What I’m wondering about is that both have resistors in series with the output, but in the first that resistor is in the op amp feedback loop. In the second it’s not.
Why would one choose to do one versus the other? I assume there are relative advantages and disadvantages to both, or that one is suitable for some purposes while the other is better for others, or… well, I’d like to hear an explanation.
I have an application note that discusses this somewhere but cannot find it right now, but iirc the version with the output resistor in the loop can deal with capacitive loads (e.g. long patch cables) a bit better.
(the resistor itself is also there to protect the output stage and more importantly protect other modules, by limiting the current in and out of this module.)
The goal of the output resistor in both cases is probably to protect the circuit for the output to be shorted to ground while it is outputting a non 0 voltage and / or to limit the current that can be drawn from that output .
From a design perspective I would say that including the output resistor in the feedback loop makes that the transfer function of the circuit is independent of what it is connected to. If any subsequent module has a near infinite input impedance, e.g. by starting with a buffer before processing the signal, that might make it easier to guarantee the working of the module within the specs one starts the design with.
However, if you plug two outputs together accidentally, then with the resistor after the feedback loop, it’ll act like a mixer set 1:1 (well, up until one of the
op-amps goes into clipping); at least a musical effect, and no sparks should result.
With the resistors inside the feedback loops, the two op-amps will be in some sort of crazy cage-match, fighting each other, and could conceivably
go into oscillation in the ultrasonics (all depending on the rest of the circuit topology) and bizarreness may happen - especially if your downstream speakers
don’t have good ultrasonic filtering, as you may then blow the tweeters.
and
However, the main difference is in how the opamps respond to downstream capacitive loading. An “outie” resistor provides very good compensation for capacitive loading (and a long patch cable is often enough downstream capacitance to create oscillation problems). And “innie” resistor does nothing for capacitive loading, and in fact can even make matters worse. This is why I apply a special capacitive loading compensation network on all my opamps when I’m serious about not creating CV errors.
and
Yes, but don’t do it indiscriminately. Sometimes, these feedback capacitors create more harm than good. Putting them in the gain stages of a cascaded filter, for example, actually renders the circuit more unstable. Also, using them when it isn’t necessary can create unwanted lags in the system, filtering out high frequencies and decreasing slew rates.
So a Muffwiggler “outie” is what the Analog Devices application engineers call “in the loop”, and their “innie” is “out of the loop”? The linked post isn’t entirely clear.
So who do I trust – opamp manufacturers, Art of Electronics, application engineers who show their work, etc, or some dude named “etch a sketch” on an internet forum?
As I read it, what he’s saying, and what several other people in that thread agree with, is that an innie resistor alone does nothing for capacitive loading, while an outie resistor does. With an innie, they say, you also need a feedback capacitor.
And as I read it, Analog is saying the same thing. They’re not saying that simply putting a resistor in the loop does anything for capacitive loading. They’re saying there is a compensation technique that involves adding a capacitor and a resistor in the loop, as well as a cap and resistor to ground after the loop. (whoops, sorry, that’s the load).
They also say:
Q: Is there a simpler compensation scheme that uses fewer components?
A: Yes, the easiest way is to use a single external resistor in series
with the output. This method is effective but costly in terms
of performance (Figure 8).
I read that as being entirely consistent with what several “Muffwiggler dudes” agree on.
(He ends up with 33 pF, vs. the 3.3 pF advocated by Sketch-n-etch. Analog supplies a formula that depends on all the Rs and Cs involved including the load. I haven’t tried applying that to see what it gives.)
