped to the cone of the driver, and a matching RF based capacitance probe is used to detect cone movement and display it on an oscilloscope. I can confirm various degrees of over and under damping, and have tested speakers that range from wanting to see as high as a 7.5 ohm amplifier output impedance (this was a large driver and quite unexpected) to speakers that are still underdamped with zero output impedance. I find that most high-end speakers fall in the range of working best with 2-4 ohms amplifier output impedance.

Rob,

Even with a fairly complex 4 ohm load, and an amplifier output impedence of let's say 3 ohms, the variations in frequency response are going to be tiny compared with those generated by a listening room. That is David's point, and measurements taken by Charles Hansen seem to bear that out.
 
It's easy enough to just stick a resistor in series with any amp's output.

I'm surprised by David Berning's findings if only because most amps have a very low output impedance and most speakers are designed to be driven by a voltage source. Possibly 'high end' speakers tend to be a bit odd?

Paul
 
Distortion, Output Z and others....

Hi,

I'm not sure that we part ways.

What you are saying is that different types and orders of distortion have different effects on the final sound. I agree with that.

The question is, do you prefer the final sound because the ear 'likes' certain types of distortion in modest amounts, or is the absense of NFB in itself the reason for the preference?

I don't think that the case for saying that NFB is a 'bad thing' is in any way proven. The problems at clipping are of course valid but I'm more concerned with performance within rated output.
[/QUOTE]

Well, there is a good body of research on the subject of harmonic distortion audibility.

Some really interresting ones are recent, such as that published recently by Hifi News whjere they had extensive listening tests to determine if adding certain types of distortion "improved" the sound (conclusion it doesn't) and Earl Geddes research on distortion audibility (which cofirms the work of Shorter and Crowhurst and extends this). Well worth reading and together with a lot of other research strongly indicative and supportive of "added HD does not per se make for better percieved sound".

Equally, since the 50's a number of negative effects of inverse feedback have been documented, throught the work of Ottala in the 70's, to list:

Blocking/Saturation on overload
Transient Intermodulation
Phase Intermodulation
Frequency Intermodulation

All these have not been particulary tested for audibility up till now, PIM and FIM are still extensively debated despite reliable and easy proof for them being present.

At this stage of our understanding of what humans can percieve and what not we must conclude that very high levels of simple, low order HD are inaudible and that equally very small levels of high order HD are not only audible but highly objectionable.

We can further observe that the application of inverse feedback produces "distortion of distortion", in other words the order of HD is is multplied, even though the absolute summed value of THD is drastically lowered. Thus largely or completely inaudible HD may be converted into HD of lower value that is audible and objectionable.

Further, inverse feedback may give rise to a number of other forms of distortion which have a so far unquantified effect on human perception.

All of the above are at least indicative that inverse feedback may give rise to undesirable sonic effects and that at the very least the use of inverse feedback is suspect and should be carefully evaluated if applied.

Some other areas of concern impact on speakers.

First, Speakers have high levels of distortion, well in the same region as "non feedback" valve amplifiers. Some of this distortion and especially the higher order components are a result of driving the speaker from a low impedance source. If we increase the output impedance of the amplifier we may get lower distortion from the Speaker.

Second, the noatble levels of thermal compression (BTW, the HFN Article on thermal compression in my view actually underscores the relevance, as the speaker tested altered it's tonal balance overall by nearly 0.7db towards "bright sounding" when driven loud) in speakers decrease as the source impedance rises.

The "lack" of electrical damping as a result can be replaced by mechanical damping in either enclosure or driver. Such a solution is at any extent preferable as mechanical problems ideally should be solved mechanically, as evidenced by the worlds first high fidelity coaxial speaker, the legendary Eckmiller designed in the late 30's....

So, we have some reasonable arguments for "non feedback", few if any relevant ones for (low output impedance and low levels of THD not offering per se any relevant or reliable improvement in sonic quality) "inverse feedback". So I think people should make up their own minds by listening. Antonio seems to have done so, so more power to him.

