D
Deleted member 375592
It could be of interest to some that crossover distortions in class-AB amplifiers can be relatively easily simulated and analyzed in the MATLAB - LTspice pair. Doing the same in h/w is way harder, I'd say.
I published the results of my studies on the MATLAB file exchange: Simulation of Crossover Distortions in Class AB 0.00002% THD - File Exchange - MATLAB Central
I apologize in advance: MATLAB is not free. Since 2016 MathWorks sells home licenses (strictly non-commercial use) for a reasonable price, currently $149.
I would like to ask for comments and advice - I surely missed a few important somethings and made lots of mistakes.
I published the results of my studies on the MATLAB file exchange: Simulation of Crossover Distortions in Class AB 0.00002% THD - File Exchange - MATLAB Central
I apologize in advance: MATLAB is not free. Since 2016 MathWorks sells home licenses (strictly non-commercial use) for a reasonable price, currently $149.
I would like to ask for comments and advice - I surely missed a few important somethings and made lots of mistakes.
Python's numpy, scipy.signal, matplotlib libraries between them probably provide all the functionality you need for this, and are free and very well supported. And Python is a useful language to learn for many purposes, not just signal processing / matrix ops. Also IIRC matlab indexes from one
rather than zero which drives most programmers crazy (!)
rather than zero which drives most programmers crazy (!)
D
Deleted member 375592
there are also free octave, scilab, etc
yes. all of them use indexes from one. I am not sure about driving most programmers crazy: that implies that they were perfectly sane prior to exposure to matlab which is a rather debatable proposition.
for those who are understandably lazy to register on matlab file exchange, .pdf attached
yes. all of them use indexes from one. I am not sure about driving most programmers crazy: that implies that they were perfectly sane prior to exposure to matlab which is a rather debatable proposition.
for those who are understandably lazy to register on matlab file exchange, .pdf attached
Simulation of crossover is very hard to get realistic. Generic transistor models just don't capture charge storage, current gain at low current, Ft changes etc well
D
Deleted member 375592
yes and no. I did not use oversimplified generic models. But there is always a danger to get yourself immersed into 2nd and 3rd order details while missing 1st order effects.
The much bigger problem I run into was a need for very low .TRAN tstep, 100MHz to satisfy Parseval's theorem requirements, setting adequate tolerances and thresholds, and understanding how to check that the results are not just a numerical artifact. I still do not understand how "normal" EE can use spice-like tools, without deep knowledge of numerical methods and DSP.
The much bigger problem I run into was a need for very low .TRAN tstep, 100MHz to satisfy Parseval's theorem requirements, setting adequate tolerances and thresholds, and understanding how to check that the results are not just a numerical artifact. I still do not understand how "normal" EE can use spice-like tools, without deep knowledge of numerical methods and DSP.