A1000 and weird colors

[webmany]

I have an NTSC A1000 with 8Meg fast memory and a gotek. When I try to run Cannon Fodder I get really strange colors on the screen. I do not get this problem with other games. I tried emulating it on Winuae and was not able to get the same output, so thinking a hardware issue. Any ideas?

This is on the commodore monitor connected to the TV Mod port


This is on an LCD TV connected to the video port


Also get the same result with Ghosts N Gobblins.

[Rotareneg]

The early A1000s didn't have EHB, perhaps that's the problem? In WinUAE, enable No-EHB Denise in Adv. Chipset to check.

*update* That's not it, just checked in WinUAE.

[Toni Wilen]

Does game's title screens ("not in any way endorsed" screen for example) also have similar color problem?

I checked game's copper list and I didn't see anything interested except it has dual playfield priority bit set (playfield 2 has priority over playfield 1) which does nothing if dual playfield is not enabled. Perhaps it makes difference if Denise is early no-EHB Denise..

EDIT: does same happen if you use real floppy drive?

[webmany]

Not able to check with a real floppy yet, I am trying to track one down at the moment (external one, but the gotek in it and use the internal one as the second drive).

I will check the title screen tonight, in Ghosts N Goblins it was messed up.

[AmigaHope]

That's weird, it looks like an inverted chroma signal.

I read somewhere that Cannon Fodder DOES use EHB in parts btw, and I'm not sure if UAE's emulation of early Denise chips is 100% accurate. Denise 8362 R5 and earlier are the pre-EHB chips, if you care to look at your motherboard, R6 and later have EHB.

Also you could try some games that definitely do use EHB like Black Crypt, Super Cars 2, Agony (the loading screens only).

EDIT:

Just saw that you're having issues with Ghosts'n'Goblins which is def not an EHB game. It could be something is not playing nice with the composite modulator. Have you tried the RGB port directly? Maybe hooking up an A520 modulator to the RGB port?

I say this because it seems more than coincidental that the things that are supposed to be blue are red, the things that are supposed to be green are blue, and the things that are supposed to be red are green, as opposed to just random colors for everything.

[mark_k]

Hmm, interesting... Are you able to test via analogue RGB connection to the 1084? My guess is, the picture will look fine then.

My theory: The hardware register bplcon0 has a "COLOR" bit. On the A1000 only, that controls whether the composite video signal is monochrome or in colour. Technically, setting COLOR to 0 probably stops the colour burst being present in the back porch of the composite video signal.

Since no other Amiga model supports that feature, various games and demos accidentally don't set the COLOR bit. (The developers never tested on A1000.)

It seems Cannon Fodder is one such game; at least, the WHDLoad info page says "bplcon0.color bits fixed"

Since your A1000 is pimped-up, you could try using the Cannon Fodder WHDLoad installer, install to RAM: and run the game from there.

If you want to check other games that might have colour issues, Google search for: bplcon0 site:whdload.de

So why do you see those strange/wrong intense colours? My guess is, enough of the Amiga ~3.58MHz clock signal leaks into the composite video signal, so even with the bplcon0 COLOR bit 0, the video decoder in the monitor sees a very weak colour burst (of the wrong phase). Thus it "amplifies" the chrominance signal which may still be present in the active part of each scan line.

Do you have any other TVs or monitors (especially old ones) to test?

[Toni Wilen]

GhostnGoblins title screen does not have color bit set but main game screen does set it.

[webmany]

I opened up the 1000 and it does have the denise without EHB. I double checked Cannon Fodder and the title screen also has the problem.

I am trying to track down a RGB cable to test using it. Currently I am using the TV MOD port.

I have an IDE adapter comming, and once I have it installed I will give whdload a try.

Thanks for all the help.

[TjLaZer]

Originally Posted by webmany

Not able to check with a real floppy yet, I am trying to track one down at the moment (external one, but the gotek in it and use the internal one as the second drive).

I will check the title screen tonight, in Ghosts N Goblins it was messed up.

GNG is a PAL game so it might not work on a NTSC Amiga.

[webmany]

Ghosts N Goblins gets the title screen, but the game does not run, but I think it it need 1Meg of chip ram, and I only have 512K.

[Toni Wilen]

Originally Posted by webmany

Ghosts N Goblins gets the title screen, but the game does not run, but I think it it need 1Meg of chip ram, and I only have 512K.

