[sdiy] They aren't sawtooths, they're ramps

[David G. Dixon]

Damian, I don't think this is necessary, actually.  I believe that all this
effect does is offset the triangle peak and the comparator switching by a
few milliseconds.  I don't believe that it has any effect on tracking, at
least below many 10s of kHz.

After looking at the simulations a bit more, it turns out that the LM318
also has a fairly long period of gradual change before switching.  It is
just much less noticeable.  I think this is just the nature of the beast.

Also, I misspoke before.  It's not the LM318 comparator output which is
spiking above/below the rails, of course, but the reference voltage at the +
input.  This is supposed to be +/-5V, but the + feedback cap is sending
spikes into it which raise/lower it to about +/-15.6V.

If I put back-to-back 5.1V zeners to ground on this input, the spikes almost
disappear, of course.  My big concern with the spikes is that they take
about 25us to settle back to the threshold level of +/-5V.  However, this
would only affect the triangle amplitude above about 40kHz, so I suppose
that it's not really a problem.  I'm not sure if it justifies the zeners,
but I've made provision for them in my layout, so the option to use them is
there.

Also, incidentally, I discovered that putting a fast switching JFET on my
saw shaper virtually eliminates the mid-saw glitch.  The best one seems to
be 2N5485.  2N4391 still makes the glitch in simulation.  I had J201 in
there from a very old simulation, and the difference is "chalk and cheese".


> David,
> couldn't you have something that would 'exaggerate' the part that
> triggers the op amp?
> 
> Say, you could have a hard-clipper based circuit that amplifies the
> signal that goes out of the bounds allowed by the clipper - like you
> can do in a normal wave shaper. You could amplify by 10 times. Maybe
> you can make that clipper very precise so that it only creates big
> spikes on the very top and bottom of the triangle wave. In that case,
> wouldn't those spikes give enough voltage difference for the switch to
> flip quickly? 'speedup spikes' as they were..
> 
> D.
> 
> On Fri, Nov 6, 2009 at 04:33, David G. Dixon <dixon at interchange.ubc.ca>
> wrote:
> > John, it's not really like that.  When the triangle approaches the
> > threshold, the trigger output starts to fall very gradually.  In fact,
> it
> > takes several microseconds to fall a few millivolts.  It's as if the
> > threshold is not very sharp.  Once the fall or rise in trigger output
> > voltage is noticeable, then it curves downward about 0.5V over about
> 150ns,
> > and then makes the final transition through zero very quickly, in less
> than
> > 50ns.  I'm not sure how much all of this contributes to tracking errors
> -- I
> > think it still tracks very well.
> >
> > On the other hand, the LM318 trigger is very sharp.  Of course, with a
> > speedup cap, it spikes well above and below the rail voltages, but I
> don't
> > think this will have any adverse effect.
> >
> >
> >> >The discrete version takes about 200ns to switch.  It takes about
> 150ns
> >> to
> >> >make up its mind, but then the actual switching takes place very fast,
> >> >within about 50ns.  However, it doesn't begin switching very swiftly.
> >>
> >> Can you speed up the "make up its mind" part by feeding some bias to
> >> the transistor so that it's closer to the switching threshold?
> >>
> >> Keep in mind that I'm virtually clueless. (Alright, Harry -- I'm
> >> totally clueless! ;-)
> >
> > _______________________________________________
> > Synth-diy mailing list
> > Synth-diy at dropmix.xs4all.nl
> > http://dropmix.xs4all.nl/mailman/listinfo/synth-diy
> >