[sdiy] op amp substitutes

[Magnus Danielson]

From: Harry Bissell Jr <harrybissell at prodigy.net>
Subject: Re: [sdiy] op amp substitutes
Date: Wed, 19 Jan 2005 15:18:30 -0800 (PST)
Message-ID: <20050119231830.30889.qmail at web51003.mail.yahoo.com>

> Arrgh !!!
> 
> (as if this did not bite-me-in-the- at ss for the
> one millionth time yesterday).

Ofcourse it didn't, you're a pro! ;O)

> Any opamp might oscillate with at some capacitive
> load...  It has to do with (basically) shifting the
> feedback so that it becomes resonant.

Indeed it will!

>  Maybe MAGNUS (hej!) will chime in with the math.

Hej Harry! Yeah, I can make a short chime in... ;O)

In the ideal model the op-amp has infinite gain, infinite bandwidth and perfect
common mode suppression. We can model this as

V    = G(V  - V )
 out      +    -

where G is infinite. Anyone that wants to play with math on op-amps just needs
to line up the expressions for V+ and V- and then assume that the op-amp is
operating in its linear negative feedback mode let V+ = V- and reduce the
formulas from that.

Real life isn't even close from this ideal model. We have lag in the op-amp, we
have limited gain and then we have an output resistance. The lag results in
what is called the dominant pole, which we can model as an RC lowpass filter.
The gain G of the op-amp will only be the DC gain or rather, the passband gain
of the op-amp. This gain reduces with frequency beyond the frequency of the
dominant pole. However, alongside with this the phase will also shift. What can
now happen is that the phase of the feedback path lags so much that it is in
-180 degrees phase into the negative pin, thus resuling in a positive gain with
the polarity of the pin, positive feedback is a buster for oscillation, but
only if the gain is beyond unity. So, one of the parameters we often find is
the gain-marginal, that is, how much additional gain we can handle before
self-oscillation. Another bad thing that could happen to us is that the phase
lags in addition to the phase lag we have, that is, how much additional phase
lag can we have in the negative feedback path before we reach unity gain and
-180 degree total phase. The maximum phase lag is refered to as the phase
margin. The better phase-margin, the better. Now, where can such an additional
phase-lag occur. An op-amp with negative feedback has no output impedance,
don't it? No, that's just the ideal model toying with you, the output
resistance together with the load capacitance will provide additional phase lag
and a little gain reduction. The unity gain frequency will move down in
frequency with additional capacitive loading, while the phase will lag more and
more. Increased loading capacitance lowers the second lag filters corner and
thus makes more and more of the additional phase lag apparent to the feedback
chain and eventually make the circuit so unstable that it oscillates insteady
of just ringing *badly*.

Now, this is why you at least want to look at a lesser than ideal model for
your op-amps. The bare minimum is fairly accurate modeling of gain, dominant
pole and output impedance. Good models also include multiple poles at higher
frequencies and frequency dependent output impedance.

The parameters to look for is phase margin and output impedance, those will
give a good hint.

I strongly recommend looking in "Designing with Operational Amplifies and
Analog Integrated Circuits" by Sergio Franco. Chapter 8 is devoted to the
problem of negative-feedback instability of op-amps.

> There are techniques to use (wish I had...) such as
> decoupling the capacitance with a small series
> resistance.  I believe (but am not positive) that
> making a voltage follower with a resistor from the
> output to the inverting input (instead of a wire) has
> benefit as well.

By putting a resistor in the feedback, a resistor to ground for slight damping,
and then shorting the feedback resistor with a capacitor helps beyond the
additional resistor. Actually, many op-amps have series-resistors on the
output already!

> Sometimes you can use a much BIGGER capacitor and
> swamp out the effect.
> 
> TL07x and TL08x are the worst offenders I can think
> of.
> My case yesterday... oscillated with 3' of
> UNTERMINATED
> MIDI cable plugged in.  Not even connected to
> anything...just 'there' (and of course its a
> capacitive load to ground...)

Naturally! ;O)

Cheers,
Magnus - There's two kinds of mathematicians, the discrete and the indiscrete ones!

> H^) harry
> 
> 
> --- anthony <aankrom at bluemarble.net> wrote:
> 
> > I was wondering just what parameter you look for
> > when you want to see how 
> > much an op amp output can take high capacitative
> > loads.
> > 
> > More specifically I was wondering if I could
> > substitute one TL034 or one 
> > TL054 (or even '74's) for four LF356's.
> > 
> > The LF356's datasheet goes to the trouble to mention
> > that it can handle high 
> > cap loads. But the '34 and '54 are FET input amps
> > too (yeah I really don't 
> > care about the inputs - it's the OUTputs...)
> > 
> > This is in the Analog Shift Register circuit. 
> > 
> > 
> > 
>