rbairos

Sorry I missed this. I've added it to the list of 2600 hardware projects:

Thanks to all of you on this thread really. You've all been so helpful on different parts of this problem throughout. I feel I can finally get ready to ship now. Just gotta turn this into a patch and email it off for testing, but otherwise, feeling hopeful.

Presented without comment... (THANK YOU!)

Do you think it might show the problems I'm seeing with the correct developer tia options, or have I fallen into a brand new category with this kernel? It did show the problem exactly for the tiny digits kernel example.

If that was the solution, but currently I can't even get any clear movie image on PAL yet. Just the tiny digits sample kernel. I think my next step is either porting the latest hardware firmware to Stella, or porting the old stella moviecart code to the firmware.

So... manually adjusting the kernel so player positions are 6 positions further to the right on startup fixes the digits test on PAL machine, (but breaks it on original Ataris). It seems to almost fix the title screen too, but the content isn't quite right either (ignoring the b/w). A couple gaps in the title screen, but the content seems to have lots of somewhat duplicated half columns. Not sure what's going on at all now.

Right. Good lead. So manually moving the players 6 positions to the right in stella as mentioned earlier in the thread, brings up the same glitch I see on the PALs. EDIT: This also comes up when I turn on Kool-aid TIA variant! On both my digit and digit-reverse bin. Exactly the same on PAL compared to Kook-aid TIA variant emulation. That's good news. However playing an MVC file is fine. This goes back to what @JetSetIlly was saying about the different kernels setups used. Im trying to play an old MovieCart-Version1 hardware cart but unfortunately its not getting past the title screen, and my firmware compiler is out of date for that setup. Sorry If I've missed something everyone's been trying to tell me on the solution, just lots of data swirling around my confused head. Cheers!

Sorry I take some of this back I don't think I need the extra sta wsync. It should be hit when I clear 256 bytes of RAM in the initial setup. Not sure what this all means. I guess it went from one consistent (but wrong) set of player positions to another (less wrong) on startup with my change. Very confusing. It must be moving back and forth the full 8 else the image would snake diagonally too. Maybe I need to pulse respx each frame next

Here's another test where I just reverse the bits of digit 3: photo 3: original photo 4: initial wsync stella reference Much better, but that digit 2 has a tail I don't see on the emulator... and the emulator doesn't have a gap between 3 and 4 as well.... Drats. Here it is on FlashBack2 :

Okay, some possible good news. Photo 1 is the original kernel. Photo 2 has an initial WSYNC before setting RESPx. Perhaps the PAL TIA's start at a different color clock compared to the NTSC TIA's and it was wrong of me to assume an initial clock value on startup. Looking better, but not sure if there are some phantom pixels yet. Might also be the NTSC TV / PAL output fuzziness. @JetSetIlly And here is stella emulator for reference:

Okay, the PAL Ataris have arrived! I realize the 6-switch has a funky PAL connector on one end, so I'll have to pick up an adapter for that. and I couldn't get the Jr to show anything on my 1977 CRT (even going through the UHF channels). I suspect I need an RCA to UHF adapter (since the Jr is plugged straight into the VHF input on that 1977 CRT). *However* plugging into my 2009 LCD, does show output from the Jr! It's black and white and static for noise, but it shows something! And it shows the original column swapping issue from above. First step: Start running the basic 'digits' kernel. Then modify it so its pulsing [EDIT] RESPx at least once per frame. and/or put in an initial wsync

No. I picked up a jr and [EDIT]6 switch woody from UK England. Arriving April 5 hopefully. Also hoping that shows the issue. From what I've read there are at least two Atari pal versions made (pal-i and pal-b?) but I *think* they use the same pal TIA.

EDIT: I ended up ordering one of these: Hopefully, will allow me to see things properly on my PC if it works as advertised: https://www.ebay.ca/itm/233079792659

That was great!

So, Im waiting for the arrival of a PAL system and spent the evening trying to figure out how to view a PAL signal. Just realized, I can just play an NTSC-formatted movie/kernel and play it on an NTSC TV right? Colors will be different, but if I output a regular 262 lines, that should output 60-Hz that I can experiment with right? EDIT: from some more reading I think it will work, except the image will be black and white (and use the Atari pal color palette). That's more than fine for testing the kernel issues.

Never noticed until just now the TV is from 1977, birth year of the Atari!

So both 7800 and FlashBack2 will work on the same card now. No reconfiguring will be needed. solution: monitor A12*A11 not just one or the other. Luckily in FlashBack2, A12 is stuck, but stuck high, and the kernel is just 512 bytes of ROM space.

I don't know. I think its separate issues. The 7800 uses A12 very differently than the 2600s Im guessing. Output wasn't just scambled, movie was playing back twice speed, very glitchy, staticy, etc. I should be getting a couple PAL systems first week of April, but feeling confident now! EDIT: That is to say, I never saw the column swapping artefact discussed above on the 7800. It's artefacts were very different.

Okay, update: I've got my hands on a 7800, saw the issues with garbled graphics, and some audio, did a bunch of trial+error testing and now have a fix. The issue is the 7800 doesn't let me ignore the A12 line like I can with the FB2 + 2600 (where I was using A11 as a proxy). Running attached update.frm on the sd-card by itself should patch it to work: (Please ignore poor CRT TV quality, think my blue died) update.frm

stay tuned... EDIT: Sorry, I guess this isn't kernel related, just 7800 so I'll post on the main moviecart thread.

Ug. Not getting too far yet. Changed the firmware to just dump in a quick loop: set data bus = rom[address & 1023] as fast as possible. In this case, Cave 1K from Thomas Jentzsch. It works fine on the modded FB2, but not the 7800. Im vaguely aware of some encryption/hash scheme that causes the 7800 to dump part of the cart on startup? The queried addresses on bootup are not what I see in the 2600. Even so, there's no logic in my test firmware. It just sets data bus as described above. I'll try with combat next.

Okay, I'm back, and now have a 7800 to test these suggestions on. Unit works fine, but MovieCart shows very garbled graphics (not just swapped columns). Similar to what others reported, including audio working (with clicks during silence) and joystick control working. Strangely, the audio comes out twice as fast, like chipmunks. I think there's still the more fundamental bus timing issue with the 7800 I need to tackle first. To start, I've modified the kernel to be much easier to debug. It's shortened 1K and shows digits 0 through 9. It's on the github if anyone's interested. I'll try to port this to the moviecart firmware soon and post screenshots.

I'm away from my development tools the next few days but I can say I'm pretty certain I was careful to transplant the same core kernel between all implementations. I knew the early 8 bit shift would be central to it working. I maybe have forgotten the initial wsync in the core.asm I recently put together last week but it should be moot as the overscan and vsync delay before restarting the frame on cycle 73. I think it is what spiceware is saying and the particular shift I'm using behaves differently on some TIAs. That's unfortunate as I'll have to replace single sta hmclr With two separate LDs and stores. Might be doable at the cost of halving the audio update rate.

But if it's on the wrong cycle how did it ever work?

Oh right. Stella still has the old title screen and setup. Firmware core.c has the new setup where title screen and movie are the same loop