Stella Gets Some New Clothes

[Trebor]

New palettes for Stella - Stella257 & Stella267.

 

Both palettes contain updated NTSC and PAL colors, SECAM retains the standard Stella settings.

 

Stella257 has the 180 degree colorburst in place resulting in 1$ and F$ being a near exact match (Visually they will appear exact to most).

 

Stella267 closer represents the standard Stella palette having a slight phase shift from the 180 degree colorburst, providing the most noticeable change to F$ - darker browns in comparison to 1$, and E$ - giving a more neutral balance between a green and brown hue, instead of a stronger green presence under Stella257.

 

There is no manipulation of contrast, brightness, or gamma for these palettes. Stella has internal controls to handle that

 

Download the palettes here: Stella2XX_Palettes_20130911.zip

 

Screen captures to follow (click pics to enlarge and remove distortion) are in this order:

Stella (Standard) --> Stella257 --> Stella267

 

Color Bar Generation (1984) (VideoSoft) (NTSC) 'Title Screen' & 'Color Bars'

[Composite mode @ Stella default values]

 

Color Test (26-09-2002) (Eckhard Stolberg) (NTSC)

[Composite mode @ Stella default values]

 

[TV mode disabled (AKA RGB)]

 

Color Test (26-09-2002) (Eckhard Stolberg) (PAL)

[Composite mode @ Stella default values]

 

[TV mode disabled (AKA RGB)]

 

Enjoy.

[Trebor]

Phase shifting on the Atari 2600 is much slower than expected in comparing both a composite mod 4-switch Woody and a RF 4-Switch Woody starting with a phase shift of about 24.7.

To complement the two posted palettes Stella257 and Stella267 with a phase shift of 25.7 and 26.7 respectively, and account for the slow phasing noted above, three more palettes have been added: Stella247, Stella252, and Stella262. This gives half degree increments above and below the 180 degree colorburst, up to a full degree.

Download the entire set here: Stella2XX_Palettes_20130913.zip
(Includes same previously posted Stella257 and Stella267)

For NTSC palettes, the greatest differences lie in the E$ and F$ color range. The further you scale upward the less and less difference between the palettes.

Screen captures follow with this pattern:
Stella247 --> Stella252 --> Stella257 (aka 180 degrees colorburst) --> Stella262 --> Stella267

Color Bar Generation (1984) (VideoSoft) (NTSC) 'Title Screen' & 'Color Bars'
[Composite mode @ Stella default values]




Color Test (26-09-2002) (Eckhard Stolberg) (NTSC)
[Composite mode @ Stella default values]


[TV mode disabled (AKA RGB)]


Color Test (26-09-2002) (Eckhard Stolberg) (PAL)
[Composite mode @ Stella default values]


[TV mode disabled (AKA RGB)]


If you're looking for 'technical perfection' then Stella257 would be the most ideal of the available selection with the 180 colorburst in place.

Enjoy.

[Thomas Jentzsch]

Somehow the colors on my TV are partially way off Stella's PAL palette. Sometimes the hue changes with the value, most notable in my example (see below) for 2x (low and high values are more red, middle values are more yellow).

 

Compared to 267 palette 2x, 4x and 8x have more red, dx, bx, 9x and 7x have more blue and less green, 5x is less blue, 3x is much more yellow (especially the middle values)

Stella's standard palette seems closer, but 3x and 5x are both much more green on my TV.

 

Is that a problem with my console setup?

 

The attached picture shows a reproduction of what I see on my TV for 2x.

[Trebor]

Excellent feedback, Thomas. Thank you very much. The issues you brought up regarding the PAL colors is two-fold:

"dx, bx, 9x and 7x have more blue and less green"...
That's due to the phase shift (26.7) selected. If you try Stella247 (24.7) you will see less green, more blue. The green travels upwards in the spectrum with the higher the phase shift. Ditto with blue (just not as much) over red (I.e. "8x have more red").

Regarding the rest of the 'red': "2x, 4x ...have more red", that was human error on my part. PAL colors 2x,3x,4x...I went under, over, and then back under again regarding color row transcription. NTSC colors are not impacted, but as you're well aware all palettes (NTSC, PAL, SECAM) are included in the same file. Sorry folks, my fault. PAL on the 2600 is trickier than PAL on the 7800...lol.

