Sorry, dimitri. BTW what flag is that? You're in Florida US?
c.f. is not newspeak it's in any English dictionary - compared with!
Look at my block diagram -
1. you can use any output stage
2. internal loop includes diff'l input, a Vas and the output stage
just like any std amplifier but UGS stable.
3. First stage can be a typical op amp or discrete diff'l plus Vas.
Tuning is simple.
c.f. is not newspeak it's in any English dictionary - compared with!
Look at my block diagram -
1. you can use any output stage
2. internal loop includes diff'l input, a Vas and the output stage
just like any std amplifier but UGS stable.
3. First stage can be a typical op amp or discrete diff'l plus Vas.
Tuning is simple.
guru: I am aware that taking global or nested feedback from the output stage is 'the usual' way of designing an amplifier. However as it turns out it is not necessary, or even desirable. With proper adjustment the natural distortion of the emitter coupled output stage is not significant. And if it was significant you could not use feedback to remove the distortion anyway. I know ..... your sine waves, simulators, THD analyzers, and so on. But all these methods use a sine wave to analyze the THD. A sinewave is the simplest signal for an amplifier to handle, it resembles a wheel turning with a constant speed. Music is NOT a sine wave. IMO good amplifier design is all about avoiding unlinearities to build up anywhere in the signal chain. Because if they do, you can not remove them later. (Again i'm not talking about removing THD from a single sine wave). That's also the reason there is no correllation between THD measurements and sonic properties or sound quality at all
To give you an idea what i'm talking about, i will ask you this question: What happens if you insert a sine wave and then several sine waves into this circuit:
To give you an idea what i'm talking about, i will ask you this question: What happens if you insert a sine wave and then several sine waves into this circuit:
Attachments
And further if you really can't keep yourself from including the output stage in the feedback loop, 
and therefore making your poles fly around uncontrolled with changing capacitive loads, then there are better ways to solve the problem, than adding the band-aid coil.
One option shown below ...
and therefore making your poles fly around uncontrolled with changing capacitive loads, then there are better ways to solve the problem, than adding the band-aid coil. One option shown below ...
Attachments
Lars Clausen said:
Sorry Lars...a properly designed coil is NOT a band-aid....
RF can be a serious problem....Thiel's network on a competently designed amp. is mandatory....methinks..
Moreover your 'solution' is no solution at all......RF still gets in unchecked, and since the NFB is not taken directly from the output.......a significant increase in distortion is to expected from this arrangement...(see Self).
Mikeks,
An integrator isn't Miller capacitance.
Sure the output stage Ft usually dictates the stability and bandwidth of the internal loop but so what. It's easy to stabilise for UGS, as is a std amplifier for it's closed loop gain. Except with internal C pref not miller as it degrades PSRR.
An integrator isn't Miller capacitance.
Sure the output stage Ft usually dictates the stability and bandwidth of the internal loop but so what. It's easy to stabilise for UGS, as is a std amplifier for it's closed loop gain. Except with internal C pref not miller as it degrades PSRR.
amplifierguru said:
Easy to stabilise..?
Not really..requires shunt capacitance to ground at the TIS output...and the stability margins obtained are not usually adequate...unless you use fairly large capacitance...which then degrades TIS linearity rather badly...bringing you nicely back to square one...
Lars,
You're full of revelations - I feel so naive, 40 odd years of designing amplifiers and you've opened my eyes!
"With proper adjustment the natural distortion of the emitter coupled output stage is not significant. And if it was significant you could not use feedback to remove the distortion anyway."
All those typical output stage distortions (typ 1%) particularly exacerbated by reactive loading and reduced to below the noise floor by feedback didn't really exist after all. Why did I bother.
"I know ..... your sine waves, simulators, THD analyzers, and so on. But all these methods use a sine wave to analyze the THD. A sinewave is the simplest signal for an amplifier to handle, it resembles a wheel turning with a constant speed. Music is NOT a sine wave."
Here's another eye opener for me. I was labouring under the misapprehension that music was a spectrum of sine waves all varying in time. I mistakenly believed that a held musical note might be very similar to a simple group of a few sine waves variant in time but holding still for a moment. Somewhat like my triangle or square wave or even half wave AB side output in response to sinewave stimulation.
I'm gobsmacked by this revelation that the music is made up of something else than sinwaves and that output stage distortions
aren't significant and can't be reduced by feedback, but your master revelation has to be -
"there is no correllation (sic) between THD measurements and sonic properties or sound quality at all"
That's it for me - time to throw in the towel and bow to your great wisdom.
Now what was it your selling - Oh I recall, switching amps. Where's my earplugs.
You're full of revelations - I feel so naive, 40 odd years of designing amplifiers and you've opened my eyes!
"With proper adjustment the natural distortion of the emitter coupled output stage is not significant. And if it was significant you could not use feedback to remove the distortion anyway."
All those typical output stage distortions (typ 1%) particularly exacerbated by reactive loading and reduced to below the noise floor by feedback didn't really exist after all. Why did I bother.
"I know ..... your sine waves, simulators, THD analyzers, and so on. But all these methods use a sine wave to analyze the THD. A sinewave is the simplest signal for an amplifier to handle, it resembles a wheel turning with a constant speed. Music is NOT a sine wave."
Here's another eye opener for me. I was labouring under the misapprehension that music was a spectrum of sine waves all varying in time. I mistakenly believed that a held musical note might be very similar to a simple group of a few sine waves variant in time but holding still for a moment. Somewhat like my triangle or square wave or even half wave AB side output in response to sinewave stimulation.
