[sdiy] Arpeggiator1 / Arduino UNO

[Nantonos]

Hello Rick,

Friday, April 10, 2015, 11:24:40 AM, you wrote:


>> On 10 Apr 2015, at 11:09, Nantonos <nantonos at epona.net> wrote:
>> 
>> Hello Rick,
>> 
>> Friday, April 10, 2015, 10:26:36 AM, you wrote:
>> 
>>> No idea, and I don't care :-) All I want is 12 discrete voltages per octave.
>> 
>> You mean that they are going into a quantizer, so you don't care about
>> accuracy as long as they are distinct?

> Sort of, yes.

>>> As far as I can tell, without actually measuring, the voltages are
>>> reached quick enough, and are repeatable.
>> 
>> Filtered PWM is a trade off between lots of ripple (but fast) and
>> little ripple (but slow and slewing). Without either a measured result
>> or a theoretical analysis of errors, the actual response is just
>> guesswork.

> Well, of course I looked at the oscilloscope, to check if the
> ripple had gone. With the 480 Hz pins there was still a ripple, with
> the 980 Hz pins none is visible, on my old russian scope.

Ah okay. I wouldn't describe that as "without measurement" :)

> It's not guesswork entirely, the 10uF/1k resistor result in an RC
> time of 10ms for rough settling. That order is good enough for note
> playing, I think, so far.

>> I would be surprised if the ENOB is even as high as 12; probably more
>> like 10. Which is fine if it suits the purpose, but describing this as
>> "16 bit" is misleading for anyone else thinking of building it.

> That is why I labeled it "~16 bit DAC" :-)

My point is that ~10 is more likely.

> Even an 8-bit (ZN426) DAC would have been good enough, it will give
> you 256 discrete voltages.

Yes, if an error of +/- 25 cents is ok then a theoretically perfect
8bit DAC will suffice.

Then you learn about INL, DNL, zero offset, gain error, thermal drift,
etc. At which point the wisdom of saving $5 comes into question and
you get a very good 12bit DAC or an okay 16bit one.


-- 
Best regards,
 Nantonos                            mailto:nantonos at epona.net