Download the corrected palettes here, just have them overwrite the older ones: Stella2XX_Palettes_20130923.zip

Here's the updated PAL charts, pattern again is:
Stella247 --> Stella252 --> Stella257 (aka 180 degrees colorburst) --> Stella262 --> Stella267

Color Test (26-09-2002) (Eckhard Stolberg) (PAL)
[Composite mode @ Stella default values]


[TV mode disabled (AKA RGB)]


Just to help appreciate the phase differences better here is the Stella247 PAL side-by-side Stella267 PAL under RGB mode:


Last (but certainly not least), Thomas, the issue with 2x I really can't say exactly. I'm not sure if you're using a CRT*. If so, it could be some color bleed coming in on the lighter end of the scale. You could try increasing the contrast and gamma, and perhaps some extra saturation while manipulating fringing and artifacts under the Stella video controls.

*Important to note, if you are using a LCD/LED/Plasma over a CRT, 'browns/tans/yellows' may lean more 'greenish'. Some 'blues' may lean more 'purpleish'. It's very evident with a 7800, and definitely evident with a 2600 too.

Here's an example, utilizing the status bottom part of the screen in Solaris.

The entire background color as well as the fuel display will come across more greenish on a HD/Flat Panel display:

http://youtu.be/WY2LrUjSpN0?t=49s

Now here is the same on a CRT:

http://youtu.be/buhHMAcRSwU?t=3s

 

This CRT has the saturation bumped up a bit, but still a very brown and gold combo:

http://youtu.be/f68IjW7_w98?t=6s

While the phase-shift will account for some of the green(er) or brown(er) being evident on a CRT or HDTV respecting the overall background color. The fuel display remains a greenish color on a HD/Flat Panel, while more golden-tan on a CRT.

[Thomas Jentzsch]

I am using a CRT (~20 year old Sony Trinitron). I also opened my Atari to adjust the phase, trying to get close to the 26.7° phase. If I push it a tiny bit too far, 2x becomes greenish immediately.

 

IIRC the new palettes come closer, especially the red part, but even compared to 267 3x still has more yellow for me. I will warm up my Atari and check again this weekend, since I need good color values for the PAL version of IXION/ULTRA.

 

I think the brightness for lower values goes much too low. Especially when compared with the gray colors. On my TV the dark colors are definitely brighter than the grays for the same value. In default TV setting, I cannot even see 02, so I have to increase brightness a bit. I remember not seeing the dark gray Tardis in Manuel's Star Fire.

 

I have written my own color test program where gray is only displayed once, the PAL colors are displayed reordered (like in NTSC) and the the first color is repeated after the last one. If anyone is interested I can post it.

Edited September 27, 2013 by Thomas Jentzsch

[Thomas Jentzsch]

Warming up my 2600 indeed shifted the phase a bit. The middle values of 2x became slightly greenish now. Actually I wonder, since the 2600 has no real yellow, why they didn't use an even larger phase shift. Then 2x could have become really yellow.

 

I could not reduce the effect at 2x I posted above by changing my TV settings. I suppose the same might happen with other colors too, but then it is not noticeable for me.

 

In general my colors seem to match the new 267 best, but the brightness is still off. On my system with default TV settings, 20..80 are darker than 06, a0 and b0 have about the same brightness, c0 and d0 are brighter. So colors are brighter than grays, especially with high saturation.

 

Also blue colors seems to be noticeable brighter than yellows. But that could be my TV.

 

When I take out all the color on my TV, 0x is always as bright as 2..d(x-2) which means the grays are all one step darker than the colors. This cannot be my TV.

 

I am using a 2600 Darth Vader, modded for composite.

 

Attached is my test program for PAL50 (I have other versions too). Use the joystick directions and fire to change the display. I plan to improve it a bit, so that it displays some values on the screen.

ColorTest (PAL50).bin

[Thomas Jentzsch]

Updated version, now with digits.

ColorTest (PAL50).bin

[Thomas Jentzsch]

I readjusted my my color phase to get rid of the yellow in 2x. I am still using the 267 palette file.

 

Besides the problems described above, my greens are still much more greenish.

• The differences start at bx, which has no green tint. 9x has a lower green tint.
• 7x seems almost correct.
• 5x has no blue tint at all, it is very greenish and actually it has a small yellow tint already, 3x looks almost like 5x, but with some more yellow. The difference between 5x and 3x is notably small, as if the system maxes out.
• From the colors listed, 5x is most different

It seems like in your palette green is starting a bit too early and then the increase is much too small.

