There is no such thing as 'pulse width' for saw waves. Only psuedo pulse widths, which each are different implementations from differing manufacturers.
I discussed this in an earlier thread (with pics and audio!), regards the saw pulse-width feature of the Korg Z1/Moss synthesis:
http://www.sunesha.nu/virusforum/viewtopic.php?t=2197
Came to the conclusion that Korg used a simple "chorus/phasing" type effect to give the saw-oscillator "motion", instead of literally altering the pulse width of the saws. Regardless the effect sounds infinately fatter than static de-tuned saws.
How did the Juno do it?
However, I also wonder what these
other types of pulse-width processing would sound like on saw waves:-
(..with graphs (1) and (2) being different, depending what positive and/or negative orientations are used. The last one (3) is just a variation of (2). You could say that (1) is a digitally-clipped saw wave. - Increase the gain, and you'll clip it further (making the saw "pulse" smaller in width), while decreasing the gain to zero will return to the original saw wave again. This could not be used as the implementation for (2) or (3), though.).
I think that these could be true pulse-width of saw-waves in the logical sense of the definition if there was one.
This is the sort of stuff that I'd like to see more of in synthesis. The more oscillator mangling there is AT SOURCE, the more diversified and dynamic the output will be.