It most likely requires "slow ram" (A500 trapdoor RAM). Does the title screen have correct colors? (It should not if problem is due to color bit not being set)

btw, could you run this test program: http://eab.abime.net/showpost.php?p=1289627&postcount=8

[mark_k]

If you post links to the Cannon Fodder ADFs you're using, I could have a go at patching it to always set the COLOR bit.

[webmany]

The title screen did not have the proper colors. I'll give the program a try.

Originally Posted by Toni Wilen

It most likely requires "slow ram" (A500 trapdoor RAM). Does the title screen have correct colors? (It should not if problem is due to color bit not being set)

btw, could you run this test program: http://eab.abime.net/showpost.php?p=1289627&postcount=8

[webmany]

That would be great.

I added the ones I am using to the zone, but any version would do:

http://eab.abime.net/zone/CannonFodder.zip

Originally Posted by mark_k

If you post links to the Cannon Fodder ADFs you're using, I could have a go at patching it to always set the COLOR bit.

[Toni Wilen]

Originally Posted by webmany

The title screen did not have the proper colors. I'll give the program a try.

Ok. So it most likely is COLOR bit side-effect.

I have early european A1000 (PAL Agnus but NTSC composite output and NTSC main crystal), I can check what COLOR bit does to composite signal with a scope and create test program that can be used to test what happens to colors. (if it is color phase shift or something more complex)

EDIT: I checked (bad quality) A1000 schematics and it looks like COLOR bit (=Denise "Burst" pin 18) controls pin 1 of MC1377 RGB to composite video encoder. Datasheet says pin 1 controls color burst duration so apparently it sets some "illegal" voltage level which causes chip to generate very quick or very low level pulse. Or something like that. Chip does not have any color burst enable/disable input signal.

[mark_k]

Maybe there could be some difference if your machine has PAL Agnus? (Different number of colour clocks per line so different colour burst phase maybe?)

Look at the MC1377 datasheet from here. Page 8: "Another means of inhibiting the burst pulse is to set Pin 1 either above 1.3 Vdc or below 1.0 Vdc for the duration that burst is not desired."

From the A1000 schematic, Denise pin 18 is buffered, then via simple resistor network which (by my possibly wrong calculation) leads to these voltages on MC1377 pin 1: low => ~1.15V, high -> ~3.8V

So the Denise BURST pin goes low for the duration of the colour burst in each active scan line. (I assume it's only in the active lines, not during the vertical blanking period.) And I assume setting bplcon0 COLOR to 0 just means the BURST pin always stays high.

I wonder if aging capacitors could be the cause of the strange colours? If that causes more of the 3.58MHz clock to leak into the output composite video signal, the video decoder in the monitor sees a very weak colour burst, so it amplifies the chrominance component of the video signal by far too much.

Or maybe other models of monitor wouldn't see that low-level signal during the colour burst period, and thus would activate a "colour killer" so the scanline would be shown in monochrome. (Back when there were both B&W and colour TV broadcasts, that feature eliminated colour artifacts appearing when viewing B&W programs.)

Some 1084 models have a "VCR" button which is supposed to improve the picture for poor-quality video signals. Maybe changing that setting could change things here?

[Toni Wilen]

I checked the signal with a scope:

if COLOR bit is zero: constant ~3.9v at pin 1.
if COLOR bit is one: high about 60.90us (4.10v), low about 2.5us (1.15v)

EDIT: Tested also composite out, color=0 seems to disable color burst. But there is lots of noise and scope does not want to sync to it, probably because it is PAL/NTSC hybrid. Or my A1000 video output isn't fully working.. Obviously this is impossible to confirm without swapping components

[webmany]

Just wondering if you had a chance to alter the ADFs.

Originally Posted by mark_k

If you post links to the Cannon Fodder ADFs you're using, I could have a go at patching it to always set the COLOR bit.

[mark_k]

I'm working on it.

By the way, it seems the WHDLoad installer requires 1MB chip RAM so that rules out using it on your A1000. (Maybe it would be possible for the installer developer to reduce the amount of chip RAM needed though. Because on the Cannon Fodder disk there are three versions of the game code assembled for different addresses.)

[mark_k]

Fully patching the game code will be a bit of a pain. I'd need to extract the RNC-crunched files from ADF, patch, recompress and write back to ADF. Those steps aren't actually that hard, the main hassle being the data isn't written continuously in the ADF (the last two bytes of each sector seem to contain the track number or something).

Actually finding where to patch is relatively simple since I can start from the WHDLoad installer source code.