Tech Flashback: The (beautiful) EPROM Chip Collection

It’s been a busy and draining week and a bit in Melbourne, and it’s consumed the vast majority of my energy just to get everything unpacked and put into order. However, things have been getting a little too quiet around here, so I just felt like I had to put something out, if only just to say I’m still here! During my time in Melbourne, the postal service was busy delivering a multitude of small parcels to my house. Almost ten had arrived by the time I got back. Sifting through all of them, the majority of them turned out to be the EPROMs that I had ordered off eBay.

It might seem odd to be trying to collect EPROMs in 2016, given that most devices don’t even use them anymore, but they are one of the most beautiful components I have ever seen owing to their open window by design thanks to their ceramic fritz-seal DIP packaging. It’s one of the few times you can see the die of an IC intentionally. They’re also getting somewhat rare, so having a few spares on hand might be a good idea. I decided to sample nine different types of chips, all of which cost under AU$10 a piece. Some were listed as new, however, it seems likely that the majority were second hand pulls which had undergone a refurbishment process. Lets meet the chips and admire their beauty!

SGS M2716

2016051822426303

This appears to be a very old chip, owing to the SGS Thomson branding which later became STMicroelectronics. This particular chip is an M2716, a 16kBit EPROM with 24 pins. Some adhesive residue from the previous labels can be seen.

2016051822436304

The die is mounted off-centre in the window. Below the chip, a wire bond to the base can be seen – this appears to be a body contact which is done differently on more modern devices which allow for the chip to be centred. The off-centre mounting could possibly affect the erasure effectiveness. Because of the low memory density, older lithography and the large die in comparison, some more patterning can be seen in the memory array area which kind of looks like a metal memory plate. The component model number can be seen in the die as well, which suggests the design hasn’t optimized the use of silicon. The wire bond pads can be seen along all four sides of the die.

Intel D87C64

2016051821596246

This was one of the pricier acquisitions in the set of chips, but it was justified. It was the only chip that presented as it was described – namely new old stock which appeared to be unused and pristine with no adhesive residues. The chip is a 64kbit array with 28-pin interface.

2016051821596247

Intel chips seem to be very peculiar, especially because they have their own pin-outs which are not compatible with many programmers, including the TL866 I own. However, that’s okay, since you don’t need a programmer to admire it up close.

2016051822036252

The die can be seen to be bonded to a conductive rear plate using some form of gold solder. This often makes the chips pricey and collectable, as people like to destroy them and melt them down to recover the minuscule amount of gold within them. Anyhow, it seems this Intel chip is much more advanced, with a smaller die that has its memory as one large nearly-square segment, and wire bond pads along two sides (mostly). The body contact is still present, and the die is slightly rotationally offset, however, it’s still much less likely to be affected by the edge shadowing of the window.

STMicroelectronics M27C64A

2016051822296289

This chip looks a lot more familiar and modern, being a 64kbit array with 28-pin standard interface that can be programmed by the TL866. This one has some evidence of labels being removed.

2016051822306293

This memory array is split into two areas, with the die centred in the window. The wire bonds are mostly along two sides, and there is no obvious separate body contact as with the prior chips. There is some text on the dies, and the array does still show some textured striped patterns.

STMicroelectronics M27C512

2016051822236274

This is a much more modern EPROM, with a 512kbit capacity and 28-pin interface. If the date code reads as I expect, this was a Week 28, 1998 device.

2016051822286287

Unlike the other devices, this one seems to be mounted against a dark background. I’m not sure if that’s a conductive surface or not. Wire bonds are groups on the two sides, but the die has again shrunk which reflects the improvement in lithography processes over time. This results in the two halves of the array being “featureless”. However, they do shimmer an iridescent colour when exposed to light because the feature sizes are so small that the pattern is acting as a diffraction grating similar to how a rainbow appears when viewing the underside of a CD/DVD.

STMicroelectronics M27C2001

2016051822376298

This particular chip is quite easily available, so I got a set of five because it was much cheaper that way. Interestingly, I got two different sorts of chips – one with a darker body, and one with a more traditional lighter body. The size of the dies were different, but both claim to be Week 34 of 2011 with the same marking codes. The chips are a 2Mbit EPROM with a 32-pin interface.

2016051822396301 2016051822416302

Ultimately, both dies look fairly similar, one appearing to be a smaller version of the other with the array split into halves and wire bonds top and bottom sides.

STMicroelectronics M27C4001

2016051822336295

This chip is another modern device, with a 4Mbit capacity and 32-pin interface. Essentially, it’s a double-size version of the above chip, dated Week 40 of 2010 with a slightly dark body.

2016051822356297

A look at the die seems to show it has a very similar configuration although the array is twice as tall.

AMD AM27C040

2016051822176270

A short break from the STMicroelectronics chips, this one is from AMD and is a much older device with a copyright date of 1990. I’m not sure whether the markings indicate a date of Week 23 of 2008, but I suspect not based on the size of the chip. It has an identical pin-out and is a 4Mbit chip with 32 pins.

2016051822186271

The die has a much “greyer” silicon colour as per the older chips, and features visible stripes in the array which is broken into two. The size of the die fills the window almost completely.

