I am modeling an amplifier with Multisim using an EF3 OPS with 6 pairs of MJL3281A/MJL1302A driven by MJE15032/MJE15033.
With a 1KHz input signal and an output voltage of 34.6VRMS into a 4 ohm load (300WRMS), the Fourier analysis of the output node shows a THD of less than 0.0001% (9 harmonics) so the closed loop behavior seems OK.
I do a Fourier analysis of the VAS (the test node is the connection of the VAS collector to the bias spreader) to get an idea of the "correction" applied to the OPS.
With 0.2R emitter resistors, the VAS THD is 0.054% with a dominant 2nd harmonic (0.026%).
Changing the emitter resistors *only* on the PNP output devices (MJL1032A) to 0.26R (empirically determined) lowers the VAS THD substantially to 0.011%, now with a very low second harmonic. (<0.0002% = much better symmetry)
The bias is set for best THD (approx. 70mA).
I thought that something may be wrong with the models but I also tried MJL21194/MJL21193 with similar results.
Looking at the data sheets, I see that these PNP devices tend to have a higher current gain which seems to explain this.
Your thoughts and/or experiences please.
With a 1KHz input signal and an output voltage of 34.6VRMS into a 4 ohm load (300WRMS), the Fourier analysis of the output node shows a THD of less than 0.0001% (9 harmonics) so the closed loop behavior seems OK.
I do a Fourier analysis of the VAS (the test node is the connection of the VAS collector to the bias spreader) to get an idea of the "correction" applied to the OPS.
With 0.2R emitter resistors, the VAS THD is 0.054% with a dominant 2nd harmonic (0.026%).
Changing the emitter resistors *only* on the PNP output devices (MJL1032A) to 0.26R (empirically determined) lowers the VAS THD substantially to 0.011%, now with a very low second harmonic. (<0.0002% = much better symmetry)
The bias is set for best THD (approx. 70mA).
I thought that something may be wrong with the models but I also tried MJL21194/MJL21193 with similar results.
Looking at the data sheets, I see that these PNP devices tend to have a higher current gain which seems to explain this.
Your thoughts and/or experiences please.
Last edited:
This is probably partly just a cancellation effect - check what happens at a different power level which uses a different part of the OS response curve. If you can plot the wing spread diagram this will tell a fuller story - perhaps the two sides of the wing spread diagram are more balanced with the Re fix?
Last edited:
Basic math never lies. VI*=watts. But the subtleties implied by nonlinear models need to be taken with the same grain of salt as CNN and Fox. To model an amplifier with that level of fidelity (accuracy) requires that the model fit be verified against real data - including process spread. And a very complete model for the entire amplifier, including the layout, power supply model, component mismatching, and any significant thermal or EM effects (half-sine currents magnetically coupling into your feedback resistor can easily produce more distortion than the minimums you report from your sim). A simulation is a very useful and insightful guide, but they do have a “noise floor” which takes real effort to extend.