Variation in screen color for 8-bit Star Raiders

[1982VideoGames]

Playing Star Raiders for the 8-bit computers, is it normal for the viewer screen with shields on to be pea-soup green on the Atari 800 and blue on the Atari 800XL?

[+orpheuswaking]

yup!

[Faicuai]

Playing Star Raiders for the 8-bit computers, is it normal for the viewer screen with shields on to be pea-soup green on the Atari 800 and blue on the Atari 800XL?

 

 

Absolutely NOT.

 

...The NTSC rendering of the shields are blue with a hint of green on them... but certainly looks blue on both my 800 and 800XL, via composite (800) and composite/s-video (800XL), at the cockpit (it that's the viewpoint you refer to).

 

F.

Edited May 25, 2011 by Faicuai

[Rybags]

Shield colour value is $A0. Color 10 is Turquoise.

 

http://en.wikipedia.org/wiki/Turquoise_(color)

 

On every PAL machine I've ever seen, it's Green Turquoise.

 

It's one of those "problem" colours that doesn't look the same across standards or even machines of the same standard.

 

IMO, the bigger problem with SR is that it uses colour $42 for the Red Alert.

 

On PAL, colour 4 isn't Red, it's more like a pinkish purple.

[Chilly Willy]

On NTSC systems, the cpu clock is derived from the color subcarrier (3.579545 MHz). There is a slight variation from system to system. The A400 has a trimmer pot on the CPU card you can use to adjust the frequency slightly. I used it on my A400 to "correct" the colors as they were off as the unit came from the store. I would hope all the systems should have an adjustment control for the system clock, but I'm only familiar with the A400 hardware as it applies to this area.

[+orpheuswaking]

All units have the POT

 

you've nver heard the saying NTSC means “Never Twice The Same Color.” ?

[Faicuai]

On NTSC systems, the cpu clock is derived from the color subcarrier (3.579545 MHz). There is a slight variation from system to system. The A400 has a trimmer pot on the CPU card you can use to adjust the frequency slightly. I used it on my A400 to "correct" the colors as they were off as the unit came from the store. I would hope all the systems should have an adjustment control for the system clock, but I'm only familiar with the A400 hardware as it applies to this area.

 

Correct.

 

There is also a color adjustment "trimmer" in BOTH the JayMiner-800 and the 800XL. With the SuperSalt Cartridge, there is a specific Color Bars pattern with a fixed-reference at the bottom that you can use in combination with the "trimmer" to adjust Color Response with overall good precision. This "trimmer" seems to be the equivalent of Altirra's HUE-START parameter (Altirra offers the BEST color-control/rendering model among ANY other emulator, by-the-way).

 

My machines are adjusted with this rather simple (yet effective) criteria, and, as a result, shields from cockpit are BLUE, the enemy-sector/in-range Alerts are a convincing and dense RED, and when switching to Galactic Map, there's a distinctive blue-bar/title section at the top that separates it from the galactic-map color itself.

 

There IS NOT green-colored shields in the original rendering of this game (NTSC), ANYWHERE, and as clearly pointed out by Dug N. himself.

 

Enjoy,

 

F.

Edited May 25, 2011 by Faicuai

[+orpheuswaking]

WTF, will everyone stop saying 'jay miner' 800... It's an Atari 800 fer fecks sake!

Edited May 25, 2011 by orpheuswaking

[Kr0tki]

There is also a color adjustment "trimmer" in BOTH the JayMiner-800 and the 800XL.

It's in all NTSC and PAL Atari 8-bit computers.

 

This "trimmer" seems to be the equivalent of Altirra's HUE-START parameter

Actually, its equivalent is "Hue step", not "Hue start". "Hue start" should be set to -57 and left there.

 

(Altirra offers the BEST color-control/rendering model among ANY other emulator, by-the-way).

Actually, Atari800 offers the exact same colour model, at least in the NTSC mode.

Edited May 25, 2011 by Kr0tki

[Faicuai]

 

Actually, its equivalent is "Hue step", not "Hue start". "Hue start" should be set to -57 and left there.

 

 

Thanks, but your description DOES NOT match my on-screen behavior.

