[sdiy] Question for those with musical ears

Mike Bryant mbryant at futurehorizons.com
Thu Apr 15 02:49:27 CEST 2021

Yes generating the frequencies exactly isn't the problem, it's what I can get away with in analysing an input signal into separate frequency bins as each bin adds more computation just to prove there's no signal there.

-----Original Message-----
From: Brian Willoughby [mailto:brianw at audiobanshee.com] 
Sent: 15 April 2021 01:43
To: Adam Inglis (synthDIY)
Cc: Mike Bryant; synth-diy at synth-diy.org
Subject: Re: [sdiy] Question for those with musical ears

On Apr 14, 2021, at 17:26, Adam Inglis wrote:
> On 15 Apr 2021, at 9:47 am, Mike Bryant wrote:
>> Consider a continuous glissando played on an analogue synth - the frequency rises smoothly between the start and end pitches.
>> But on a digital synth there will always be discrete steps between successive frequencies as the frequency is gradually stepped between the start and end pitches.  
>> Question is, what is the minimum step (in cents) needed such that the best musical ears cannot tell the frequency isn't rising smoothly as on the analogue synth, but in many discrete steps.  
>> It may be that the rate of change can effect it so if so please assume a glissando starting at middle C rising at 1 second per semitone.
> So, you're kind of talking about the auditory equivalent of a video “frame-rate”, are you? 
> As in, How slow can the frames per second get before you notice a ‘flicker’?

There are a few challenges here.

First, sound does not quite have the equivalent of "persistence of vision" ... there is not the same kind of filling in of the gaps.

Second, small changes in amplitude can be perceived as pitch changes, as well as the converse.

Finally, many of the good digital/analog hybrid synths designs (I'm thinking of the Matrix-12/Xpander) have analog resistor-capacitor filter networks that smooth out these discrete steps. If you have a VCO, the pitch doesn't have to step. One cool thing Oberheim did was make this filter switchable, so that envelopes can have instant attack when needed, but still get smoothed out at other times. A DCO doesn't have this option, but there are ways to synthesize pitches that are not discrete, and thus the glissando does not need to be stepped. A digital sine wave generator can smoothly drift between frequencies (although I suppose that a 64-bit float engine would have a step size of 0.000000000000376 cents if I did the math correctly).

That said, I seem to recall that a 16-bit DAC is good enough that a human won't be able to detect the pitch error. Problem is that there's a big difference between not being able to discern the difference between two similar frequencies, and not being able to hear the step as the frequency changes in a single oscillator.

I'm sure there has been some psychoacoustic research into this exact question. Have you tried search the library of the Audio Engineering Society? (AES)


More information about the Synth-diy mailing list