NFB is NFB, trying to hide the local feedback and adding icing doesn't change the fact that you might still have 100dB of overall NFB. I've tried to understand the local good, loop bad arguments for years but I can only see marketing to those who think feedback is bad. Have I missed something?
Ultra large amounts of over-all nfb is not necessary to get vanishingly low thd. One approach which - historically, at least - has been associated with IC VFB OpAmps - and I think JC would say sound bad from critical listening view.
i just told you how to use small amounts of neg fb to reduce harmonics to invisible levels without using large amounts of over-all neg fb.
[Any down side could be s/n. And, if ultimate s/n is also required then you have to eliminate the local degen and do heroic transistor (esp fets) matching to get distortion down. ... and/or parallel many devices.]
THx-RNMarsh
i just told you how to use small amounts of neg fb to reduce harmonics to invisible levels without using large amounts of over-all neg fb.
[Any down side could be s/n. And, if ultimate s/n is also required then you have to eliminate the local degen and do heroic transistor (esp fets) matching to get distortion down. ... and/or parallel many devices.]
THx-RNMarsh
Last edited:
Not only NFB is NFB, but it was mathematically shown (by no one else, but mr. Bode) that splitting a constant amount of loop gain across multiple feedback loops leads to a net loss in NFB benefits.
In practice, local feedback can’t be avoided for various reasons (a classic reason is improving the slew rate performance of a long tail pair feeding a Miller compensated gain stage, even the Miller compensation is local NFB) but the strategy described by mr. Marsh and enthusiastically supported by his peers and cronies is definitely suboptimal. Long story short, yet another load of marketing BS.
You seem the one hung up on it. There is this strange idea that having local feedback and then a soupcon of overall feedback like a chef adding salt at the end of making a dish. My point is that the total feedback is the same so why the obsession with small amounts of overall feedback other than flooby marketing fluff?
My power amps have over 100dB of loop feedback at DC. Why is that bad?
My power amps have over 100dB of loop feedback at DC. Why is that bad?
I have no idea why that is bad all by itself..... without any other details. Only you suggest some one thinks it is bad.
Usually to get very high fb levels at highest freqs requires a lot more complication in the circuitry. My suggested way can utilise very simple circuitry to get same results or better.... esp at high freq.
THx-RNMarsh
Usually to get very high fb levels at highest freqs requires a lot more complication in the circuitry. My suggested way can utilise very simple circuitry to get same results or better.... esp at high freq.
THx-RNMarsh
That would be reduce 3H and cancel 2H. The argument against GNFB historically is the so called reentrant distortion. A useful input would to compare the two methods.
Yes. And, I think that has been done. mathematically anyway. Hasn't it? I wouldn't be surprised.
The only complaint I would have is that it is hard to have very high over-all neg fb at all freq.... usually only at low freqs. The phase shift with a falling nfb vs freq leads to incomplete cancellation... the familiar rise in thd with increasing freq. various techniques have been used to shift that effect for better cancellation/reduction in thd.
Something about distortion rising with freq is more disturbing than a constant level of distortion with freq. increase.
-RM
The only complaint I would have is that it is hard to have very high over-all neg fb at all freq.... usually only at low freqs. The phase shift with a falling nfb vs freq leads to incomplete cancellation... the familiar rise in thd with increasing freq. various techniques have been used to shift that effect for better cancellation/reduction in thd.
Something about distortion rising with freq is more disturbing than a constant level of distortion with freq. increase.
-RM
Last edited:
The distortion of a GNFB amplifier has nothing to do with cancellation. Maybe that's a property of servos an input coupling cap is outside the loop and the feedback can work down to DC.
I seen the second argument before, if say an amp that has 1ppm THD at 20K and only drops monotonically with lower frequency "sounds" better than one that has constant -60dB someone is going to have to prove it.
You can't hear 20kHz. 2HD at 20kHz is 40kHz You definately can't here that. 19/20 IMD wraps down to where there might be 100dB of GNFB which will supress the IMD tone. 50ppm at 20kHz does not seem to be to be an issue esp if that means you get sub 10ppm from 20Hz to 10KHz.
If its' just personal preference then fine.
And my favorite Ubu:
YouTube
The Modern Dance is in my top 5 albums of all time...love these guys.
Howie
The amp made in Thailand has more than 60 dB of GNFB from zero up to 20 kHz. Main reason for increased higher harmonics is decreasing GNFB with increasing frequencies. This amp uses unique kind of two pole compensation to extend GNFB up to high frequencies.
Local feedback has reentrant distortion too. Or maybe by redefining local feedback as 'not feedback' (e.g. degeneration) the 'no feedback' crowd can also redefine reentrant distortion as 'not distortion'?scott wurcer said:
In some cases you can get the best outcome by using some local positive feedback to increase loop gain for GNFB. It all depends on where the distortion arises, the signal handling capability of each stage, frequency response etc.
I call 2 opamps 'low parts count'. Complexity is low. Skill of designer to get it to work is somewhat higher.
Edit: Damir: Very impressive. Your amplifiers continue to intrigue me. One day when I need the power...
Last edited:
- Status
- Not open for further replies.