Stubborn Oscillation

[DF96]

Please don't feed the trolls! In this discussion we already have one person who thinks feedback can achieve perfection, and another who fears that any phase shift will render feedback useless.

[TheGimp]

DF96, I suspect a lack of math background.

I'm slowly working backwards through various levels I no longer remember (I just ordered a college Algebra book) until I can understand it. Then I'll work forwards again until either I once again understand FB, or I expire.

Either way that page on the site was nonsense.

[Globulator]

Originally Posted by DF96

I have never measured the PSRR of an SRPP, but in the usual balanced configuration with both cathode resistors unbypassed it will be around -6dB. If the low cathode is bypassed then it will be better, maybe -9dB. Are you saying that valves and resistors do not obey circuit theory? If your measurement differs markedly from these figures then there is something wrong with your measurement technique.

This illustrates perfectly how theory can be wrong. The theory and maths you are leaning on here simply doesn't cover the reality of what goes on inside an amplifier.

Think of an SRPP (or any stage with the output at B/2) driver - powered off 400V swinging 200Vpp, centred at 200V. So you say, PSRR is -6dB. Well at 200V you'd be right, but you'd be wrong at 100V where the PSSR would be -12dB, and wrong at 300V it would be -2.5dB.

So in reality your steady -6dB PSRR of theory is a variable AM ripple (and PSU hash) dependent on were the signal happens to be, with (in the example) a difference of 12-2.5 = 9.5dB between the extremes.

I know it's small (usually smaller) and many say it doesn't matter, but is this hi-fi right? Surely we can only claim hi-fi if the signal is passed unmolested wherever the music makes it go?

Originally Posted by DF96

another who fears that any phase shift will render feedback useless.

I don't recall saying that - could you refresh my memory?

Questioning convention is how new discoveries are made - it's just science guys. My concerns about phase shift is that the corrective signal is out of phase with the original: mainly it's only a few degrees - but the ear's phase response is pretty good, and it is still out.... and if it's out, it can't work perfectly as it relies on correcting the signal, not a phase shifted version of it.

Using Thorsten Loesch's ideas about the futility of GNFB over for instance the transformer (that many people disagree with) my SE amp now sounds fantastic, but also brutally powerful and tight, and clear even with complex music - without any GNFB at all, perhaps that's why I always question the line that GNFB is always needed and always the best solution

BTW: Judging by some comments on here we are almost reaching the religious fervour of the high priests of global warming, it's just a discussion to gain ideas guys

[MelB]

Originally Posted by Globulator

I know it's small (usually smaller) and many say it doesn't matter, but is this hi-fi right? Surely we can only claim hi-fi if the signal is passed unmolested wherever the music makes it go?


I don't recall saying that - could you refresh my memory?

Questioning convention is how new discoveries are made - it's just science guys. My concerns about phase shift is that the corrective signal is out of phase with the original: mainly it's only a few degrees - but the ear's phase response is pretty good, and it is still out.... and if it's out, it can't work perfectly as it relies on correcting the signal, not a phase shifted version of it.

Using Thorsten Loesch's ideas about the futility of GNFB over for instance the transformer (that many people disagree with) my SE amp now sounds fantastic, but also brutally powerful and tight, and clear even with complex music - without any GNFB at all, perhaps that's why I always question the line that GNFB is always needed and always the best solution



BTW: Judging by some comments on here we are almost reaching the religious fervour of the high priests of global warming, it's just a discussion to gain ideas guys

Two thumbs up!! Me thinks GNFB improves the measurements perhaps but not the listening enjoyment.

[DF96]

Originally Posted by Globulator

Think of an SRPP (or any stage with the output at B/2) driver - powered off 400V swinging 200Vpp, centred at 200V. So you say, PSRR is -6dB. Well at 200V you'd be right, but you'd be wrong at 100V where the PSSR would be -12dB, and wrong at 300V it would be -2.5dB.

So in reality your steady -6dB PSRR of theory is a variable AM ripple (and PSU hash) dependent on were the signal happens to be, with (in the example) a difference of 12-2.5 = 9.5dB between the extremes.

You are using a different, possibly personal, definition of PSRR. By PSRR I mean (and I believe others mean) the degree to which the output of a stage rejects signals on the supply rails. -6dB means the output noise/ripple is half the voltage of the noise/ripple on the supply rail. How this relates to likely signal levels is a separate issue. To calculate PSRR you just need to know how to calculate the circuit. Your figure, more related to signal-noise ratio, also requires knowledge of the PSU noise and the signal level. Your figure is useful, but it isn't PSRR.

This illustrates that there is more to measurement than measuring: you also have to correctly handle your raw data.

Questioning convention is how new discoveries are made - it's just science guys. My concerns about phase shift is that the corrective signal is out of phase with the original: mainly it's only a few degrees - but the ear's phase response is pretty good, and it is still out.... and if it's out, it can't work perfectly as it relies on correcting the signal, not a phase shifted version of it.

As I have said, NFB cannot achieve perfection. NFB is always roughly in phase with the input signal, even when the amp imposes nearly 90 degrees of phase shift. Otherwise it ceases to be NFB - this is what happens at frequency extremes and is how instability occurs. Either do the maths, or draw yourself a diagram of signal vectors. Failing to grasp basic theory is not how science proceeds.

[Globulator]

Originally Posted by DF96

You are using a different, possibly personal, definition of PSRR. By PSRR I mean (and I believe others mean) the degree to which the output of a stage rejects signals on the supply rails. -6dB means the output noise/ripple is half the voltage of the noise/ripple on the supply rail. How this relates to likely signal levels is a separate issue.

It's only a separate issue if you listen to the amp while the input is grounded.
As soon as you start listening to music is becomes a PSRR issue, and as I tend to use my amps for music it is something I must consider! BTW it is not a level issue, it's a voltage divider issue.

The problem with PSRR theories and formulas is that they are incomplete simplifications of the real physics (or even the simulation) inside the amplifier. The same is true of GNFB.

Simplifications, approximations and rules of thumb can only take you so far IME.

[DF96]

I am not talking about listening to the amp, but calculating what you might hear when you do listen. PSRR is Power Supply Rejection Ratio - the ratio by which the power supply noise is rejected. You can refer the result to the input or the output, but output-referred PSRR avoids having to calculate gain too.

It is you that made it a level issue, but introducing signal level. I was talking about a ratio, calculated from a voltage divider.

All calculations are necessarily approximations, but an approximate value can be better than total ignorance. Anyway, perhaps you can tell me what is wrong with the usual calculation for SRPP PSRR? Is there some flaw in the derivation? What aspect of "real physics" has been overlooked?

[Globulator]

Originally Posted by DF96

I was talking about a ratio, calculated from a voltage divider.

Ah - at last we are talking about the same thing!!

The simplistic PSRR calculation uses a constant ratio, which assumes that the voltage divider is fixed.

It isn't.

[DF96]

It is, for the usual definition of PSRR. If you want to see how PSRR varies with signal level then that can be done, but that is another issue. Don't change the subject.

You may, of course, have your own private definition of the meaning of PSRR but don't then complain that other people's calculations, based on the normal definition, are wrong.