[Trebor]

Let's try a different approach with the PAL 2600 color palette, Thomas. Here's a normal/standard PAL color palette. We have a begin degree point of 57. The phase shift is 25.7 degrees, so the colorburst is right at 180 degrees with color 1 and 15 for all intents and purposes is a match (There's is a very, very, very small difference between them).

 

Color 0 (Grayscale) is out of the picture as we know for the 2600 PAL palette it makes up 0x, 1x, Ex, and Fx. That is without any doubts.

 

With these colors, label the row order of the Atari 2600 PAL color palette (2x thru Dx) as you see it or as close as possible. The 'brightness' issue, I have a feeling you're looking for the alternate set of columns. I was on the fence about either going pure black or pure white. You cannot have both without hacking the palette and I want to avoid that. So instead of columns 0x, 2x, 4x, 6x, 8x, Ax, Cx, Ex (For a pure black and darker scale)...I think at least from your Sony TV you rather 1x, 3x, 5x, 7x, 9x, Bx, Dx, Fx (For a pure white and lighter scale).

 

Once we sort through the "right" color order for PAL on the 2600, I'll create both the pure black and pure white sets with the different phase shifts so it covers either preference regarding 'brightness'.

 

[Thomas Jentzsch]

The 'brightness' issue, I have a feeling you're looking for the alternate set of columns. I was on the fence about either going pure black or pure white. You cannot have both without hacking the palette and I want to avoid that.

What do you mean with that "alternate set of columns"?

 

And I don't see (or want) pure white, because no TV can display that. And I know it only can get as black as the TV can do. What I am saying is, that compared to the colors, the gray is at least one (no color) to two or three (color enabled) steps too bright. I verified my eyes by reducing the TV brightness to the minimum, with and without color saturation reduced to zero too.

 

Summary: Gray (0x, 1x, Ex and Fx) is darker than the colors. And colors with high saturation become even brighter compared to gray with identical x value.

 

As far as I understand the CRT, this seems pretty logical. To display black, you reduce all three electron guns to the minimum. If you want to add color, at least one electron gun has to become stronger. So even the darkest color has to become brighter than the darkest black possible.

 

In theory one could make the black brighter, by not reducing the three guns to the minimum possible. But why should one do that and reduce contrast? Also Sony's Trinitrons are famous for their high contrast.

 

So instead of columns 0x, 2x, 4x, 6x, 8x, Ax, Cx, Ex (For a pure black and darker scale)...I think at least from your Sony TV you rather 1x, 3x, 5x, 7x, 9x, Bx, Dx, Fx (For a pure white and lighter scale).

Not sure if I get you here, too.

There is absolutely no difference between x0 and x1 (or x2 and x3 etc.), if that's what you mean. Also 0x, 1x, Ex and Fx are 100% identical.

[Thomas Jentzsch]

Let's try a different approach with the PAL 2600 color palette, Thomas. Here's a normal/standard PAL color palette. We have a begin degree point of 57. The phase shift is 25.7 degrees, so the colorburst is right at 180 degrees with color 1 and 15 for all intents and purposes is a match (There's is a very, very, very small difference between them).

 

Color 0 (Grayscale) is out of the picture as we know for the 2600 PAL palette it makes up 0x, 1x, Ex, and Fx. That is without any doubts.

 

With these colors, label the row order of the Atari 2600 PAL color palette (2x thru Dx) as you see it or as close as possible.

OK (counting rows from 0) and focusing on the more saturated, darker colors:

2x = between 1 and 2 (~1.50)

4x = 3

6x = between 4 and 5 (~4.50)

8x = between 5 and 6 (~5.50)

Ax = between 6 and 7 (~6.25)

Cx = between 7 and 8 (~7,25)

Dx = between 8 and 9 (~8.25)

Bx = 9

9x = between 9 and 10 (~9.75)

7x = 11

5x = 13

3x = between 13 and 14 (~13.25) (unexpected and barely visible small difference to 5x, as if the green maxes out)

 

So the gaps between the colors I see are pretty small around blue and become larger towards red and green.

Edited September 28, 2013 by Thomas Jentzsch

[Trebor]

Referring to 'alternate columns' I was referencing the 'brightness' scale. Colors change moving up and down the chart with the scale (0-F), but utilizing the same scale (0-F) 'brightness' changes when moving left and right in the chart.

