[sdiy] Midi2CV accuracy(was MIDI2CV alternatives)

[Richard Wentk]

At 16:41 06/10/2004 +0200, Maarten Halmans wrote:
>Hi all,
>the accuracy and midi2converter stuff is keeping me busy for some time
>now.....What is from a musical point of view an acceptable accuracy when
>working with polyphonic midi2cv converters? Very small differences in pitch
>between two sources are audible, a difference 0.5 Hz at 440Hz is audible
>right?

Not very. You'd get beating at 0.5Hz, which most people wouldn't find too 
obviously objectionable. 1Hz would start getting problematic.

It's usually assumed that for the mid-range, human pitch discrimination can 
hear differences of around 2Hz. But that's a monophonic situation playing 
two discrete frequencies. When playing polyphonically 0.5Hz would be a 
comfortable kind of resolution to aim for. But you'd more usually see that 
specified as a ratio in cents, rather than as an absolute frequency.

>So the desired accuracy is even better than 0.1% or am I missing something?

It depends how picky you are. It's usually assumed that hardware with 1 
cent accuracy - that's 1% of a semitone - is exceptionally good. 1% of a 
semitone on a 1V/oct converter is 833uV, which is a fairly relaxed kind of 
spec and seems to be adequate even for polyphonic use.

The main advantage of using more bits and higher resolutions is that you 
can do things like break out of equal temperament and work with custom 
tuning schemes.

In my experience 1% semitone accuracy is fine for ET polyphonic use. (Note 
that we're talking about the accuracy per step here, from every possible 
error source, not a nominal DAC specification.) But there are good reasons 
for using more bits. One is that famous extra accuracy. Another is that you 
can sum in pitch bend and other modulation sources internally and send the 
result to a single DAC. This is less analogue design and trimming hassle, 
and sometimes also cheaper than using multiple DACs.

Some of the commercial converters include digitally generated envelopes and 
LFOs that can be added to each voice, and this is always easier to do 
digitally than with multiple DACs of lower resolution.

Richard