What's new

obcd

Professional
Joined
Oct 28, 2016
Messages
543
Reaction score
246
Location
Belgium
I brought it up in another threat and Metallic suggested to start my own threat about it. I thought I might as well do it as some brain storming about it wouldn't hurt.

The problem we have with this gaelco hardware is that it regulary starts to freeze. It drives me nuts as sometimes all works well for several months.
Suddenly the problem comes back ans shows up 2 or 3 times / day.

The last time, I simply reflashed the bad unit and everything was well again. (You can flash one unit with the contents of the other linked one)
I am still not certain that this was the solution for the issue, as before, I have switched the cpu case between the 2 cabintes to conclude that this also made both units functional again. I switched the 2 power supplies, and yes, things where good again for a couple of months.

So, on the other threat, Metallic says that the gaelco hardware uses (nor) flash chips that slowly loose their contents. This could perfect explain why it got fixed using my latest reflash, but doesn't explain why switching it from one cabinet to another is a solution as well, unless the position to the earth magnetic field can also influence it.

Some smart guys will tell me I have a bad wire connection somewhere, but in such case, the problem would remain with the machine, even if I change the unit.

At one point, I was so desperate that I was looking for a pc solution with an emulator. I know that demul (originally a dreamcast / naomi (2) emulator) is able to emulate that gaelco hardware as well. The original hardware also has steering handle force feedback and a signal for 2 industrial async motor speed controllers that make the seat move upwards and down again. (Left and Right).

As demul was closed source, it was a dead end for the purpose of replacing the original hardware.

Gaelco had another football simulator game that used (nearly) identical hardware. As that's a standalone cabinet, it isn't having the option to be reflashed.

There are some headers on the boards for something that looks like Jtag's for the 2 hitachi sh4 processors on it. Maybe that's a possibility for reflashing.

The gaelco football game likely has another bootloader as it doesn't show the black screen when you power the machine with the service button pressed like the ATV track boards do. At that moment, they can be flashed with the other board.

Maybe, a tool on pc to flash such board using the link interface would be handy for people who don't have 2 unit's. The interface is full duplex rs422, so building an interface between that and a pc serial port shouldn't be that hard.

I really don't know well how to start. According to Metallic the system runs in Mame but doesn't show anything on screen yet.

I bought a second hand unit on ebay, but it's not solving the stability issues either.

As cherry on the cake, those systems have poorly designed digital inputs with the wires arriving directly on the inputs of a HC buffer. So occasionally, one of those buffers dies likely due to EMC from perhaps the motor controllers.

Maybe the emulator wasn't demul... It was maintained by a russion group I once contacted on irc.

Any suggestions upon this are welcome. I just noticed there isn't a question yet, so here it comes.

What are my best options to keep this machines alive?
Are there still many around or have most people given up on those?
Anybody else with experience servicing this beasts?
 
I have no experience with ATV Track, but I can give some details on the flash technology. There are primarily two types of flash memory storage, NOR and NAND. They were both invented in the 80s by Toshiba; but each every very different pros and cons.

ATV Track uses NAND flash, which has a weakness: endurance. It has a life span and can develop "bad blocks". When NAND is used in a system design there is typically a memory management component that detects memory errors. Each "block" of memory is a series of "pages". Each page has ECC data stored to verify the data has remained valid (error correcting code). Even brand new nand chips from factory have bad blocks, this is actually expected and specified in datasheets.

@MetalliC mentioned it is able to self recover data. I'm guessing the system will recover single bytes when it detects an ECC error. But there are limits to how many bytes ECC is able to correct.

The nand memory manager (this is typically software, like part of an operating system), by specification should mark blocks as bad if too many errors are detected, and re-locate the data from one block to another block. It is entirely possible that ATV Track does not support being able to re-assign data to a new block. You end up in this circle of fixing the data, then the same pieces of data go bad again.

Don't have any explanation to how swapping the boards around fixes the problem.
 
The nand memory manager (this is typically software, like part of an operating system), by specification should mark blocks as bad if too many errors are detected, and re-locate the data from one block to another block.
about right, but after such recovery there is almost none matching with original blocks, that's why I've said it does full reprogram. (and that's why MAME have 2 sets /dumps of this game, where actual file data exact the same, but dumps totally different if compare rom binaries)

Don't have any explanation to how swapping the boards around fixes the problem.
perhaps @obcd was saying about recovery link mode - if two cabs / pcbs linked, using JP (or DIP?) you may put them into special mode, where one board will transfer it's ROM contents to another PCB, which will write full game into its ROMs. quite nice feature btw.
 
Last edited:
In order to enable flashing, you had to power the board that needed to be flashed with the service button pressed. The screen remained black in that mode.

Next you can go to the settings menu of the working game ans select the option export program. I really should make a sniffer and capture communication between the 2 boards when a transfer is being established. The exporting board shows info on screen like exporting maps, exporting textures and a percentage indicator showing how far it came exporting that stuff.

So I assume there is some kind of bootloader. Let's hope that one berver get's corrupt. The gpu is a powervr as they advertise it on screen. It has a heatsink but no fan. There is a small 40 mm fan in the game board case that cools the primary voltage regulator. (The one converting 5V into 3V3)

I am still suprised about the simplicity of this hardware. There is a giant fpga between the powervr gpu and the 2 sh4 cpu's, so maybe it's not that simple after all.
 
