[sdiy] MOTM-800 clicks. Arg!

[jhaible at debitel.net]

> I have heard this fast attack problem on many synths, also digital.
> I guess all such configurations show the same behaviour, if the
> envelope is really fast enough.

Yes. It can be explained directly from signal theory, and would
even occur in perfectly ideal modules.


> Another question: we could ask what kind of thing this VCO-VCA-Env
> thing is modelling. If you create sound by applying a impulse to 
> a 2nd order resonator, you will see that the rise time depends
> on frequency and resonance. Perhaps a very simple ideophon
> model. It will never happen that the wave starts at 100%, the
> differential equation will always start with 0 (given the initial
> conditions say so). I've just done this for a composition
> I'm going to present this Saturday or Sunday.
> 
> The usual synth circuits divide the action
> of a resonator in two parts, frequency and time transient,
> which are not correlated. Thus it is not surprising that things
> do not work as supposed (in the sense of said modelling).

That's a very interesting way too see it - I never thought about that,
but it makes perfect sense. Maybe that's also the reason why synth
that only have a single envelope generator (shared by VCA and VCF)
often produce a more "natural" behaviour. Of course that's a big 
limiation as well - every time I played with a 1-ADSR synth, I wished 
I had a second ADSR. But often, when I _have_ 2 ADSRs, I wish the
2nd would "follow" the first one somehow. Would be easy to implement
with two VCADSRs: One set of knobs for common A. D, S, R values,
second set of knobs for (relative) offset A, D, S, R values on the
second EG.


Linked parameters can often be very nice. I've been playing with
VC ADSR parameters on the PolyKorg Clone a lot recently. I think
with full polyphony available, the behaviour of release becomes
even more important. (No need to model everything exactly, when
the release is cut off 8 notes later, anyway.)
I found that the usual prolonging of the Release time for
a "damper pedal" effect can be much improoved if the Decay
time is adjusted accordingly as well. But I disgress.


Back to the Attack click problem. Many non-modular synths have
a small time constant (RC-filter) built-into the VCA, in order
to limit the transient speed of the EG. With a nonlinear VCA
CV input, this would even be different from ordinary 
filtering, and different from ordinary slew limiting as well.
Just think of the famous Vactrol-based VCAs!
On a modular, you want more degrees of freedom, and you want
to _allow_ sound that sound unnatural. This can mean your
emvelopes can be faster than you really want them for a certain
application. Referring to Martin's example of a 2nd order resonator,
no think of a "real" bass drum, for instance. If you hit the base drum,
the hitting may produce a transien that is faster than the damped
oscillation after the impact, but _how_much_ faster can it possibly
be? (on a real bass drum?). Will certainly depend on the strength
of the impact, damping from the beater's material, and so on.
But will it be as fast as the 1 ...2ms of a "good" synth ADSR?
1ms coresponds to a quarter-cycle (i.e. zero-to-maximum) of a 250Hz
resonator. But this zero-to-maximum excitation will only happen
when you really synth your VCO to the ADSR such that it starts with
a zero-crossing. Think of it: If your ADSR strikes at the VCO
wave _maximum_, whis coresponds to a drum membrane that is sort-of
"plugged". If you have a 1ms transient, it's like plugging a
membrane that's tuned to 250Hz (approximately).


JH.


-------------------------------------------------
debitel.net Webmail