Ciao T
 
Hi,

There was some research done years ago by Harold Leak on just when distortion becomes audible. If I can find the link I'll post it but his conclusion based on subjecting listeners to the same circuit was that anything above 0.1% THD could be reliably and repeatedly detected. This lead to the naming of the Leak 'Point One' amp.

There was a LOT of research done since, which has comprehensively invalidated this "more than 0.1% THD are audible" position.

However, I am arguably blaming Mr. Leak for setting of the still ongoing THD Wars.

More to the point, how would Mr. Leaks listeners have been able to hear 0.1% THD through the much higher THD from the Speaker?

Ciao T
 
Hi,

PS: You missed a good BBQ yesterday.

I know, but I had to be somewhere else specifically to ensure a certain review continues without hitches and problems, I very much regreted this, as I'm suspecting I missed great food, which I'd have preferred by far to the expensive and not very good Chicken and Mushroom Pie and Beer at the Crown Inn (staff was friendly at least)....

Ciao T
 
Last edited by a moderator:
We can further observe that the application of inverse feedback produces "distortion of distortion", in other words the order of HD is is multplied, even though the absolute summed value of THD is drastically lowered. Thus largely or completely inaudible HD may be converted into HD of lower value that is audible and objectionable.

Where can we observe this Thorsten? I would be most interested in seeing evidence of this. Any links to studies backing this up with real world examples?

Further, inverse feedback may give rise to a number of other forms of distortion which have a so far unquantified effect on human perception.

Sorry but that is pure conjecture surely?

All of the above are at least indicative that inverse feedback may give rise to undesirable sonic effects and that at the very least the use of inverse feedback is suspect and should be carefully evaluated if applied.

Again, without some case studies, it is purely conjecture. I do understand that global feedback loops in most valve amps are taken after the output trannie and this causes potential issues but I'd love to read something that verifies your statements Thorsten.

Some other areas of concern impact on speakers.
Ciao T

I think you will agree that we have been discussing the impact on speakers, but pointing out that this could well be merely the result of overdamping transients because of ultra low output impedences rather then because of the NFB loop itself. It would be interesting to rig up a variable output impedence buffer and observe the effects.

So, we have some reasonable arguments for "non feedback", few if any relevant ones for (low output impedance and low levels of THD not offering per se any relevant or reliable improvement in sonic quality) "inverse feedback". So I think people should make up their own minds by listening. Antonio seems to have done so, so more power to him.

I agree about listening but I don't think you've presented a viable case for Antonio's removal of the NFB loop as being anything other than a wild gamble that may fortuitously result in subjectively preferable results. And without measurements highlighting the before and after performance, it is impossible to say whether Antonio is hearing the increase in distortion. It is only possible to say that distortion from the amplifier will undoubtedly have increased considerably, and that it most certainly will not have reduced!
 
Hi,

Where can we observe this Thorsten? I would be most interested in seeing evidence of this. Any links to studies backing this up with real world examples?

You can observe this using a spectrum analyser of suitable design (EMU 1616M and for Example RMAA will do) and with a suitable circuit. Or you could read Crowhursts 1950's AES Articles, which can be obtained from www.aes.com


Sorry but that is pure conjecture surely?

Well, if you consider Otala's, Crowhursts and others articles in JAES (all peer reviewed and backed by measurements) conjecture, be my guest.

Again, without some case studies, it is purely conjecture.

Annecdotal evidence is legion, including Martin Colloms Article, among many.

I'd love to read something that verifies your statements Thorsten.

Clearly Martin Colloms fails to convince you. The link between Inverse Feedback and TIM as demonstrated by Dr. Otala, as well his demonstrations of PIM as result of the use of NFB fail to convince you, I won't add much more, sorry. If you wish to reject what has been basically established for many decades as fact, be my guest.

I think you will agree that we have been discussing the impact on speakers,

To discuss amplifiers alone, without considering the speaker they drive is ranking in irelevance right up with certain current political figures.

but pointing out that this could well be merely the result of overdamping transients because of ultra low output impedences rather then because of the NFB loop itself. It would be interesting to rig up a variable output impedence buffer and observe the effects.