Gaelco had another football simulator game that used (nearly) identical hardware. As that's a standalone cabinet, it isn't having the option to be reflashed.
right, main PCB is the same, but different filter? board where IO connectors and few ICs.
I think you may reflash Football main PCB to ATV track via link (if replace football filter board with the one from ATV). if I not mistaken, client-side flasher is in initial boot loader, which is inside FPGA firmware, which is the same in 2 these games.

Maybe the emulator wasn't demul... It was maintained by a russion group I once contacted on irc.
right, I'm one of developers.

in general, most probable causes is obvious bad capacitors or bad power, or bad NAND Flash ROMs.
Anybody else with experience servicing this beasts?
there was one (Belgian ?) guy, who serviced Gaelco Footbal and contacted us years ago, but I'm not sure if I have his contact data.
and another guy from Italy, but I haven't heard anything from him quite long time, not sure he is still in arcade tech business.
 
Last edited:
I probably was that belgian guy. I even made some pictures and sent them over.
Indeed the filter board was different on the football game.
I think I tried if I could get it into flashing ready mode and it didn't work.
Maybe it was because the football filter board used a different input for it's service button?
Besides, I didn't wanted to transfer that board into an atv track as there was no way back.
At least, the export option is missing in that football game.

Gaelco also had rolling extreme using lookalike hardware. I think you guys told me that the flash chip memory map and size was different on that hardware board.
 
I probably was that belgian guy. I even made some pictures and sent them over.
:D nice to meet you again

Indeed the filter board was different on the football game.
I think I tried if I could get it into flashing ready mode and it didn't work.
Maybe it was because the football filter board used a different input for it's service button?
exactly. that's why in previous message I've said you have to replace filter board.
IPL and flashing mode loader located in FPGA firmware, which is same in these games.

basically IPL does:
- basic SH4 init
- RAM test
- is it passed OK ? no -> go to dead loop
- check if service pressed ? yes -> go to loader mode
- read 1st 4 bytes of NAND
- is it "MBR\01" ? no -> go to loader mode
- read boot sector from NAND and run it

loader mode code doesn't do flashing itself, its small routine which download chunk of data via link DMA and run it
Gaelco also had rolling extreme using lookalike hardware. I think you guys told me that the flash chip memory map and size was different on that hardware board.
Smashing Drive, and yes, this game PCB is similar in general but different in details.
 
Last edited:
Good to hear the bootloader isn't part of the nand so it should basically keep working.

Is the mame implementation of this hardware mature enough to study the flashing routines in an attempt to understand communication during the proces? Obvious I would prefer demul for that but assume that will stay a wet dream.

Besides that, I still have some chihiro work to finish, so I probably can't start with this right away. (Won't get bored in the next 5 years I assume)
 
Dumb question, but have you contacted Gaelco? I know the odds aren't in your favor since they've made darts games exclusively for maybe 10 years, so the staff possibly isn't around more. Worth a shot in any case!

I just checked my emails and I contacted them in 2010. Jose Luis Fernández (joseluis at gaelcodarts dot com) was the head of the spares / repair dept back then. They still produced Tuning Race (2005) motherboards at that time.
 
Is the mame implementation of this hardware mature enough to study the flashing routines in an attempt to understand communication during the proces?
NAND write/program is implemented, as it required to run one of dumps which have recoverable errors. but 2x PCB link/comm. not imlemented in both MAME and Demul.

congrats with good WC result btw :thumbup:
 
Thanks, but I am not much of a soccer enthousiast. In my opinion, they are much overpaid for what they are doing.
I have played tennis and could respect players doing that on top level, as you are on your own in that sport.
I also can appreciate rally and other auto (moto) sport drivers as those sports don't tolerate a single mistake.
If you play tennis and you miss a ball, you pick it up and play further. If you drive a car and mis interpret a curve, it's game over and you can only hope for the best.

I was in contact with gaelco before, but it was a "hard to find error" as sometimes the board ran fine for several days. So, sending it over to get it back some weeks later with still the same problems didn't seem like an interesting option at the moment.

Besides that, the machines are mechanical also not an example how things should be made. The steering bar is huge so people are able to stress the mechanics real hard up to a level that the mounting holes in the corners of that section simply break off. The piece holding the steering bar bearings already broke a couple of times as well. During warranty period, we also had several mechanical failures until they sent us a more robust steering construction.

I won't mention the 26 inch hantarex polo vga monitors they use in those....(They might hear me)

Gaelco's next game (Tokyo cop) never really attracted people the way those ATV tracks do, but never the less I am afraid that upon the next serious repair cost it will be "game over"

I seem to remember they didn't actually design the atv track game themself but had the work done by some japanese company (could be wrong about that)

Damned, again no question.... So here it comes.

Are there any options (maybe in the nearby future) besides the original hardware to keep the game alive?
I assume pcb linking is not a high priority for most.
Did they encrypt the data needed for flashing when they sent it over the RS422 from one board to another?
 
I've never tried sniffin data on this game but i believe chances of linking data not being encrypted are very high. Its not helping much or stopping bootlegs at all. Protection is usually set in other parts of the PCB.
 
Back
Top