[sdiy] Henry, 256 Bw limited!

[thx1138]

On 12/27/09 9:44 AM, "Antti Huovilainen" <ajhuovil at cc.hut.fi> wrote:

> On Sun, 27 Dec 2009, karl dalen wrote:
> 
>>> Grant Richter <grichter at asapnet.net>
>>> It has always puzzled me, but the number of bytes in the
>>> waveform is involved in the tone color.
>> 
>> I assume that the more bits you insert the smother/mellow it will sound
>> up to a certain point of bits. And as pointed out a square wave of any
>> bit width sound the same but every other wave does not.
> 
> It sort of works like that, but not exactly.
> A simplified explanation is that an abrupt transition in level, as happens
> for every sample when the wacetable oscillator has no interpolation, has a
> spectrum that falls off at 6dB/oct asymptotically. This falloff starts at
> the inverse of the transition repetition rate (this would frequency *
> wavetable length / 2 = nyquist frequency).
> 
> Below that point, the spectrum is determined by the wavetable contents.
> This way you can have both very sharp sounding waveforms (square,
> sawtooth), and also mellow sounding (sine) when the filter cutoff is below
> the nyquist frequency.
> 
> As there is no interpolation, the spectrum detail gets mirrored above
> nyquist (you can work out the maths, but it's generally not worth it).
> Take simple 8 sample waveform as example: The harmonics are a1, a2, a3 and
> a4. The resulting spectrum is then a1, a2, a3, a4, a4, a3, a2, a1, a1, a2,
> a3, a4, a4, a3 ...
> 
> Superimposed on that mirrored spectrum is the 6dB/oct falloff rate. There
> is also some additional falloff near multiples of nyquist due to the
> step-like "interpolation" (the official term is zero order hold). The
> effect on sound is minor when the waveform is long as ear is not very good
> at determining exact spectrum shape at high frequencies.
> 
> 
> Now for the real reason some wavetable synths sound so gritty:
> PPG waveforms were simply miscalculated. If you take normal 128 entry 8
> bit sine, it really sounds quite clean as long as filter removes the
> mirrored harmonics. PPG sine on the other hand evens looks gritty when
> the wavetable contents are viewed. This is either due to a bug in the
> software they used to calculate the wavetable contents or deliberate
> design decision. I'm betting on "happy accident".
> 
> Antti
> 
> "No boom today. Boom tomorrow. There's always a boom tomorrow"
>  -- Lt. Cmdr. Ivanova
> _______________________________________________
> Synth-diy mailing list
> Synth-diy at dropmix.xs4all.nl
> http://dropmix.xs4all.nl/mailman/listinfo/synth-diy
Hi All,

The original PPG samples were 8bit later on 12-bit. The first 16 bit samples
were done using a Sony PCM F1 ECL based A/D Front end and 16-bit D/A ECL
back end. This is one reason the samples sounded so Gritty early on. When
we designed the 16-HDU unit we were able to make a great deal of
improvement.

We opted to keep the SSM 2044 filters as they sounded great and we did not
have so much Analog design issues.

When we switched over to the MC68000 based design on the Realizer we also
used a TMS32010 DSP for Filters,oscillators etc. As this unit never really
made it to production, the end of PPG came soon.

Wolfgang and I never agreed upon a grand plan and as a result the best I
could do was at least get the HDU, PRK, EVU and PPG 2.3 into solid
production. 

Much of the software was chaos and undocumented. A real fright to make
changes and improvements but things worked well enough to make some really
fun sounding products that are still great to listen too.

I have made my own PPG incarnation using DSP56307 and MC68340 but never took
it too production. It has sinced been scrapped and re-used for other
projects and will never rise again I am afraid.

I want to do this project on an ARM11 and DSP5672x based machine under Linux
and will cobble a project together perhaps when I get my other projects
finished.

Best regards,

Terry Shultz