Just How Vulnerable To Accidental Erasure Are EPROMs Anyway?

On the scale of things worth worrying about, having to consider whether your EPROMs will be accidentally erased by some stray light in the shop is probably pretty low on the list. Still, losing irreplaceable data can make for a bad day, so it might just pay to know what your risks really are.

To address this question, [Adrian] set out to test just how susceptible to accidental erasure some common EPROM chips are. An EPROM, or “erasable programmable read-only memory”, is a non-volatile memory chip that can be programmed electrically and then erased optically, by exposing the die inside the chip to light at a specific wavelength, usually in a special chip erasing tool. But erasure can also happen in daylight (even if it takes a few weeks), so [Adrian] cooked up an experiment to see what the risk really is.

He exposed a selection of EPROMs with known contents to UV and checked their contents. Three of the chips had a simple paper or foil label applied, while one had its quartz window exposed to the UV. As expected, the unprotected chip was erased in just 30 minutes. The covered chips, though, all survived that onslaught, and much more — up to 780 minutes of continuous exposure.

So rest easy — it seems that even a simple paper label is enough to protect your precious retro EPROMs. It’s a good data point, and hats off to [Adrian] for taking a look at this. But now we can’t help but wonder: what would a little sunscreen applied to the quartz window do to erasability? Sounds like a fun experiment, too.

45 thoughts on “Just How Vulnerable To Accidental Erasure Are EPROMs Anyway?

  1. I want to see how long they would last with one of the silver “floppy disk write protect” labels over the window, like I used to use a couple of years ago. (OK, nearly 40 years ago).

    Actually I have some 27c256 with checksums written on the silver label still, maybe I should check for myself ;-)

    1. I’m willing to bet some have degraded. I’ve repaired a few pieces of early 80’s tech where the EPROM suffered bit rot. One was an early LED scroller sign. The EPROMS were MOS (pre CMOS). I don’t know if that played a factor. When they were cold they’d sometimes work, but once they got hot, the sign would crash. Those MOS EPROMS were total power pigs too, drawing 700mA each, so they warmed up pretty quick.

      Luckily I found another sign from the same company that hadn’t been running for years on end. I copied it’s EPROMS on to their CMOS equivalent, and it started working reliably again.

      1. You can also try to read them at reduced voltage (4V or so), the threshold voltage of the read amplifiers is a fraction of the supply voltage, so if they are “barely” readable, this can give you the headroom you need.

        Programmers do the opposite thing and often verify at 6V supply, to be sure “it’s really in there”.

  2. Not all EEPROM are created equal. I’ve left some of them in the direct sunlight for few hours and not a single bit was erased. After reading the datasheet and buying the correct wavelength UVC LED it only took a couple minutes at a few mW power to completely erase them.

  3. I had some on a project near a window and had errors after several months. Not having labels didn’t matter when I was iterating and burning new revs regularly, but once I started working on something else there was trouble. Of course it wasn’t a big deal to just burn new chips and put lids on those.

      1. yep, unfiltered Xenon light (Xenon light on 90’s photography flashes are UV filtered, thus unuseable as-is).

        Seeing the comments of the video this particular device is a Zax Quick-EII (https://web.archive.org/web/20201023140659/https://reverb.com/item/28499982-zax-quick-eii-ep-rom-eraser) but I remember using big party xenon strobe lights with the UV window removed.

        A modern, portable unit that erase eeprom in seconds would definitively be a killer device in some circles.

  4. Only did this stuff a few years ago a few times with that “Minipro” model in the background on an XP machine that has since died. I have the little disc of software that came with it. Since then I have tried and nothing on the disc works or any of the scary choices on the web either. What to do? I’ve got XP on a slim case with a Soundblaster sound card in it as a tape and vinyl transcriber. It’s worth a try.

    Squirrel TV, me too. Too hot, they’ve gone outside for now.

    1. If by minipro you mean the TL866 it won’t program many eeproms without modding. The Vpp in the application shows 21V as the max programming voltage but the device only puts out about 16V. You can still get the job done with an external power supply.

        1. The schematic is online, as well as an open source firmware replacement. The programmer has two boost converters that create the programming voltages. Voltages are set by switching in different resistors in the voltage divider that feeds the internal comparator on the boost converter. Changing one or more of those resistors might be enough to get you there, although the entire boost converter should be reevaluated.

