Jump to content

Photo

Artifacting - isn't it weird?

artifacting ntsc gtia ctia

46 replies to this topic

#26 falcon_ OFFLINE  

falcon_

    Space Invader

  • 39 posts

Posted Tue Oct 23, 2012 3:00 PM

Hmmm, that's a good solder point to know about IMHO. I'll file this away in my
drawer of developer theory. ;-) I already liked the 'LUM' project (anyone else?) and
your thought means the other side is available for 'special video input' too.
There are curious possibilities here...<devious laff...>

&falcon

An interesting hardware note: the CADJ input can be driven at color clock frequencies, giving you independent control of a pixel's color. Just lift the pin and supply a voltagr level.

Bob



#27 jacobus OFFLINE  

jacobus

    Dragonstomper

  • 661 posts
  • Location:Canada

Posted Tue Oct 23, 2012 8:01 PM

As requested, photos of all models including a CTIA 800. The 400 has a monitor output, but the 600 XL is using the RF modulator.

400.jpg 400
CTIA 800.jpg CTIA 800
GTIA 800.jpg GTIA 800
1200 XL.jpg 1200 XL
800 XL.jpg 800 XL
600 XL.jpg 600 XL
Atari 65 XE.jpg 65 XE
128 XE.jpg 130 XE
XEGS.jpg XEGS

#28 dmlloyd OFFLINE  

dmlloyd

    Chopper Commander

  • Topic Starter
  • 188 posts

Posted Wed Oct 24, 2012 8:20 AM

Wow, that's a greater variation than I was expecting. I think I have an idea for improving the test...

#29 dmlloyd OFFLINE  

dmlloyd

    Chopper Commander

  • Topic Starter
  • 188 posts

Posted Wed Oct 24, 2012 8:45 AM

Okay here's another version which displays color bars for comparison.

Attached Files



#30 FujiSkunk OFFLINE  

FujiSkunk

    Quadrunner

  • 7,309 posts
  • Behold the Fuji!
  • Location:Planet Houston

Posted Wed Oct 24, 2012 8:55 AM

For what it's worth, I seem to get different colors than jacobus on my 130XE. I base this on how the image looks through my PC's video capture card. I don't change the hue, but I do play with the contrast and saturation a bit, trying to match what I see when the 130XE is connected to its own monitor.

In fact, I seem to get different colors in different games, though there may be other factors at work. Two games I know that use this trick are David's Midnight Magic for the pinball table, and Flight Simulator II for the cockpit instruments. I get greens and purples in both games, but not quite the same shades of each. I wonder if David's Midnight Magic using a not-quite-black background has something to do with that.

#31 Mclaneinc OFFLINE  

Mclaneinc

    River Patroller

  • 4,582 posts
  • Location:Northolt, UK

Posted Wed Oct 24, 2012 10:14 AM

Playing around in Altirra and I noticed that standard artifactig in NTSC allows you to change the phase / colours but when you put it in high artifacting it completely changes the colours.

Seems odd to me?

But then again I'm not a TV engineer or have tech knowledge like Phaeron does :)

#32 dmlloyd OFFLINE  

dmlloyd

    Chopper Commander

  • Topic Starter
  • 188 posts

Posted Wed Oct 24, 2012 10:22 AM

FujiSkunk, you can change the color qualities of artifacting by choosing a different color value for the background, which will effectively mix the effect of luma artifacting with the normal color output of the GTIA. For my test I chose black and white to provide the sharpest contrast possible, which will allow a better identification of the relative phase shift of the luma output.

#33 FujiSkunk OFFLINE  

FujiSkunk

    Quadrunner

  • 7,309 posts
  • Behold the Fuji!
  • Location:Planet Houston

Posted Wed Oct 24, 2012 10:34 AM

FujiSkunk, you can change the color qualities of artifacting by choosing a different color value for the background, which will effectively mix the effect of luma artifacting with the normal color output of the GTIA. For my test I chose black and white to provide the sharpest contrast possible, which will allow a better identification of the relative phase shift of the luma output.


That would explain the differences in David's Midnight Magic then. Any idea why I see different colors for the cockpit instruments of Flight Simulator II on my 130XE? As far as I can tell it's just black and white there. Okay, never mind; it turns out Flight Simulator II uses a very dark grey for its background. Interesting how a slight change in the "mixing" color dramatically affects the final product.

