N64 SRAM/FlashRAM results

Just wanted to post my results with implementing SRAM/FlashRAM support into my savegame dumper.

Forum love

The most active retro forums seem to be Benheck and Assemblergames. The first one is closed for new members and on the second it took one week until my thread got unlocked by a moderator. No comments after one week and today the thread got deleted.... Kind of disappointing, because I would have appreciated some external input. 

PIN mirroring

My first idea was to mirror AD0-AD15, ALE_L, ALE_H, VCC, GND and switch \READ, \WRITE.

The problem with this idea is that the 64drive will not start with the flashram connected to the mentioned PINs and if you connect them while the 64drive is already running you will just read crap. I also realized that the XPloder64 was interfering with the reading/writing commands. So I had to desolder the cartridge slot.

Full PIN switching

Because the PIN mirroring was not working I modified the gameshark cartridge so I can switch AD0-AD15, ALE_L, ALE_H, \READ and \WRITE during runtime. VCC and GND can of course be mirrored. This is the hardware setup:

Ugly PIN switch setup

Kind of messy, but works in general.

Writing the dumper

I thought this would be an easy task, because the N64 port of ScummVM already has FlashRAM support. It is based on hkz-libn64. Sadly neither the SRAM nor the FlashRAM routines were working for me at all. So I wrote my own routines based on the code from hkz-libn64 and LaC's n64 hardware dox. Against MESS and the 64drive SRAM/FlashRAM emulation these routines work perfectly. I used the following test to verify the functionality of the routines. I filled a buffer (sie of SRAM/FlashRAM) with values from a pseudo random generator and wrote the buffer to the SRAM/FlashRAM. Then I set the same SEED again for the pseudo random generator and filled the buffer with zeros. Now I read the SRAM/FlashRAM into the buffer and compared it to the results from the pseudo random generator. If this works it is very likely that my routines work.

The results

As mentioned above the routines work great against MESS and the 64drive, but not so much against an actual SRAM, FlashRAM chip. There are some strange read/write errors.

Strange SRAM/FlashRAM read errors

I am still not sure if this a problem of my software routines or because of bad wiring. I rewired several times and resoldered some parts, but I was still getting very strange results. I was able to back up F-Zero X (SRAM), 1080 (SRAM) and Pokemon Stadium (FlashRAM) which tells me that it is working in general. But it is not working perfect so I declared the code routines as experimental and I can not ensure that it will work properly for your games. Games I had no luck with are Zelda OOT (SRAM), Zelda MM (FlashRAM), Jet Force Gemini (FlashRAM) and Paper Mario (FlashRAM). It could just be a bad connection problem, but simply neither have the time nor the urge to create a custom board with full soldering for all 20 important connections with switches. These are all games anyway which I would start from the beginning if I ever wanted to play them again. So I backed up the three games which were most important to me.

The final dumper

Here you can see some screenshots of the dumper with experimental SRAM/FlashRAM support and can download it here.

The main menu

SRAM quick Hexview

FlashRAM copy feature with CRC

Future work

I am only partially happy, because it is no definitive solution, but neither a full soldering setup nor switching to the official N64 setup are solutions I think would justify the effort. The only think I might try is to check if the XPloder64 is compatible with GSuploader. But most likely it wont be compatible, because the Gameshark was from Datel and the Xploder64 from Blaze. The only Xploder64 software seems to be X-Killer 0.6, which would involve setting up a Windows 98 PC again. Maybe I am bored on Christmas...