The data retention is based on when the data was last written to the memory location, not when the chip was manufactured. So in theory, if you play your game forever, you "data" will be retained forever. Where this may not apply is in regards to settings. They are written only when changed so if you never change your settings (3 ball/5 ball etc.), you may need to update the setting in about 45 years.

Yes, solutions exist:

http://lockwhenlit.com/anyPinlist.htm

simple answer is No.

We had a discussion and solution in an older thread. It either requires the simtek version of nvram or some additional modification of the board besides removing the Dallas chip.

Will this one fit between the pins? Of course, I don't know if you consider $0.54 cheap in single qty.

http://www.digikey.com/product-detail/en/SN74LVC2G00DCTR/296-13257-1-ND/484537

Others like ST, NXP also make one but no one seems to stock them?

Not sure if it will cause any issues but you could solder a 3 volt battery cell across the power and ground pins if the original CMOS RAM before removing it to preserve high scores and bookkeeping.

It's a little like what slot techs do when the mpu reports a low battery before it drops below 2 volts. They apply 5 volt to the board while changing the battery.

And, once you have done that layout and run auto-route, you can then go back and start moving Address and Data around to make for the shortest possible traces and start reducing vias by re-arranging address and data (A1 to A4, D0 to D7 etc.). After all, in this slow world of Pinball CPUs, it doesn't matter what cell of memory or what data bit goes where (unless you want to read it back on an EPROM programmer). As long as the same cell gets written and read from, and the same data bit that is written gets read back, its all fine and dandy for the CPU which stored the data there in the first place.

Obviously, the above is not really necessary when doing a 6264 to FM16W08 adapter since the pins are adjoining their dip counterpart, but can come in handy when doing a 5101 to FM16W08 adapter. It can still be tested after assembly, just can't expect to get data that makes sense when viewed in a hex reader.

No, the old fashion way was with sticky black tape and an Exacto knife.

FWIW, technically, you probably shouldn't simply ground the unused Data Lines of the FM16W08. I know it has been done before and probably without too much ill effect on the IC but if the part is "read" without previously having zeroes written to it. You can't assume it is initialized with all 00's especially if its coming from China. If any of those data cells contain a logic 1, you will be driving the outputs directly to ground. Some of the ones I purchased already had data in them even though they were sealed in reel tape.

Of course, if you are sure they are NOS parts, which may or may not come initialized to 00 (not really sure myself), or you have previously run them through a test routine ending with the data of all 00's, then I suppose it is safe to use them that way.

The boards I made for myself don't have limiting resistors and they seem to work fine. But then, I never removed the FM16W08 to see if the output drivers for those data bits are fried! Once a write has been performed on the cell, obviously, the data will be forced to all zeros for all future reads.

I gotta ask: Why are you thinning the traces between the solder pads? I know it is for required clearance but why not just use smaller pads or thinner traces? We are not talking about a lot of current here. And its not exactly a noisy circuit running at 0.5-1MHz.

But please run me a few spare blank boards so you can sell them along with the Nand gate to me so I can build them myself for my Gottliebs! LOL. Remember, I am a cheap Bastard! Can't justify $29.95 either and too lazy to do my own OSHpark run if I don't have to!

No, I would not leave them floating. It would just be a good design practice to either pull them up or down through four 4.7K-10K resistor so as not to strain the output buffers.

As I said, in theory, it should only be a one time write and only if the cell contained a "1. After that, no strain on the part. Of course, if it contains a 1 and is never written to, it would happen every read. But, since the 5101 is usually used exclusively for storage of data, it shouldn't be a problem.

Certainly not worth the time, or cost to add a 4 piece resistor to the design since it hasn't reared its ugly head in the past. Just wanted to mention it.

Looks good. If you do a run of blank boards, I would love to buy a few with the discrete SMT components. I can supply the RAM, headers, and labor!!!!

Like I said, will save me the time of doing a run for myself on OSH park.

The jumper selects whether a 6116 or a 6264 RAM is used. The mpu board doesn't use the extra memory space of the 6264 but it must have the connection for Vcc be moved to use it in the board.

So, by moving the jumper to change from the 6116 to 6264, it allows you to use the more readily available ( and currently) cheaper choices of 6264 either in dip form or the sufrace mount packages on simple adapter boards.