|
Post by ClodBuster on Jan 22, 2019 6:21:46 GMT
I once wrote a post about GameBoy games creating read errors even though the contacts were in perfectly clean shape. It turned out the common cause are ROM chip pins lifting from the PCB's contact patches. The GB cartridge PCB is rather thin and thus sometimes prone to flexing. The solution to the problem is to resolder the pins, which I did to successfully repair a broken Link's Awakening DX.
I wouldn't be surprised if some HuCards suffer problems due to a similar cause. HuCards are thin, and so are their internals.
I also think there might be something as simple as several ROM revisions distributed for the same game over time. That concept is nothing new, we know this from other consoles, prominent examples being Shinobi on the Genesis (reworked sprites for e.g. Spiderman) or Terranigma on the SNES (the German market received two revisions which differ slightly in the translation, with the final weapon being either wrongly named "Elfenbeinstab" in 1st rev. or correctly "Terralanze" in 2nd rev. ingame). Often times though, revision changes might be so subtle that they aren't noticeable in most games on regular play.
What I do not buy into are stories about sudden "bitrot". That sounds way too esoteric to my ears. Leave that problem to the Laserdisc afficionados.
|
|
|
Post by soop on Jan 22, 2019 9:26:27 GMT
I once wrote a post about GameBoy games creating read errors even though the contacts were in perfectly clean shape. It turned out the common cause are ROM chip pins lifting from the PCB's contact patches. The GB cartridge PCB is rather thin and thus sometimes prone to flexing. The solution to the problem is to resolder the pins, which I did to successfully repair a broken Link's Awakening DX. I wouldn't be surprised if some HuCards suffer problems due to a similar cause. HuCards are thin, and so are their internals. I also think there might be something as simple as several ROM revisions distributed for the same game over time. That concept is nothing new, we know this from other consoles, prominent examples being Shinobi on the Genesis (reworked sprites for e.g. Spiderman) or Terranigma on the SNES (the German market received two revisions which differ slightly in the translation, with the final weapon being either wrongly named "Elfenbeinstab" in 1st rev. or correctly "Terralanze" in 2nd rev. ingame). Often times though, revision changes might be so subtle that they aren't noticeable in most games on regular play. What I do not buy into are stories about sudden "bitrot". That sounds way too esoteric to my ears. Leave that problem to the Laserdisc afficionados. you've just given me an idea. Dave, it might be worth removing one of the non-working PCBs and trying to reflow it with hot air. There's usually glob-top chip, and I don't know how you'd go about removing the blob
|
|
|
Post by dshadoff on Jan 22, 2019 13:43:49 GMT
If it were as simple as a pin lifting, there would be a consistent pattern to the differences - they would appear throughout a ROM (which could be less than the whole dump, but would be on a strict aligned memory range, of an average 256KB of data), and would be regular in appearance - either replication for another region of the ROM (in the case of an address line lifting), or a consistent bit being set compared to the "regular" rip. I *did* see patterns of this type, but only when the card contacts were dirty; cleaning them resolved this type of issue.
HuCards are also packaged in a much different way compared to a regular cartridge; other systems' cartridges are essentially a PC board onto which one or more ROM/PROM/EPROM chips are soldered, often with some address-selector logic to choose which ROM is active. PC Engine HuCards don't contain ROM chips, but rather the ROM dies themselves (the part that is packaged inside of a chip). And they put a blob of resin over top of it for protection. There is nothing that you can reflow; the connections between the board and the chip die in those days were done with a special machine, and were basically finer wires than a human can manage.
I agree that there could, in some cases, have been multiple versions of the same card, if there were re-issues. But because of the costs involved in mask-programmed ROMs in those days (which dropped through the 90's), this would not be likely unless there was a serious bug to be addressed, on a very large mass-market title - and we may have heard of such a thing. And fixing that bug (assuming that it's code) would not likely show up as as a single byte of difference, or a 20-30 byte block with no other changes; they would either be several blocks, or a large-scale shift of code slightly. But because of the way that programming was done on this machine, it likely wouldn't spill over 8KB block boundaries.
So, while R-Type II, PC Genjin 2, and Street Fighter 2 (and maybe Darius Plus) would qualify as "mass marketable enough for a re-issue", I doubt the others were. And of the titles which could have been re-issued, the differences don't really match my expectations...
I've made arrangements to get additional copies of the other cards, so we'll soon see what they show.
Dave
|
|
|
Post by soop on Jan 22, 2019 13:51:29 GMT
HuCards are also packaged in a much different way compared to a regular cartridge; other systems' cartridges are essentially a PC board onto which one or more ROM/PROM/EPROM chips are soldered, often with some address-selector logic to choose which ROM is active. PC Engine HuCards don't contain ROM chips, but rather the ROM dies themselves (the part that is packaged inside of a chip). And they put a blob of resin over top of it for protection. There is nothing that you can reflow; the connections between the board and the chip die in those days were done with a special machine, and were basically finer wires than a human can manage. Interesting... Any idea how Hit Japan are making their cards? Someone got one and it was connected the same way. Maybe they got their hands on some kind of machine
|
|
|
Post by dshadoff on Jan 22, 2019 14:29:12 GMT
Back in the day, the chip dies were mask-programmed PROMs; they could only be programmed once, and they were covered in a blob, and put on a thin board which was then inserted into the plastic carrier we know as a HuCard.
