A few naive questions

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

As a newcomer to this group, and having been out of audio electronics for a number of years, I would be interested to have answers to the following questions

1. What is the advantage of ‘no feedback’ amplifiers, surely feedback has a number of advantages – stabilise the gain, reduce the distortion, reduce the output impedance ? Having looked at t number of the submitted designs, I ask why the use of un-bypassed emitter resistors is not considered as negativefeedback, it always was in my earlier days ?

2. What has happened to the ideas put forward by by Edward M. Cherry (Assoc. Prof. at Monash University) ?

3. Why are amplifiers now being specified / designed with a ‘flat’ frequency response curve going up to (say) 1 mHz ?

Your comments on the above would be very enlightening.

Richard
 
Gandalph said:
What is the advantage of ‘no feedback’ amplifiers, surely feedback has a number of advantages – stabilise the gain, reduce the distortion, reduce the output impedance ?

I see no advantage in "no feedback" amplifiers...they aro not true with the input signal, but as we know , not everybody is looking for accuracy .


Having looked at t number of the submitted designs, I ask why the use of un-bypassed emitter resistors is not considered as negativefeedback

Yes they use lots of un-bypassed emiter resistors and even emiter followers that have near 100% negative feedback.
Why are amplifiers now being specified / designed with a ‘flat’ frequency response curve going up to (say) 1 mHz ?

In that case I agree , even if we can't ear higher than 20Khz the wide bandwidth help to diminish the phase shift across the audio band and generaly a with bandwith design, is a signal of better engineered amplifier ( but not always ... :) ).

My target is usually 200Khz..

Regards
 
Hello,

It's more of a philosophical issue than technical.
If you go back in time 30 years or so, and read the engineering papers on the topic then, you'd find there were some companies using feedback as a 'patch' to hold together some rather poor designs, giving the poor designs some impressive specifications but poor sound.

Some purests thought this was bad and started a movement towards less feedback is better.

Eventually coming to no feedback.

Distortion in an amplifier greater than 0.1% is audible and realistically cannot be achieved without some feedback.

I'm of the *feedback is good* camp myself, so long as the design is good.

As for bandwidth, I can hear to 45KHz. So can most people, though they don't realise it. So, bandwidth to 100KHz or so is worthwhile IMO.
Beyond 100KHz it becomes a numbers game IMO.
 
Gandalph said:
1. What is the advantage of ‘no feedback’ amplifiers

Having no feedback reduces some problems to triviality, so FOR THOSE PROBLEMS it has advantages - for other, equally important for proper functioning of an amplifier, not only does no feedback have no advantages, but often it has considerable disadvantages. Perhaps the most important advantage is stability, especially with, shall we say, unflattering loads. A less important and not as pronounced, but often quoted advantage, is a more favorable harmonic spread in the distortion that is there (note I am talking only about the relative magnitude of the harmonics, not the RMS sum absolute value). For various reasons which have already been discussed on the forum many times, (global) feedback can generate amounts of higher order harmonics that are disproportionate in comparison to the low order harmonics (this phenomenon is reduced when feedback loops are kept short, eg. local feedback in a single stage). It should be noted that this particular notion of 'disproportion' exists as a result of the way we hear - mathematically speaking, the feedback amps can be far more linear.
In most cases, when 'no feedback' is cited, it is understood that no GLOBAL feedback is employed. It is arguably near impossible to design an amplifier with no feedback mechanism whatsoever.
It's a whole different story that the good and bad of feedback is often vehemently debated by people who would not know global, local or indeed any feedback from a cow's rear end :)


2. What has happened to the ideas put forward by by Edward M. Cherry (Assoc. Prof. at Monash University) ?

I have to admit I have never heared of E. M. Cherry's ideas, so no comments here.

3. Why are amplifiers now being specified / designed with a ‘flat’ frequency response curve going up to (say) 1 mHz ?

This is a side effect of another requirement. Phase angle, mentioned by another poster is actually only a relatively minor problem here. Certain distortion mechanisms (including high order distortion generation in global feedback amps) produce distortion that rises with frequency, sometimes very sharply. Designong an amp as high bandwidth is a form of 'postponing the evil day' as far as possible outside of the audible frequency band. Think of it as making sure that 20kHz THD figures are as acceptable as the 1kHz ones :)
 
Re: Re: A few naive questions

ilimzn said:
I have to admit I have never heared of E. M. Cherry's ideas, so no comments here.