Actually, I worked extensively with variable output Impedance in the 80's, including the use of negative impedance. I found that using open loop, current output (high output Impedance) amplification to drive dynamic treble and midrange units was by far the best solution, however it demands an active speaker.

I struggeled long and hard to get a bass section that matched the midrange and treble, ending up with Inverse Feedback around the woofers. Using inverse feedback there tought me much about how it works, what it's limitations are and so on.

I agree about listening but I don't think you've presented a viable case for Antonio's removal of the NFB loop as being anything other than a wild gamble that may fortuitously result in subjectively preferable results.

I never presented any arguments for it being other than a gamble. In his case (and in many others) the gamble worked out, but there is no unversal case for removing NFB, nor is there any for having it.

I noticed however that after Antonio presented his experience ("I removed NFB and liked the results") load of people jumped on to declare that thuis should never be done and cannot work and pretty much that Antonio should put it back and not speak about it again.

That is an approach I have noticed here on several occasions when views and experiences are presented that seem unpalatable to some.


And without measurements highlighting the before and after performance, it is impossible to say whether Antonio is hearing the increase in distortion. It is only possible to say that distortion from the amplifier will undoubtedly have increased considerably, and that it most certainly will not have reduced!

First, if the open loop amplifier produces 1.2% THD @ 10W I would consider it pretty low Distrotion, at least for normal listening levels using 96db/W/m efficient Speakers. As in a PP Output stage 3rd HD dominates and this goes down with the square of the reduction of level we can expect around 0.12% 3rd HD from the amplifier at an SPL from the speaker of 96db and comparably or more 3rd HD from the Speaker.

Further, the levels of this distrotion are well below anything suggested as audible by current research (eg. Geddes/Lee) which I would recommend to you for study, so I think we are reasonably safe to conclude that harmonic distortion levels are sufficiently low in Antonio's case not to play a material part in the equation.

As to reduced/increased, usually when adjsuting an amplifiers feedback levels towards zero we find HD to go up and TIM and PIM to go down.

So we trade an increase in one form or type of distortion (harmonic), which we have good evidence is in fairly high levels at low orders quite inaudible, for a reduction in other types of distortion (PIM, TIM, FIM) for which we have some formal and much annecdotal evidence as to them being quite objectionable.

In the case of Antonio the differences overall only amount to 4.8db, which is not much, but as MC found himself, as little NFB as that can cause audible consequences that are percieved negatively.

Ciao T
 
Hi,

You can observe this using a spectrum analyser of suitable design (EMU 1616M and for Example RMAA will do) and with a suitable circuit. Or you could read Crowhursts 1950's AES Articles, which can be obtained from www.aes.com

I don't have access to the AES site I'm afraid but, whilst I was aware of higher order HD being attenuated less than lower order by NFB loops, I have never read anywhere that they create higher order harmonics from nothing. I can't find anything on this.

Well, if you consider Otala's, Crowhursts and others articles in JAES (all peer reviewed and backed by measurements) conjecture, be my guest.


Sorry but I consider anything that has a "so far unquantified effect on human perception" to be conjecture when it comes to audibility.

Clearly Martin Colloms fails to convince you.

Absolutely. he's a hifi journalist. On top of which his "experiments" are fatally flawed, as one would of course expect.

Annecdotal evidence is legion, including Martin Colloms Article, among many.

Annecdotes aren't evidence Thorsten. Again, the big mistake being made by Colloms and others is that they fail to take into account the increase in output impedence. Without isolating that aspect, it is impossible to take a meaningful view of the audibility of the NFB loop

The link between Inverse Feedback and TIM as demonstrated by Dr. Otala, as well his demonstrations of PIM as result of the use of NFB fail to convince you, I won't add much more, sorry. If you wish to reject what has been basically established for many decades as fact, be my guest.

