The mystery behind the "other" GTIA modes

[Rybags]

Most of us are familiar with some of the extra software driven modes like APAC, HIP/TIP/RIP etc.

 

But, how about modes based upon using GTIA GPRIOR settings in ANTIC modes like 4 (OS Gr. 12, multicolour character 40x24) etc.?

 

Well, to work out just what goes on, we have to look deeper into the technical descriptions of ANTIC and GTIA.

 

ANTIC sends graphics data to GTIA at double the system clock rate. It uses 3 bits of information on 3 pins (AN0, AN1 and AN2). These bits are decoded by GTIA, and used to generate the playfield.

 

The high bit is simply used to tell GTIA to either display graphics, or to do the other stuff like syncing or preparing for graphics to be displayed.

 

For the purpose of this discussion, we'll just look at values with AN2 set to "1", except for value "000" which means "display background"

 

That leaves AN1 and AN0, which can assume one of 4 values. In the normal course of events (except GTIA modes and hires modes like GR.0 or the translation occurs:

 

Pixel value (bin)	AN2-AN0 value   Resultant display
00						000				   Background colour (COLBK)
01						100				   COLPF0
10						101				   COLPF1
11						110				   COLPF2						
11						111				   COLPF3 (note:  only applicable to multicolour charmodes)

 

Anyway - that's all reasonable and simple, but what about if we use a GTIA mode, in this case the 16 colour, single luminence mode?

 

It gets a little more complex. In GTIA modes, we have a case where two sets of data are taken and combined together. So, two alternate sets of AN1-AN0 data are combined:

 

Pixel value (bin)	AN2-AN0 pair values   Resultant display
0000					 000,000					 COLBK
0001					 000,100					 COLBK
0010					 000,101					 Colour 1
0011					 000,110					 Colour 2 (or 3*) 
0100					 100,000					 COLBK
0101					 100,100					 COLBK
0110					 100,101					 Colour 1
0111					 100,110					 Colour 2 (or 3*)
1000					 101,000					 Colour 4
1001					 101,100					 Colour 4
1010					 101,101					 Colour 5
1011					 101,110					 Colour 6 (or 7*)
1100					 110,000					 Colour 8 (or 12*)
1101					 110,100					 Colour 8 (or 12*)
1110					 110,101					 Colour 9 (or 13*)
1111					 110,110					 Colour 10 (or 15*)

* = alternate colour available if high bit of character is set

 

As you can see, you lose out somewhat in the colour department. The problem being that GTIA modes only combine what it sees on alternate transmissions on AN1, AN0. Therefore, background and Playfield 0 are treated the same, so we have some bit combinations made redundant.

 

What is the use of this mode then?

 

Well, it is useful in that you can use a kernal to alternate between GPRIOR = C0 and 00 hex.

Then, you have a display which alternates between 80 pixels wide which you use to select your colour, and 160 pixels wide which you select one of four luminences (or 5 including chars with hi-bit set).

 

Of course, we can do this more easily by just doing it the normal way, with Gr. 11 alternating with 15 - but this method has the advantage that it is in character mode, so a moving display (ie game) will be possible.

 

[JayoK]

Nice technical explanation - I really like that!

 

Where did you get the technical specs of Antic and GTIA?

 

Ta.

[Heaven/TQA]

I leave that to Fox & Eru & Tebe & other polish coders who definitly might find such things useful?

[eru]

What you describe is a text-mode version of CIN, known for like 10 years

http://lepix.sourceforge.net/#cin

Well, actually CIN is slightly different, as it interlaces two screens to avoid 'empty lines' visible where we use GTIA C0 mode.

But the principle is the same.

[SeaGtGruff]

What would the advantage be of using this instead of using GRAPHICS 0.9, 0.10, or 0.11 (which is what I call GRAPHICS 0 with GPRIOR set to GTIA modes 9, 10, or 11)? With those you get all 16 luminances, or all 9 register colors, or all 16 hues, so you don't lose out on the available colors.

 

Michael

[Rybags]

The "advantage" of this mode is that it's character based (ie games) and has true 160 pixel resolution on alternate lines.

 

I was actually preferring to recreate the mode where you get 160 pixels where you set GPRIOR =00 then 80 then 00 again to get the paletted 4 colour mode but for some reason couldn't get it to work properly, so I started playing around with this one instead.

[+CharlieChaplin]

Hmm,

so is there a) a difference in visible colors or b) in practical use when one combines Gr. 9 + Gr. 12 instead of Gr. 11 + Gr. 15 (CIN afaik)... ?!? If I remember correctly CIN uses Gr. 11 + Gr. 15 and thus produces 16 colors with 4 hues = 64 colors. How many colors are there for Gr. 9 + Gr. 12 ?!? And does this combination of gr. 9 + 12 use less CPU time than Gr. 11+15 ?!?

 

-Andreas Koch

[Rybags]

It's visually practically the same as CIN (Gr. 11 interleaved with 15, with 15 using the 4 shades of white).

 

But due to the way ANTIC talks to GTIA, you get a different mapping of colours than normal (refer to table in my first post).

 

CPU time? Well, most systems that implement modes where GTIA's GPRIOR register is changed on a scanline basis use a single DLI which establishes a kernal for the duration of the display, ie - all of the cycles are used within the displayed portion.

 

As I said before - the advantage of this mode is purely for games where character-based movement of objects is desired).

