I'm working on a design that uses the LM3886 as the differential stage in an instrumentation amplifier. Obviously, there are some theoretical advantages to using the input buffers to provide most of the gain. I don't know much about external compensation (I've only used unity-gain stable opamps before), but I've been playing around with spice simulations and it looks like the LM3886 can be configured for 4V/V if a lead-lag compensator is used. The phase margin is roughly the same as 10V/V without external compensation.
Is this a bad idea? Bound to oscillate? Assuming it is stable, would you expect any actual measured improvement over a 10V/V configuration?
Is this a bad idea? Bound to oscillate? Assuming it is stable, would you expect any actual measured improvement over a 10V/V configuration?
It is possible to keep LM3886 stable at gains below 10. See this posting (#47): Chip-amps suited as power stage in a composite amplifier, LM1875/TDA2050 excluded.
Use a resistor between the two input pins
Further, see posting #22 in the same thread. These results were concerning use as the power stage in a composite amplifier. The output offset of the LM3886 may increase.
If this is useful for your instrumentation amplifier use I cannot say but else try your spice configuration in practice.
Use a resistor between the two input pins
Further, see posting #22 in the same thread. These results were concerning use as the power stage in a composite amplifier. The output offset of the LM3886 may increase.
If this is useful for your instrumentation amplifier use I cannot say but else try your spice configuration in practice.
Adding a resistor between the inputs artifically raises the noise-gain, and its noise-gain that determines stability.
However this method increases distortion proportionately to the drop in gain also. And the noise gain is typically also dependent on the source impedance for standard non-inverting amp configuration.
This trick is used to measure distortion in very high performance opamps, since it can be used to magnify the distortion up to a level that's easy to measure.
Noise gain is a measure of how much the output voltage affects the differential input voltage through the feedback network, its basically the reciprocal of the feedback network attenuation.
However this method increases distortion proportionately to the drop in gain also. And the noise gain is typically also dependent on the source impedance for standard non-inverting amp configuration.
This trick is used to measure distortion in very high performance opamps, since it can be used to magnify the distortion up to a level that's easy to measure.
Noise gain is a measure of how much the output voltage affects the differential input voltage through the feedback network, its basically the reciprocal of the feedback network attenuation.
10V/V is just about the limit for a 3886 (20dB). I'm using 12V/V in an everyday application and it is stable without external components (I do add Zobel and compensation afterwards as they help when the chip is stressed).
Many thanks Mark for the good explanation.
While it is not an ideal measure, it is a way around instability problems in situations where less gain in the chip-amp is not a major issue.