[sdiy] Voltage Feedback Resistors and Circuit Stability

[David G Dixon]

Yes, adding feedback capacitance is the standard way, but if you've already
got all the PCBs for a design made without these capacitors in place, then
it is nice to know that decreasing the values of the feedback (and input)
resistors is one way to eliminate ringing.

  _____  

From: Synth-diy [mailto:synth-diy-bounces at synth-diy.org] On Behalf Of Richie
Burnett
Sent: Saturday, November 14, 2020 2:29 AM
To: synth-diy at synth-diy.org; Bernard Arthur Hutchins, Jr
Subject: Re: [sdiy] Voltage Feedback Resistors and Circuit Stability


[CAUTION: Non-UBC Email]In my experience smaller feedback resistors
generally give better op-amp stability. With a large feedback resistance
there is more potential for stray capacitance between the virtual earth
(inverting input) node and ground to cause a phase lag. Too much phase lag
degrades phase margin and turns negative feedback into positive feedback,
ultimately leading to oscillation. Just like the case of capacitive loading
on the output that Bernie mentioned. However the virtual earth node can be
much more sensitive to stay capacitance than the op-amp output which is
generally a stiffer voltage source.

Of course the proper solution to the problem described above is to add some
capacitance across the feedback resistance. This introduces some phase-lead
into the feedback path to compensate for the undesirable lag from the stray
capacitance. This improves the phase margin and moves the op-amp further
away from instability/oscillation. This process is commonly used in things
like photodiode amplifiers to prevent the capacitance of the photodiode from
destabilising the op-amp.

-Richie,

Sent from my Xperia SP on O2

---- Bernard Arthur Hutchins, Jr wrote ----



First, the question was about :
<https://synth-diy.org/pipermail/synth-diy/2020-November/174435.html> [sdiy]
Voltage Feedback Resistors and Circuit Stability.   "Feedback resistance"
makes almost no sense.   It is a feedback RATIO that maters, as a design
objective (e.g., setting gain) and in any (if any at all) op-amp stability
issues.  Typically, this ratio is set by a series resistive voltage divider
Vout/Vin=R2/(R1+R2) where R1 is connected to Vin, and R2 goes (usually) to
ground.  A ratio of 1/11 (0.090909.) is obtained by R2 = R1/10.  R2 might be
10k with R1=100k, or R2 might be 1k with R1 = 10k, etc.  If Vin is the
output of an op-amp, good practice suggests that the SUM of R1 and R2 should
be at least several k (for the op-amp to drive) and less than about a meg
(to avoid stray signal pickup). ELSE which values do you have the most of.  




All this freedom goes out the door IF THERE ARE CAPACITORS IN THE DIVIDER
LOOP!  In this case, if you scale the resistors by a factor B, you must
scale the capacitors by 1/B. I can't recall a familiar example of this in a
feedback case, but in an input attenuator case, one is very familiar: the
case of an R1 =100k, R2= 220 ohm attenuator into our original OTA
integrators  (S-V VCFs).  Originally no capacitors were used.  Then we
started to use shunting "phase-lead" capacitors across the 100k R1.  The
needed shunting capacitor was an inconveniently small (rare, and comparable
to stray) 3pf or so. This is why we changed R1 down to 10k, R2 down to 22
ohms, and C up to a more agreeable 30 pfd. 




Finally, keep in mind that "instability, in general, is intuitively
associate with high gain (like positive feedback in PA systems).  In the
case of linear op-amp applications, NEGATIVE feedback is used to restrain
the extremely large gain of otherwise open-loop devices.  Low gain circuits
have MORE (presumed negative) feedback.  To the extent that actual feedback
slides slightly less negative, a high-frequency oscillation may kick in.
Maximum (negative) feedback (100%) is at unity gain - hence the near
universal unity-gain internal compensation.  Such an op-amp "follower" may
oscillate if asked to drive a long scope cable (capacitor) for example
(phase shift inside the  loop) while being perfectly well-behaved if the
gain is perhaps 4!
- Bernie



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://synth-diy.org/pipermail/synth-diy/attachments/20201114/3abf77f5/attachment.htm>