Capacitive loading of op-amps

[Arthur Harrison]

AH:


> >It also occurs to me that adding a large capacitor directly to the
> >output of an op amp will essentially cancel much of the loop's AC
> >gain,

JH:

>Yes. Loop gain being more or less a LP function, it will
>shift the cutoff frequency downward, i.e. less loop gain
>at a certain frequency.
>Now, loop gain (or let's speak about loop bandwidth, as
>the DC gain is not affected by the capacitor on the
>output !) isn't a benefit by itself. Why build a buffer with
>high loop gain / bandwidth and tight feedback at all ?
>(1) It will help for a well defined closed loop gain (i.e. relative
>wide band voltage buffer), and
>(2) it will also reduce the output resistance of the buffer.
>
>You need (1) and (2) for amplifiers. But for a buffered
>auxiliary voltage, you don't need (1), as the "amplifier
>input" doesn't change at all. You still want (2), i.e. a stiff
>voltage with low internal impedance over a wide frequency
>range. The opamp with reduced loop gain (because of the
>cap) will not provide such a good wide band low impedance
>anymore. But who cares - that part is taken by the cap.
>The opamp takes care of the DC and low frequency part.
>So, for high frequencies, all is ok. (It even would be ok
>with just a resistor divider and capacitor, no opamp)
>For DC, everything's fine too. (cap is irrelevant here)
>If there is no resonance peak in the closed loop gain
>(i.e. external pole and internal pole from compensation
>wide apart), everything should be nice and smooth  in
>between, too.

AH:

> >sort of rendering the circuit as a precision emitter follower
> >without a Vbe offset.

JH:

>Emitter follower is different. You don't have an external feedback
>loop and no high output impedance of the "open loop" either.
>Can't see how this would fit together.
>
> >For this simpler approach, however, another consideration is that a
> >huge slug of current will be drawn through the op amp's output
>transistors
> >upon power-up.  Most op-amps are capable of withstanding an
> >intermittent surge such as this, although it may cause a failure in
>some
> >cases when the capacitor is large, has an exceptionally low ESR
> >(e.g., tantalum), and the supply' s rate of rise is high.
>
>The current will be limited by the internal current limiting that's
>build into most opamps. It's not like a single transistor, where
>you must take care how long a certain high current can last
>without burning something. The opamp will simply go into
>limiting, and behave like a current source rather than a voltage
>source until the cap is charged.
>The only thing I'd worry about is what happens when the circuit
>is switched *off*, i.e. supply voltage goes down, but the capacitor
>is still charged. I never had problems with 10u caps and the usual
>cheap opamps myself, but I think that there is some limit.

AH:

All points well taken.  From a reliability standpoint, I still prefer output
protection for the op amp during power-on and power-off.
(You never know when a synthesizer will be used in a life-saving
capacity!)

I suspect that loop dynamics for all transient load conditions really
would benefit from the high AC gain, but admittedly, would have to
do a physical comparison to be sure.

I suppose a Hg relay to switch a load and a storage scope would be
the way to tell.

-Art