 

In the SuperSalt color-bar pattern, there are two (2) key bars to pay attention two: the SECOND bar from top-to-bottom and the LAST bar. When you rotate the trimmer, EVERYTHING (from second to last bar) moves towards RED or towards GREEN, and what reproduces this effect in Altirra's color panel is HUE START (totally the opposite to what you describe). In fact, this can also be verified by running the SuperSalt image in ALTIRRA itself, and then playing with Hue Start (it will output the same effect as real HW).

 

Furthermore, It is not necesarily set at -57, because a combined and very careful adjustment of BOTH the Start and the Step would be required to eventually match YOUR hardware (that's the goal), adjusted via the SuperSalt. Most of the machines come with an "implicit" Hue-step (hard to change) which may be linked to GTIA's own output, and/or the mumbo-jumbo of resistors and discrete parts from which an analog signal eventually emerges out of the machine.

 

Actually, Atari800 offers the exact same colour model, at least in the NTSC mode.

 

This is not the case. Atari800 color adjustment *S*U*C*K*S* big-time (it may implement the same model, but not the same interface, as I saw prior to 2.2.x release), but it was no match to the overall precise control that Altirra provides with NO NEED of external files to adjust color. The only thing that may need is a tool for reproducing NON-LINEAR changes in the steps, but that would require real-time readings from colorimeter for true match, and results would "stuck" to your particular's screen behavior (brightness distribution, primaries purity and a whole slew of things that would bog down anyone trying to just have some fun).

 

I have been able to verify final results on my colorimeter-calibrated Bravia 52" screen, when comparing color-tuned ALTIRRA samples with high-quality captures from the JayMiner-800 and the 800XL. You can throw at it ANYTHING you want (Fractalus Splash + auto-Droid, Frogger auto-intro, FroggerII splash and demo screens, Preppie splash and play-field, RiverRaid auto-scrolls, BlueMax2001 intro-screen, Nautilus splash+demo, Star-raiders, Stealth full-demo, PolePosition demo, SuperSalt color-bars, etc.) and it just comes out SWEET, frame-by-frame, color-by-color, once properly tuned with just its control panel.

 

Cheers,

 

F.

[Faicuai]

WTF, will everyone stop saying 'jay miner' 800... It's an Atari 800 fer fecks sake!

 

The problem with this is that:

 

1. There are OTHER things called Atari800 (which is an emulator, for instance).

2. "Atari 800" (as well as 800XL, etc.), are words and terms frequently used, abused or simply "confuzed".

3. It is precisely Jay Miner's work (and his 9+ crew) what made Atari's 8bit line what TRULY is, and what in essence remained unchanged for the next 5-6 years. Even StarRaiders was born out of this team.

4. As an example, a 130XE simple feels/look like "low-quality, Commodore-droid", while (in contrast) the original JayMiner-800 OOZES quality, craftmanship, made right here at home (U.S.), and, ultimately, *THE* machine to have (only the 1400XL would potentially change that equation, though).

 

Of course, there are no rules (or directives) written ANYWHERE on how each and everyone may choose to denominate its most appreciated collectibles, or which collectibles to ultimate have! :-)

 

Cheers,

 

F.

[Shamus]

Quite frankly I don't get all the animosity towards Atari800, especially when it manages to get so many things right. I think the NTSC emulation that's in there now (courtesy of Blargg) is the best thing I've seen on any emulator, period.

 

Altirra may be all that and a bag of cheese doodles, but I (and many others, I'm sure) wouldn't know because it's tied to Windows.

 

(No offense meant to you Phaeron, I'm glad you decided to share your source code. If I had the time, I would work on porting it to Qt but as it is I hardly have time to work on my own projects. )

[kskunk]

Playing Star Raiders for the 8-bit computers, is it normal for the viewer screen with shields on to be pea-soup green on the Atari 800 and blue on the Atari 800XL?

The author addresses this question on his blog, with pictures:

 

http://dougneubauer.com/starraiders/

 

I originally wanted the color of the shields to be a dark blue/greenish color (Like the picture on the Left), but depending on the color settings of your television, the shields will appear as green (Right).

