Bitrot happens

[+Mitch]

Just a general reminder to dump/back up your protos even if you don't choose to release the BIN files.

 

Case in point, I recently received a 7800 proto that wouldn't work. After dumping it I found that one of the two EPROMs was partially bit rotted. Fortunately, between the one good EPROM and what I could get off of the second EPROM I was able to confirm that it was the same as the final released version so nothing was lost. If this had been a one of a kind proto I would have been out of luck.

 

Mitch

[Shawn]

Scary shit!

[RevEng]

Definitely a good PSA for those hanging on to those protos.

 

In a dire case it might be interesting to see if bit-rotted EEPROMs can be recovered by looking at the analog output levels. It's my understanding that EEPROM bit-rot is caused by electron leakage. Maybe there's still a detectable difference between 1's and 0's in some cases, even though it may not meet digital logic thresholds. Or maybe there's a difference in rise/fall time for 0's that have become 1's due to leakage, vs. bits that were always 0's.

[+Gemintronic]

This is going to be a more prominent problem as electronics age. I hope we all collectively take a hint and support projects like the Retrode.

 

Back up often. Back up to multiple kinds of media. Doesn't matter if it's your thesis or The Music Machine.

[NIAD]

I've dumped all the protos and one-of-a-kind carts that I have in my collection and shared them with the CV community already, so I not as worried about losing anything of importance with these items. Where I have been lazy of late is with a number of Press Kits and Press Release Flyers that are extremely rare (only 1 or 2 are known to exist) and it's equally important to back these up by scanning them.

 

As theloon mentions, backups are EXTREMELY important and not just one backup to an external drive, but to different media such as DVDs, CDs, etc. I have even taken things a step further by storing the extra backups I make on DVD discs in a safe deposit box at my bank in case something were to ever happened to my house such as a fire or flood. After all the time and money spent acquiring these rare items, I couldn't live with myself for being to lazy to make an extra backup and store it in another location... just in case.

[Keatah]

Definitely a good PSA for those hanging on to those protos.

 

In a dire case it might be interesting to see if bit-rotted EEPROMs can be recovered by looking at the analog output levels. It's my understanding that EEPROM bit-rot is caused by electron leakage. Maybe there's still a detectable difference between 1's and 0's in some cases, even though it may not meet digital logic thresholds. Or maybe there's a difference in rise/fall time for 0's that have become 1's due to leakage, vs. bits that were always 0's.

 

That detection is done on the chip. The x-y (exact memory cell) is selected, patched in, and the level of electrons is detected by an amplifier so to speak. The pins that fit into the sockets on the boards are just the outputs of a 0/1 gate and will be either one or the other. You'll never see analog there. Very much like a modern CCD or CMOS image sensor. Lots of A/D going on in the chip die.

 

Modern forensics allow you to place a pin tip probe and directly read the cell level. Do it all the time.

 

Backups are of course king when crunch time happens. It is important to verify your backups by whatever method you used to make them in the first place. Simple file copies require a bit-by-bit checksum or comparison against the original. Images of whole disks require verification by the program that created them. The image must be error free, because this modus operandi and its tools are completely unforgiving. 1 single bit error can fuck up your entire 40GB image. In any case, a test run of being able to read and restore the backup is vital - to be sure your procedure is correct.

Edited July 13, 2014 by Keatah

[+-^CrossBow^-]

I only own two protos, but I believe I dumped them both many years ago and I believe Tempest has the bins for them as he wanted them for comparison? Perhaps he can confirm that. They were a 2600 parker protos, one of Star Wars the arcade game and the other was for Gyruss. Star Wars was definitely a WIP version with differences from the release, but the Gyruss was listed as RLS 1 on it and should have been the same as the released final.

[Foxsolo2000]

How was the Star Wars game different from the finished product?

[CPUWIZ]

Same goes for most homebrews and many many commerical games, not just protos. And the bad thing is, all the commercial games that use EPROM's are very rare. Every Ultravision title is made with EPROM's, for instance.

[Keatah]

I want to add one more thing to post #6. If a bit is known wrong, sometimes it is possible to change the voltage levels applied to parts of the chip in order to eke out the last errant cell. But you have to strongly suspect you've got a bad one in the first place.

 

A variation on this IS purposely changing the voltage levels at the A/D converter and seeing if you get consistent reads. Similar to purposely trying to throw off the alignment of an old school disk drive - by having the head approach the target track from a higher track or lower track. Looking for that hysteresis or wiggle.

 

I tend to over-use the A/D term, we all know what that is though. The more exact phrase is output buffer transistor array, typically 8 bits wide for the classic cartridges we mess around with. The cell is the front end of the converter, and 8-bit output array is the back end. Together they measure the charge level (electrons*) in each cell, and swing bang-bang hi-lo. The addressing lines determine what cells are (in packs of eight) connected to the data lines for placement onto the bus. Hope that clarifies things more.

 

*Electrons!! Sub-atomic physics is so confused and bloated at the moment, some prominent scientists are wanting to redo the whole damned theory of matter and the universe at large. And we all know the truth will be stranger than anything that's come before.. We shall see!

Edited July 14, 2014 by Keatah

[Keatah]

One more item, then I'm done. A weak cell wouldn't show at the data lines, not in rise/fall time, nor in skew time (1 bit becomes available before the others). Never happens in a correctly functioning chip.

 

Understand that a weak cell or borderline cell might take several extra ns to trigger its bit in the 8-bit array latch. But the latch waits much more time before passing it to the data bus. Thus, it allows for straggler bits to come in later. The window of arrival of raw bits from storage is so huge that nobody arrives late.

 

Why would one bit come into the latch sooner or later than another? If it's weak, the read voltage that goes into the cell would have to go up through more of its rise time prior to cascading the cell gate into conductivity (the front end I mentioned earlier). Filling the glass faster or slower before tipping it over. Not filling faster or slower because you're pouring faster or slower, but because the size of the glass is different. And the size of the glass is the size of the trapped charge in the gate. When it finally cascades it's turned on and you got the bit from the front end!

 

In the lab most all of these points are measurable.