Actually it's not established fact as I am sure you are aware. The audibility of TIM that was claimed by Dr Otala is open to conjecture and hasd been for the past 30 years and again any objective evidence fails to take output impedence into account. As far as I am led to believe, general scientific thinking is that TIM is not an issue in any sensibly designed amplifier - the darker side of the "audiophile industry" would rather we believed differently. The fact that nothing has been proven in over a quarter of a century makes me naturally sceptical.

I never presented any arguments for it being other than a gamble. In his case (and in many others) the gamble worked out, but there is no unversal case for removing NFB, nor is there any for having it.

I noticed however that after Antonio presented his experience ("I removed NFB and liked the results") load of people jumped on to declare that thuis should never be done and cannot work and pretty much that Antonio should put it back and not speak about it again.

That is an approach I have noticed here on several occasions when views and experiences are presented that seem unpalatable to some.

I don't think many did. I pointed out that he had increased distortion and output impedence - but also pointed out that this could sound nice. An amp like Antonio's lends itself to minimal amounts of NFB anyway, but there is always the possibility/probability that the designer would have produced a more linear design had he not anticipated having the NFB loop. I also pointed out that in the opinion of some, he has not removed the audible distortions of the NFB loop, but instead ended up with more appropriate speaker damping due to the increased output impedence. I'm glad you have said yourself that removal of the feedback is nothing more than a gamble. Out of interest, what now happens to the gain on the amp in question?

First, if the open loop amplifier produces 1.2% THD @ 10W I would consider it pretty low Distrotion, at least for normal listening levels using 96db/W/m efficient Speakers. As in a PP Output stage 3rd HD dominates and this goes down with the square of the reduction of level we can expect around 0.12% 3rd HD from the amplifier at an SPL from the speaker of 96db and comparably or more 3rd HD from the Speaker.

Further, the levels of this distrotion are well below anything suggested as audible by current research (eg. Geddes/Lee) which I would recommend to you for study, so I think we are reasonably safe to conclude that harmonic distortion levels are sufficiently low in Antonio's case not to play a material part in the equation.

I've read a lot of Earl's work , particularly his papers on waveguides. Interesting but even that wouldn't explain why 0.005% of say 9th harmonic would sound worse than 1.2% of 2nd would it?

MC found himself, as little NFB as that can cause audible consequences that are percieved negatively.
Ciao T

That's where we differ. I don't think he discovered anything of the kind, and would not consider his article for Stereophile to be credible evidence in this case. That seems to be profoundly lacking when it comes to the negative effects of NFB.
 
Last edited by a moderator:
I have never read anywhere that they create higher order harmonics from nothing.
I guess the hypothesis that the 2nd harmonic generated by the amp is fed back and turned into 4th on its second time around the loop etc.

The problem is that there is little or no 2nd harmonic on the actual output output of the amp, so very little or less 4th can be generated. Push 2nd down to negligible and 4th et al drop off the map.

But is this is a reasonable way to look at feedback?

Paul
 
Hi,

I guess the hypothesis that the 2nd harmonic generated by the amp is fed back and turned into 4th on its second time around the loop etc.

Not quiet. There are of course many different ways of looking at NFB.

We could look at the process of NFB as one of sucessive approximation by which the looped system attempts to match the output to the input (or with a suitable voltage divider to a determined multiple) but, due to limits of bandwidth and loop gain never achieves, it's "hunting" operation produces errors of it's own.

These will take a number of forms, including "noisemodulation" artefact (what Hawkesford calls "Fuzzy Distortion") and due to the underlying mathematical principle there is also a multiplying operation on the harmonics, while the amplitude of the main harmonics is reduced.

Seeing Feedback loops operating at very low speed (like in industrial or military applications) is highly instructive for their intuitive understanding.

But the view that the 2nd HD the NFB loop attempts to cancel by injecting it into the nonlinear amplifier input in opposite phase is being re-distorted to produce 4th and higher HD and equally that the 3rd HD being re-injected is being re-distorted to produce 5th and higher HD is certainly a valid, if simplified view of the systems operation.

The bottom line is that feedback, be it positive or negative is a design tool. Just like any other tool, if it is sufficiently well understood it can be used where required, however if applied where un-needed it's negative consequences may outweigh it's use.

