Wide-Boy64 AGB ターゲットボード専用 (TS2) Problem when connected to a retail modified GBA

[voodooween]

Hello @msinger and @RWeick!

I am contacting you here about the WideBoy TS2 that we discussed through email earlier.

So, I own a Nintendo TS2 WideBoy :

And unfortunatelly that type of Wide-Boy only runs when used with a Game Boy Advance TS2 Board and as you can see on the 2nd picture, the TS2 WideBoy has no CPU whatsoever, so it takes everything from the TS2 Board (cartridge slot, cpu, sound etc etc).

So I tried to create my own "brain" by following the pinout that his present on the TS2 Board.

But the result wasn't the one I expected, the picture is crooked/twisted

When it should look like that :

I have tried with "AGB-CPU-01" and "AGB-CPU-10" and exact same behavior, I also tried recently to shield the DCK, SPL and SPS lines but same result again.

So I don't know if my TS2 Wide-Boy is the problem (damaged/something else) or if it would only work with a TS2 Board/Prototype GBA.

My version should be the same as the one that was previously dumped (I took a picture with my rev : 109b, unfortunately the other owner won't be able to do the same since you need to have a GBA connected to the TS2 WideBoy to let you access the menu with the 2nd controller).

Is anybody as an idea on why it looks like and if it's possible to make it work with a retail GBA or if it's not.

Thanks!

[msinger]

I think what I wrote earlier by email, about maybe the pixel clock clocking in each pixel two times may be wrong. By looking at the pictures, I think there is just one additional pixel per row. Could this be right? This reminds me about the DMG display. On the DMG Game Boy, the pixel clock is used to clock the first 159 of the 160 pixels of a row. The very last pixel of each row doesn't have an edge on the pixel clock line. The last one is clocked in by the H-blank signal. The original DMG display however doesn't have a problem when it received 160 pixel clock pulses per row. It works with 159 and with 160. But repro display modules, like the one from Benn Venn or Funny Playing won't work with 160 clock pulses. Maybe there is something similar here. Maybe the prototype AGB CPU is issuing one pixel clock less than the later CPU revisions. The original AGB display maybe doesn't care about this extra clock pulse that the newer CPUs are emitting, just like the DMG display doesn't care. But the WideBoy may have a problem with it, shifting each line by one pixel to the right. But of course, I don't know. We sadly can't measure the amount of clock pulses the prototype is generating, if we don't have one ourselves.

About the cables being shielded: Those cables look like the more "modern" PATA aka. IDE cables to me. The ones that have 80 individual strands of thin cables in stead of 40. Those PATA cables have every second cable connected to ground to give at least some shielding. They could have used such cables for the WideBoy. (Maybe not actual PATA cabes, since there are some unconnected pins, I know, but this 80 pin style with ground wires.)

If you want to dump the PROM, you could use any small microcontroller that has 3.3V IO voltage. The PROM is socketed luckily. You could also buy the same FPGA board that I used and connect the PROM there. There is also a smaller one, called IceStick. It has 16 IOs, so it should also work.

If you would buy one of those FPGA boards, then you could also use it to mess around with the pixel clock. We could configure it to remove the last clock pulse from the pixel clock line to test my hypothesis, see how it affects the image.

[voodooween]

Do you have a type of a cable you would recommend me using, to eliminate a possible problem and which FPGA board should I buy?
The same as you used on your website : ICE40HX8K-B-EVN Or the IceStick that looks to be ICE40HX1K-STICK-EVN?

Edit : I ended up buying "ICE40HX8K-B-EVN" it looks to be a little bit more useful than the stick version.
I am not expert with FPGA (never did anything with FPGA) so I think I'll need a little bit of help on how to dump that chip.

[msinger]

I don't know where to get such cables. It was just an idea, but now I think it is unlikely that the cables are the problem. The picture on the screen would most likely move around, when the signal quality is the issue. But it looks like it is very stable with a constant shift to the right for each row.

