The case for two pole compensation

Two pole compensation seems superior to single pole in almost every way.  Some of the gurus (Self, Sloan, etc) talk about it in their books, but beyond this, it seems like its an obscure technique.  Anyone have any opinions on two pole compensation?  From an engineering perspective, it seems like a no brainer.  Here is a white paper on the topic discussing loop response and op amp selection in general. 

http://sparkoslabs.com/audio-op-amps-gain-seek-bandwidth/

Hi again...  Back a few decades ago when I was tweakier than I am now I even considered the implications of having the open loop gain track the closed loop gain, so loop gain margin would be constant over the bandpass... but that was when I was over-designing phono preamps (that had a falling closed loop response). Even the 5534 a VERY GOOD opamp lost some love with the tweakers for it's unconventional compensation (split-pole?).

Ideally there is no nonsense (pole) in the audio bandpass for the open loop transfer function too, but that just isn't practical when you start with more than 100dB open loop gain.  Can you push your 2 pole comp roll-off higher so open loop is flat up to 20 khz?

The meat on your 2-pole bone is to do a series of distortion  (and phase shift) measurements of otherwise identical opamps (or better yet the same opamp), but with the only difference the compensation scheme. 

If it is better, how much... ?

JR

Open loop rolloff beyond 20KHZ would mean either a really high bandwidth amplifier, or (gasp) 3 pole compensation. 

Which I have played with.  In addition to the two pole scheme, I played around with inductively shunting RE resistors in the input stage diffamp - the idea was lower noise, but the result was a 3rd pole from the inductive shunting.  Open loop response showed a phase dip in excess of 180 degrees from the 3rd pole.  Closed loop was infact stable, however presumably, there would be some closed loop gain(s) that would not be stable.  And at that point, I reached the extent of my compensation knowledge.    A third pole is enticing, but Im not sure of its ultimate affect on closed loop stability. Perhaps it would work out unless one was designing a high gain (several 10's of dB) amplifier stage. 

Like a mic preamp?

JR

Yes.  Like a Mic preamp.    lol

Exactly two poles is stable.

There's always more poles "above" your intended frequency range. They add lag far below their corners. 10 degrees here, 15 degrees there, a "safe 160 degrees" can become 180.0 degrees real easy.

Don't forget any poles the user has connected to the amplifier. (Such as hundreds of feet of cable.)

9dB/octave has been a real good guide since Black invented his oscillator (which was supposed to be an amplifier).

6dB/oct is a whole lot easier to get.

If you think your knowledge is light, there are FAT BOOKS about stability of systems.

One of Black's early short papers:
http://www3.alcatel-lucent.com/bstj/vol13-1934/articles/bstj13-1-1.pdf
10MB PDF

sparko311 said:

Open loop rolloff beyond 20KHZ would mean either a really high bandwidth amplifier, or (gasp) 3 pole compensation. 

Click to expand...

Only if you need a lot of NFB to reduce open loop distortion. An alternative approach is to design for modest open loop gain with low distortion. You can then set the dominant pole at 20KHz and use less NFB for the same closed loop gain and hence maintain stability. The other advantage of this approach is that the amount of NFB is the same over the entire audio band whereas in an op amp with a dominant pole at a few Hz the amount of NFB at 2KHz is 20dB less than at 20Hz and another 20dB less at 20KHz with obvious consequences for the distortion.

One further wrinkle is to arrange the open loop gain to track the closed loop gain thus keeping the amount of NFB and the stability margin the same as the closed loop gain is varied. This is the TransAmp technique.

It is interesting to note that these techniques were used in the V76 amplifier.

Cheers

Ian

PRR said:

One of Black's early short papers:
http://www3.alcatel-lucent.com/bstj/vol13-1934/articles/bstj13-1-1.pdf
10MB PDF

Click to expand...

IMO NF gets a bad rap from the audiophools. Were it not for NF, the frequency response of thousands of cascaded telephone repeaters would be unmanageable.  NF=good, while there is no free lunch and everything has design considerations.

JR