 

Ultimately the idea is not to gear a palette towards just one television CRT type. I don't want to hack colors or removing just 'colors' from one row and place it under the 'grayscale' of another row due to the behavior or/and preference of one television or model type, or to account for CRT factors that need to be applied after a base palette is established.

 

Here's an example. Same exact base palette for both captures. Left is a raw, base, typical NTSC palette (A similar thing to be achieved here for the PAL 2600) and then the same palette with some CRT effects applied:

 

 

Brighter/stronger colors, difference in grayscale are all lumped under CRT effects like bloom, phosphor effects, gamma correction, saturation...after the technical base palette is established. The base can be tweak for different end results until content, but right now I'm looking for the best matching base sticking with calculated palette values. (I.E. Hue Begin Degree of 57 (For PAL), Phase Shift Degree of 24.7, 25.7, 26.7, etc.). The colors and grayscale produced from there while keeping Brightness and Contrast neutral serve as the base.

 

Thanks for the feedback on the chart. It is very much appreciated. Since you found many values to fall between two different rows at the 25.7 phase shift, let's try PAL 26.7:

 

[Thomas Jentzsch]

Referring to 'alternate columns' I was referencing the 'brightness' scale. Colors change moving up and down the chart with the scale (0-F)...

I cannot confirm that for my hardware. If I change the brightness, the colors do not change or move around.

 

Ultimately the idea is not to gear a palette towards just one television CRT type. I don't want to hack colors or removing just 'colors' from one row and place it under the 'grayscale' of another row due to the behavior or/and preference of one television or model type, or to account for CRT factors that need to be applied after a base palette is established.

OK, we can concentrate on the colors first and I really hope more people with a PAL setup who can support us here. And later we can adjust the relative brightness of the colors/grays.

 

...but right now I'm looking for the best matching base sticking with calculated palette values. (I.E. Hue Begin Degree of 57 (For PAL), Phase Shift Degree of 24.7, 25.7, 26.7, etc.). The colors and grayscale produced from there while keeping Brightness and Contrast neutral serve as the base.

I understand that approach and in theory it should deliver optimal results. But at least my hardware does not support that theory. The phase shift degree does not seem linear. And if you look at e.g. the existing z26 and Stella PAL palettes, they seem to indicate the same. Of course to confirm that, we would need more people testing and reporting. ANYONE?

 

Thanks for the feedback on the chart. It is very much appreciated.

Same here for your work.

 

Since you found many values to fall between two different rows at the 25.7 phase shift, let's try PAL 26.7:

I will do again, but I doubt the result will be much better. Because of the very non-linear phase shift I am observing.

 

Also, since PAL only has 12 colors (+ gray), as long as your graphic has 15 colors (+ gray), there is not even a theoretical chance for a perfect match. Therefore, shouldn't the phase shift for PAL be larger than for NTSC?

Edited September 28, 2013 by Thomas Jentzsch

[Thomas Jentzsch]

Just to get the theory right, here is my understanding of it:

 

After 360° the colors repeat. So if you want 15 evenly spaced colors, you need a 360°/15 = 24° phase shift. But the Atari setup instructions we know indicate, that 1x and Fx should be calibrated to be close together. So effectively we have just 14 NTSC colors. Which means 360°/14 = 25.7°. I don't know if and how many Atari console have set up like that. From my disassemblies I know that Fx was used by the programmers (though less often than other values). Which only makes sense if Fx was different than 1x. So maybe we are looking for something like 14.5 colors (24.8°)? More people testing their consoles would help for sure.

 

For PAL the values have to be different, if the consoles were set up identically. Then we would need a phase shift between 360°/12 = 30° and 360°/11 = 32.7°.

 

But is the shift really linear? And to make things more complicated, we cannot ignore that TVs may not display linear input in a linear way. I am not sure if my console or my TV are the reason for my results.

 

And on top of it, there are the PC graphics cards and monitors why must not be faithful to the input too.

Edited September 28, 2013 by Thomas Jentzsch

[Thomas Jentzsch]

One small missing note: On my TV the darker values of 3x and 5x (x0, x2, x4) are 100% identical, above 5x tends more towards blue and 3x more towards yellow.

[Trebor]

I believe I'm not explaining myself correctly or we're using different terminologies referring to the same thing. If you look at the pictures I posted for the NTSC palette you'll notice that if you move up and down the chart (Rows 0-F) the colors change for each row. If you move left and right in the chart (Columns 0-F) the brightness for the rows change. Am I making sense now?