The HX8K board is good. We just need to get the tools (IceStorm, Yosys and Nextpnr) to run on Windows. Most important is IceStorm, which is used for programming the binary into the serial EEPROM on the board. They don't seem to have precompiled binaries for Windows. I'll try to figure out how to build them. If all fails, you could still boot an USB stick with a small Linux distro.

You need to solder in some pin headers on the board. On is already there out of the box (a female one), the other three, you have to get by yourself. I soldered in three male pin headers for the missing ones, since I find it more convenient, but that's up to you.

[voodooween]

Oh okay, I have never really used something else outside of Windows.

I should receive the break out board in a few days I think (I ordered it from DigiKey).

[msinger]

Oh, and I would recommend getting boards and cables like these, if you don't have any yet. Makes it easier to connect the PROM to the board.

[voodooween]

I fortunately already have that type of stuff.

[voodooween]

So, I tested the TS2 WideBoy again yesterday and found something curious about the image orientation of the GBA signal.

IMG_7918.mov

I have tried a 100pF capacitor to the "DCK" line and it kinda worked, but only when I am touching the capacitor, if I solder the cap and leave it alone, the image is crooked again.
And I also can twist the image if I touch/jiggle the cable from the TS2 side.

Then I tried to short "R36" on the GBA side (it's directly connected to DCK from the GBA CPU) and it worked when I shorted it myself with a pair of tweezers, but when I bridged R36 with a blob of solder, crooked again.

[voodooween]

Yes, the normal GBA WideBoy has a different pinout because it also has to work with the IS-AGB-CAPTURE

They both use the same controller.

[voodooween]

Okay, I just tried pin 36 to SPL and 32 to P8 and it works!

[msinger]

Congratulations! Now you have a working WideBoy AGB TS2. 🎊

I was thinking about audio. You could try connecting the L and R audio output of the GBA to pins 1 and 2 like in my WideBoy AGB schematics. In your tracings, it looks like pins 1 and 2 go to some resistors that could be the analog part. So maybe on the WideBoy CGB those pins are the same as on the AGB. If this doesn't work, you could connect the audio lines to pins 24 and 49 of the N64 edge connector (or figure out where those are connected on the CN2 connector first, so you don't have to solder at the edge connector directly). It is not guaranteed to work. I think the code running on the N64 has to enable those analog audio inputs from the cartridge connector first before you can hear them. But I don't know much about the N64, so I'm not sure. It may be possible to patch this into the N64 code that resides in the flash chip.

I love how you added the pin layouts at the top of your tracing pictures. Helps a lot! 👍 Does this mean the two rows are swapped on the WideBoy compared to the TS2 board? Also, when I was suggesting to move SPL and P8 up by one it somehow sounded logical in my head. Now after thinking about this, it must have been a pain in the ass for them to just move the even pins in the cable without moving the odd ones too. Have you tested the P9 button (I don't know which one this is) if it works too? Could you post a final pinout of the 40-pin connector for clarification?

[voodooween]

I tried pin 1 and pin 2 for the audio and nothing unfortunately, I haven't tried pin 24 & 49 yet.

And I actually can't believe I haven't spotted that there was an error in the TS2 pinout, because I looked at the back of the TS2 board for easily 10 minutes straight and there was a ground point that was marked as --- on the pinout... But since everything is mirrored with the pinout, the connector and the TS2 WideBoy, I guess my brain must have been lost... hahaha.

You can see that the SPL line is actually grounded on the 2nd picture...

And for the pinout I think the "real" pinout must be that one. Pin 40 is a 3,3V, but I left it as ---. I didn't wanted to power the GBA with that 3,3V, I think using double A's batteries is the best choice imo.

P9 always worked, only P8 was a problem.

[msinger]

I can't believe they labeled it wrong. That is so nasty of them. 😠

Yes, I wouldn't connect 3.3V either. The GBA needs to generate its own voltages, since it also needs 5V because of the DMG/CGB compatibility.