I'm gobsmacked by this revelation that the music is made up of something else than sinwaves and that output stage distortions
aren't significant and can't be reduced by feedback, but your master revelation has to be -
"there is no correllation (sic) between THD measurements and sonic properties or sound quality at all"
That's it for me - time to throw in the towel and bow to your great wisdom.
Now what was it your selling - Oh I recall, switching amps. Where's my earplugs.
Hi Mikeks,
Miller C .....parasitic and effectively multiplied across to the input of the stage. Loop feedback C doesn't resemble this in any way shape or form.... oh, IMO. Does it? I've become very uncertain of myself since Lars.
"Easy to stabilise..?
Not really..requires shunt capacitance to ground at the TIS output...and the stability margins obtained are not usually adequate...unless you use fairly large capacitance...which then degrades TIS linearity rather badly...bringing you nicely back to square one..."
What's TIS. But I think i see your point.
Scenario: UG output stage driven by complementary Vas, driven by input stage - all in loop with gain. For UGS stability apply C at Vas output to ground. Due to symmetry of Vas push pull and local emitter degen there is little distortion generated as long as there is adequate stage current to slew to max required simply as I =CwVo (w =omega =2*pi*f).
Simple.
Miller C .....parasitic and effectively multiplied across to the input of the stage. Loop feedback C doesn't resemble this in any way shape or form.... oh, IMO. Does it? I've become very uncertain of myself since Lars.
"Easy to stabilise..?
Not really..requires shunt capacitance to ground at the TIS output...and the stability margins obtained are not usually adequate...unless you use fairly large capacitance...which then degrades TIS linearity rather badly...bringing you nicely back to square one..."
What's TIS. But I think i see your point.
Scenario: UG output stage driven by complementary Vas, driven by input stage - all in loop with gain. For UGS stability apply C at Vas output to ground. Due to symmetry of Vas push pull and local emitter degen there is little distortion generated as long as there is adequate stage current to slew to max required simply as I =CwVo (w =omega =2*pi*f).
Simple.
Hi A.G.,
I've found that for some posts, you must take the word "sonic" and replace it with "imaginary" or "imagined" for the true meaning to be known. I won't say which posts though. This should be an entry in some sort of "audiophile dictionary".
But seriously though, I have a couple of questions regarding your block diagram of the multi-loop amplifier in post #60 of this thread.
Am I correct in assuming that the input stage also behaves as an integrator in this design, such that the combination of two integrators, plus the zero introduced by the feedback network, results in a two-pole compensation approach? (I'm aware that there wil also be a pole in the loop gain contributed by the feedback loop as well, at about 3 MHz or so).
Secondly, what is the target unity loop gain frequency of (a) the inner integrator loop containing the output stage, and (b) the overall loop?
I've found that for some posts, you must take the word "sonic" and replace it with "imaginary" or "imagined"
for the true meaning to be known. I won't say which posts though. This should be an entry in some sort of "audiophile dictionary".But seriously though, I have a couple of questions regarding your block diagram of the multi-loop amplifier in post #60 of this thread.
Am I correct in assuming that the input stage also behaves as an integrator in this design, such that the combination of two integrators, plus the zero introduced by the feedback network, results in a two-pole compensation approach? (I'm aware that there wil also be a pole in the loop gain contributed by the feedback loop as well, at about 3 MHz or so).
Secondly, what is the target unity loop gain frequency of (a) the inner integrator loop containing the output stage, and (b) the overall loop?
Hi Andy c,
I think I need that dictionary!
The nested integrator is defined to just permit max output from the first stage output signal , at 100KHz. As the input and feedbach combined 100KHz 12dB/octave linear phase filter define response to be -3dB at 100KHz, this provided 3 dB up sleeve.
Yes it's 2 pole + feedback zero starting at 150KHz.
Clear as mud?
I think I need that dictionary!
The nested integrator is defined to just permit max output from the first stage output signal , at 100KHz. As the input and feedbach combined 100KHz 12dB/octave linear phase filter define response to be -3dB at 100KHz, this provided 3 dB up sleeve.
Yes it's 2 pole + feedback zero starting at 150KHz.
Clear as mud?
Mikeks,
we seem to have a communication problem. Please refer to my scenario above. Think of complementary diff'l input stage, single endedly feeding a complementary full swing voltage amplifier stage (VAS) in turn feeding a unity gain output stage. A viable amplifier. Depending on the closed loop gain this may need stabilising C and I would apply it from VAS output to ground OR alternatively increase emitter degeneration OR a combination of the two to optimise slewing rate and define OL bandwidth rather than depend on the vagaries of individual devices.
I don't feel like drawing pics right now.
we seem to have a communication problem. Please refer to my scenario above. Think of complementary diff'l input stage, single endedly feeding a complementary full swing voltage amplifier stage (VAS) in turn feeding a unity gain output stage. A viable amplifier. Depending on the closed loop gain this may need stabilising C and I would apply it from VAS output to ground OR alternatively increase emitter degeneration OR a combination of the two to optimise slewing rate and define OL bandwidth rather than depend on the vagaries of individual devices.
I don't feel like drawing pics right now.
amplifierguru said:
Okay, so that establishes the time constant of the output stage/integrator. But that integrator is itself a feedback amplifier with an open-loop gain, say, A1 and a feedback factor, say, B1. What I was asking in (a) was essentially "what is the frequency for which |A1*B1| = 1?". The reason I asked this is as follows: Suppose that the inner loop had a phase margin of exactly 90 deg. Then the closed-loop gain of the integrator would deviate from its ideal characteristic by -3 dB in magnitude and -45 deg in phase at the unity loop gain frequency of the inner (integrator) loop. This will in turn affect the stability of the outer loop. So that led to question (b), namely "what is the unity loop gain frequency (|A*B| = 1) of the outermost loop"?
Thanks
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.