You're understanding is correct, and the rearranging of color rows is what makes PAL 2600 trickier than say PAL 7800's.

None of this can be exact and there will be slight discrepancies due to slight differences in components (Whether display or console related), and once you start factoring the "pot" of the system, it throws another level of complexity. I covered much of this with the 7800 for both NTSC and PAL. 2600 NTSC is very simple once the NTSC 7800 palettes were established, for the most part, it is just half the columns (8 instead of 16).

Here's what I can provide on my observations with the 7800 and 2600 NTSC systems:

Both start (As soon as you hit the power button) with ~24 degree phase.

The 7800 phase shifts ('warms up' as some like to call it) much faster than the 2600. It goes from 24 degrees (As soon as you hit the power button) to ~24.7 within seconds. It hits ~25.7 within a couple of minutes. In approx. 20-30 minutes you are at ~26.7. Some systems will hover around that phase shift, while others go much higher to at least ~27.7 degrees with more consistent run time.

When I made my opening post I presume the 2600 behaved similar to the 7800 in phase shifting. That is why I originally just posted 25.7 and 26.7 varieties for the 2600 as that was believed to be all that was needed for the most part to cover the majority of reasonable amount of system runtime. However, real world observations proved differently. The 2600 phase shifts much slower. It takes a more time.

Deduced from the above, my presumption was that PAL 2600 systems behave a similar way to the NTSC 2600 regarding phase shifting, staying mostly in the 24.7 - 25.2 degree range of shift. I thought it was further confirmed when you mentioned too much green too quickly as with the higher the phase shift the quicker and more green sooner is evident.

However, I could be completely wrong about PAL 2600 and especially as it relates to the 'newer' manufactured models like Vader and especially the Junior. If you look back at the Solaris pictures/links I posted earlier, the last one with the strongest browns and the more saturated colors came from a Junior; which could be an indicator of a higher phase shift or just someone who likes their display that way and made the adjustments via display rather than the default of the console by itself.

Further, if the former is true, this would also coincide with your statement and findings of saturation/colors and the 26.7 palette coming close® to what you see on the Vader. I also have a theory with NTSC 7800's that some newer models (A3/X Serials) may phase overall higher than many older models (EP, AT Serials) due to cheaper components and perhaps overall 'sloppier' manufacturing.

There's a variable resistor that impacts 7800 colors too and overtime not only do you see the phase shift but the same changes made with the pot adjustment for a 7800 (Influence of blue over red and green) can change automatically as well. You can see that work and examples here: http://atariage.com/forums/topic/216413-pot-auto-adjust-phasing-color-shifts/

Certainty regarding 2600 PAL palette(s) will be easier with a greater sample of data to review. A variety of consoles, variety of consoles on CRT displays, and the interest from more in the PAL community to assist would be beneficial. Having at least one or two more PAL consoles or/and at least one or two more PAL displays from the community to provide some input, indeed would be great; but if not, the best can still be achieved with the technical and observed data possessed.

If worse comes to worse, the idea is to get feedback from 25.7/26.7 and if needed bring in 24.7 PAL palettes. Have the rows labeled and where there is indecision refer back to where one color is a definite match and build from there in a pattern that's established. Hypothetical example:

PHASE SHIFT 25.7
2x = F or E
3x = 2 or 3
4x = E or D
5x = 4
6x = D or C
7x = 5 or 6
8x = C
9x = 8

PHASE SHIFT 26.7
2x = F or E
3x = 2
4x = E or D
5x = 3 or 4
6x = D
7x = 5 or 6
8x = D or C
9x = 8 or 9

With the above data the PAL palette established would then be following patterns and reasoning:
2x = F
3x = 2
4x = E
5x = 4
6x = D
7x = 6
8x = C
9x = 8

Again, the above is a small hypothetical example, but I think it demonstrates what can be accomplished best.

We then apply the 57 degree start point with the variety of phase shifts to stay as objective/neutral as possible. Factoring in how (non)linear the phase shifting may or may not be on some or even many console is something that can be tweaked after the fact. I've literally created several hundred of 7800 palettes, and a few hundred have been posted (That's of good ones - Ones that are 'wrong' or impossible, brings the total to well over 1000).

Nonetheless, much of that variety is due to not have the any video controls in the ProSystem emulator that many like to utilize. Thankfully with Stella, you can tweak many video options and CRT effects.

