Reprogramming already-written PROMs

12 posts / 0 new
Last post
Offline
Last seen: 4 years 4 months ago
Joined: Feb 19 2020 - 18:31
Posts: 6
Reprogramming already-written PROMs
I'm considering making an Apple-1 replica, and so I've looked into the availability of 256x4 bipolar PROMs like the MMI 6301. I may or may not succeed in finding blank ones (and finding or making a programmer), but here are a few blue-sky ideas about taking PROMs that have already been programmed and refurbishing them, that is, rewriting them with new data. Take them with a grain of salt. :-) 1. First blow all the fuses with a programmer, then use a FIB machine to selectively deposit metal at the fuses to reconnect a subset of them. 2. Same idea, but make a mask with holes in it, then sputter metal through this mask onto the fuses 3. Use a laser welder to reconnect fuses (this may or may not work depending on the geometry of the blown fuse) These all require decapping the chip, which is simpler with a ceramic one. I think a ceramic sandwich can be re-fritted to regain the hermeticity; but if you're willing to give up hermeticity, then something simple like epoxy would work. Also, I'm not sure if there is passivation over the fuses, but at least FIB can drill through the glass if needed. These techniques are a stretch at the hobbyist level, but a commercial chip-reverse-engineering shop should be able to execute them: it's just a matter of cost. Or, perhaps, a university group with the appropriate equipment could be coaxed into taking on the task. Best not to throw away any programmed 6301s, just in case.
Offline
Last seen: 1 month 4 weeks ago
Joined: Jul 5 2018 - 09:44
Posts: 2587
It would probably be easier

It would probably be easier and cheaper to try to get a Chinese manufacturer to make new "reproduction" chips and then have them marked as MMI 6301.

 

 

Offline
Last seen: 5 months 5 days ago
Joined: May 30 2019 - 00:29
Posts: 51
Wow. That shound like a

Wow. That shound like a really heavy job. To do this you would need to:

 

1) Decap the chips. Usually by removing some plastic with a CNC, then dripping them with hot concentrated nitric acid (or some other acid mix). You have to be careful not to remove the Al metal which connects things to the outside. I don't know if the nichrome fuses of the MMI 6301 would be damadged by it - it depends on the passivation that covers the chip (silicon nitride?).

2) To redeposit metal (given that you can find out about the actual composition of "nichrome" specified in the datasheet) would require you to remove the current metal. To get down to it you would need to remove the passivation in such a way that the Silicon beneath isn't damaged. Probably acid or dry etching with a litography mask. How you are going to make such a mask is beyond me - but maybe through reverse engineering from photos of the chip.

3) After etching of the passivation you would need to deposit the fuses. They are not simply a square of metal, but then to have a "thin" section that is adjusted so that it "blows" as the proper voltage (and resulting current) is applied. I attach a photo as an example of how they may look (but this is from some other chip).  I could not upload the example file, but you can find one here of a 5300/6300: https://www.righto.com/2019/07/looking-inside-1970s-prom-chip-that.html

 

All-in-all it would probably be "easier" to make the chips all the way from the wafer level.  But you would need to reverse-engineer a MMI6301 from a decapped chip to get all its internal workings mapped out. I don't know if its a n- or p-MOS based transistor process that was used, but others has managed to get the internals mapped out like the visual6502.org.

 

Still, with a reverse-engineered design for the 6301, you would be really lucky if you managed to get some old process documents from MMI. The reason is that there are a LOT of parameters that needs to be adjusted to make such an IC from the bottoms up. You would probably need to be a few people with a free accessible cleanroom and a couple of years to pull all this together.

 

So.. maybe its easier to find some other PROMs and remark them.

Offline
Last seen: 4 years 4 months ago
Joined: Feb 19 2020 - 18:31
Posts: 6
I was able to find some blank

I was able to find some blank 6301s on eBay, so I guess draconian chip-surgery measures won't be needed. Now they just need to be programmed.

Fabricating new PROM chips would be an interesting effort; I wonder if any of the MOSIS processes would support a fused ROM. Masked ROM would, of course, be no problem, provided a high-voltage (5 V) process is used. There seem to be enough alternative parts in the 256x4 bipolar PROM space, though, that it might be some time before they're completely unavailable.

The other hard-to-find chips, the Signetics serial memories and character generator, might benefit from a reimplementation too, either at the chip level or at the FPGA level. I'm really just after a functional Apple 1 in the original board form-factor, so FPGA stand-ins would be fine so long as the behavior is identical to the original chip. If it's reasonable to use original chips, though, naturally that's preferred.

Offline
Last seen: 3 days 4 hours ago
Joined: Jun 5 2008 - 07:26
Posts: 478
I don't know about the

I don't know about the current state of Apple 1 chip supplies, but I have found a couple of other chips that were extremely difficult to find.

The TMS2501 character generator used in the Scopewriter couldn't be found anywhere, but a friend in Eastern Europe obtained a Soviet clone chip for me.

The early Apple II keyboards used a hard to find encoder chip, MM5740AAE.  A lot of those that could be found, turned out to be defective.  Never found a solution or alternate source for those.  I did envision using a 40 pin AVR as a replacement, but the rewiring necessary required stacking sockets, which created physical clearance problems.

Unfortunately, many of these chips use unusual voltages that make using ordinary modern day FPGAs a bit complicated.  The -12 volt clock that runs all around the Apple 1 motherboard is an example of something that would have to be worked out.

 

regards,

Mike Willegal

 

Offline
Last seen: 4 years 4 months ago
Joined: Feb 19 2020 - 18:31
Posts: 6
> [MM5740AAE] I did

> [MM5740AAE] I did envision using a 40 pin AVR as a replacement, but the rewiring necessary required stacking sockets, which created physical clearance problems.

