dimitri said:Dave, please, what would you like to say/ask?
Mikek, Miller compensation is _dominant_ pole compensation, "not to single-pole miller compensate" is stupid idiomatic expression, because two-pole compensation is not the Miller one
No.....i was particularly carefull not to fall into that trap......
Miller feedback, is at the heart of both schemes......whether it initialy results in a single or double pole characteristic, is of sublime irelevance.
pole vaulting revisted
"Miller feedback, is at the heart of both schemes......whether it initialy results in a single or double pole characteristic, is of sublime irelevance."
I always thought difference in two pole and one pole response made a large difference in phase margin since two poles are changing the phase twice as fast as single pole. Second order systems ring and settling time is affected. Most of the text I've read advocate moving the other poles as far away from the dominant pole as possible to get as close to a single pole response as possible for the best transient response and greatest phase margin. I believe the dominant pole is usually the one from second voltage gain stage were the largest amount of gain is. Discussing this in the in the absence of the contributions of the of the other poles is nonsense. The assumption of current sources charging the compensation cap is also an over simplification since many amplifiers include resistors in addition to current sources define the open loop gain.
The only thing ridiculously sublime is that all this is being discussed for such an ideal and over simplified model. This seems
to be another example of the difference between the pontificator of some particular pet theory on amplifier design and someone who as actually designed or even simulated an actual amplifier design with real transistor models instead of simplified model of a current source, transconductance stage and a capacitor. The real world is a lot more complicated than some academics would like to imagine. Links to contentious threads instead of actually addressing specific questions at hand is something I would expect for a troll rather than someone interested in actually investigating a topic to inform or learn.
http://grove.ee.iastate.edu/docs/papers/20000807.pdf
"Miller feedback, is at the heart of both schemes......whether it initialy results in a single or double pole characteristic, is of sublime irelevance."
I always thought difference in two pole and one pole response made a large difference in phase margin since two poles are changing the phase twice as fast as single pole. Second order systems ring and settling time is affected. Most of the text I've read advocate moving the other poles as far away from the dominant pole as possible to get as close to a single pole response as possible for the best transient response and greatest phase margin. I believe the dominant pole is usually the one from second voltage gain stage were the largest amount of gain is. Discussing this in the in the absence of the contributions of the of the other poles is nonsense. The assumption of current sources charging the compensation cap is also an over simplification since many amplifiers include resistors in addition to current sources define the open loop gain.
The only thing ridiculously sublime is that all this is being discussed for such an ideal and over simplified model. This seems
to be another example of the difference between the pontificator of some particular pet theory on amplifier design and someone who as actually designed or even simulated an actual amplifier design with real transistor models instead of simplified model of a current source, transconductance stage and a capacitor. The real world is a lot more complicated than some academics would like to imagine. Links to contentious threads instead of actually addressing specific questions at hand is something I would expect for a troll rather than someone interested in actually investigating a topic to inform or learn.
http://grove.ee.iastate.edu/docs/papers/20000807.pdf
He who knows not, and knows not that he knows not is a complete @#*&$, leave him....
The point of two pole compensation is to keep foward path error as small as possible, for as much of the audio band as posible.
A zero then introduced to cancel one of the poles, so that the two pole roll off reverts to a single pole slope long before unity gain frequency.
Thus as far as stability, phase margin, ringing etc...etc... are concerned, we are effectively dealing with a single pole response...since roll-off changes from two pole to single pole before Ft.
I am familiar with the reference you've cited....indeed i have carefully read, analysed, and verified, (or not as applicable), all of Schlarmann/Geiger published work.
It obvious from the stuff you've written above, that you have difficulty understanding even the reference you've cited........i don't think you've even made any attempt to to examine the later with any attention to detail, beyond a casual perusal...
Rubbish.
Fred Dieckmann said:
The point of two pole compensation is to keep foward path error as small as possible, for as much of the audio band as posible.
A zero then introduced to cancel one of the poles, so that the two pole roll off reverts to a single pole slope long before unity gain frequency.
