I have an amplifier I'd like to do a LARGE signal frequency sweep on...  to ideally get the gain and phase response.   Anyone have a suggestion on how best to do this?

Thanks.

Need a way to do it in simulation.    I dont understand your points.   The signal shape will be the same...  say a 5Vpp sinewave...   just need to get gain and phase vs frequency...    like a network analyzer would do in th lab.

??   Hmmm.... ok.   well- how about we ASSUME that my circuit doesnt distort the large signal and that the output sine wave is still a perfect sinewave?  Then any ideas?  Are you telling me that people who design audio or other power amps only do smalll signal bandwidth analysis?  I find that hard to believe.    thoughs?

kevinvinv wrote on Aug 18^th, 2009, 5:32am:


I suppose, the problem in fact is not a problem.
Kevinvinv, if you assume or expect that the output is a perfect sinewave, you are speaking of a "small signal" analysis.
What means "small" ? 1 volt ? Or 5 volts ?
There is no voltage limit ! "Small" means only, that the small signal analysis is applicable because distortions are not yet relevant.
And this can be very different for different circuits. A transistor without negative feedback allows perhaps only 10mV without remarkable distortions - and an opamp with feedback perhaps 5 volts.
Problem solved ?

kevinvinv wrote on Aug 17^th, 2009, 2:09pm:


Hi Kevin,
There isn't a good way to do this that I'm aware of. The only thing I know to to is put in a large sin wave, measure output gain and phase, repeat for the next frequency. You probably want to automate this... There is a swept sin source, but how fast do you sweep it before you aren't measuring what you really want to? For distortion measurements, I never measure before the fourth cycle of the sin wave. You could get by with a quicker measurement here, but I do not recommend the first cycle.

This is especially a problem with circuits that use active biasing (bias depends on signal level). Try running a sin wave near the full power bandwidth through the circuit and simulate for a dozen cycles. I've often had circuits that decrease in amplitude over the first few cycles, sometimes over the first 6-10 cycles. If you just simulate one or two cycles, you can miss this effect (you are in effect simulating a burst, not a 'continuous' signal). Guess what you measure on the bench when your silicon comes back. Some biasing schemes can sustain for a brief period but after several cycles, errors build and reduce the slew rate.

You've gotten some theoretical responses so far; mine is a practical one, just like you'd measure it on the bench. Even though you are measuring in the large signal region, your output will be pretty much a sin wave. Bandwidth will decrease as you approach slew limit. When the waveform is really distorted, you're not measuring anything useful, but that's past where you're interested in. Remember, when the circuit hits slew limit, the AC response drops like a rock, not like a one pole circuit.

Steve Smith
formerly of Comlinear Corporation