Thanh, sounds like you tried this in a real amp? I believe it was meant to be a thought exercise. At least one must consider another way to provide stability if the Cdom cap is removed.
Experimenting with compensation in real power amp circuits can often lead to smoke unless current limiting is used in the power supply.
Another thought question: where is the real Miller cap, and what is unusual/interesting about it? Why is it called Miller capacitance?
Experimenting with compensation in real power amp circuits can often lead to smoke unless current limiting is used in the power supply.
Another thought question: where is the real Miller cap, and what is unusual/interesting about it? Why is it called Miller capacitance?
thanh said:
Thanh,
if You have one proper designed amplifier, with Miller type compensation, and You want to modify it, than You have to redesign the compensation. It's not as easy, to move the Miller capacitor to the other location.
The Hood type compensation use larger value capacitor, due the missing Miller effect...
sajti
PB2 said:
Miller capacitance is between the collector, and the base of the transistor (same as the input, and output of inverting amplifier). The most important, if You want to transform it into the input of the stage, You have multiply the vale with (1+stage gain). So with 100pF and A=100 resuslts 10nF equal input capacitance
sajti
Thank you Sajti,
Yes, the interesting point is that the effective input capacitance is multiplied by the stage gain.
For those who're new to this: The Miller effect was first documented by John Miller in the 1920s while investigating the characteristics of tube amplifiers. I believe that the main point was that the devices own internal parasitic capacitance provided a feedback path, and that while this parasitic capacitance was small it had much more influence than expected because it is multiplied by the voltage gain of the amplifier.
We see the same thing in transistors with the capacitance of the reverse biased BC junction.
Figure 7 on this page provides an overview:
http://carcino.gen.nz/tech/elec/millereffect.php
Here's an interesting link, BTW nice control theory page, about half way down he has a link to a pdf with the original writeup by John Miller:
http://64.233.161.104/search?q=cache:TrngNR6DzMAJ:web.mit.edu/klund/www/+%22miller+effect%22+history&hl=en
This answers part of the question that I posed, the open loop HF response rolls off because of junction capacitances with the miller capacitance in the stage with high voltage gain dominating because of the voltage gain multiplication factor. However, this does not answer why the distortion is frequency dependent.
Yes, the interesting point is that the effective input capacitance is multiplied by the stage gain.
For those who're new to this: The Miller effect was first documented by John Miller in the 1920s while investigating the characteristics of tube amplifiers. I believe that the main point was that the devices own internal parasitic capacitance provided a feedback path, and that while this parasitic capacitance was small it had much more influence than expected because it is multiplied by the voltage gain of the amplifier.
We see the same thing in transistors with the capacitance of the reverse biased BC junction.
Figure 7 on this page provides an overview:
http://carcino.gen.nz/tech/elec/millereffect.php
Here's an interesting link, BTW nice control theory page, about half way down he has a link to a pdf with the original writeup by John Miller:
http://64.233.161.104/search?q=cache:TrngNR6DzMAJ:web.mit.edu/klund/www/+%22miller+effect%22+history&hl=en
This answers part of the question that I posed, the open loop HF response rolls off because of junction capacitances with the miller capacitance in the stage with high voltage gain dominating because of the voltage gain multiplication factor. However, this does not answer why the distortion is frequency dependent.
PB2 said:
Hi,
Would that not be because as frequency rises, the cap impedance goes down. That has two effects I think: first, it shunts the output Z of the a gain stage therefore giving less gain thus less for the feedback (assuming global fb) to work its magic. Second, the stage itself could have more distortion working in an increasingly non-linear load (the cap is also level dependent).
Dimitri?
Jan Didden
Jan, if you discussed intermediate stage (VAS) I am doubtful that it distortion characteristics will be frequency dependent, while the gain is not frequency dependent. Thereby I could not find frequency dependent distortion mechanism in intermediate stage, although it sure exists it the output stage.
I sorted out my old piles and found 'Programmable audio attenuator' in WW, May 80. Are your hobbies, apart from audio, still include reading science fiction?
I sorted out my old piles and found 'Programmable audio attenuator' in WW, May 80. Are your hobbies, apart from audio, still include reading science fiction?
dimitri said:Jan, if you discussed intermediate stage (VAS) I am doubtful that it distortion characteristics will be frequency dependent, while the gain is not frequency dependent. Thereby I could not find frequency dependent distortion mechanism in intermediate stage, although it sure exists it the output stage.
