[sdiy] SAW core VCO flyback time

[Mattias Rickardsson]

On 30 August 2016 at 19:21, Richie Burnett <rburnett at richieburnett.co.uk> wrote:
>> You can see the way that increasing amounts of flyback time affect the
>> harmonic structure on this page:
>>
>> http://electricdruid.net/timbral-evolution-harmonic-analysis-of-classic-synth-sounds/
>>
>> The bit you want is titled "Ramp-down/Triangle/Ramp-up" in the Phase
>> Distortion section abut half-way down.
>
> A very detailed analysis!

Yes, thanks for this! :-)

>> Up to about 10%, the effect is pretty much a soft roll-off of the higher
>> harmonics...

Very expected, as it looks almost exactly like a slightly
lowpass-filtered sawtooth.

> And a reset time that is 10% of the overall period would be a *very*
> sluggish sawtooth reset!

Not necessarily, it could also be a very high frequency for the poor
oscillator! When having a slow reset time, you have to decide how high
your oscillator should be able to go. For higher frequencies the reset
time starts to get very real. The sawtooth ending up a bit
triangle-ish isn't too odd here. 10% is not that extreme.

> ... The point where the high-frequencies start to get rolled-off (-3dB
> point) is at approximately half the reciprocal of the reset time.  So for
> reset times less than about 25us that puts the 3dB point beyond 20kHz where
> any audible effect on the spectrum is marginal at best.

This is an interesting observation, thanks!

> BTW, I think it's fascinating how the analogue designers strive to achieve
> perfection in the form of "instantaneous" sawtooth resets, low-distortion
> sinewaves, and symmetrical triangle waveforms without the little nipple at
> the peak.  Whilst digital synthesiser designers strive to model the
> imperfections of old analogue oscillators because they are somehow perceived
> to be the reason for their unique character.  Taken to the limit, analogue
> oscillators will become perfect and sterile, whilst VA models will inherit
> all of the "character" ;-)

"Virtual digital" vs. virtual analogue. ;-)

On 30 August 2016 at 19:40,  <mskala at ansuz.sooke.bc.ca> wrote:
> I was holding off on saying it, but if you're serious about analog design,
> do you even *want* a super-fast reset?

In fact no! I just want it fast enough to produce the different
waveshapes needed in a good manner. But speed isn't everything here,
other tricks could also do. For instance, when converting a sawtooth
to a triangle and wanting to avoid the glitch, it could be better to
clamp the output triangle wave at the top during reset instead of
fighting for the fastest possible sawtooth reset.

Btw, the sawtooth can also be clamped at the end while performing the
reset, take a look at Yamaha's good old VCO III:

http://therogoffs.com/cs80/manuals/CS80_Service_Manual/42%20-%20M%20Circuit%20ICs.jpg
http://www.cs80.com/vco.html

> It seems to me prioritizing that
> is going to force you to use wider-bandwidth op amps everywhere the signal
> is going; think harder about stability; think harder about EMI (both
> radiated and conducted into the power supply); and it's all going to be
> wasted anyway once someone runs it through the average audio patch cable
> that attenuates such frequencies.  I think if you're generating an audio
> sawtooth wave with significant energy above 100kHz (and probably lower)
> then your design may not be the best possible for its purpose.

Indeed - considering that the high end starts to fall a couple of
decibels around 25 us reset time, perhaps a good reset time for
full-range audio signals would be something like 10 us or so. Shorter
times and you just introduce EMI problems and other unwanted stuff.

/mr