Thus as far as stability, phase margin, ringing etc...etc... are concerned, we are effectively dealing with a single pole response...since roll-off changes from two pole to single pole before Ft.
Fred Dieckmann said:
I am familiar with the reference you've cited....indeed i have carefully read, analysed, and verified, (or not as applicable), all of Schlarmann/Geiger published work.
It obvious from the stuff you've written above, that you have difficulty understanding even the reference you've cited........i don't think you've even made any attempt to to examine the later with any attention to detail, beyond a casual perusal...
Fred Dieckmann said:
Rubbish.
Dave said:
It's still oscillating. Well not at the moment, because I put it on the shelf and will get back to looking at the problem in due course. In the meantime I built a couple of gainclones because I needed something to listen to. Tweaking those should keep me going for some time.
James
Don't you mean that the point of 2-pole compensation is to maintain a high, uniform feedback factor across the audio band. Unless you are suggesting the pole/zero implementation is beneficial for the forward path on its own. I think you meant to say that increasing the feedback at the higher audio frequencies reduces the closed-loop output error. This is quite different from what you actually said.
I think Dave has a good question. Why don't more amps use pole-pole-zero compensators in the forward path? See if you can figure this out and explain it without dropping a reference to yet another from your plethora of articles and journals.
Are you a librarian?
Dave said:
Hi dave,
i often wonder about this extremely valuable point you've made....
Indeed, most amps. using 2-pole-comp. employ incorrect values......infact only one commercial design with correctly choosen values springs to mind.......one of Randy slone's designs...can't remember which.......
(C1=150p;C2=300p;R=1K).
choose the wrong component values, and your design will probably oscillate...
traderbam said:
Different way of saying precisely the same thing.....
2-pole comp. maximises feedback across as much of the audio band as possible.....which merely means it allows the maintenance of a high open loop gain across as much of the audio band as possible......which put differently, is the same as saying it helps reduce foward path error, (the difference between input and feedback signal), for as much of the audio band as possible.
traderbam said:
No...
I know what you are saying and I know what I am saying and they are not quite the same.
Let me ask you a teaser. Are you suggesting that using this double pole-zero method will improve the sound quality? If so, why? What assumption are you making that leads to this conclusion? It needs to be a bit more than "more feedback is better"...there are numerous people who have found the opposite to be the case in practice.
Let me ask you a teaser. Are you suggesting that using this double pole-zero method will improve the sound quality? If so, why? What assumption are you making that leads to this conclusion? It needs to be a bit more than "more feedback is better"...there are numerous people who have found the opposite to be the case in practice.
Five-pole compensation
For those who haven't noticed: there is a relatively new thread about an op-amp with a ridiculously high open-loop gain. Looking at its gain versus frequency plot, it appears to have five-pole compensation...
By the way, why do you all want to keep loop gain constant over the audio band? I understand you want as much loop gain as possible at 20kHz, but what is wrong with having even more at lower frequencies? Why should the low-frequency distortion be as high as the high-frequency distortion?
For those who haven't noticed: there is a relatively new thread about an op-amp with a ridiculously high open-loop gain. Looking at its gain versus frequency plot, it appears to have five-pole compensation...
By the way, why do you all want to keep loop gain constant over the audio band? I understand you want as much loop gain as possible at 20kHz, but what is wrong with having even more at lower frequencies? Why should the low-frequency distortion be as high as the high-frequency distortion?
I think the theory is that is provides more feedback over the audioband without raising gainbandwidth product above the final zero and therefore won't compromise stability.
This is different from just increasing total gain or open loop bandwidth which will increase gain bandwidth product and possibly affect stability.
toll for the pole troll
"Miller feedback, is at the heart of both schemes......whether it initialy results in a single or double pole characteristic, is of sublime irelevance."
And the clarification:
"A zero then introduced to cancel one of the poles, so that the two pole roll off reverts to a single pole slope long before unity gain frequency."
