LM3886: Stability at high frequencies and output level

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I'am making a LM3886 amplifier for fun and when it works well, it goes to a friend.

I pushed the amplifier to its utmost output at high frequencies (50 khz and more) and found that the Cc capacitor (see datasheet) makes the amplifier rather more unstable than stable.

I also found that the Cf capacitor and Rf2 resistor of course lowers the -3 dB point, but when you increase the input with the same amount, it has no effect on the stability (nor positive or negative): the amplifier starts to oscillate - given a high frequency - at the same high output level.

Has anybody the same of another experience?

In the last case it is always better to use an passive filter at the input rather than an active filter in the feedback network. So I think I use an input filter with a resistor of 1 kohm and a capacitor of 1 nF (-3 dB point at 160 kHz).

Now I'am thinking about how to avoid the electrolytic feedback capacitor Ci without high output offset voltage and without using a DC servo.

Marc.
 
If you are happy to add an op-amp, this technique (see attached pic) is a good way of getting the low distortion of the inverting mode, together with a low output offset.
The extra components to ensure stability under all operating conditionns aren't shown. The schematic was simply a means of checking the SPICE model of the LM3875 that TI have recently released.
 

Attachments

  • LM3875 SPICE INVERTING.JPG
    LM3875 SPICE INVERTING.JPG
    33.6 KB · Views: 379
Hi Marc,

I share your observations, the datasheet circuit to me always looked like a bad prototype layout fixed up to "make it work", somehow. In a spice analysis, using an AC model derived from curve fitting of open-loop gain and phase as given, it's close on the edge of oscillation like the real thing.

For most any chipamp in non-inverting mode, I prefer to stabilize HF gain with parallel capacitors to the main feedback resistors, like 20k//470pF and 1k//1n for a gain of 20x (26dB).

Input and output must be filtered against HF ingress and the 2nd input roll-off capacitor must terminate exactly to the GND point of the 1n cap.

For stabilizing offset I see no other way than matching input pin impedances and/or injecting a quiet offset nulling current somewhere... and hope that drifts and aging don't spoil the party...
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.