 

Of course, the other advantage of using any mode which combines the attributes of Gr. 11 and 9 (or 15) is that Player-Missiles are still fully usable, unlike modes like HIP which can't use them properly due to Graphics 9 not allowing display of players.

[Tezz]

What you describe is a text-mode version of CIN, known for like 10 years

http://lepix.sourceforge.net/#cin Well, actually CIN is slightly different, as it interlaces two screens to avoid 'empty lines' visible where we use GTIA C0 mode. But the principle is the same.

Hi Eru, are there any examples/demos of the IN8 & HR2 modes around?

[eru]

What you describe is a text-mode version of CIN, known for like 10 years

http://lepix.sourceforge.net/#cin Well, actually CIN is slightly different, as it interlaces two screens to avoid 'empty lines' visible where we use GTIA C0 mode. But the principle is the same.

Hi Eru, are there any examples/demos of the IN8 & HR2 modes around?

Well, IN8 is just an interlaced GR8 (320x200, 3 colors), as seen in NUMEN (intro and a woman with a snake in the demo).

 

As for HR2 (320x200, 5 colors), I kind of stopped working on it a while ago, but just a few days ago I took a look at it again.

I think it has a great potential for hi-res on Atari, but needs some more work.

I have lots of conversions into HR2, but I really need to work on the conversion algorithm to make them look better.

 

So, here go two pictures.

 

First, by Boris Valleyo. BMP-original, D0 - no dithering, D3 - my bad bad dithering (I need to apply some 'real' stuff like Floyd-Steinberg):

http://eru.nutki.com/a8/boris9901.bmp

http://eru.nutki.com/a8/boris9901-D0.xex

http://eru.nutki.com/a8/boris9901-D3.xex

 

Second, by Royo (same naming):

http://eru.nutki.com/a8/r7.BMP

http://eru.nutki.com/a8/r7-D0.xex

http://eru.nutki.com/a8/r7-D3.xex

 

I wanted to work on it, and add hi-res to Lepix, but there are so many things happening, I just don't have time for everything

And more people can use a nice game like Yoomp! than a new mode in a graphics program, right?

 

EDIT: btw, if you watch interlaced pictures on an emulator, try to make sure your monitor refresh rate fits the emulator. On my Mac I have 60Hz refresh, and when I watch interlaced Atari in PAL (50Hz), it looks terrible. When it's NTSC (60Hz), it's a different world. Also, taking an interlaced screenshot (Atari800WinPlus has it) produces roughly what you would like to see.

Edited November 4, 2007 by eru

[Tezz]

As for HR2 (320x200, 5 colors), I kind of stopped working on it a while ago, but just a few days ago I took a look at it again. I think it has a great potential for hi-res on Atari, but needs some more work. I have lots of conversions into HR2, but I really need to work on the conversion algorithm to make them look better.

I've not seen HR2 before, your early work is impresive. I agree it has great potential. Hopefully it will apear in Lepix soon

 

And more people can use a nice game like Yoomp! than a new mode in a graphics program, right?

Yes yes!

[peteym5]

Has anyone ever tried to use APAC mode and two character sets instead of 2 graphic 8 screens? That is graphics 9 & 11. (GPRIOR +64 & +192). I did some experimenting and a machine language DLI routine can easily run in the background to drive APAC mode over a part of the screen. I had a Turbobasic program randoming drawing characters on the screen. I can release the source code if anybody is interested.

[Rybags]

APAC in character mode is totally feasible, except:

 

- to get full detail you'd need 8 character sets. So, it's more costly in RAM usage.

- more DMA penalties in character mode means your kernal routine needs to be tightly written.

 

I wanted to use the mode where you use GR.0, then have one scaline of colour mode (like Gr. 11) followed by a line where you wait a specific time (about 20 or so cycles after return from WSYNC) then change to GPRIOR=00, which gives the mode where PF0-PF3 determines the pixel colour, and there are 160 pixels width (like Gr. 15).

Problem is that I get it running fine on a real machine, but the two versions of A800Win+ that I try it on don't give the same results.

[SeaGtGruff]

The "advantage" of this mode is that it's character based (ie games) and has true 160 pixel resolution on alternate lines.

Okay, I'm not trying to be a pest, I'm just trying to understand.

 

Where does the 160-pixel resolution come from, if these are GTIA modes? As I understand it, the pixels in one of the GTIA modes are offset by 1 color clock from the other GTIA modes, so that when you mix them, the combined pixels produce a 160-pixel display, rather than an 80-pixel display. So are you saying that this is *not* the way it works when GTIA is used with GRAPHICS 0?

 

Also, do the emulators correctly replicate this offset behavior?

 

Michael

[Rybags]

The mode I described in the first post is 160 across on alternating lines because it's Antic Mode 4 (Gr. 12).

 

Modes which are based on Gr. 10 (GPRIOR=$80) have their pixels shifted one colour clock right, that's likely due to palette lookup delay.

 

Emulators (well, the good ones) replicate most of the behaviour. But not all - however they are usually accurate enough to run virtually all games/demos as good as a real machine.

 

The "combined pixels" 160 across is an illusion. Usually achieved by interlacing two screens and having the graphics modes swapping scanlines on each frame.

[peteym5]

What I did was alternate between 16 color and 16 luma with 2 different character sets. Ending up with pixel wide resolution. I was not looking to include antic 4 or the 9 color mode that is 1/2 pixel off. I tried not to do anything that resulted in noticeable flickering. I had not thought of a good game design that will take advantage of this yet.