I did not know anything about Cherry, but a search with Google gave result:
5. The Feedback Lie

Negative feedback, in an amplifier or
preamplifier, is baaaad. No feedback at
all is gooood. So goes this widely invoked
untruth.
The fact is that negative feedback is
one of the most useful tools available to
the circuit designer. It reduces
distortion and increases stability. Only
in the Bronze Age of solid-state amplifier
design, back in the late ’60s and
early ’70s, was feedback applied so
recklessly and indiscriminately by certain
practitioners that the circuit could
get into various kinds of trouble. That
was the origin of the no-feedback
fetish. In the early ’80s a number of
seminal papers by Edward Cherry
(Australia) and Robert Cordell (USA)
made it clear, beyond the shadow of a
doubt, that negative feedback is totally
benign as long as certain basic guidelines
are strictly observed. Enough time
has elapsed since then for that truth to
sink in. Today’s no-feedback dogmatists
are either dishonest or ignorant.

For more info about 10 Audio Lies
see this topic we had some time ago:

Audio Lies - diyAudio discussion

;)
 
Cherry's work involved applying feedback spanning 2 consecutive stages... "nested" feedback, rather than Global feedback... between the Input and output stages of the amplifier.

"No Feedback" has been turned into a markeing tool and is clearing a misnomer. It is Impossible to design any practical amp with Zero ( Local ) feedback. Even if there is No external emitter resistor in a voltage gain stage, there will Always be an intrinsic emitter resistance of the device itself. This is what yield a finite voltage gain in a common emitter amplification stage.

Yes, as rightly pointed out, the Emitter follower ( buffer ) present in most "Zero Feedback" amps has 100% feedback !

"Zero Feedback" I believe is supposed to refer to feedback between the over all input and outputs of the amplifier.

WHY SOME CONSIDER FEEDBACK BAD :

As posted earlier in this thread, there were (are ? ) many shoddy amp designs, which take the quick and easy route : Get as much gain as possible, dont bother about liniarity. Then put ooodles of feedback, abd as long as the amp does not oscillate... you will have vanishingly low distortion.

Such amps measured Very well ( distortion in Parts per million) but sounded not so great.

Engineers began to scratch their heads ....

Mati Otala ( apologies if the spelling is wrong) proposed that amplifier sound was affected not so much by 'static' (harmonic) distortion, but by Triensient Inter Modulation (TIM) Distortion. TIM is caused when the signal transient is too fast for the feedback loop to handle....

Still no 'Perfect' sound....

Today's amp designers seem to have thrown out the 'baby with the bath water' and design amps with no feedback at all.

Some of these are stated to sound closer to the proverbial 'perfect amp' even though objective tests such as distortion ( and in some cases a ruler flat 20 to 20 KHz response have taken a hit.

But still no "Perfect Sound"....

( ON THE SIDE: I believe the distortion measurenments have been inspired by Fourier, who showed mathematically, that any wave shape can be deemed to be composed of sine waves. Hence if an amp could perfectly reproduce sine wave ( ie with zero distortion) the same amp could reproduce Any wave form... including music..... )

To reproduce the waveform, the different frequency sine waves also Must maintain their phase relationships... ie how much later one sine wave starts or ends, with reference to another ( typically the fundamental, or the lowest freq sine wave in the group. )


WHY DC TO LIGHT FREQ RESPONSES ?

When an amplifier output begins to roll off and is is 3 dB down at a particular freq, it also introduces a 180 degreee phase shift at the same freq.... which in turn will mess up the signal, because the sequence of the the sine wave components in the output will not be the same as that in the input signal.

Clearly then, the -3dB comes not only with a lower HF penelty, but also with a HUGE phase distortion.

As a practical example, to "almost" perfectly reproduce a 20KHz square wave, you will typically require an amp with a flat freq response to 10 Times the 20 Khz freq, or 200 Khz.....

Cheers
 
Actually the feedback question is more complex than expressed here. There is also higher order distortion generated from lower order linearities and FM distortion generated by non-linear input stages within a feedback loop with low open loop bandwidth. Real audio designers take negative feedback very seriously.
 
A really good thread........ particularly liked Ilimzn's and John's responses.....

I concur that global negative feedback introduces stability issues with 'unflattering' loads. This happens because of the existence of a pole frequency beyond which phase shift reverses negative to positive feedback. OTOH, GNFB gives low source impedance so important with modern drivers to confer slam and impact.