It's the ('ideal') 2600 PAL base I would like to establish first without trying to factor the other plethora of possibilities as this starts to venture into something similar as CRT effects.

[Thomas Jentzsch]

I believe I'm not explaining myself correctly or we're using different terminologies referring to the same thing. If you look at the pictures I posted for the NTSC palette you'll notice that if you move up and down the chart (Rows 0-F) the colors change for each row. If you move left and right in the chart (Columns 0-F) the brightness for the rows change. Am I making sense now?

Perfect sense.

 

You're understanding is correct, and the rearranging of color rows is what makes PAL 2600 trickier than say PAL 7800's.

There is no requirement to work with unsorted color rows. See my test program above.

 

Deduced from the above, my presumption was that PAL 2600 systems behave a similar way to the NTSC 2600 regarding phase shifting, staying mostly in the 24.7 - 25.2 degree range of shift. I thought it was further confirmed when you mentioned too much green too quickly as with the higher the phase shift the quicker and more green sooner is evident.

I think there was a misunderstanding. When I talked about phase shift I meant the e.g. 25.7°.

 

Further, if the former is true, this would also coincide with your statement and findings of saturation/colors and the 26.7 palette coming close® to what you see on the Vader.

Why so low values? I my last post I explained that in theory for PAL the value has to be larger than 30°. Or did I get something wrong there?

 

Again, the above is a small hypothetical example, but I think it demonstrates what can be accomplished best.

If we have multiple testers when can define an average palette based on their reports. And we can define additional palettes which cover the full range.

 

Thankfully with Stella, you can tweak many video options and CRT effects.

I we find the correct parameters, then eventually Stella may be able to adjust the palette via the GUI, by just changing those parameters.

 

It's the ('ideal') 2600 PAL base I would like to establish first without trying to factor the other plethora of possibilities as this starts to venture into something similar as CRT effects.

I that's the goal, then I still don't get why the ideal PAL base does not have higher phase differences. Or are the 12 PAL colors not going full circle?

[Thomas Jentzsch]

Here the new list:

 

2x = ~1.5

4x = ~2.75

6x = ~4

8x = ~5

Ax = ~6

Cx = ~7

Dx = ~8

Bx = ~8.75

9x = ~9.75

7x = ~10.75

5x = ~12

3x = ~12.5

[Trebor]

It's going to take me some time, but the game plan is this...

 

Roll out 27.2 and 27.7 degree phase shifts for NTSC and PAL for the 2600. They will be setup with the grayscale previously released along with corresponding 'brightness' (AKA x0, x2, x4, x6, x8, xA, xC, xE for both colors and grayscale).

 

I will also put out an alternate palette set for NTSC and PAL containing the same grayscale for all the phase shifts (24.7 thru 27.7 with 0.5 degree increments) with the alternate color columns. In other words, grayscale will remain: x0, x2, x4, x6, x8, xA, xC, xE; but the corresponding color 'brightness' will be x1, x3, x5, x7, x9, xB, xD, xF.

 

From there Thomas, you can pick and choose which PAL palette matches your configuration best. Out of the 14 PAL palettes that will be present, if one of the phase shifts (I.E. 27.2 or 27.7 alternate palette set) nails it, that's fantastic.

 

If not, and it take a few rows from one set, a few rows from another and even if a third or fourth set needs to be brought together, I am pretty certain, especially with the feedback you have provided with PAL 25.7 and 26.7 degree phase shifts, that we should be able to create exactly (or a close as possible to exact) what your Vader system and Sony television are providing regarding 2600 PAL palette.

[SeaGtGruff]

How are you calculating the NTSC colorburst? Hue 1 in the NTSC palette looks off to me. The axes of the YIQ color space are rotated 33 degrees from the axes of the YUV color space, but the degrees aren't counted from the Q axis, they're counted from the U axis (AFAIK). In other words, what you're calling 180 degrees for the NTSC colorburst should coincide with the -U axis-- in which case the V value should be 0.

 

The following illustrations should help explain what I mean. The first is a graticule for a vectorscope with the I and Q axes marked, as well as the positions of the six basic colors. The second is a screenshot of the YIQ "color cube" in the ColorSpace program (from Couleur.org); the "camera" position and color cube orientation are set so the colors line up with the vectorscope graticule. The third shows the YIQ color cube overlaid on the vectorscope graticule. The fourth is a screenshot of the YUV color cube in the ColorSpace program; the camera position and color cube orientation are set so the YUV color cube matches the orientation of the YIQ color cube. In all cases, the NTSC colorburst is "due west" (marked 0 degrees on the vectorscope graticule), which corresponds to the -U axis of the YUV color space.