One workaround might be to use an SMD AVR mounted on a small PCB sized for a 40-pin DIP. Adding pins along the edges would allow plugging into the original socket (or PCB footprint) with low overall profile. These pins work better than 25-mil square posts (which would damage a socket):

http://oshchip.org/products/Flip-Pins_Product.html

> Unfortunately, many of these chips use unusual voltages that make using ordinary modern day FPGAs a bit complicated. The -12 volt clock that runs all around the Apple 1 motherboard is an example of something that would have to be worked out.

Yes. The usual sources of level-translation parts are completely useless, since they generally max out at 5.5 V, whereas the Apple 1 clock spans 17 V. Also the 2504 and 2519 don't have ground pins, complicating the power situation. But I think it's doable. It might go something like this:

- generic small FPGA with 3.3 V I/O and 1.2 V Vcore, for example Lattice's ice40
- starting from +5V and -5V pins: a negative LDO of -5V to derive local ground, followed by two positive LDOs of 3.3V and 1.2V.
- clock (need only one of the two): make a CMOS inverter from two discrete FETs with very wide Vgs, e.g. Toshiba's SSM3K15ACTC and SSM3J15CT
- data in: feed direct to FPGA with ~2k series resistor
- data out: same Toshiba PMOSFET in open-drain

This is quite a lot to fit on a PCB some 5mm on a side, but I think it works if one uses both sides.

By the way, thanks for all your nice documentation on the Apple 1.

Offline
Last seen: 1 week 3 days ago
Joined: Apr 1 2020 - 16:46
Posts: 996
New PROMS not likely

Peter - nobody has a NiCr process anymore. Since recently, TSMC offers a SiCr option for some of their CMOS processes.

With 1K/sq SiCr and some NMOS and PMOS devices that can work from 15-20V it is perfectly possible to design SiCr fuse PROMs. Been there, done that. But it is not as trivial as it sounds. You need to know exactly what you do.

Then, consider costs. You need access to a CADENCE Composer/Virtuoso seat. And even a multiproject shared reticle engineering / sample run of a 180nm process will set you back by $50K or so.

I think the best route to make a "perfect" replica is to use still available PROMs of any compatible type and date code , grind off the top surface, and print (by a rubber stamp) the desired MMI logo, type and date code. Much like these Chinese chip counterfeiters do.

 

 

 

Offline
Last seen: 4 years 4 months ago
Joined: Feb 19 2020 - 18:31
Posts: 6
PROM options

> Since recently, TSMC offers a SiCr option for some of their CMOS processes.

That's interesting. There must be current applications for this material, either for PROM fuses or for something else needing a high-resistivity metal.

Still, I take your point: there are other ways to get to working PROMs. I've been thinking about a programmer for my MMI 6301s (microcontroller, DAC, high-voltage op-amp, source-follower buffer) but it's a bit of a science project that may involve trial and error. I did spend some time doing an FPGA stand-in for this part. Let me start a separate thread on that.

Cheers,
Peter

stynx's picture
Offline
Last seen: 11 hours 29 min ago
Joined: Apr 19 2012 - 09:41
Posts: 223
 

 

Signetics MMI TI Harris Raytheon AMD National Intel --------- --- -- ------ -------- --- -------- ----- 82S129 6301-1 24S10 7611-5 29661 27S21 74S287 3621

 

Maybe you could try to find an equivalent chip?

Offline
Last seen: 1 month 4 weeks ago
Joined: Jul 5 2018 - 09:44
Posts: 2587
stynx wrote: 
stynx wrote:

 

Signetics MMI TI Harris Raytheon AMD National Intel --------- --- -- ------ -------- --- -------- ----- 82S129 6301-1 24S10 7611-5 29661 27S21 74S287 3621

 

Maybe you could try to find an equivalent chip?

Easier said than done.  I don't think any of those other chips are much more readily available than black 6301 are.  They've all been out of production for 25+ years.

 

stynx's picture
Offline
Last seen: 11 hours 29 min ago
Joined: Apr 19 2012 - 09:41
Posts: 223
The 74S287 were produced by

The 74S287 were produced by Tesla as well. Tesla was a Eastern Block company that produced clones of western chips.

Most of the Tesla 74S287 on ebay are probably NOS. They may be a bit slower (60ns) than the 6301 though.

The chipdates are: I=82 ... W=88 X=89

Some of the 74S287 were therefore produced until 1989. The remaining stocks seem to be very big.

These here https://www.ebay.de/itm/132779689355 seem to be from December 1989 (the newest date i have found).

Therefore around 30 years old ;-)

-Jonas

Offline
Last seen: 1 week 3 days ago
Joined: Apr 1 2020 - 16:46
Posts: 996
SiCr PROMS

Hi Peter -

 

you are right. SiCr is a thin film that is deposited in a sputter tool. Its major use if for precision, low TK, resistors in analog circuits. The 1KOhm/sq version of SiCr can also be used to make fuse proms if the process allows for handling 15-20 Volts.

The lower ohms/sq SiCr could also be pressed into service as a PROM, but the blow transistors must be twice the size.

The higher ohms/sq SiCr such as 10KOhm/sq can't be blown electrically.

 

I did design SiCr PROMs on a proprietary process, not on TSMC. They are much smaller and much more reliable than Poly fuse PROMs, which tend not to work very well in deep submicron processes.

 

And then there are metal fuses. They need passivation openings over them, as in the vintage NiCr PROMs. The achilles heel if in a non-hermetic package.

 

Be aware I have close to a thousand 256 x 4 PROMs in stock. I can't yet tell which are still virgin and if they still can be blown reliably after all these decades, but USPS botched the delivery of my Data I/O programmer I scored on Ebay, so I still lack of any equipment to try this out.

 

Log in or register to post comments