Thus as far as stability, phase margin, ringing etc...etc... are concerned, we are effectively dealing with a single pole response...since roll-off changes from two pole to single pole before Ft. "
And we are back to the desirability of a single pole response at the unity gain point. Also worth noting is the fact that most amplifiers do not have a real single pole response at the unity gain frequency and are venturing into the beginning of the second pole response, giving rise to consideration of the phase margin. These are basics in any good engineering text, as well as op amp application notes and data sheets. This is not some obscure mechanism hidden in papers by Ph.Ds and unknown to competent amplifier designers who actually build amplifiers.
All of this seems to be a far cry from being "ridiculously sublime" to me.............
It seems to me that dropping broad statements with little or no detail concerning the implications, is trolling so you can drag out the involved explanations and qualifications to support the initial statement after the usual confusion it creates. In this particular case I can write your next reply in advance.... "
"I said whether it INITIALLY results in a single or double pole characteristic"!
If there is a good reference to make your initial point, why not include it. Put the abstract of what it said with the link. Those interested in the topic can read it if inclined to, instead of trying to figure out the concept from a three or four sentence post. The need for lengthy clarifications and explanations after the fact, as a result of the brief and vague initial post, seems to indicate the intention to start an argument. You can then come in with the details and references to prove you were right when no one can understand what your initial point was. Initial post meant to inquire, illuminate, or to promote actual discussion does not require this frustrating game of 20 questions, to see what you meant. I have no doubt you have read every paper on frequency compensation and slew rates ever written. I do question whether you ever analyzed and built an actual amplifier or ever plan to.
Or do you just post to provoke arguments and drag out the appropriate AES or IEEE paper to prove you were right? I have received Email from several people that feel that is your exact intention.
"Miller feedback, is at the heart of both schemes......whether it initialy results in a single or double pole characteristic, is of sublime irelevance."
And the clarification:
"A zero then introduced to cancel one of the poles, so that the two pole roll off reverts to a single pole slope long before unity gain frequency."
Thus as far as stability, phase margin, ringing etc...etc... are concerned, we are effectively dealing with a single pole response...since roll-off changes from two pole to single pole before Ft. "
And we are back to the desirability of a single pole response at the unity gain point. Also worth noting is the fact that most amplifiers do not have a real single pole response at the unity gain frequency and are venturing into the beginning of the second pole response, giving rise to consideration of the phase margin. These are basics in any good engineering text, as well as op amp application notes and data sheets. This is not some obscure mechanism hidden in papers by Ph.Ds and unknown to competent amplifier designers who actually build amplifiers.
All of this seems to be a far cry from being "ridiculously sublime" to me.............
It seems to me that dropping broad statements with little or no detail concerning the implications, is trolling so you can drag out the involved explanations and qualifications to support the initial statement after the usual confusion it creates. In this particular case I can write your next reply in advance.... "
"I said whether it INITIALLY results in a single or double pole characteristic"!
If there is a good reference to make your initial point, why not include it. Put the abstract of what it said with the link. Those interested in the topic can read it if inclined to, instead of trying to figure out the concept from a three or four sentence post. The need for lengthy clarifications and explanations after the fact, as a result of the brief and vague initial post, seems to indicate the intention to start an argument. You can then come in with the details and references to prove you were right when no one can understand what your initial point was. Initial post meant to inquire, illuminate, or to promote actual discussion does not require this frustrating game of 20 questions, to see what you meant. I have no doubt you have read every paper on frequency compensation and slew rates ever written. I do question whether you ever analyzed and built an actual amplifier or ever plan to.
Or do you just post to provoke arguments and drag out the appropriate AES or IEEE paper to prove you were right? I have received Email from several people that feel that is your exact intention.
Hey Fred,
I can't work out why you waste so much text on bickering with mikek. It is obvious to me that you are in a higher league both in terms of experience, engineering knowledge and the desire to impart knowledge to others in a form which they can absorb. Mikek on the other hand "appears" to me to be rather a rather young contributor who's self-assurance is rather larger than his competence, who's ego requires continual point scoring and who's experience is borrowed from the writings of others.
