[sdiy] Interesting blog with Roland Alpha Juno details

[rburnett at richieburnett.co.uk]

The DCO sawtooth waveform definitely has 256 discrete levels...  I 
checked an old recording I had of a very low-pitched sawtooth waveform 
with the VCF wide open.  There's quite a lot of background noise as 
you'd expect from a digital/analogue hybrid of that era, but by 
averaging consecutive cycles and differentiating the recording you can 
clearly see 256 evenly spaced steps along the ramp's slope.  The 
frequency spectrum also clearly shows every 256th harmonic of the 
sawtooth waveform is missing, which is symptomatic of a sawtooth that's 
been quantised to 256 discrete levels.  (The quantisation "noise" or 
error is essentially a sawtooth waveform at 256x the note's fundamental 
frequency, but with opposite polarity, and with 1/256th of the 
amplitude, so every 256th harmonic gets cancelled out in the resulting 
sawtooth spectrum.)

You have to play a note below 78.125Hz before this artefact falls below 
20kHz though, and most bass patches have a filter that's closed to some 
degree so you probably wouldn't notice this artefact unless you 
specifically went looking for it :-)

Proving what the sample-rate (clock rate) of the NCO is by looking for 
aliasing will take a bit more effort.  I know I've seen evidence of 
aliasing in the spectrum of very high-pitched notes on the Alpha-Juno 2 
before, but don't seem to have any recordings immediately at hand.

-Richie,



> On 2024-11-01 17:24, Tom Wiltshire wrote:
>> So...can you have a guess what resolution the sawtooth output is?
>> 
>> The clock to the DCO chip is the same 12MHz clock used for the 8032 
>> microprocessor. So if it's 12MHz sample rate, it's not half bad! It's 
>> like an early version of the Novation Peak!......