Re: [stella] TIA Audio Polynomials

Subject: Re: [stella] TIA Audio Polynomials
From: Adam Wozniak <adam@xxxxxxxxxxxxxxxx>
Date: Sat, 22 Nov 2003 14:17:17 -0800 (PST)
On Sat, 22 Nov 2003, Eckhard Stolberg wrote:
> How did you get this frequency? I opened up my NTSC VCS,
> and the clock crystal reads 3.579575 MHz. So assuming that
> the TIA produces two samples per scanline, that would be
> one sample every 114 TIA clocks. This should give a sample
> rate of almost 31400 Hz. My PAL VCS is only running at
> 3.546894 Hz, so the sample frequency should be lower too.

I measured it.

I sampled C=4, F=7.  That should be divide by 2, divide by 8.
Frequency analysis (audacity, view/plot spectrum/spectrum/512/hanning)
of the sample shows a peak at 1966 Hz.
1966 Hz * 2 * 8 == 31456 Hz

Any number of things might have affected this.  Temperature is a factor,
both in the console and the audio card I used to sample.  Clocks
on sound cards generally aren't really all that accurate.  I've
seen and measured sound cards off by anywhere from 0.05% to as
high as 3%.  Percent difference here is 0.09%, so it's well within what
I might expect from a modern sound card.

> > C=6
> > divide by 31, with an average 52% duty cycle
> Is that 16 HIGHs and 15 LOWs, or 15 HIGHs and 16
> LOWs?

16 high 15 low.  Duty cycle is usually given as the percent time spent
on or high.

> > C=7
> > LFSR [5 3] (tia doc says 5 bit poly / 2; this is incorrect)
> > C=9
> > LFSR [5 3]
> So you are saying that C=7 and C=9 are actually identical?

The TIA doc says they should be different, but I cannot hear a difference in
my sample, nor can I see a difference in the sample values.  As far as I can
tell they are the same.

> > C=11
> > set last 4 bits to 1
> Do you know what it is that has it's last 4 bits set to 1?
> It seems that the output of C=11 is all HIGH just like C=0.

C=0 and C=11 difficult to do anything with, really.  My short program
ran each mode for one second.  That may as well be a flat signal with a
constant DC bias.  Since the sound card filters out any DC components,
they both just look like a brief pop followed by silence in my sample.
I imagine Berzerk voice uses one of these, but I'm not sure which.

Will code for food.
"The dinosaurs are not around today because they did not have a space program."
  -- Arthur C. Clarke

Archives (includes files) at
Unsub & more at

Current Thread