Placement of compensation capacitor

[rtarbell]

Okay, here's a question regarding your basic three stage power amplifier and how to compensate it. Normally, the compensation capacitor is placed from the output of the first stage to the output of the second stage (and thus it envelopes the VAS stage).

==> The slew rate depends upon the bias current in the differential input stage, and the bias current in the VAS stage.

What if we arranged it so that one end of the compensation cap is connected to the output of the driver stage? In this capacity, the bias current of the VAS stage does not determine part of the slew rate of the amp; the bias current of the driver stage determines it (and the driver stage current is usually higher than the VAS stage current, which means more current to discharge Ccomp, which means higher slew rate).

As seen on the picture, instead of connecting the floating end of Ccomp to the collector of Q3 (as is traditionally done), why not connect it to the emitter of Q4?

[EWorkshop1708]

Not sure why it's not been tried.

The logic makes sense.

I'll try it.

[darkfenriz]

As far as I remember long lost mr. Evil experimented with this, but from different reasoning.
I don't think you will improve slew, rather you will make it more asymetric. Not sure though.

regards

[traderbam]

Hi rtarbell,
I'd say the primary reason this is not usually done is because of stability. Q4 introduces extra phase shift that reduces the phase margin of the miller stage. In either configuration, the slew rate is primarily limited by the supply current of the stage driving Q3's base, rather than the collector current of Q3, so including Q4 in the loop doesn't help that much.

Consider the extreme case of the beta of Q3 being infinite. You'll see that the first stage effectively charges and discharges Ccomp...like an op-amp integrator. The fact that Q3 beta is finite only adds to the burden of the first stage.
BAM

[jcx]

Dr Ed Cherry proposed this long ago, in fact he prefers enclosing the whole output stage

http://www.diyaudio.com/forums/showt...ht=#post228259

http://www.diyaudio.com/forums/showt...rry#post281394 (and next post)

MikeS has some objection to enclosing output Qs (apparantly objecting to bjt outputs) which he has yet fully to share with us in the "how many ways to stabilize" thread:

http://www.diyaudio.com/forums/showt...197#post647197


compensating from a buffered vas allows the use of 2 pole compensation with with unequal C, potentially removing the Miller multiplication of the compensation C seen at the vas input at audio frequencies - without buffering the vas would be loaded by the larger C, R to (ac) gnd on it's output. limiting audio frequency gain improvement from the 2 pole compensation

[lumanauw]

Fig A is the most common compensation, miller cap, between B/C of T1 (VAS transistor). The value is pF.

DougSelf said that Fig B is "sub optimal", since it is loading collector of T1 by big capacitance (nF value here)

I'm not thinking voltage/current phase. But fig B seems to be a lowpass filter (in conjuction with internal impeance of T1 collector).

If Fig B is actually a lowpass,

Has anybody tried fig C? Inductor there (in the position of base stoppers) should be a lowpass too.

[Johan Potgieter]

The C placed as in A in Lumanauw's sketch is universally used but can pose stability problems since it remains active in a manner of speaking. The method B is more docile. It is indeed a simple low-pass filter affording the desired phase correction. I disagree somewhat with Douglas Self (something that I have never done on any other occasion!), in that for optimal design this compensation should apply ideally outside the audio band, He is correct in saying that it gives rise to distortion, but that is only relevant if compensation starts within the audio band (and how it does in some circuits!) I would also recommend reading of Dr Cherry's relevant articles.

In that sense inductors could indeed be used as in C but I have never seen it - perhaps cost?

In a circuit of my own I use topology B, with a capacitor of only 220 pF required, starting effect at about 25 KHz only. If I try to get similar results by A, I get oscillation, or at least a sharp peak in response at about 3 MHz.

But there is a preferable way that I have mentioned before on another thread, which was expounded in an article by Dr John Ellis in Electronics World March 2003, called "Audio Power Amplifier Feedback Compensation". There he showed the superiority of a phase-lead-input-lag type of system, which avoided the disadvantages of the commonly used Miller C.dom. (Lumanauw, I think that I posted a copy of that article to you.) In general, the advantages of also using a capacitor in
parallel with the feedback resistor must not be overlooked.

[lumanauw]

Hi, Johan,

Thanks again for the article Is it possible, that people not use C, because when the amp is headed to capacitive load, and when all transistors are bipolars, this capacitive properties will ramp back via base, and when it meets inductor there, the whole amp will oscilate?

[Johan Potgieter]

Lumanauw,

I will have to do a Spice analysis regarding that, when I have time. The next couple of days are somewhat tied up - will come back to you.

Regards.