The non-linearities introduced by GNFB can lead to less than monotonically decreasing distortion spectra; but if the loop is very fast (which implies a swift VAS and a judicious mix of lag comp and phase lead) then the audible effects of this are much reduced.

There is still art in the management of the many compromises of amp design. It's a shame an audiophile amp is so named; if it were called a musical instrument more care would be taken with design.....

Cheers,

Hugh
 
Gandalph said:
3. Why are amplifiers now being specified / designed with a ‘flat’ frequency response curve going up to (say) 1 mHz ?
.... because "more is better", see my title at the left
:D

One reason is marketing but an another reason is that if those amps are current feedback the distortion is extremely low. No harm in that.

My QRV-06 produce 0.00046% and http://mirand.dk has also pretty good performance.

My QRV-06 is good for DC-25 MHz
 
AX tech editor
Joined 2002
Paid Member
AKSA said:
[snip]I concur that global negative feedback introduces stability issues with 'unflattering' loads. This happens because of the existence of a pole frequency beyond which phase shift reverses negative to positive feedback. [snip]Hugh


Hugh,

Technically correct, but I'd like to note that those "issues" can be completely and satisfactory taken care of by using straightforward, well-known and accepted design methods.
I just thought I'd put the "issues" thing in perspective. I'm sure you don't mind ;) .

Jan Didden

PS Seen my sig ?
 
Why is that in most feedback amp design, the stability consideration comes the first (and usually the only one)?

Is there any designer that design an amp with exact poles and zero's at frequencies that he wanted (to tailor the "sound") while the stability consideration is not becoming more important than these open-loop curvature?

Maybe this sounds crazy or imposible, but as I see designs of "high-class" amps, they put poles and zero's everywhere in the CCT. I think this "tailored" how the final amp will sound. And to these amps, the stability is no.2 compared to open loop curvature. The drawback of this approach, while the designer can make a "nice sounding" OL curvature, it will be in small bandwith, because it is done after the stability problem fixed.
 
.:rolleyes: Feedback
Has anybody analytically proved that feedback makes the output any closer to input signal? For any music type of signal?
It is weel known, that feedback reduces non-linear distortion of a purely static memoryless system, rises freq. response of some (Nyquist criterion) perfectly linear systems, and reduces noise added to a wire with gain.
But what if these three interact? Term 'conditional stability' is born but not very much investigated.
We have some examples that feedback may make things worse when:
-any of stages gets overdriven, even by distortion component of the other stage or even noise
-the amp has low slew
-any of stages gets saturated
-one of above causes non-stability for a while
Anyone denies?

regards

P.S. I would sugget talking about dB and kHz rather than feedback/zero feedback situation
 
Cherry and nested feedback..?

For no technical reason I don't get the idea of nested feedback. Probably it can improve parameters, but...


Will you like such a buffer:
 

Attachments

  • nes.jpg
    nes.jpg
    6.5 KB · Views: 417
rozak said:

Can somebody explain to me why use feedback in this amp?

class A
0.007% dist.
150kHz flat friq. res.
1Meg -3db ---------

With feedback, the output impedance will be lower and more constant.
In your schematic , the output impedance is 1/Gm of M3/4 , and this parameter change a lot with the current , that is changing during the output voltage ( and current) swing...
 
darkfenriz said:
.:rolleyes: Feedback
Has anybody analytically proved that feedback makes the output any closer to input signal? For any music type of signal?

Do a null test using a music signal (any CD from your collection), you will see (and ear ) that when you increase the feedback factor , the residual error signal become smaller and the output mimics the input signal.

If in this null test , you achieve to ear only thermal noise , you have in your hands the "wire with gain" and is time to look at the speakers and source components...
 
rozak said:
If amp need some correction - ok with global feedback.
Can somebody explain to me why use feedback in this amp?

class A
0.007% dist.
150kHz flat friq. res.
1Meg -3db ---------

To compensate the outputimpedance of ~0.6ohms ? Still, a nice circuit ! And, some people like amps with such high outputimpedance...
Feedback is not only used to reduce THD.

Mike
 
Tube_Dude said:


With feedback, the output impedance will be lower and more constant.
In your schematic , the output impedance is 1/Gm of M3/4 , and this parameter change a lot with the current , that is changing during the output voltage ( and current) swing...

I know about 1/Gm and PSpice know.
Douglas Self can use PSpice and I can use PSpice.

Hot follower is well known and proven and works wery well
whithout help of global feedback.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.