Edited by FujiSkunk, Wed Oct 24, 2012 10:36 AM.


#34 JD6502 OFFLINE  

JD6502

    Star Raider

  • 71 posts

Posted Wed Oct 24, 2012 4:47 PM

I'm not exactly a super tech guy, but the story I heard back in the day was that artifacting stems from the fact that the NTSC standard was never intended to include color. It was literally added after the fact, taking advantage of gap between the video and audio carriers, two additional signals were tacked onto the back end, using phase differences to determine the exact hue. This tends to smear the signal, and white areas will have a bluish tint on the left side and reddish on the right. Fine patterns of black and white (like herringbone jackets) could create some eye popping results, so people would turn the color down when watching shows that hadn't made the leap to color.
On a computer the tint of an artifacted color would depend on how it combined the chroma and luma - even a slight variation up or down the carrier would move chroma enough to change the tint, I would think.
'Course my knowledge of this stuff is pretty thin, I'm sure I'm over simplifying a complex bit of engineering.


#35 Bryan OFFLINE  

Bryan

    Quadrunner

  • 10,316 posts
  • Cruise Elroy = 4DB7
  • Location:Coming Soon to Michigan!

Posted Wed Oct 24, 2012 7:16 PM

I'm not exactly a super tech guy, but the story I heard back in the day was that artifacting stems from the fact that the NTSC standard was never intended to include color.

All color television was a hack of the existing B&W standards, and was designed to be backwards compatible. It sounds a little like you're confusing the color encoding methodology with the way a color signal is mixed into the picture. The color information is in-band in the B&W picture but is at a high enough frequency that it doesn't affect the picture on B&W TV's much. It's funny that saturated color information is often quite visible on NTSC color TV's as "dot-crawl."

#36 SeaGtGruff OFFLINE  

SeaGtGruff

    Quadrunner

  • 5,558 posts
  • Location:Georgia, USA

Posted Wed Oct 24, 2012 7:25 PM

Okay here's another version which displays color bars for comparison.

Just a thought-- If the color bars are for comparison with the artifacted colors, then it might be good to rearrange things as follows:

Even bit positions:
<line of artifacted color (even bits)>
<line of color bars>
<line of artifacted color (odd bits)>
Odd bit positions:

That way the color bars can be easily compared to both lines of artifacted colors.

#37 JD6502 OFFLINE  

JD6502

    Star Raider

  • 71 posts

Posted Fri Oct 26, 2012 4:36 PM

All color television was a hack of the existing B&W standards, and was designed to be backwards compatible. It sounds a little like you're confusing the color encoding methodology with the way a color signal is mixed into the picture. The color information is in-band in the B&W picture but is at a high enough frequency that it doesn't affect the picture on B&W TV's much. It's funny that saturated color information is often quite visible on NTSC color TV's as "dot-crawl."



#38 JD6502 OFFLINE  

JD6502

    Star Raider

  • 71 posts

Posted Fri Oct 26, 2012 5:38 PM

Oops, sorry about that. Haven't gotten the hang of the quote thing.
Just wanted to say that I'm not confused so much as uninformed. Also I wasn't very clear. What I think I read about artifacting was that because adding color was done as a hack, it means that color is "appended" on the tail end of the luma signal. This causes a tiny time delay between the display of luma and chroma. Because the electron gun is still moving, luma and chroma are slightly out of sync, causing the color to bleed slightly into the next pulse. Ordinarily this isn't noticeable, but shows up strongly in tight patterns of white and black. At least that was my impression.
What I wonder is - Do the differences in artifacted colors between different G/CTIA chips come from differences in timing between the luminance and color signals, causing either more or less shift?
Or am I completely out to lunch?

#39 Gunstar OFFLINE  

Gunstar

    Gunstar

  • 8,148 posts
  • Location:Kellyville, Oklahoma

Posted Fri Oct 26, 2012 6:43 PM

I've got a few things to talk about with artifacting, being an artist, I used to use artifacting for color in high-res on my 8-bit. But first, I want to point out that I have two 1200XLs and both of them output the green and purple scheme like the 65XE in the pictures above, not the blue and orange as shown for the 1200XL above. The 130XE I used to have ouput the blue/orange scheme. If I turn the tint way to one side or the other, I can get my 1200XLs to display the blue/orange that I prefer (in Ultima games for instance). One of my 1200XL's has the PAL Antic, the other is stock. The monitor I use is a Commodore 1084S. I used to use a TV or Commodore 17XX series monitor for a while with the 130XE when i owned it..