  5. Google Translate Amharic to English quote.

    “I’m really grateful she was released this morning and thank those who worked hard for her release, and we’re glad she’s alive,” he said.

  6. Xrays work also. I used to work at a place that tested EPROM die before packaging them, and they replaced the big UV “ovens” that took 30 minutes with a stepping Xray machine system that took about a minute to do the whole 6″ wafer. Yes, it was a long time ago.

      1. I headed that some old OTP ROMs and micro’s have EPROM’s but cannot be erased due to fact that they are in a non-windowed package. Maybe X-rays can erase these.

        1. There may be configuration information in EPROM that is set by the manufacturer, stuff like package pin count or amount of working memory. If there is an ADC on board, trim information on the reference voltage. Other than that, yes, you should be able to erase windowless packages.

  7. One other consideration is how hard the device was programmed. Spec sheets would indicate a minimum duration write pulse with the Vpp applied, and some programmers would just go with that, while ‘smarter’ ones would perform shorter writes but then read back and repeat until the ‘0’s were solid. End of the day it’s just about how much charge did you drive onto the floating gates. It’s going to be a product of Vpp and time. So did your burn just barely achieve enough for a ‘0’ bit? Or did you put it well over the threshold? The effect is of course cumulative, so that’s why if it didn’t take the 1st time you could often get it on the 2nd try.

    Point being, a marginally burned EPROM is going to start losing its 0’s a lot quicker than a solidly burned one.

  8. I seem to recall from back in the day that 2 weeks of sunlight would do it, so maybe a couple months near a window would also be a problem. We had neither in my lab, and the boards went into opaque enclosures, so it wasn’t a big concern.

  9. Someone used an eprom as a cheap camera sensor. Exposure times were astronomical high but it worked quiet well.
    Just write all zeros or all ones and start exposure through a lens and the object will be “written” to eprom.
    IIRC there had some corrections to be made for non linear memory mapping.
    Unfortunately I cannot find the article anymore.

      1. He’s talking about eprom, completewith a window on top. Sounds feasible, but I don’t recall seeing it done before.

        As for dynamic ram, it actually goes back to the Cromemco Cyclops in 1975. It was going to be a project, and when the Altair came along, it was changed to a digital interface.

  10. I ran across some windowed Microchip microcontrollers from the 1990’s not too long ago. They had been collecting dust on shelves for over 20 years, exposed to whatever was going on in the room at the time. Every one that I tested was happy as a clam with the stuff I had put on them way back when. Not that there were many bits to erase, but I suspect those things will still be there after the big solar EMP event wipes out the rest of our electronics.

  11. I just cover the quartz windows with Kapton tape… It’s roughly the same color as Amberlith so I’ve always assumed it filters about the same wavelengths of light. Well enough, at least. Not like I go exposing EPROMs to direct sunlight just to see if I can accidentally corrupt them… 🤷‍♂️

    The few UV EPROMs I’ve seen in my NEO-GEO MVS cartridges didn’t have any protection on the windows, so I guess it isn’t a huge concern, even in a production setting?

  12. I remember a lot of devices that used EPROMS as logic without a micro. Some commercial arcade game boards were full of them.

    I also remember, now accurately, predicting the 15-20 year data retention (quantum tunneling I think?) data sheet limit was going to come back and bite (byte?) them. I saw a real scramble of game restorers trying to find the data for their corrupted EPROMs.

    I opened my big mouth once and bragged about wirewrapping a programmer into my homebrew when I wanted to rewrite my boot EPROM. I wound up with boxes of EPROMS from guys restoring arcade games hoping to recover the weak data. I worked out some algorithms and was moderately successful. The interesting thing was many could be restored right to the same EPROM. The part didn’t wear out, the data did.

    Building the interface (Intell 8255 port with a transistor switch for Vpp) and figuring out the timing for programming the 25C32 was my first digital project for my homebrew. I think it was the first one to use a single 5V supply where as the 27Cxx needed +12v and the earlier ones needed a negative supply as well.

    That 5 x minimum retention pulse algorithm I didn’t see pop up in the data sheets until the EPROMS started hitting 128k & 256k.

  13. There may be configuration information in EPROM that is set by the manufacturer, stuff like package pin count and if there is an ADC on board trim information on the reference voltage. Other than that, yes, you should be able to erase windowless packages.

  14. In my industrial experience with old CNC machine often problem is that glue on the stickers get old, and windows is not covered… So please make backup or check stickers :)

Leave a Reply to JohnJCancel reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.