That's ok but don't let your normally poiniant posts become banal because of it.
I can't work out why you waste so much text on bickering with mikek. It is obvious to me that you are in a higher league both in terms of experience, engineering knowledge and the desire to impart knowledge to others in a form which they can absorb. Mikek on the other hand "appears" to me to be rather a rather young contributor who's self-assurance is rather larger than his competence, who's ego requires continual point scoring and who's experience is borrowed from the writings of others.
That's ok but don't let your normally poiniant posts become banal because of it.
Regarding me liking the Frequency response to be flat well beyound the audio bandwidth has to due with Phase shift. the Closed loop response of the Amp must have a -3 dB point beyound 100KHz to avoid phase shift with in the upper octaves of the Audio range. It is true that Low frequency Phase shift is more noticable to the ear than High frequency phase shift is, I still like to remove any phase shift in the 20-20,000 Hz bandwidth to less than 1 deg. IMHO this is very importent to avoid colorations of the midrange and smearing if the related upper harmonics in relation to the fundamental frequency. The use of DC Blocking capacitors is another Evil of allot of Amps. these Coupling capacitors are most often selected way too small and have the -3 dB points set quite High 5-10 Hz. this will introduce low frequency phase shift. Low frequency phase shift is more of a problem that high frequency phase shift as the wavelength is longer and small timing delays seriously muddy the Bass and lower Mids. Add to this the absorption effects of capacitors and the sound can go to mud real quick with large tonal and pitch changes. IMHO the sonic signatures of alot of Audio Circuits can be traced to the Designers Choice of capacitors. DC or Direct coupling along with no or minimal compensation capacitance and these tone colorations are greatly reduced allowing attention to be directed to the remaining circuits.
It is a common misconception that placing a cap around the feedback resistor is a cure all for capacitive loads. in actuality this is not always true. Consider that this cap if too large will couple some of the load capacitance back into the capacitive sensitive inverting input. I admit at times a slight amount of capacitance is needed to trim the Transient response. in these cases i prefer to be critical of the layout and arrange the resistor values around the feedback network so as the Parasitic Board capacitance will do the Trick. I think that a properly designed Amp will only require a few pF(1-5pF) of capacitance to work properly, this is obtainable on the PC board. if more intrusive methods are needed and a additional capacitor is needed then this should be a series R/C network around the Feedback resistor. if the gain stage is not unity gain stable then a Resistor in series with the Cap is Mandatory so as to prevent the Amp from reaching unity gain too soon from External compensation capacitance.
The Above can be used in both IC Op Amp and Discreet component Designs. Discreet has the advantage of allowing the designer access to internal critical nodes not available in op amp's. This brings up the question of at what point is the compensation applied and of what type. the Previous discussion has listed about every point possible and some compelling arguments support each method. In my view i totally avoid the cap around the Primary feedback resistor and like to place compensation at the Differential output where it meets the input to the second VAS stage in the form of a series R/c network from this node to the proper supply rail. This will simply roll off the HF gain of the Diff amp prior to the VAS stage. I my self don't like pole splitting as that secondary pole can come back to haunt you unexpectedly. A nice 1 st order roll off will provide ideal transient behavior. these methods are only required if the stages after the Differential are slower than it is. With most Power Amps this is indeed the case the larger geometry Higher power transistors following the Diff amp have more capacitance and reduced Ft's compared to the smaller geometry of the input stage Devices.
Resistors in series with the Base of emitter followers are used allot and are one of my own tools to fight instability in output and driver stages. Common collector stages do not like to be driven from a low impedance source as at some point the semiconductor will become inductive the resistor provides a real impedance for the driver and output stages. One place where a series base resistor is helpful is on the Vas stage, While the Impedance of a Diff Amp is quite high compared to emitter followers this is still a place where added series base resistance can be used as a compensation tool.