I sorted out my old piles and found 'Programmable audio attenuator' in WW, May 80. Are your hobbies, apart from audio, still include reading science fiction?
Well, I thought that the miller cap being nonlinear with level, AND needing more charge/discharge current with freq, would lead to higher HD with frequency. But I have no proof for it.
Om WW May80: I sold 500 sci-fi pockets last year to a 2nd hand bookstore (De Slegte for my Dutch friends). Moved on to other interests. BTW, do you know I have troubles following my own math in that article? I wonder how many really read it
Jan Didden
dimitri said:jan, you mean _external_ Miller cap or p-n junction? As the cap is low value, we can use extra high quality cap here (vacuum)
Actually, I was thinking about the intrinsic p-n junction cap, but now realise you guys may have been talking about the external brute force
comp cap. That one, of course, has no (almost none) freq dependent HD.Jan Didden
Jan,
You've hit all the points, lets ignore the external Cdom cap for the moment. Even if the CB junction capacitance was ideal it would have an impact on frequency response just as you state, and of course this leads to less feedback and higher distortion in the closed loop case. But let's talk open loop for now.
You also seem to be suggesting that the CB junction capacitance is nonlinear, yes this is very true and it is the answer I was looking for. A reverse biased junction makes a fairly good varactor diode, so the large signal swing at the collector of the VAS results in modulation of the Miller capacitance by the signal, it is a time varying capacitance. At low freq stage gain is set by resistive loads since the reactance of the nonlinear capacitance is high in comparision, this reactance becomes more significant as frequency rises and distortion increases as a result.
A linear external cap can reduce this effect since the capacitance variation is lower on a percentage basis, however we loose HF gain also. The distortion primarily results from the VAS nonlinear input capacitance presenting a nonlinear load to the diff amp modulating the gain of the diff amp. Driving the VAS with more of a voltage source helps, this is why an emitter follower is sometimes used between the diff amp and VAS. A cascode configuation at the VAS also nearly eliminates this problem because it holds the VAS collector voltage constant, and the cascode load is a common base configuration with its base at ground potential from an AC perspective.
There are output stage HF nonlinearity considerations but for amplifiers with high enough current gain in the output stage the VAS junction Miller cap is the main contributor. This assumes high gain in the VAS, really any stage with high voltage gain will contribute. Running stages with high bias currents tends to improve the situation since there's more current available to charge the capacitances.
Thinking back to the Tiger amplifiers we see that with the output devices configured in common emitter mode that the Miller capacitance will present a nonlinear load to the drivers, however the VAS does not have as large output voltage swings since there's gain in the ouput stage. It's a tradeoff.
You've hit all the points, lets ignore the external Cdom cap for the moment. Even if the CB junction capacitance was ideal it would have an impact on frequency response just as you state, and of course this leads to less feedback and higher distortion in the closed loop case. But let's talk open loop for now.
You also seem to be suggesting that the CB junction capacitance is nonlinear, yes this is very true and it is the answer I was looking for. A reverse biased junction makes a fairly good varactor diode, so the large signal swing at the collector of the VAS results in modulation of the Miller capacitance by the signal, it is a time varying capacitance. At low freq stage gain is set by resistive loads since the reactance of the nonlinear capacitance is high in comparision, this reactance becomes more significant as frequency rises and distortion increases as a result.
A linear external cap can reduce this effect since the capacitance variation is lower on a percentage basis, however we loose HF gain also. The distortion primarily results from the VAS nonlinear input capacitance presenting a nonlinear load to the diff amp modulating the gain of the diff amp. Driving the VAS with more of a voltage source helps, this is why an emitter follower is sometimes used between the diff amp and VAS. A cascode configuation at the VAS also nearly eliminates this problem because it holds the VAS collector voltage constant, and the cascode load is a common base configuration with its base at ground potential from an AC perspective.
There are output stage HF nonlinearity considerations but for amplifiers with high enough current gain in the output stage the VAS junction Miller cap is the main contributor. This assumes high gain in the VAS, really any stage with high voltage gain will contribute. Running stages with high bias currents tends to improve the situation since there's more current available to charge the capacitances.
Thinking back to the Tiger amplifiers we see that with the output devices configured in common emitter mode that the Miller capacitance will present a nonlinear load to the drivers, however the VAS does not have as large output voltage swings since there's gain in the ouput stage. It's a tradeoff.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.