STMicroelectronics M27C801

2016051822106256

This chip has been cleaned so much that it’s practically lost its markings. This chip is an 8Mbit 32-pin chip, and it appears to be dated Week 27 of 1997. As a large array, the shape of the window had to be enlarged from its regular circular shape to a rounded rectangle to accommodate the size of the die.

2016051822126258

The die is super-sized compared with most of the other chips, being split into two parts, in a very tall array arrangement.

STMicroelectronics M27C160

2016051822086254

The largest device I purchased was this 16Mbit behemoth with 42 pins, which is not compatible with the TL866 programmer. This array appears to be dated Week 38 of 2003.

2016051822076253

A large rounded rectangular window exposes the chip which is divided into four segments, and is physically not quite as big as the 8Mbit chip above. The symmetrical design with four segments may be because the device is a 16-bit output device, which is sort of like welding two (shrunken) 8Mbit 8-bit devices side by side.

Beware the Recycled?

After receiving all of the chips, the first thing I was interested in is finding out whether any of them still contained their original data. Sadly, but as expected, they were all erased suggesting that they had undergone some refurbishment process. Whether this process is kind to the chips or not is up to debate, but from what I can see, there are some things to be careful of.

On all the chips I had received, with the exception of the Intel and a few of the ST’s, many were improperly packaged in non-ESD safe bubble wrap and foam and had bent and dirty pins, some severe. It could be possible that the chips would be damaged in transit because of this.

Then I also had some questions about the texture of the surface finish on the M27C2001 and M27C4001’s. I tried to glue a label using super-glue to the chip, and by a chance discovery, the glue bonded to the surface and scraping it off with a knife revealed something interesting!

2016052216436318

speed-grades-from-1999-datasheetThe M27C2001 had been re-marked. They turned an old Week 15, 2001 chip into a “new” Week 34, 2011 chip. Interestingly, they did the reverse of what I had expected, re-branding a 100ns chip as a 120ns chip.

This seems to be a deliberate choice for inventory reasons – it turns an otherwise “potentially common” chip according to the 1999 datasheet into one which might be limited supply and worth a little more. That being said, you could always substitute a faster chip with few ill consequences, but those companies looking for like-for-like replacement might be willing to pay more for newer like-for-like stock without knowing that it has been remarked.

The other M27C4001 did have its complete surface layer peel off using the same super-glue + paper + wait to set + knife technique, but underneath, there wasn’t too much untoward. The remnant markings seemed to match the remarked printing – the reason for the remarking may be due to the refurbishing and cleaning process which may have damaged the original markings. It still isn’t a legitimate reason for it – so be careful with the parts you get, because they might not be what you were expecting!

The M27C64A proved to be even more interesting – it failed to provide its chip ID and refused to program at the specified Vpp. For all intents and purposes, even though it looks like the real deal, it doesn’t seem to work quite as such. However, jacking up the Vpp to 21v seems to allow it to program (well in excess of the 14v absolute maximum) although the chip ID seems permanently damaged. It also seemed more susceptible to erasure, so it might have been erased during the refurbishment process.

Conclusion

I didn’t expect to have been a victim of component re-marking, which is an issue amongst others with component recyclers which have been caught substituting components for military applications. When buying these “second hand” parts regardless of the description, one has to be careful as they might not be getting what they expected – it could be counterfeit, it can be a different specification or age, or it could be an inferior “compatible” product marked as a branded one.

Despite this, I have now a collection of EPROMs to play with, and it was nice to admire their beauty.

There are a lot of different EPROMs which aren’t represented here – in fact, this collection is very STMicroelectronics centric because they were available for cheap. Maybe I should spend some more money and collect some other manufacturers – e.g. Fujitsu, National Semiconductor, Mitsubishi etc. Maybe if someone would like to donate some to be featured on the site, I’d be happy to take care of them!

About lui_gough

I’m a bit of a nut for electronics, computing, photography, radio, satellite and other technical hobbies. Click for more about me!

This entry was posted in Computing, Electronics, Photography and tagged , , , , , , . Bookmark the permalink.

4 Responses to Tech Flashback: The (beautiful) EPROM Chip Collection

  1. Mark says:

    I once wondered about that “body contact” that you mentioned and asked a sales engineer for the chip maker. I was told that it is actually a connection to a separate diode chip that is used by the programming voltage circuit. The chip technology at the time could not do it on-chip, so they used an off-chip diode.

    • lui_gough says:

      That could be quite plausible, since many of the chips used split-rail programming where the Vpp would be as high as 21v, although commonly 12.5v with the data and Vcc at TTL logic levels (5-6.25v). How the diode would help may have to do with preventing Vpp (at the higher voltage) from driving Vcc. It would be interesting if anyone has old TI EPROMs, as I’ve heard some of them use three-rail programming, whether they have two diodes or something?

      Thanks for that, glad to know that has intrigued others as well.

      – Gough

      • Mark says:

        I have some old three-rail eeproms… somewhere… They were 27xx’s. The single supply ones were 25xx.

        I was TI’s first volume customer for their 2532 EPROMs. They went into some oil well monitoring equipment. Each controller used 8-16 of them (at the time the standard price was over $100 a chip… we got them for a lot less, but still a bunch ‘o bucks each)

Error: Comment is Missing!