It is a common misconception that placing a cap around the feedback resistor is a cure all for capacitive loads. in actuality this is not always true. Consider that this cap if too large will couple some of the load capacitance back into the capacitive sensitive inverting input. I admit at times a slight amount of capacitance is needed to trim the Transient response. in these cases i prefer to be critical of the layout and arrange the resistor values around the feedback network so as the Parasitic Board capacitance will do the Trick. I think that a properly designed Amp will only require a few pF(1-5pF) of capacitance to work properly, this is obtainable on the PC board. if more intrusive methods are needed and a additional capacitor is needed then this should be a series R/C network around the Feedback resistor. if the gain stage is not unity gain stable then a Resistor in series with the Cap is Mandatory so as to prevent the Amp from reaching unity gain too soon from External compensation capacitance.
The Above can be used in both IC Op Amp and Discreet component Designs. Discreet has the advantage of allowing the designer access to internal critical nodes not available in op amp's. This brings up the question of at what point is the compensation applied and of what type. the Previous discussion has listed about every point possible and some compelling arguments support each method. In my view i totally avoid the cap around the Primary feedback resistor and like to place compensation at the Differential output where it meets the input to the second VAS stage in the form of a series R/c network from this node to the proper supply rail. This will simply roll off the HF gain of the Diff amp prior to the VAS stage. I my self don't like pole splitting as that secondary pole can come back to haunt you unexpectedly. A nice 1 st order roll off will provide ideal transient behavior. these methods are only required if the stages after the Differential are slower than it is. With most Power Amps this is indeed the case the larger geometry Higher power transistors following the Diff amp have more capacitance and reduced Ft's compared to the smaller geometry of the input stage Devices.
Resistors in series with the Base of emitter followers are used allot and are one of my own tools to fight instability in output and driver stages. Common collector stages do not like to be driven from a low impedance source as at some point the semiconductor will become inductive the resistor provides a real impedance for the driver and output stages. One place where a series base resistor is helpful is on the Vas stage, While the Impedance of a Diff Amp is quite high compared to emitter followers this is still a place where added series base resistance can be used as a compensation tool.
...ironically, if i had said this, traderbum, and dieckman would probably dispute it out of pure juvenile pique....
Do you ever shut up?
Do you ever build anything, or are you just entralled by the sound of your own voice?
Have you always been this annoying, or did you have to work hard at it for years like the rest of us.
Enquiring minds want to know. It might be a more enjoyable read than the constant stream of references to old JAES and IEEE articles.
Cheers.
Jocko
Do you ever build anything, or are you just entralled by the sound of your own voice?
Have you always been this annoying, or did you have to work hard at it for years like the rest of us.
Enquiring minds want to know. It might be a more enjoyable read than the constant stream of references to old JAES and IEEE articles.
Cheers.
Jocko
Putting up resistance
Most interesting post ppl. You have obviously built quite a bit of stuff and know your stuff. I like the stick "extra" resistors in the circuit as well. One point I believe you didn't raise is the RF resonance that lead and trace inductance in conjunction with device capacitances can cause, and the effect of these resistors on damping them. I have seen capacitors across the feedback resistor mentioned as a path for RF to get in the front end diff pair. Most designers filter the input side of the pair but miss this path. Maybe Jocko and I won't get as many questions over our "put a resistor on the gate" advice that we make such a point of. Please post some more good stuff like this.
Most interesting post ppl. You have obviously built quite a bit of stuff and know your stuff. I like the stick "extra" resistors in the circuit as well. One point I believe you didn't raise is the RF resonance that lead and trace inductance in conjunction with device capacitances can cause, and the effect of these resistors on damping them. I have seen capacitors across the feedback resistor mentioned as a path for RF to get in the front end diff pair. Most designers filter the input side of the pair but miss this path. Maybe Jocko and I won't get as many questions over our "put a resistor on the gate" advice that we make such a point of. Please post some more good stuff like this.
- Status
- Not open for further replies.