[sdiy] Better testing techiques for audio circuits

[Phil Macphail]

On 4 Apr 2012, at 19:03, David G Dixon <dixon at mail.ubc.ca> wrote:

>>> However, I find that it does capture a lot of the common nuisances, 
>>> like spurious oscillations, etc, and it does predict the 
>> THD of sine 
>>> shapers and things with remarkable accuracy.
>> 
>> I'd be interested in how you confirmed the accuracy.  Did you 
>> run the real circuit into a distortion analyser?
> 
> Yes.  Multisim predicts the THD of a sine shaper when the transistors are
> perfectly matched (as Harry mentioned earlier, the SPICE models of
> transistors are ideal).  In real life, this tends to be the minimum
> attainable THD.  You can turn on "Use component tolerances" but this doesn't
> always give very realistic results.  My attitude is, design for the ideal
> case, but ensure that you can live with the variances, and use precision
> components when you really need them.  You get an interesting perspective on
> component tolerances when you tune and calibrate 120 VCOs!
> 
> In any case, I think that absolute values are meaningless with this sort of
> stuff.  What I'm looking for are relative values (for component
> optimization, for example), and Multisim (or any SPICE model) 

It should be remembered that SPICE simulates in the time-domain, while THD is a frequency-domain term. With the non-linear time-stepping used in modern simulators this transformation can be 'problematic', so the precision of the result will be orders of magnitude less than the resolution.
Intermodulation is handled well by standard BJT models and can be simulated at higher levels, but requires a translation to the frequency domain too... This all becomes quite trivial with harmonic balance, but that's another story,

Phil.