 

 

As you can see, the NTSC colorburst is mostly yellow but with a distinct greenish tinge. In your NTSC palette it has an orange tinge rather than a greenish tinge. In fact, your hue 1 looks a lot like hue 2 from a palette I created based on RGB values calculated by a Visual Basic program I'd started working a while back:

 

 

I used a lower "brightness" than you did, based on some posts from the old stella mailing list, but I think yours probably looks closer to what a TV might display (since TVs have a higher gamma setting than computer monitors do). I also used a phase step based on the field service manual, which says that hue 15 should be as close to hue 1 as possible, so I set them equal (a phase step of about 25.71429 degrees). I personally disagree with that-- as Thomas implied, programmers wouldn't have used hue 15 if it weren't different than hue 1-- and on my Ataris they've always been different, but I was trying to make a palette that conformed with the field service manual. Based on my interpretation of the GTIA specs-- which I may well have misunderstood-- I think the phase step should be about 27 degrees, which puts hue 15 about 2-3rds between hue 1 and hue 2 (i.e., hue 15 = hue 1.6667):

 

 

Anyway, it looks like your hue 1 is off, which I'm guessing may be because you're shifting (rotating) the YUV color wheel such that the colorburst isn't coinciding with the -U axis?

 

[Thomas Jentzsch]

I personally disagree with that-- as Thomas implied, programmers wouldn't have used hue 15 if it weren't different than hue 1-- and on my Ataris they've always been different,

The problem here is, we do not know how well the Atari's were calibrated back then. I wouldn't be surprised, if the phase step was varying a lot. So depending on that, the programmer might have assumed pretty different values for Fx (somewhere between green and orange).

 

But maybe the consoles were better calibrated at Atari than at 3rd party companies? This is something we could verify by looking on Atari disassemblies. If they do not use Fx we have at least some evidence.

[Thomas Jentzsch]

5 minutes later, there goes my theory.

 

Sprintmaster and Sorcerer do use Fx. Sorcerer even together with 1x!

[SeaGtGruff]

5 minutes later, there goes my theory.

 

Sprintmaster and Sorcerer do use Fx. Sorcerer even together with 1x!

 

Yeah, on my Ataris (two 2600s, a 7800, a 600XL, a 800XL, two 65XEs, and two 130XEs) they were always different. But as someone pointed out in another thread (8-bit programming forum), the phase shift actually affects the pixel on a sub-color-clock level, by which I mean that if you do set hue 1 and hue 15 to be identical, there should be (as I understand it) a "line" between them (similar to the way you get lines of "extra" color between two different hues) if you display them side-by-side because of their phase shifts (delays) being different-- hue 15 being 360 degrees from hue 1, if you see what I mean. But that doesn't account for the fact that whenever I displayed them next to each other *vertically*, they were always different. In my experience, hue 15 has almost always been somewhere between hue 1 and hue 2, although on some of my Ataris it has been somewhere between hue 14 and hue 1. Sometimes it even seems to jump around on the same machine-- for example, when I recorded my 7800 to DVD and played it back frame-by-frame on my computer, one frame might show hue 15 with a greenish tinge, then the next frame might show it with a brownish tinge! This may be a natural consequence of tiny variations in the phase step from frame to frame, with the variance being most noticeable in hue 15 because it's 14 times the phase step.

[Thomas Jentzsch]

Since we do not know what is "right", I suppose Trebor's approach with different phase steps is the only option we have.

 

Anything between 360°/13.5 and 360°/15.5 seems possible for NTSC.

[SeaGtGruff]

Since we do not know what is "right", I suppose Trebor's approach with different phase steps is the only option we have.

 

Anything between 360°/13.5 and 360°/15.5 seems possible for NTSC.

 

Yes, that's the closest we'll get to having a color pot adjustment in Stella-- which may eventually happen one day.

 

I just want to be sure the NTSC palette is starting from the right value for hue 1. Hue 1 in Trebor's NTSC palette looks like it would be "correct" for hue 1 in the Atari 800's PAL palette, which is different than hue 1 in the NTSC palette.