Another thing I feel needs mentioning, is that it's possible to get more than just the blue/orange or green/purple colors from artifacting, there is a high-res art program for the 8-bits, I forget the name now, which had a variety of patterns to choose from, that with the positioning of pixels and artifacting, one was able to have about colors total, IIRC. I think I remember an option to change pixel offset when drawing that allowed to mix artifacting too, creating multiple colors. I remember besides the usual blue/orange that I could get a yellow, a light green, a purple and orange/red shades, all slightly different from the green/purple and orange/blue artifact colors. Lines drawn when doing this would be thicker/thinner depending on the color you wanted to acheive. I will have to find this program to share here.

The last thing is that I always thought that not only the computer's circuitry, but also the tv/monitor had to do with causes of artifacting. I heard once that the artifact colors observed had to do with the dot density and pitch of the electron gun to the screen as to the color of artifacts seen too on many tv/monitors.

Edited by Gunstar, Fri Oct 26, 2012 6:48 PM.


#40 Bryan OFFLINE  

Bryan

    Quadrunner

  • 10,316 posts
  • Cruise Elroy = 4DB7
  • Location:Coming Soon to Michigan!

Posted Fri Oct 26, 2012 8:51 PM

The last thing is that I always thought that not only the computer's circuitry, but also the tv/monitor had to do with causes of artifacting. I heard once that the artifact colors observed had to do with the dot density and pitch of the electron gun to the screen as to the color of artifacts seen too on many tv/monitors.


That's a misconception based on the idea that the colors come from positioning the high-rez pixels over specific phosphors. While it's true that lower resolution tubes can have trouble showing the color in small details distinctly, the artifact colors are actually being decoded in the circuitry and sent to the RGB guns and are not dependent on any precise gun alignment within the tube.

#41 Bryan OFFLINE  

Bryan

    Quadrunner

  • 10,316 posts
  • Cruise Elroy = 4DB7
  • Location:Coming Soon to Michigan!

Posted Fri Oct 26, 2012 9:10 PM

Oops, sorry about that. Haven't gotten the hang of the quote thing.
Just wanted to say that I'm not confused so much as uninformed. Also I wasn't very clear. What I think I read about artifacting was that because adding color was done as a hack, it means that color is "appended" on the tail end of the luma signal.

The color signal is mixed with the luma signal in real time. It's literally like mixing a constant 5KHz tone with your favorite song, then playing it back through a 5KHz notch filter so you can hear the song without the tone, or recover the tone without the song. Can you perfectly extract the 5KHz signal? Not really, there will be some cross contamination but it'll be good enough. That's how color is added to video. It's not appended, it's in there all the time.

This causes a tiny time delay between the display of luma and chroma. Because the electron gun is still moving, luma and chroma are slightly out of sync, causing the color to bleed slightly into the next pulse. Ordinarily this isn't noticeable, but shows up strongly in tight patterns of white and black. At least that was my impression.
What I wonder is - Do the differences in artifacted colors between different G/CTIA chips come from differences in timing between the luminance and color signals, causing either more or less shift?
Or am I completely out to lunch?

When patterns of white and black cause the luma signal to oscillate at certain frequencies (or harmonics thereof), the TV is likely to think it has found a color signal. In the previous example of a song and a 5KHz tone, you're going to have the hardest time separating them when the song contains frequencies near 5KHz already. Likewise, when the luma patterns approach 3.58MHz, the TV will start displaying color patterns even though none were intended. This is artifacting.

The reason for the differences between CTIA and GTIA occur because CTIA shifts the high rez pixels over by 1. This results in the two colors being swapped.

#42 Great Hierophant OFFLINE  

Great Hierophant

    Stargunner

  • 1,181 posts
  • Scribbler
  • Location:Massachusetts

Posted Thu Aug 1, 2013 11:26 PM

It seems that different systems, and even different revisions of the same system, can produce different color artifacts in the 320x192 mode. Ultima II-IV uses the artifact colors in its overworld. Ultima I, by contrast, does not, it uses a non-artifiact color mode on the overworld. The most notable graphics affected are the water and the trees.

