[sdiy] 1v /oct with ADC question

[mskala at ansuz.sooke.bc.ca]

On Wed, 30 Nov 2016, Vinicius Brazil wrote:
> However, in order for the ADC deviation not to be perceived as detuning,
> this deviation has to be less than 1%.
>
> (1/12) / 100 = 0.0008333volts

Bear in mind that this is an ADC, measuring the voltage from a pot.  It's
just supposed to know which semitone the pot has been turned to.  I don't
think the original poster really needs the boundaries between those
semitones to have single-cent accuracy.  Even if it were a DAC generating
a control voltage, single-cent accuracy seems excessive; very few VCOs are
capable of tracking anywhere near that closely, and making the voltage
input much better than the VCO is wasted effort.  Sub-millivolt precision
like that will also be lost at the first ordinary (not expensive special
low-offset) op amp it goes through, anyway.

However:

> 5 Volts / 0.000833 Volts = 6000 steps, so you need a 13-bit converter.
>
> 12 bits is already satisfactory.

If you *do* need single-cent accuracy, 12 or 13 bits in your converter
isn't going to be good enough by itself, because ADCs and DACs are
frequently not perfectly linear.  A voltage that ought to correspond to
the number 5000 might read out as 4997 or 5003.  The specification for how
many counts the value might be off depends on the individual ADC chip and
often gets worse when there are more bits.  Not all 12-bit converters are
equally linear.  If it matters to you, there's no solution except reading
the fine print on the data sheet carefully and making sure you know its
consequences in your application.

-- 
Matthew Skala
mskala at ansuz.sooke.bc.ca                 People before principles.
http://ansuz.sooke.bc.ca/