Power rail stability could simply limit output power, or it might appear as one of the 'signal envelope' effects. Some people regard higher power for brief transients as an advantage, provided that the stated spec is clear. '50W continuous, 80W on brief peaks' might be better than '50W continuous, 50W on brief peaks'.
Think about the time-domain signal representing (for example) a note blown on a flute. There is the fundamental, of course, but there's also an envelope that begins and ends the note which spans the note's duration. If a different note is blown the same way, the fundamental will change, but the envelope won't. The analogy is to amplitude modulation of a carrier.
DF96
For me, this thread has been more a lesson in human nature and the persistence of its reptilian origin than anything else.
Why can’t we achieve it perfectly? Are you conceding or buttressing the arguments I put forward on page 43 and 45? Please review my opinion about striving for mediocrity on page 55.
Unhelpful to whom: Those trying to make a great sounding amp or those supporting the nay saying position? How did you maintain a straight face while writing this? You have not provided any measurements, only things to measure. What amount of emphasis on each parameter is perfect to make a truly transparent amp? Are you suggesting that transparency might mean different things to different people? If they do mean different things to different people, then where is the audible standard and how should it be defined? Why can’t you provide percentage or absolute numbers for them and why should I believe your acceptable parameters give me a transparent amp? Your descriptions about what issues to look for are ambiguous in nature and don’t quantify the numbers to make it valid by your own definition. If you don’t see the problem and how my framework can address it after this post, you will join BigE, nezbleu and auplater on my ignore list.
For me, this thread has been more a lesson in human nature and the persistence of its reptilian origin than anything else.
Last edited:
It doesn't have to be perfect to be irrelevant when compared to the imperfections of other things in the recording chain, like microphones and speakers.
We can certainly achieve amplifiers that, when the input and output are subtracted, can stay within 0.01% of perfection with any musical signal. Probably much better if you just read Self and Cordell and apply what they teach. I contend that any and all amplifiers that do this are indistinguishable in a properly conducted listening test. IMO, it's actually far easier to design and build said amplifiers, then it is to set up said listening test.
Indeed it is. When designing for a fashion market, there's much more to be considered than the actual sound, thus the need for peeking during evaluation.
I think that word encompasses a huge amount, much of which has nothing to do with technology, measurements or even sound quality!
Nelson Pass calls it "entertainment," and he's absolutely right.
My ex-wife used to say, "It's lots of fun believing things that aren't true."
A brief play with a simulator demonstrates that there is much more interaction in play - the distortion figures can easily degenerate by a factor of 100 or more, by introducing a power supply much closer to the real thing. This is because NFB used at conventional levels starts to lose its ability to make amends, because it has to deal with both trying to correct non-linearities in the circuit, and the power supply modulating - unless the design is truly well sorted out the effects will be audible.
The advantage of an amplifier much greater in power than it nominally needs to be is that the voltage rails are far more stable while operating at normal listening levels - then the NFB largely only has to "worry" about circuit non-linearities ...
yldouright, you need to be careful when you bandy around the word "transparent" - subjectively it makes sense, and objectively it can be measured; the real argument, as you're saying, is determining what you measure, and under what conditions. Excessive measured distortion in some areas means nothing to the ears, but almost unmeasurable artifacts - using standard, current techniques - are the kiss of death, for audio systems being acceptable for long term listening.
The real conversation needs to be about isolating those subtle distortion artifacts which have so much subjective influence, and learning how to measure them properly ...
Last edited:
negative feedback is very "stupid" doesn't know where errors measured by the feedback network and differencing stages "comes from", doesn't "know" how much or little "work" it is doing
successful use of negative feedback does require that power, outputs, every amplifier stage be sized to signal and load demands to operate in the "small signal linear" range - no deadband, slew limits, clipping...
...just competent engineering
I do put test sources in power supply sims - PSRR can be high over audio, certainly over rectified mains ripple frequency
http://www.diyaudio.com/forums/solid-state/240712-cfa-topology-audio-amplifiers-135.html#post3650824
high loop gain feedback, circuit/compensation tricks give better than 80 dB PSRR over audio, add LCR passive filtering if you are having problems because you didn't block line noise up front, or didn't properly snub poorly behaving diodes
Yes, FB is "stupid" - the only thing it can do is try to correct the output, which is deviating from the ideal for "unknown", to it, reasons.
Competent engineering will resolve the issues, but 20 years ago virtually every amplifier I heard was incompetent - because of power supply issues - so what does say about the "quality" of the engineering at the time?
Rather than separating the issues in a sim, I prefer to test as a lump sum - does the complete contraption work well, or not, when operating close to what happens in the real world?
Regarding PSRR numbers, I did an interesting exercise recently when I investigated the PSRR behaviour of an opamp circuit - NOT a somewhat unrealistic one using ideal components, pretending to be an opamp, but one made of flesh and blood, errh, silicon transistors, all the way. And the PSRR was very non-linear - that is, the feedback ratio had a huge influence on what the numbers were at various frequencies, the shape of the curve changed dramatically ... a single PSRR number, or curve guaranteed nothing, in predicting what the behaviour would be when used in a particular way ...
Competent engineering will resolve the issues, but 20 years ago virtually every amplifier I heard was incompetent - because of power supply issues - so what does say about the "quality" of the engineering at the time?
Rather than separating the issues in a sim, I prefer to test as a lump sum - does the complete contraption work well, or not, when operating close to what happens in the real world?
Regarding PSRR numbers, I did an interesting exercise recently when I investigated the PSRR behaviour of an opamp circuit - NOT a somewhat unrealistic one using ideal components, pretending to be an opamp, but one made of flesh and blood, errh, silicon transistors, all the way. And the PSRR was very non-linear - that is, the feedback ratio had a huge influence on what the numbers were at various frequencies, the shape of the curve changed dramatically ... a single PSRR number, or curve guaranteed nothing, in predicting what the behaviour would be when used in a particular way ...
Last edited:
If this had simply read "And this one time at band camp," it would have been a lot more humorous.
- Status
- Not open for further replies.