 

- KS

[SeaGtGruff]

Actually, its equivalent is "Hue step", not "Hue start". "Hue start" should be set to -57 and left there.

 

 

Thanks, but your description DOES NOT match my on-screen behavior.

 

In the SuperSalt color-bar pattern, there are two (2) key bars to pay attention two: the SECOND bar from top-to-bottom and the LAST bar. When you rotate the trimmer, EVERYTHING (from second to last bar) moves towards RED or towards GREEN, and what reproduces this effect in Altirra's color panel is HUE START (totally the opposite to what you describe). In fact, this can also be verified by running the SuperSalt image in ALTIRRA itself, and then playing with Hue Start (it will output the same effect as real HW).

I may be able to help clear this up.

 

The schematics for the TIA (2600) show that the color delay adjust shifts hues 2 through 15, whereas hue 1 always has the same color phase as color burst. Thus, the only way to adjust hue 1 on the 2600 is to use the TV's tint control.

 

The schematics for the GTIA (800) show an arrangement similar to the above, but there's an extra cell that's applied for hue 1. So on the 800, the color delay adjust does shift hue 1 as well.

 

Michael

[Kr0tki]

 

Actually, its equivalent is "Hue step", not "Hue start". "Hue start" should be set to -57 and left there.

 

 

Thanks, but your description DOES NOT match my on-screen behavior.

I must admit, that's a huge surprise. It seems to contradict the information from both the GTIA spec and the 400/800 Field Service Manual. Both suggest that the colour pot is meant to adjust the "delay" or "step" between consecutive hues. The Field Service Manual says you have to adjust the pot so that hue 1 and hue 15 are "same color (golden rod)". This corresponds with setting "Hue step" in Altirra to 25.7 (= 360/14), regardless of the "Hue start" setting. (The 5200 Field Service Manual uses a different expression, the bars should be "within one shade of each other", but the image on the next page still says "same color".)

 

Furthermore, there are multiple screen captures from different NTSC Ataris available (mainly from here). Despite being captured on computers with variously-adjusted GTIAs, all of the screen captures have hue 1 _exactly_ (and I mean that) equal to the hue of the colorburst signal (greenish-yellow, or "Hue start"=-57). Based on that, I assumed to this day that adjusting the colour pot does not change hue 1 (it's always equal to the colorburst). What you said, directly contradicts the screen captures and my assumption. (Concidentally, all of the captures seem to have been made with TV cards with a default hue/tint setting - otherwise the hue 1 would not be equal on all of them.)

 

An important question follows: Do all NTSC Ataris behave the way Faicuai's unit does?

 

And another: Faicuai, does your colour pot adjusts only "Hue start", or also "Hue step"? Are you able to change hue distance between the two bottom bars in SuperSalt? The Service Manual says they should be "same color".

 

Furthermore, It is not necesarily set at -57, because a combined and very careful adjustment of BOTH the Start and the Step would be required to eventually match YOUR hardware (that's the goal), adjusted via the SuperSalt.

-57 is the angle of the colorburst in the YIQ colourspace, and all NTSC screen captures so far show hue 1 as -57 deg in YIQ.

 

Most of the machines come with an "implicit" Hue-step (hard to change)

Contradicts with both the GTIA spec and the Field Service Manuals, as explained above.

 

Actually, Atari800 offers the exact same colour model, at least in the NTSC mode.

 

This is not the case. Atari800 color adjustment *S*U*C*K*S* big-time (it may implement the same model, but not the same interface, as I saw prior to 2.2.x release), but it was no match to the overall precise control that Altirra provides with NO NEED of external files to adjust color.

You might need to reconsider your opinion, 'cause from the date of release of version 2.2.0 the above statement nicely fits in the definition of "load of bollocks".

 

The only thing that may need is a tool for reproducing NON-LINEAR changes in the steps, but that would require real-time readings from colorimeter for true match, and results would "stuck" to your particular's screen behavior (brightness distribution, primaries purity and a whole slew of things that would bog down anyone trying to just have some fun).

I tend to agree, IMHO implementing all of the possible deficiencies of a particular TV unit or Atari unit would be a bit too pedantic.

