zipper noise was: digital control

[Martin Czech]

:::Here's a question... I haven't ever built a glide circuit, so I
haven't been 
:::able to try this yet, but if you put a glide circuit on
your DAC output, 
:::would it cover up the zipper noise?  I was figuring
this would work well in 
:::non-critical situations (like a filter sweep)

This zipper noise is really sampling sidebands. A lowpass filter can
damp them, glide is lowpass filtering/integration.

The 7 bit resolution is not a big problem for most purposes (except
frequency).

The problem is the steps and the limited MIDI bandwidth.

It would be a good idea to use a cheap 8 bit DAC (not for pitch of course,
depends on the total error of the dac, 1/2 LSB is too much when using
8 bit) and an analog network that interpolates between samples. For a
sweep just two points (samples) would be needed. The question is  how
the analog stuff can determine the time constant. Is it possible to
guess the needed time constant from the incomming midi data? I guess an
additional controller & dac would be needed.

A primitive implementation of that is a simple glide on/off controller switch.

A reasonable sequencer like LOGIC enables us to draw controller data
on the screen, so why not add some  controll voltage lowpass filtering
information.  I guess the "real" controller data can be sparse, ie.
only some points in time would be needed. The auxilary controller for
integration time is not often used, too. This would really improve the
MIDI bandwidth situation, which is a problem if one has many tracks with
much controller movement.  Some sequencers have an option to throw away
most of the controller data, but I don't know if this is done in a clever
way (never throw away minima or maxima). Of course this would allow for
all kinds of "lfo" and "envelope" curves, too.

A good idea to have moving modulation, instead of simple fixed lfos
running, which tends to be boring.

The auxilary dac could be a simple CMOS switch network, to switch in /
out fixed resistors for the first order lowpass. This way offset would
be no problem.

Perhaps 2,3, or 4 reasonable time constants will do the job, the zipper
noise will go away, and smooth movements can be achieved.

Of course, an industrial product would use a dsp and higher resolution
DACs.

I have never heard of a midi/vc converter with such a feature.

One problem remains: If you want to record real time controller
movements, this data has to be preprocessed in some way. E.g.
determining the important samples, adjusting slope time.

m.c.