Hi!

There were several postings regarding module suggestions, which show
basically all for the same thing: a voltage controlled audio delay.

I experiment a lot with VC delays and similar techniques and would like
to share my experience (I write this while a boring conf call, I hope I
don't mix something up...):

We have to differ to kinds of fields of application:
1.)Delay in the sense of "echo".
The delaytimes begin with ca. 10ms and may end somewhere
above 100ms and or more than a second. The modulatable range has
not to be very wide: 75% to 150% of the base delaytime are
usually enough. This is already the maximum, the most delays
provide!

2.)Delays for Chorus and Flangers.
The minimum delay time should be quite short (0.1ms). The maximum
must not be longer than 50 ms. The modulatable range is enough
with 75% to 150%

3.)Delay for Resonance-based synthesises like Karplus-Strong.
The minimum delay time has to be very (!) short (0.01ms).
The modulatable range (especially for tonal synthesis
applications) has to be much(!) wider:
10% to 1000% of the base delaytime would be great.
25% to 400% are a must to have a useful tonal range of four octaves.

Also we have three different techniques to build an audio delay:

A.) BBDs (Bucket Bridge Devices).
The delay time is direct depending from the Clockrate. Very high clock
rates are technically not really feasible to my knowledge (this may be
wrong -> Dieter ). Low clock rates sound awful, not only because the
clockrate comes to the audiorange, but also because the hold time of the
single "buckets" is reached, and the signal is distorted heavily.
So to get a delay for Karplus-Strong stuff we would have to use only a
short Bucket bridged row. This would urge us to use low clockrates for
the longer delay times, which cannot be recommended. So BBDs are not
useful for Karplus Strong.
For Delay in the sense of Echo, BBDs have a too bad noise to signal
ration. Delays above 250 msec become worse and worse.
BBDs are nice and smooth sounding for Chorus and Flangers

B.) Digital memory based delay, A/D-converter has a constant clock:
This is the type most modern delays work like. The A/D works with a
constant clock. The change of the delay time is achieved mainly by
changing the size of the memory, which is used. If this change is done
while working with an audio signal, the audio signal will have spikes,
like a "jumping" CD. The old Roland delays simply muted the signal when
changing the delaytime. Very modern dsp-based delays can cover this by
smoothing algorithms, which are difficult to develop.
Small delay time changes can be achieved by modulating the reading clock.
This kind of Delays is good for "echo"-like delays and for
chorus/flange. It is unusable for karplus-strong stuff.

C.) digital memory based delay, one clock for ecerything (A/D, memory
reading, and D/A).
This is a quite old fashioned digital delay type. The memory size is
constant. All changes of delay times are achieved by changing the
clockrate.
The clock may come from a simple highspeed VCO (not a PLL-based one, but
a CEM3340 with small C-value).
The main disadvantage against the type B is, that the signal quality
gets worse the lower the clock frequency is.
The main advantage is, that the delay time can be changed in a quite
wide range. Also the change can be done in realtime without producing
spikes or similar unwanted side effects.
This is what is required for Karplus-Strong synthesis (or similar). Also
Chorus and Flanger are ok with this. "Echo" like delays ar not that
goood, because of the bad quality of longer delays (lower clockrates...).


So I am not sure, whether Doepfer simply can build one "delay"-module
for everything we usually subsume under "delay-effects".


Regards, Florian



--
Florian Anwander |ConSol
Tel. +49(89)45841-133 |Consulting&Solutions Software GmbH
Fax +49(89)45841-111 |Franziskanerstr. 38, D-81669 München
email: florian.anwander@... | http://www.consol.de