 

I have been able to verify final results on my colorimeter-calibrated Bravia 52" screen, when comparing color-tuned ALTIRRA samples with high-quality captures from the JayMiner-800 and the 800XL. You can throw at it ANYTHING you want (Fractalus Splash + auto-Droid, Frogger auto-intro, FroggerII splash and demo screens, Preppie splash and play-field, RiverRaid auto-scrolls, BlueMax2001 intro-screen, Nautilus splash+demo, Star-raiders, Stealth full-demo, PolePosition demo, SuperSalt color-bars, etc.) and it just comes out SWEET, frame-by-frame, color-by-color, once properly tuned with just its control panel.

Is it true that on a properly-calibrated screen the hue of the colorburst signal equals -57 deg in the YIQ colourspace?

 

I've seen your settings here. Note that setting "Hue step" to anything different than 25.7 causes hues 1 and 15 to not be "same color", thus not following the service manual.

[Kr0tki]

The schematics for the GTIA (800) show an arrangement similar to the above, but there's an extra cell that's applied for hue 1. So on the 800, the color delay adjust does shift hue 1 as well.

Thanks so much! That indeed clears some confusion. Not being an electronics guy, I'd like to ask you, does the DEL line also shift the colorburst signal?

 

On a related note: GITA Die Shots.

[Rybags]

CAD3 aka DEL is the colour adjust input.

[SeaGtGruff]

The schematics for the GTIA (800) show an arrangement similar to the above, but there's an extra cell that's applied for hue 1. So on the 800, the color delay adjust does shift hue 1 as well.

Thanks so much! That indeed clears some confusion. Not being an electronics guy, I'd like to ask you, does the DEL line also shift the colorburst signal?

I'm not an electronics guy, either, and I can understand only portions of schematics drawings. Furthermore, the GTIA schematics that I have (which are the ones in the GTIA.PDF document) are difficult to read, partly because the scans are so bad, and partly because each sheet is split across two pages of the document.

 

But as can be seen in the portion of the schematic I posted, the BURST signal does go through the same cell as hue 1 before reaching the COL pin, so I guess DEL does shift the color burst as well. I'm not sure what that means for the TV display. If the color burst is a referent so the TV can properly decode the colors on that scan line, and if the color burst is shifted, will the TV just display hue 1 as though it were not shifted? But then why feed the color burst and hue 1 signals through the extra delay cell? As far as I can tell, there's no other place in the schematics where the BURST line is shown going to the COL pin without first being fed through that extra delay cell.

 

Michael

[Rybags]

Colourburst is the "reference" so that the TV can lock on to get the phase base.

 

So, every other colour subsequent to that will be based on it's phase offset from that base.

 

Shifting the colourburst as well should in theory mean that colour 1 remains constant, since it would be shifting the same amount as the colourburst.

 

Of course with PAL we have to throw the swinging colourburst into the mix which confuses things a bit further.

[Bryan]

Boy what a strange thread. If the question is: Should all my Atari's produce the same colors?, then the answer is yes provided they are of the same video type (NTSC/PAL).

 

The GTIA color cells are supposed to work like a multi-tap delay line, each one shifting the color carrier by the same amount of time. Since a precision unclocked delay line is difficult to implement in a MOS chip, they provided the DEL input which tweaks the speed of the cells a bit. Since the cells are connected serially, any timing errors are added as you move along, so the idea is to set the DEL voltage to get the best average accuracy among the cells so the last cell isn't producing some bizarre color.

 

Oh, and the proper names for the 800 and 400 are Colleen and Candy.

Edited May 26, 2011 by Bryan

[SeaGtGruff]

Here's something else I don't understand about the GTIA documentation versus the field service manual.

 

The NTSC color signal frequency is about 3579545 Hz, so it's about 279 nS.

 

The GTIA documentation lists two sets of color delay timings for the hues, depending on whether the DEL pin is 7.0 V or 5.0 V:

 

Hue 1 = 167 nS (7.0 V) or 190 nS (5.0 V)

Hue 2 = 188 nS (7.0 V) or 225 nS (5.0 V)

Hue 3 = 209 nS (7.0 V) or 260 nS (5.0 V)

...

