32X colors

[saxman]

By doing some careful color matching between the 32X and the Genesis, I have concluded the following:

1) The 32X uses a linear color ramp.

2) The green seems to need a slight boost (less than a single color step) over the red and blue to come close to a Genesis color.

It's the second point that has me wondering: Have the individual R/G/B channels been measured on the Genesis VDP? I know there have been posts in the past regarding voltage measurements and true color approximations. But do we know if each channel has the same measurements? If any of them are off (like the green), then that may affect my understanding of 32X color approximations.

Edit:
In fact, examining a photo taken of my TV screen (I know, that's not 100% scientific, but still serves a purpose here), the Genesis colors seem to have a matching blue/green with a weaker red, while the 32X colors seem to have a matching red/green and a stronger blue.

Edit 2:
I've attached the TV picture, along with a modified picture from Ares showing that if red and blue are differentiated by the same amount (thus creating a constant) and the green is shifted down from matching blue to matching red on the 32X, the result looks pretty darn close.

[TmEE co.(TM)]

I did measurements on 32X a while ago, using a test ROM ChillyWilly made, but I have had no luck locating it on this forum but it must be somewhere around here.

one 32X color ramp

32Xvideo0.jpg (73.16 KiB) Viewed 1070 times

My measurements found that the ramps are indeed linear and that blue color is just about 1x step lower in absolute magnitude compared to red and green. Using 27 ohm resistor instead of original 33ohm on R34 will fix it, or alternatively soldering 150 ohm resistor in parallel with it which brings the output to equality also. Then all the ramps have equal magnitude and the yellow tint I have seen on every 32X output disappears. It is strange that you have hardware where green is the different color from others, do you happen to know if IC6 in your hardware is 315-5781 or 315-5781A (or perhaps something else entirely) ?

MD measurements are found here : http://gendev.spritesmind.net/forum/vie ... php?t=2188

MD ramp

MDcolorRamp.jpg (100.07 KiB) Viewed 1070 times

[saxman]

Okay, yeah I looked for such a thread and couldn't quite locate it. I'm sure it's around somewhere.

To make absolutely sure I wasn't crazy or didn't have some weird fluke, I tried it again today and did a whole slew of value comparisons and combinations.

From my comparisons, 32X green 0x1F matches Genesis green 0xE. I don't have any trouble matching any of the greens based on the assumed steps of 8.226. They're always spot on with the Genesis color measurements done all those years ago.

Both the reds and blues are pretty accurate when they're dark. But at the very top, they're about one step too bright (i.e. brighter than the brightest reds and blues on the Genesis). So for example, matching the 32X to the brightest Genesis cyan and magenta would mean using 0x1E1F00 and 0x001F1E.

For yours, is it the blue that's more accurate (meaning 0x1F matches Genesis 0xE), or is it the green and red?

As for the chip, I have no idea what's inside my 32X. I may attempt taking it apart to find out. I find it odd that mine is apparently different than yours. (How many units have you tested?) For whatever it may be worth, my unit is MK-84000 (no A), serial P51013943. It's in very good condition with not many signs of wear from previous use.

[TmEE co.(TM)]

There is one more thing to consider though which is RGB vs CVBS. The latter is subject to variety of errors arising in the analog tolerances during the conversion process. Phase shifts cause hue changes, especially with NTSC and luma itself is also subject to tolerances in the resistors in the chips. I have not actually ever tried to characterise any differences there, since I basically never use CVBS or YC/S-video, only RGB. I'm now actually curious and I think there are ways to try to map out any level based nonlinearities but I'm not completely sure how to go about testing for such a thing...

There's also room for some connection related deficiencies between MD and 32X itself, and a slightly poor connection on any of the AV conectors can have a slight effect on absolute level on the incoming RGB signals, and same with the ribbon cables inside the 32X since analog video passes through them. 32X itself only switches together the RGB signals from the 2 sides, and then after the switching it goes to encoder that then outputs CVBS+YC. The CVBS signal from MD is used though, but to align the chroma subcarrier with one in 32X to have same fixed dot-crawl/jailbars pattern instead of randomly moving one.

I have noticed blue being consistently lower on 3 units, two euro ones (VA0 and VA1) and one USA unit (VA1) and all of them had A version of IC6. The plastic I have at hand (with insides that don't match its original inners) say 140006665 serial number and MK-84200-50 model number. I have also heard reports from other people seeing the same yellow tint on 32X side. With the fix I mentioned, things level out and all the individual colors track the others in identical way. Perhaps you have the non-A VDP in your 32X and it has slightly different analog performance on the RGB output, leading to the difference you see.

MD RGB output doesn't have any color channel imbalance, other than prominent jailbars in blue channel on unmodded units which arise from poor routing of the high impedance video signals from VDP, all 3 channels are equal on every chip revision I have measured things on from all the VDP versions from MD1 to the final one in Genny3 VA2. Only exception is the ASIC found in MD2 VA2, this seems to be USA only and I don't have one at hand to play with.