Pokey's pseuso sine wave

[Irgendwer]

[Edit: Sorry for the typo in the title, maybe a moderator is able to apply the change to "pseudo"...]

 

This is nothing new, but I found the principle behind this sound generation nowhere explained, so I explored the subject by myself - maybe it's of some interest.

 

Back in the 80's, there was this German "Atari music programming language" production called "Masic".

 

http://www.atarimania.com/utility-atari-400-800-xl-xe-masic_12533.html

 

AFAIR on the second side of the disk was a sample song, using a kind of Pokey sine-wave like voice I never heard before or after again:

(Jump to 2:21!)

 

 

Just inspecting the Pokey registers, I discovered the following:

The first and second voice are configured by AUDCTL to be timed with 1.79 MHz.

Then the frequency selection utilizes the beat effect https://en.wikipedia.org/wiki/Beat_(acoustics) to produce the audible tone. Ideally the both carrier frequencies are out of the human perceivable frequency range, but forming the desired sub-frequencies you have to trade. Additionally, coming from a square wave, you aren't able to construct a perfect sine, but have some jitters. Both effects are recognizable in the demo tune above.

 

I've create a small Turbo-Basic test program, demonstrating the effect:

10 REM PSEUDO SINE-WAVE POKEY OUTPUT
20 BASEFREQ=62:PFRQ=1773447
30 POKE 53768,96
40 POKE 53761,173
50 POKE 53765,173
60 POKE 53760,BASEFREQ
70 FREQ=BASEFREQ-1
80 ? ABS(PFRQ/BASEFREQ-PFRQ/FREQ)/2;"H
z"
90 POKE 53764,FREQ
100 GET KEY
110 FREQ=FREQ-1:IF FREQ>0 THEN 80
120 END

The "BASEFREQ" is the first of the two clock dividers. In general a smaller value will reduce the noise but limits the frequency selection, while a bigger one has the opposite effect.

 

Here an example of the "476.7331 Hz" wave:

 

Pokey output:

sample.wav.zip

 

Have fun experimenting!

Edited October 5, 2018 by Irgendwer

[Rybags]

Pretty sure that's what's known around here as the triangle wave.

Also possible is the sawtooth which also uses 1.79 MHz mode and the high-pass filter.

 

I also discovered like you (I might have done a spreadsheet) that the resultant frequency of the waveform produced is the difference between the 2 high pitched voices.

There's only a fairly narrow band of useful frequencies and most of the usable notes are 1 AUDF value apart but I think there are some 2 apart that are useful as well.

The values outside the useful range tend to include useless carrier noise.

 

As I put it, it's a sort of variable repeating PWM that's generating the waveforms.
They'd be somewhat more useful and include less unwanted noise if there was some low-pass filtering available external to Pokey.

[R0ger]

Yep, sadly there is only few usable frequencies.

[AtariNerd]

Pseus-pseus-pseudio! (Forgive me, but I had to do it.)

[makary]

Interesting. I believe Miker did some experiments using Music ProTracker and similar technique. The tracker has a really nice feature, a direct access to the frequency tables. So you can define there appropriate 'couples of frequencies' to get triangle/quasi-sine or saw-like instruments and use them within a tune.

 

[+DrVenkman]

Pretty sure that's what's known around here as the triangle wave.

 

I don't think he's referring to the component parts (which do indeed seem to be triangle waves), but rather the (for lack of a better term) pseudo-sine wave produced by the beat effect:

 

[Rybags]

Yeah, the resultant is closer to a sine than a triangle. But both wave types sound almost identical. If anything the triangle probably preferable since it's a bit less muted.