I never found musch use for looped inverse feedback.

BTW, Mick I have worked with NFB in situation where the output impedance is not affected (example - Preamplifiers) and the observations on the impact of NFB where not essential remain.

Ciao T
 
I guess the hypothesis that the 2nd harmonic generated by the amp is fed back and turned into 4th on its second time around the loop etc.
Paul

Hi Paul,

That's at odds with Rod Elliot who claims on his site

Feedback does not - repeat does not - cause the signal to travel from the output, back into the inverting input, and continue through the amplifier several (or multiple) times. By not doing so, it does not (and can not) create additional harmonics that did not exist before the feedback was applied.

I'd love to know who's right! I certainly don't feel that adding or reducing feedback on my amplifier universally improves or damages the sonic performance - the results are dependent on the speaker and quite possibly the room and cabling to a lesser degree. I do know that the feedback loop will proportionately attenuate the lower order harmonics a great deal more than the higher ones, but the higher ones are at such low levels anyway. Yes I also appreciate that the threshold for hearing higher HD is lower, but not that much surely, especially given the low levels of HD displayed post NFB loop?
 
Hi,

That's at odds with Rod Elliot who claims on his site

Well, self taught, self publishing Mr. Elliot with no formal electronic education (I challenged him on that once) vs. Norman Crowhurst, widely published luminaire on electronics and audio.

BTW, Crowhurst never claimed that there where "multiple travels around the loop", he merely observed that applying looped NFB to a nonlinear amplifier generated new harmonics, not present in the original amplifier. There are many valid views explaining why, depending on the degree of abstraction.

Take your pick.

You can read a reprint of Crowhurst Glass Audio, Vol 7-6, pp. 20 through 30....

BTW, forgive me for saying so, but anyone quoting Elliot as authority qualifies him or herself as belonging to the genus ignotum per ignotius. RE is a decent collector and republisher (usually without bothering to reference his sources) of other peoples designs and is highly opinionated, not much else.

Ciao T
 
BTW, forgive me for saying so, but anyone quoting Elliot as authority qualifies him or herself as belonging to the genus ignotum per ignotius.Ciao T

Obviously you don't see eye to eye -an ego thing? You on the other hand quoted Martin Colloms.
 
Hi,

Obviously you don't see eye to eye -an ego thing?

Not really, the man is pretty ignorant of anything beyond basic electronics and is smallminded enough to vigerously attack anything he does not understand, which sadly is a lot.

Worse, he is a hypocrite who will criticise in others what he himself does commercially (after all, HE sells a tube preamp that measures quite bad compared to anything Op-Amp based he has published, but unsurprisingly sounds much better, so as long as HE is doing it's fine, he can use others peoples work on his pages most of the time without reference, but dare anyone "steal" material he in the first place pinched, beware)....

http://sound.westhost.com/vp103.htm

http://sound.au.com/shame.htm

You on the other hand quoted Martin Colloms.

I did not quote him, I mentioned his article as one of the more high profile examples of educated listeners commenting on certain issues audibility.

You would be really well advised to change your sources from those available free on the web of a million lies to ones that may cost a little (or even require an AES membership) but are somewhat more trustworthy, at least, though the JAES has mcuh to answer for as well....

Ciao T
 
Thanks for the advice Thorsten. One thing.

As Crowhurst came up with a mathematical model for the propagation of high order HD some half a century ago, could you point me to measurements of a real world device that backs up his theory? Something that is generating the levels of high order HD that would be audible at normal replay levels? Or examples of amps that suffer in this department.

I definately detect unnatural hostility to the Mr Elliot - almost of feline proportions , so suggest we move on from him. As for Mr Colloms, as I pointed out, was what he heard the result of the lessening of feedback with the Cary or the interaction of the amp with the speakers? Did he ever stop to ask that question? I can't take anyone seriously who goes into print claiming the latest Naim CD players set remarkable benchmarks. Nice guy undoubtedly, and a very experienced listener though.
 
Last edited by a moderator:

Latest posts

Back
Top