Hue 15 = 461 nS (7.0 V) or 680 nS (5.0 V)

 

So the difference between each hue is either 21 nS (7.0 V) or 35 nS (5.0 V).

 

21 nS / 279 nS * 360 d = 27 degrees, which gives colors that do look close to the 800's palette (see the screenshots I posted last night in the "Emulation" forum).

 

35 nS / 279 nS * 360 d = 45 degrees. (WTF?)

 

So if my understanding is correct, the GTIA documentation seems to indicate that the phase shift for each hue is 27 degrees.

 

On the other hand, the FSMs for the 2600, 5200, and 800 all say the same thing-- adjust hue 15 so it's the same as hue 1, which gives a phase shift of about 25.7 degrees. Aside from the fact that I dislike the idea of setting hue 15 to be the same as hue 1, since that gives us only 120 or 240 unique colors (depending on the graphics mode), I've seen several people suggest over the years that perhaps the reason the FSMs say to do this is because it's an easy adjust to make if you aren't using a vectorscope.

 

And just to add even more confusion to the mix, the documentation for the (unreleased?) CGIA chip clearly show a phase shift of 24 degrees, giving 15 hues that are equally-spaced around the Y'IQ/Y'UV color wheel (i.e., hue 15 ends up 24 degrees shy of hue 1, such that hue 1 is halfway between hue 15 and hue 2).

 

It's like the different folks at Atari over the years couldn't make up their minds what the phase shift should be.

 

Personally, I think 24 degrees gives too many greens. And 25.7 degrees takes 8 or 16 colors away from us. On the other hand, 27 degrees actually gives a fairly good fit for Red=Hue4, Blue=Hue8, Green=Hue12, and Yellow=Hue15:

 

Hue 4 = 180 - 3 * 27 = 99 degrees (off by -4.5)

Hue 8 = 180 - 7 * 27 = -9 = 351 degrees (off by +3.9)

Hue 12 = 180 - 11 * 27 = -117 = 243 degrees (off by +2.3)

Hue 15 = 180 + 360 - 14 * 27 = 162 degrees (off by -5.1)

 

Michael

[Mclaneinc]

Guys (& ladies?) can this skip being a emulator bashing part of the thread and just argue on the electronics / clarification.

 

Both emulators mentioned are beyond bashing, without Atari800 there may never have been other real tries at a Windows and continues to be a brilliantly researched multi platform bit of coding. Altirra has already proved itself a gold medal in its emulation and research.

 

This should all be about getting the right figures and structure..

[Faicuai]

 

 

It's like the different folks at Atari over the years couldn't make up their minds what the phase shift should be.

 

Personally, I think 24 degrees gives too many greens. And 25.7 degrees takes 8 or 16 colors away from us. On the other hand, 27 degrees actually gives a fairly good fit for Red=Hue4, Blue=Hue8, Green=Hue12, and Yellow=Hue15:

 

Hue 4 = 180 - 3 * 27 = 99 degrees (off by -4.5)

Hue 8 = 180 - 7 * 27 = -9 = 351 degrees (off by +3.9)

Hue 12 = 180 - 11 * 27 = -117 = 243 degrees (off by +2.3)

Hue 15 = 180 + 360 - 14 * 27 = 162 degrees (off by -5.1)

 

Michael

 

In this particular juncture, I would have to agree that you have NAILED the key issue.

 

The JayMiner-800 responds more along your preference above (and what also seems natural to me...) HOWEVER, if you pay close attention to the 800XL series, it turns to GREEN, especially at the second and last bar of the reference/calibration pattern. It can be seen in most of your suggested screen-shots, as well. Something else happened later...

 

Furthermore, folks are also "confuzed" about brightness levels, etc. I have seen 130XE catpures that show NO TONAL separation, whatsoever, on the high-end of the color-pattern (luma-wise). Sort of an overshot of contrast.

 

At the ends, it boils down to emulating the HW as it IS (and not exactly what it "should"), which is why we are all on this game.

 

F.