The water is obviously intended to be blue in these games, and most systems offer blue as one of the "primary" artifact colors. Ideally, the trees should be green, which is what they are in the original Apple II. But in emulators like Atari800Plus and Altirra, brown is only available with blue. It appears that the colors are only ideal on the Atari 800 for these games, as demonstrated in the screenshots above.

#43 Fletch OFFLINE  

Fletch

    Dragonstomper

  • 985 posts
  • Location:Pennsylvania

Posted Thu Aug 1, 2013 11:38 PM

I never understood why Micro League Baseball looked different on various machines until I found this thread. I had always assumed it had something to do with the adjustable pot on the motherboard, never thought about artifacting variances.

#44 Synthpopalooza OFFLINE  

Synthpopalooza

    Stargunner

  • 1,242 posts
  • Location:knoxville, TN

Posted Fri Aug 2, 2013 12:08 AM

One other thing I have noticed ... the colors are usually some variation of blue-brown, red-cyan, or magenta-green, but only if the PF2 register is 0. Change it to other colors and you get color blending on the artifacts. This doesn't happen with Altirra (the colors stay the same no matter what PF2 is), although it does happen with Atari800Win.

In interlace or flicker modes, you can use artifact belnding to make a 16 color rainbow. As an example, here's a screenshot of the color palette from my ICE IRG font editor. The screenshot was takein in Atari800Win. This is Super 0 (Graphics 0 with changes made on CHBAS, PF1, and PF2 every VBLANK), and the rainbow artifact effect is achieved by alternating blue and yellow every frame on the PF2 (background) register. This effect only works properly on a real Atari or on Atari800Win, and only with magenta-green artifacts. For other artifacts (blue-brown or red-cyan), a blending of different colors from opposite sides of the color wheel (red-cyan or green-magenta respectively) will produce this effect.

ice super 0 palette.png

The effect can also be achieved on a scanline basis in Graphics 8, by alternating each scanline between the two colors (blue-yellow, red-cyan, magenta-green)

Edited by Synthpopalooza, Fri Aug 2, 2013 12:10 AM.


#45 oky2000 OFFLINE  

oky2000

    Stargunner

  • 1,916 posts

Posted Fri Aug 2, 2013 6:35 PM

Yeah, most of the hi-res games on the coco2 relied on the two color mode producing four color with artifacts. Orange and Blue.
With the coco3 and an RGB monitor these games are black and white.


This the exact reason why the Tandy machines in the US were great but the compatible Dragon machines in PAL land looked monochrome...AKA ZX81 style shit....and it tanked like a lead submarine with the tubes flooded haha :)

#46 videogamecritic OFFLINE  

videogamecritic

    Combat Commando

  • 4 posts

Posted Thu Mar 20, 2014 7:35 PM

Glad I stumbled up this topic.  Could this be the reason that Microleague Baseball has the wrong colors on my 27" Toshiba?

The grass is pink, the fence is green, and the dirt is blue-ish.

 

I was thinking there might be something wrong with the game (floppy) but now I think it's a result of the interlacing used to render the graphics.  It was pretty innovative for the time.



#47 MrFish ONLINE  

MrFish

    River Patroller

  • 3,549 posts
  • Location:1010-1010

Posted Sat Mar 22, 2014 10:59 PM

Glad I stumbled up this topic.  Could this be the reason that Microleague Baseball has the wrong colors on my 27" Toshiba?

The grass is pink, the fence is green, and the dirt is blue-ish.

 

I was thinking there might be something wrong with the game (floppy) but now I think it's a result of the interlacing used to render the graphics.  It was pretty innovative for the time.

 

Here's an Antic Mode E hack for Microleague Baseball: MLB (Color Version)

 

The greens are very close in Atirra in NTSC mode, so it only looks like three colors. PAL mode looks correct in Altirra though. It also looks good in Atari800WinPLus with the XFormer palette for NTSC.


Edited by MrFish, Sat Mar 22, 2014 11:00 PM.






Also tagged with one or more of these keywords: artifacting, ntsc, gtia, ctia

0 user(s) are browsing this forum

0 members, 0 guests, 0 anonymous users