In 3 stage amplifiers the secondary pole of the output limits the possible NFB . It needs to be eliminated from the feedback loop , how?. In 77 Cherry patented a method that Pioneer used unsuccessfully in two models early 80's with the following description in the service manual.
In this method , the high frequencies from the output are suppressed from the feedback , the low pass adds an extra pole to provoke oscillation , a forth stage is added to add an extra pole to stabilize . It replaces a low secondary parasite pole of the output with a higher one in the forth stage, but as it needs the second pole on the feedback it inverses on the overall loop and the frequency response becomes high pass. Pioneer who produces hifi amps re feedbacks withe less NFB to flatten the response. The author published his version where he equalizes the response by claiming it is delay time compensation, that on other amplifiers can also be applied.
Another method is an error of application after the idea of John Lensley Hood .
This is related by Douglas Self in his book . The initial idea of JLH is to resolve the problem of slew saturation. Self did try it out in vain despite direct contact with the author. The indirect result , the feedback from the output if it gets its second unwilling pole hence decreaisng its level lower than the VAS , the VAS, with it's collector -base Miller capacitor making it low impedance , it can take the relay via the capacitor and continue to fill the feedback with a cleaner single pole. Of course it would be miracle that the transition frequencies are adjusted by chance. Our friend LKA he did this type of feedback , again for different reason , but unlike Self he made it work . By analyzing his circuit that I came up to the following conclusion.
This is a standard model amp just look at the bode plot.
This what happens if a low pass filter suppresses the high frequencies from feedbacking.
This is the result by adding the VAS high frequencies to the feedback .
I call it 2 way as speakers and not two pole , because there is one pole on the output and one zero on the VAS exactly at the same frequency canceling each other. Isn't it nice hē , isn't it?
Now this is suitable for VAF where 10k ohm feedback resistors can be used, It will be lower impedance version that I will think of tomorrow for CFA version.
Hayk
In this method , the high frequencies from the output are suppressed from the feedback , the low pass adds an extra pole to provoke oscillation , a forth stage is added to add an extra pole to stabilize . It replaces a low secondary parasite pole of the output with a higher one in the forth stage, but as it needs the second pole on the feedback it inverses on the overall loop and the frequency response becomes high pass. Pioneer who produces hifi amps re feedbacks withe less NFB to flatten the response. The author published his version where he equalizes the response by claiming it is delay time compensation, that on other amplifiers can also be applied.
Another method is an error of application after the idea of John Lensley Hood .
This is related by Douglas Self in his book . The initial idea of JLH is to resolve the problem of slew saturation. Self did try it out in vain despite direct contact with the author. The indirect result , the feedback from the output if it gets its second unwilling pole hence decreaisng its level lower than the VAS , the VAS, with it's collector -base Miller capacitor making it low impedance , it can take the relay via the capacitor and continue to fill the feedback with a cleaner single pole. Of course it would be miracle that the transition frequencies are adjusted by chance. Our friend LKA he did this type of feedback , again for different reason , but unlike Self he made it work . By analyzing his circuit that I came up to the following conclusion.
This is a standard model amp just look at the bode plot.
This what happens if a low pass filter suppresses the high frequencies from feedbacking.
This is the result by adding the VAS high frequencies to the feedback .
I call it 2 way as speakers and not two pole , because there is one pole on the output and one zero on the VAS exactly at the same frequency canceling each other. Isn't it nice hē , isn't it?
Now this is suitable for VAF where 10k ohm feedback resistors can be used, It will be lower impedance version that I will think of tomorrow for CFA version.
Hayk
Attachments
Last edited:
Phase lead compensation adds a zero and a higher f pole, The zero counteracts the output stage pole. The result shifts the 2nd pole to a higher f thus allowing more NFB around the output stage.
I didn't apply any Lead compensation. The amount of compensation depends upon the closed loop gain, lower it is less effective is the lead compensation. By this it gave just a little bit of extra NFB. Here instead of increasing a little bit the second pole from the output stage , is to cancel it totaly.
The effectiveness of the feedback depends on the ratio of open loop gain to closed loop gain, mostly around the output stage. You have to look at how the loop gain around the output stage changes with different compensation schemes. When the loop is bypassed at the VAS stage to improve phase margin, some of the desirable loop gain is reduced. It’s more subtle than it looks.
The output starts going open loop at the transition frequency, Cherry, by using the Zobel 8 ohm 100nf fixes it at 200khz. As the output impedance of the VAS now lowered by the feedback, the output stage works in open loop but hard driven. I would fix the transition frequency to that of the output LR.
This is the low imoedance CFA version.
Frankly, I don't like it.
If it is to add an active element , why not add it directly in the input stage, to feed itself. Same as the servo takes care of DC , it could also take care of HF.
Ah , hard work is waiting for me.
Hayk
Frankly, I don't like it.
If it is to add an active element , why not add it directly in the input stage, to feed itself. Same as the servo takes care of DC , it could also take care of HF.
Ah , hard work is waiting for me.
Hayk
Attachments
Last edited:
Re Fig 8.2 post 1
This form of compensation network can put some bends or peaking in a Tian plot. RC networks connecting the mirror transistor collectors to the negative rail could be used to help straighten this out.
This form of compensation network can put some bends or peaking in a Tian plot. RC networks connecting the mirror transistor collectors to the negative rail could be used to help straighten this out.
Instead of using the VAS of the amplifier, I created another independent VAS to feedback the high frequencies. By this, the IPS as it feedbacks itself ,changed its frequency response. Hence it is no more 2 way feedback. I continue this subject of limiting feedback frequency at
Comparative VAS+cascode stage
Comparative VAS+cascode stage
Rather off topic, but...
Oddly, while on the subject, we say "back driving" but "feeding back" - phrasal/cmpound verbs can be pretty random.
Its "feeding back", the verb is feed, "to feed back" is a phrasal verb, the word feedback is only a noun.
Oddly, while on the subject, we say "back driving" but "feeding back" - phrasal/cmpound verbs can be pretty random.
In this model, the crossover takes place at 1Mhz. The second OS pole is fixed at 8Mhz. The VAS stops integrating at 1Mhz to go flat . By this the closed loop response becomes first order with maximum PM, see 90° phase shift instead of 180° habitually. The frequencies bellow 1Mhz get feed backed via L1 issue from the output but above, the driver stage pours through C3 uncompensated feedback. Notice the total absence of the second pole 8 Mhz despite 100db NFB.
