Re: [stella] Re: 2600's TIA

Subject: Re: [stella] Re: 2600's TIA
From: Kevin Horton <khorton@xxxxxxxxxxx>
Date: Mon, 27 Aug 2001 22:44:07 -0500
At 16:28 8/26/01 -0400, you wrote:


> applied voltage, hence, affecting its speed. The way the NES does it (And
> TV boy, using a Nintendo-patented technology... tisk!) is use a 6x
> colourburst and run it thru a 6 bit walking ring counter, then trigger on
> both the rising edge and falling edges of the clock to get 12 different phases.

By 6x, I assume you mean clock rate?  That would be the logical way to do
the TIA, but then it wouldn't be a drop in replacement.

Yes, 21.4727272Mhz. The TVBoy and NES both use this. Interestingly, this exact same circuit is used to generate the required waveforms when in PAL mode. Only there, the units are crystalled at 26.601712Mhz which is exactly 6x the PAL colourburst frequency.

I was thinking about this, and I had a better way of generating the signals. They could use a divide by 6 counter and use the clock line input itself (it must be 50/50 duty cycle for this to work) as address lines to a small ROM lookup table.

The CLK input line would be A0, and the divide by 6 would generate A1,2, and 3. the ROM would be 12 bits wide, and 12 words deep. Each bit line could be used to generate the required phases by "storing" a square wave (6 0's then 6 1's in a row); each off by 30 degrees (or exactly 1 ROM location).

This same idea could of course be condensed down into logic, too. This is more likely than a lookup table which is "expensive" in regards to chip real estate.

Yeah, the full CMOS version will probably have an RGB option.  But at this
point, I have no idea how RGB works.  I had assumed that only the intensity
of each color was encoded, via voltage.  But I once scoped out the RGB pins
on my video card and saw an ac wave superimposed on each signal.  So I gave
up for the time being.  Do you how it's supposed to work?

If you're seeing a large AC signal, that tells me you didn't ground your scope, so you're seeing the AC line on it due to the open ground loop. Fix that and check again.

Anyways, There's a couple types of RGB. There's RGB+S which is RGB and separate synch, then there's RGB with synch on green. It is just that- the synch info is on the G line. The synch is of course the standard old synch as used on NTSC. long ~60Hz for vertical and short ~15Khz for the horizontal. There is of course no colour burst or phase information on the waveform- it is a purely analog signal; the brightness of the colour is directly related to the signal level. More voltage = brighter. I dunno if black level is sent like on NTSC; if it is, then it would work the same.

- Archives (includes files) at Unsub & more at

Current Thread