[Kr0tki]

I'm not an electronics guy, either, and I can understand only portions of schematics drawings. (...) But as can be seen in the portion of the schematic I posted, the BURST signal does go through the same cell as hue 1 before reaching the COL pin, so I guess DEL does shift the color burst as well.

I envy you. I have found the "BURST" label only after you told me it's there

 

I'm not sure what that means for the TV display. If the color burst is a referent so the TV can properly decode the colors on that scan line, and if the color burst is shifted, will the TV just display hue 1 as though it were not shifted?

Exactly! As Rybags pointed out - since hue 1's relation to the colorburst remains unchanged, the generated colour would also remain unchanged. That's why Faicuai's account suprises me so much.

 

Anyway, it's easy to verify. Anyone here has an NTSC Atari?

 

But then why feed the color burst and hue 1 signals through the extra delay cell?

Since the pot allows to change the phase shift relation between the colorburst and the luma signal, it would in theory allow to adjust hues of the artifact colours. Think about that!

 

Of course with PAL we have to throw the swinging colourburst into the mix which confuses things a bit further.

Phaeron recently resolved the issue of the PAL colours, and as far as I see it's 100% spot on. See Altirra 1.9.

 

Here's something else I don't understand about the GTIA documentation versus the field service manual.

 

The NTSC color signal frequency is about 3579545 Hz, so it's about 279 nS.

 

The GTIA documentation lists two sets of color delay timings for the hues, depending on whether the DEL pin is 7.0 V or 5.0 V (...)

I believe the GTIA spec gives these two timings only as examples, and not necessarily any of the two (5 or 7V) should be treated as an "official" or "best" value. However, judging by the colour names given in the table in the same GTIA spec, the angle of 27 deg fits those names well.

 

So if my understanding is correct, the GTIA documentation seems to indicate that the phase shift for each hue is 27 degrees.

 

On the other hand, the FSMs for the 2600, 5200, and 800 all say the same thing-- adjust hue 15 so it's the same as hue 1, which gives a phase shift of about 25.7 degrees. (...)

Yes, they contradict each other. The names of colours in the GTIA spec suggest that hue 15 should be between hue 1 and hue 2. The list of colour names was later published in the Atari 400/800 Hardware Manual, so it can be considered official.

 

And just to add even more confusion to the mix, the documentation for the (unreleased?) CGIA chip clearly show a phase shift of 24 degrees, giving 15 hues that are equally-spaced around the Y'IQ/Y'UV color wheel (i.e., hue 15 ends up 24 degrees shy of hue 1, such that hue 1 is halfway between hue 15 and hue 2).

Let's just ignore the CGIA spec and its 24 degrees. For many reasons:

1) It contradicts itself. The 24 deg angle doesn't match the list of colour names that is included in the same document (in _two_ places);

2) The chip doesn't contain a DEL line, it's phase shifts were hardcoded. That's IMHO a significant difference, so we can't apply the CGIA spec to the GTIA in a straightforward way;

3) It's an unreleased prototype, we don't know if it existed in a physical form;

4) To top it all, the CGIA was for the 5200, so it's kinda off-topic

Edited May 26, 2011 by Kr0tki

[SeaGtGruff]

Furthermore, folks are also "confuzed" about brightness levels, etc. I have seen 130XE catpures that show NO TONAL separation, whatsoever, on the high-end of the color-pattern (luma-wise). Sort of an overshot of contrast.

I've seen this also-- if not no hue difference, then very little. If I remember correctly, I've read in old [stella] mailing list posts that the very light colors are not a result of too much luma per se, but rather too much saturation. I don't know if that's actually the case, but the reason the hues look indistinguishable from each other is because one or more of the RGB values is too high, so it gets clipped by the TV. This also happens at the low end of the luma settings.

 

You notice this right away if you try to compute the RGB color space values for the Atari's colors, because you occasionally end up with RGB values that are less than 0 or greater than 1. These too-low or too-high values are (as I understand it) actually being output by the Atari, but the TV clips them so they're between the legal range of 0 to 1. You get hue shifts at both ends because of this. For example, if the TV clips the R value but not the G and B values, the displayed color won't have the expected amount of red in relation to green and blue.

 

Michael