Lets not get lost in the weeds here, folks.
Dotneck -- we don't know the parasitic capacitance of your setup, so an oscilloscope and trial/error of a square wave response (don't get confused by Gibbs effect!) to get an acceptable first-order response is good_enough. You should be able to ballpark the overall capacitance (which is your starting point) you need based on the document I gave you.
Dotneck -- we don't know the parasitic capacitance of your setup, so an oscilloscope and trial/error of a square wave response (don't get confused by Gibbs effect!) to get an acceptable first-order response is good_enough. You should be able to ballpark the overall capacitance (which is your starting point) you need based on the document I gave you.
Making the capacitor across the feedback resistor larger than needed for stability reduces the bandwidth of the transfer from input to output without reducing the loop gain at 20 kHz and its harmonics, so it shouldn't result in any increase of the THD at 20 kHz.
The LM4562/LME49720 doesn't need any feedback capacitance, unless it's driving a capacitive load. It's unity-gain stable.
Unless you're having issues with the LM4562/LME49720 in situ, I wouldn't worry about feedback capacitance. And if you do have issues, I'd find the appropriate capacitance value either by simulation or by experimentation. As the TI app note says, "add capacitance until you get the desired transient response". If you have to add much more than 10-22 pF, I'd definitely take a look at what's going on using an open-loop measurement/simulation of sorts.
Tom
Unless you're having issues with the LM4562/LME49720 in situ, I wouldn't worry about feedback capacitance. And if you do have issues, I'd find the appropriate capacitance value either by simulation or by experimentation. As the TI app note says, "add capacitance until you get the desired transient response". If you have to add much more than 10-22 pF, I'd definitely take a look at what's going on using an open-loop measurement/simulation of sorts.
Tom
Last edited:
I have had troubles with 4562 oscillation in the past, although that was probably due to my rather poor wire-wrapping. I've also read of guys experiencing wild oscillation when putting in 4562s in older circuitry. In any case, I figured the WIDEST bandwidth I'd ever need would be ~200K Hz---enough to pass a 20KHz square wave----which a rim shot or a cymbal crash rise time might approach. I certainly don't need more than that, so I ran an op amp through my SiMetrix simulator. With feedback values of ~10KΩ, I'd be OK with ~82pF (-3db @ 210KHz); with ~20KΩ in the feedback loop, I'd need ~39pF (-3db @ 200KHz). So those two values probably oughtta cover all the circuits in the Dolby 363 with an LM4562 replacement. Along with a 100nF cap across the rails, the oscillation tendencies should be killed. Eh?
Last edited: