02-11-2021, 08:42 PM
(02-11-2021, 07:51 PM)ural Wrote:(01-20-2021, 09:11 PM)mgd Wrote: Update2:
I just learned that the NerdSEQ uses 12 bit DACs for the CV outputs (apparently a design decision made a long time ago). Unfortunately that reduces the NerdSEQ's usefulness for microtonal scales and tuning as well as exotic tunings
are you sure? (micro)scale is pure voltage and cv out put is pure voltage. even now you can fine tune (out) each individual step of pattern. what is necessary - to implement save functioon of those "finetunings" as (micro)scales for sake of time.
I am sure. Here is why:
All CV the NerdSEQ creates comes from the DACs. The number of possible voltages and their values is determined by the number of bits the DAC does convert. To illustrate this assume we'd have a 2 bit DAC that creates voltages between 0V and 1V. 2 bit means we have 4 possible values, e.g. 0V, 0.333V 0.666V and 1V. There are no other voltages this 2 bit DAC could create.
Now with a 12bit DAC we have 4096 possible voltages (say 4000 for simplicity) and the output range spans 10V. Therefor we have 400 steps to cover 1V or about 2.5mV per step on average (assuming the DAC is perfectly linear which most DACs aren't). 2.5mV are pretty exactly 3 cents (1 cent is 0.83333mV).
Another way to come to this is here:
1V change equals 1 Octave change in tone and that is also 1200 cent. As outlined above we 400 possible values that we can represent which equally distributed means 1 CV value every 3 cent.
As Thomas remarked elsewhere in this thread and I think also in the manual:
Not all possible 4096 values are supported but "only" a bit less than 4000. That means that with perfect DACs we have a distance between possible neighbouring CVs of 3-4 cents. And this means that for a given tuning the worst case w/r to tuning error is somewhere between 7-8 cent (a bit less than twice the average distance).
The average error is smaller which makes it suiteable for a lot of scales. But w/r to microtuning there are limits as described above.
Just for completeness sake:
16 bit DACs would add another theoretical 16 values per (current) step and thus reduce the distance between neighbouring voltages to something like 0.2 cent. However it is my understanding that currently the internal representation of notes in NerdSEQ could not deal with that additional precision and and change may or may not be practically impossible within the current NerdSEQ HW (memory constraints).
Kind regards,
Michael