The 'All in one' Antic/GTIA or Antic/CTIA chip

[carmel_andrews]

i can't remember where i heard or read about it but i clearly remember seeing/reading or hearing that Atari were planning to do an 'all in one' antic/g/ctia graphics chip (therefore replacing the 'two chip' design)

 

Was i just imagining it or is/was there an element of truth in this

 

Additionally, i remember from the disk version of the hardware reference manual (Bob DuHamel) that it was possible to have Atari's working with 'multiple' Antic/gtia and pokey chips, apparently you could theoretically better Atari's graphics/sound capabilities by doing this, i.e higher rez., more colours, more pmg's/sprites, stereo sound and multiple sound channels etc

 

Would it be possible these day's to use the concept of a 'multiple' Antic/gtia and pokey chips in a one chip design for each... i.e 5-6 antic chips on one chip, 5-6 gtia chips on one chip etc, etc...again with added hardware capability like more colours, more sprites, own ram (64-512 K) totally reprogrammable Antic chip, programmable gfx resolutions (like the later MSX gfx chips), more sound channels, stereo sound/voice synth/adsr and all that

 

Also, going back to the other thread, the NUMEN demo, it made mention of a 'bug' in the GTIA chip which allowed for interlacing and more colours, in regards to these 'new gfx modes', what would be the highest possible resolution, 600/400, 480/320 or PC stylee rez (1280/1024) and what of the colour range, would that be limited to 512/4096 (ST/e/Amiga stylee) or PC stylee 16/32bit colour range...would it be capable of putting up more then 4 hardware PMG's (i remember a gfx type in demo published in Atari User that claimed to display 80 hardware sprites)...Also Page 6 published a couple of gfx type in demo's called doing the impossible, one of the techniques explored was 'Vertical screen splits' which the STe hasd built into in shifter chip, but this was a A8 type in graphics demo, could these other capabilities be possible on the 'undocumented' gfx modes

 

Additionally, i wonder whatever happend to the 'Apac' graphics mode (I believe it was capable of generating 4096 colours and all that)

 

Lastly, in regards to the GTIA, i recall hearing or reading that Atari didn't officially support the GTIA modes 9, 10 and 11, if this was so, what was the point of the GTIA chip in the first place (esp. if your not going to support it's additional capabilities)

[Rybags]

Here's some info on it. It is known as CGIA (Colour Graphics Interface Adaptor):

 

http://homepage.ntlworld.com/kryten_droid/...ari_hw/cgia.htm

 

The CGIA was designed to reduce the cost of machines that used Atari's ANTIC and GTIA chips. The GTIA section was address decoded to be at an address compatible with the games machines, so that was probably the priority. Perhaps the 800 series machines would have had this altered address and require software modifications? The CGIA chip made it to the prototype stage, but was never deployed commercially.

 

A Google Search on "Atari CGIA" (all words) also finds mention of it in the Atari 8 bit FAQ and there is a downloadable PDF datasheet at

 

http://www.retromicro.com/files/atari/8bit/cgia.pdf

 

 

I fail to see how multiple ANTICs could possibly improve the Atari's graphics - for starters the architecture of the 8-bit machine sees nearly every possible machine cycle dedicated to DMA on a widescreen "Badline", so 2 ANTICs could never co-exist on a system with memory running at the CPU clock speed.

 

The design of CTIA is such that it only really accepts non PMG data from one source, and that occurs with 3 bits of input data (AN0-AN2 pins). The 3 data bits cover the basic possibilities of 4 playfield colours, background colour, vertical blank, and horizontal sync with set or clear 40-character mode (hires).

 

The AN0-3 lines are sampled once per colour clock - in hires modes, the interpretation is slightly different, with a bit pair used to determine whether the luminance value of PF1 or PF2 is used for two pixels within the colour clock (with the colour usually coming from PF2).

 

More info:

 

http://homepage.ntlworld.com/kryten_droid/...ari_hw/gtia.htm

 

 

 

GTIA expanded on that capability by allowing two successive sets of inputs to be combined to allow either extended palette (GR. 10), lum only (GR. 9) or colour only (GR. 11).

 

 

The way I see it, integrating a second GTIA would be the easiest option, if only to allow an extra set of PMGs. Some sort of video mixing would have to occur with extra hardware. Collision detection between existing and extra PMGs would not be supported. Both GTIAs would presumably have to receive the exact same AN0-AN2 data, so that the pictures generated by both were in sync.

 

But by doing that, all we would achieve is adding an extra 4 PMGs to the system - but if we rely on ANTIC to do PMG DMA (as we do in 95%+ of cases), both GTIAs will be generating the same graphics for onscreen sprites, although positions, sizes and colours could be different.

 

Of course, there's always the option to load data manually with the CPU for either/both GTIAs.

 

Other graphics modes:

 

The A8 has a maximum resolution of half a colour-clock by one scanline. That is set in stone, no demo or game on standard hardware can or will ever get over that. What that translates to is a maximum of 352 pixels by 240, in widescreen, maxed out DLIST (actually 239 lines is maximum for hires modes due to the rolling screen bug).