I don't know Hit-Japan's source, but I suspect that these are bit-addressable flash memory (something like 29F040 - thought those are old now) with some sort of voltage-conversion circuitry to allow it to be 5-volt compatible. If this is the case, then programming them is a matter of sending erase and write codes. I had actually made a Tenonokoe work-alike about 18 years ago using the 29F040 which could save data into Flash, using the system's power supply. Unfortunately, not all systems had enough power supply to consistently program the data (as I recall, Duo's were weak, but PC Engine/TurboGrafx were OK).
In any case, all sorts of things are made in China these days, and lots of it is stuff you wouldn't even imagine. The technology to package a chip die on a board substrate has long ago been made mainstream, and I'm sure you've seen these blobs on other consumer devices. But maybe HuCards were only recently thought of.
I would have a minor concern that the voltage-conversion circuitry might have some shortcuts taken (because I've seen discussions/arguments on other forums about other peoples' implementations), but that's just a random thought without any basis at the moment.
Dave
|
|
|
Post by soop on Jan 22, 2019 14:39:46 GMT
Back in the day, the chip dies were mask-programmed PROMs; they could only be programmed once, and they were covered in a blob, and put on a thin board which was then inserted into the plastic carrier we know as a HuCard. I don't know Hit-Japan's source, but I suspect that these are bit-addressable flash memory (something like 29F040 - thought those are old now) with some sort of voltage-conversion circuitry to allow it to be 5-volt compatible. If this is the case, then programming them is a matter of sending erase and write codes. I had actually made a Tenonokoe work-alike about 18 years ago using the 29F040 which could save data into Flash, using the system's power supply. Unfortunately, not all systems had enough power supply to consistently program the data (as I recall, Duo's were weak, but PC Engine/TurboGrafx were OK). In any case, all sorts of things are made in China these days, and lots of it is stuff you wouldn't even imagine. The technology to package a chip die on a board substrate has long ago been made mainstream, and I'm sure you've seen these blobs on other consumer devices. But maybe HuCards were only recently thought of. I would have a minor concern that the voltage-conversion circuitry might have some shortcuts taken (because I've seen discussions/arguments on other forums about other peoples' implementations), but that's just a random thought without any basis at the moment. Dave I know GBA knockoff carts are packaged the same way now you come to think of it. hmm, interesting. Do you reckon there's a company in China that can just fabricate hucards? Do we know anyone that's produced their own in that way? I know Magical Chase was reproduced, but no idea of the process. (sorry for derailing the thread slightly btw, it's just very interesting)
|
|
|
Post by spenoza on Jan 22, 2019 15:07:43 GMT
What about more mundane sources of electromagnetic radiation, or simply stray magnetism?
|
|
|
Post by dshadoff on Jan 22, 2019 20:26:05 GMT
soop: I don't know of anybody making HuCards, but there must be many who are capable. The missing part is the plastic carrier, whose biggest cost would be the moulds (notice that Hit-Japan is re-using other HuCard plastic carriers).
spenoza: I can't see how magnetism would do anything, but normally they talk about high-energy waves (like X-rays or cosmic rays) which can penetrate, or high-energy particles (like electron beams). And potentially a slow decay of P-N junctions in semiconductors over a long period of time. It isn't an area that I've studied much, so I can't say what the true "expected lifetime" of such a device is; only that the semiconductors' physical lifetime exceeded the manufacturer's planned lifetime.
|
|
|
Post by dshadoff on Jan 24, 2019 1:06:27 GMT
Update: I found another R-Type 2 lying around that I didn't know I had, and found that it matches the internet file. -> So my original card had errors. Such a weird pattern to have errors in though - it really looked like assembler changes. Follow-up #2: I received another Street Fighter 2 card, and it extracted as the exact same image I had from the other card. However, because it's an esoteric memory map, I'm not ruling out these possibilities (and will keep an open mind about the other image I have which doesn't match): 1) Potentially the UperGrafx extracts the banks in a different sequence than other image extractors 2) There may even be a bug in the UperGrafx extractor for this specific card. 3) Could be multiple versions... but I doubt it, given the type and scope of mismatches I saw. Dave
|
|
|
Post by dshadoff on Jan 27, 2019 22:44:58 GMT
So I'm going through my US cards today (alphabetically), and found that my Ballistix card seems to be degenerating.
I am seeing Bit 1 (that's the "2"-value bit) low on all of the $xx6D and $xx6F locations, for a total of 75 instances of difference from the original card. This is a little too specific to be a pin problem (and indeed cleaning it made no difference); rather this looks like a pattern of cells are dying, or perhaps address-decode logic is getting slow on specific combinations...
I mentioned earlier that ROMs of this era were all mask-programmed, and that isn't actually correct. There was a migration toward PROM (like EPROM, but program only once) around this time period. It is possible that PROMs were programmed on devices which weren't able to make stable-enough charges to last 30 years. Even a microsecond of programming time (on a given byte) could make a difference in survivability.
So, that means that another possibility is that a PROM programmer was a little light on the program cycle on certain cards.
|
|
|
Post by soop on Jan 29, 2019 10:15:13 GMT
Interesting. If you find any pattern as to the manufacturers/time period that might help narrow things down.
Also, that's good in a way, since it should mean that the majority of hueys should be safe
|
|