[sdiy] buffer opamp configuration [was: 1/8th inchquestions]

[sbernardi at attbi.com]

With all this talk about opamp stability with capacitive 
loads, I found this online simulation at Analog Devices:

http://www.analog.com/techSupport/designTools/interactive
Tools/stability/stability.html
> René Schmitz wrote:
> 
> > At 23:46 24.10.02 +0100, you wrote:
> > >jhaible wrote:
> > >
> > >> I've read RAP's app note now, and apparently he concentrates on _one_ 
> method
> > >> to increase stability: increasing noise gain, with various circuits. But
> > >> this also
> > >> means more noise - sometimes more noise than necessary.
> > >> I normally use the other method, with resistor at the opamp output, and two
> > >> feedback paths, as described in previous mails in this thread.
> > >
> > >That's the most common method I've seen for driving capacitive loads - a cap 
> and series resistors in parallel in the feedback loop to get a gain roll off at 
> higher frequencies, with the output taken from the junction of the series 
> resistors.  It would appear that this works for both inverting and non-inverting 
> (gain<1) configurations, but *not* for unity-gain non-inverting - in this case a 
> capacitor parallel with the feedback resistor would have no effect on the gain > at higher frequencies .. it's always 1 (I think!).  Hence the alternative 
> technique of increasing the noise gain to increase stability becomes necessary I 
> guess.
> >
> > That isn't quite what was meant by two feedback paths. See the attachment.
> >
> 
> That's what I described (but not very well).  Thing's are so much easier with 
> pictures ;)  I wasn't thinking about the capacitive load properly - hence I was 
> considering the series 2 resistors in the feedback loop as 1 resistor.
> 
> >
> > >It seems unity-gain buffers are tricky, but I can't see any other way with a 
> multi output MIDI2CV - I don't want to add trimmers for each output.
> >
> > You need to imagine the loading cap also, then you see how it works:
> > The oscillation occurs because the finite open loop impedance together with a 
> capacitive load forms a lowpass which shifts phase and consumes up the phase 
> margin. (I.e. the feedback is taken at the output of this RC-combination.) For > low frequencies we can neglect the effects of the cap, and see that the gain is 
> really unity. For higher frequencies the second feedback path takes over. With 
> the feedback taken directly at the opamp output the phase shift doesn't take 
> place. The input structure doesn't play a role here.
> >
> 
> Ok, I'm starting to understand (thanks to your help).  I've attached a pic of my 
> own to make things easier (the cap on the front is just the S&H).  Also in my 
> pic I have shown the load resistance, but I assume this is high enough to be 
> ignored when thinking about oscillation.
> 
> So you are saying that oscillation occurs because at some frequency the phase of 
> the feedback signal becomes enough out-of-phase to cause the feedback to be 
> positive, not negative.  Without the 100R resistor, the phase shift is only due 
> to the lowpass filter formed by combination of the internal output resistance of 
> the opamp and the capacitive load, and is not too bad for medium-speed opamps > where the gain rolls off at higher frequencies.  But adding the 100R resistor as 
> short circuit protection changes the nature of the RC filter, causing the 
> frequency at which the phase shift becomes a problem to fall - causing even a 
> slower opamp to oscillate.  The 330pF cap in the inner feedback loop stops this 
> oscillation because it provides a more direct feedback path for higher 
> frequencies.
> 
> Sorry if I'm being slow to understand here .. all this stuff is very new to me.
> 
> Seb
> 
>