 

There are programs and demos out which claim "4096 colours" - they achieve this effect by alternating through 3 frames. Each frame contains only Red, Green or Blue - so such modes rely on TVs phosphor "persistance of vision" and our eyes to perceive more colours.

The most common variant just uses 3 seperate Graphics 9 images.

 

The mode mostly used in NUMEN, I believe, is Graphics 10 - the well documented VSCROL "feature" is utilised to generate square pixels.

 

The "GTIA bug" which many people refer to is usually referring to the fact that Graphics 10 pixels are drawn half a pixel to the right (probably delay thanks to having to access the palette?). This is utilized in HIP/RIP/TIP modes, the trick is to have 2 screens defined - one with mode 10 on even lines, the other with odd - Graphics 9 is usually used on the other lines.

 

Official support for GTIA? Hardly anyone touched the extra modes - Lucasfilm Games were among the first back in 1983 with Jaggi Lines.

 

If you were to look at games using GTIA features from 1980-1990, I wouldn't be surprised to see less than 2 dozen out of 2 or so thousand commercial releases in that time.

 

What was the point? There seems to be multiple stories around. Some say that GTIA was intended from the onset, but not released initially due to bugs. The Operating System has always had built-in support for it (have to check RevA though [?])

[DamageX]

IIRC, moving the GTIA mode line over one half-pixel is accomplished by setting the GTIA mode bits after a normal line has already started being displayed.

 

In another old thread here someone reported adding a second ANTIC+GTIA pair to their A8, with it's own memory. Then the luminance output of one GTIA chip could be combined with the color output of the other, so gfx 11 on one plus gfx 10 on the other would be 256 colors. I'm not sure how the separate memory was accessed but if both ANTICs were setup with similar display lists then they would access memory at the same time so a conflict between the 6502 and 2nd ANTIC would be avoided.

 

Kind of messy, for not much gain IMO. I like the "XE Video Card," or whatever it was called, better.

[Bryan]

The GTIA bug is a timing/clocking issue that slipped through testing (the GTIA was already late). I have seen a couple GTIA's that slip back into alignment with the other modes if you heat the chip a little.

 

-Bry

[Rybags]

With a second ANTIC and dedicated memory, you'd probably need extra logic for the CPU to access it.

 

Something along the lines of passing the DMA signal from the ANTIC only if it's memory is mapped into the CPUs field of view at the time.

 

I can't see how you could sync two ANTICs though - you'd have to rely on both achieving the same timing after a reset, which might be a bit tricky.

 

As for increasing clockspeed for higher rez - no go. ANTIC is hard-wired for 114 machine cycles per scanline, and 262/312 scanlines per frame. No changing that through software.

 

Interestingly, the POKEY 16 kHz mode conforms exactly to one scanline, but was probably intended to allow it's timers to be used as DLI alternatives if the programmer desired.

[Gunstar]

APAC mode wasn't 4.096 colors, as was stated that was done by page flipping 3 screens of Red, Green and Blue colors with 16 shades each. There was never a name for this "mode," but I called it RGB2; 80x192. The graphic display program that let you take GIF (and other formats but I forget which) or draw your own with RGBpaint and display them with 4096 colors was called COLORVIEW. There was another mode, that I called RGB1 that used 3 screens of R, G, B with a resolution of 160x192 and used 4 shades on each screen of red&green&blue that displayed 64 colors. Also, there was APAC(Any Point, Any Color-from 256 choices) "mode" is, as stated above 256 colors by alternating modes 10 and 11 on every line, and it can also be interlaced to double the vertical resolution; 80x96&80x192. A display program for this mode is called APACVIEW. There was a basic paint program that was created, called RGBpaint, that let you draw pictures for the 4.096 color "RGB" mode, but it was extremely slow and painful to use becuase it page-flipped on the fly while you are drawing. For the 256 color graphics there was RedRat software's paint program Technicolor Dream and the PD APACdraw and some others. A german paint program called Escalpaint that uses special techniques that I have no idea how they work, but you get an interlaced picture that can display 4096 colors on-screen from a "virtual" pallette of over 32,000. My guess is that it flips through R,G,B but does it on a line by line basis instead of full pages. Escalpaint is much faster than RGBpaint. Then came along some of the newer software modes like HIP/TIP/RIP.

J.D. Potter was the man behind the APAC and RGB modes and the picture viewers and paint programs for these modes as stated above.

 

program list of what I use/remember: APACdraw&APACview, RGBpaint&Colorview, Escalpaint, technicolor dreams, DEGASview, GIFincode

Edited November 11, 2006 by Gunstar

[Fröhn]

The A8 has a maximum resolution of half a colour-clock by one scanline.

Only true for a NTSC system. On PAL an A8 hires pixel is 5/8th of a color clock.

Edited November 11, 2006 by Fröhn