Teardown, Repair & Test: AY/Generic UV EPROM Eraser

With many older computers and their peripherals, the firmware code is stored inside a ROM chip of some sort. These were unlike the later EEPROMs and Flash chips which could be electrically erased and reprogrammed. Instead, these chips were often either one-time programmable (OTP) ROM chips which were permanent (meaning that a firmware upgrade required a chip swap), or an EPROM.


EPROMs are eraseable, however, to do this requires a relatively intense source of UV light. These EPROMs are typically made in a ceramic DIP (or PLCC) package (sometimes CERDIP for short) and are no longer widely manufactured to my knowledge. These packages are made by epoxying two halves of ceramic, with the top part containing a quartz or fused silica window that can admit UV light into the chip underneath. In most applications, post-programming, the window is covered with a silvered self-adhesive label to prevent any stray light from getting in and potentially slowly erasing the chip.


These chips are relatively unique, as they allow you to look inside the chip itself. As a result, these are some of the most beautiful chips to see, as you can literally “see” the patterns which form the memory array, and bits and pieces around it which form the row/column matrix decoders etc. Not all chips are of the same capacity, and depending on the vintage and technology, they can have markedly different die sizes and designs. The chips also have a hefty weight. The macro picture above is of the Signetics 256kbit chip which is from 1988 (one year older than I am).

Sources of UV light that will erase such EPROMs in a timely manner are relatively limited, with the most common and most recommended source being one that uses a UV germicidal low-pressure (254nm) lamp. As this wavelength of UV-C is very dangerous for direct exposure, such devices are typically rare and expensive. Other alternatives such as using sunlight often have slow and inconsistent results.

But Wait. There’s a Cheap Version?

To my surprise, when I went looking on eBay, I found a cheap and nasty eraser for about AU$19. In fact, this particular unit seems to have flooded the market for UV EPROM erasers and is resold by umpteen sellers. It’s not very expensive at all. I promptly ordered one, as I would love to be able to erase my existing EPROMs and overwrite them, but the unit that arrived had a bung power switch that was stuck in the OFF position.

I nagged the seller, and all I got back was a lousy AU$5. I suppose that’s better than nothing, so I set off to repair it myself, and while I was doing that, I thought I’d give it a review as well, so here goes.


This unit is often sold without a brand, with the words “durable, fast” attached to the name, but it’s branded AY in my case. The unit is made of a recycled plastic body, of the same teal green weirdness that was the recycled plastic rulers they used to give away to students. It’s light, feels hollow, and features faded printing on the top and a broken mains power switch. It also has a timer, stolen from a box fan.


The erasing chamber is accessible from the front, as a plastic drawer with no handle, but a hole where a handle might be. This hole goes right through to the inside, which isn’t a particularly good idea as UV-C is harmful to skin and eyes. Even though the instantaneous dose for a few seconds of accidental exposure won’t immediately blind you, it is going to cause some damage still. As a result, I patched the hole on both sides with black electrical tape just to be safe.


The drawer itself is not particularly big, but enough to squeeze a few chips side by side if you should have the luxury of having a few chips. The drawer itself is actually “melted” to the front lid to attach it, another sign of quality construction. The non-proven ESD properties of recycled plastic does concern me somewhat.


The unit has plastic feet, with no rubber, and a QC passed sticker which makes me wonder what sort of QC was passed in the first place.


Suspiciously thin flex is used, although not thin enough to pose a major issue, and of course, comes fitted with a two parallel prong plug as it’s a non-compliant piece of “imported” equipment.

Repair Teardown

Taking the unit apart was no big hassle – four Philips screws later and I’m in.


The insides were, as expected, remarkably simple but also remarkably dusty. I wondered if the plastic and factory conditions may have contributed. Anyway, they decided to save every penny they could when making this device – for example:

  • They reused a timer from a box fan as a timer for the eraser.
  • They used a power switch in series with the timer switch that is somewhat redundant, as the timer switch has a permanent ON setting as well. The switch was so cheap as to break in shipment.
  • They have no proper strain relief for the cable, and instead, it’s just knotted inside the unit.
  • The lamp holders are made of plastic, with the pins directly soldered to wires because they didn’t want to pay for the connectors.
  • The electronic ballast itself is made on a cheap single-sided PCB by hand in an electronics 101 style fashion, with a design that’s vaguely similar to that of most CFL bulbs, minus any fuse protection.
  • There is no effective UV lamp reflector or shielding that stops the other components from being exposed to UV-C.
  • The UV-C lamp appears to be the cheapest unbranded 4W tube they could get, with score marks on the glass.
  • The PCB is secured by three, rather than the four screws it was designed for.

As UV-C is known to degrade plastics, this design is a ticking time bomb. The plastics will become brittle, flaky and craze due to exposure. Worse still, the PVC insulation on the cables can get brittle, meaning that the insulation may fail and cause a short circuit. I suppose for occasional usage, this isn’t likely to be a major issue, but for actual heavy use, it’s likely that the unit wouldn’t survive too long.


After cleaning the tube of external dust, it was discovered that the glass inside has specks which appear to be the thermionic emission mix on the filaments already “falling” off, suggesting a poor quality tube. I suspect the whole unit will never last long enough for it to be replaced anyway, hence the “disposable” sort of construction.


Removing the PCB also showed that the front drawer was on a slide which did not actuate any interlocks whatsoever, so it is possible to open the draw during operation and get exposed to UV-C. Not a very safe arrangement.

2016042122284031 2016042122274030

Even the PCB showed issues – namely a lack of terminal points for connections of wires which had to be soldered directly to the underside of the PCB. Again, I will reiterate that there is no primary side fuse – so those diodes will probably have to become the fuse when something goes wrong. Cheap capacitors are used, as expected, and the underside shows that the PCB is so cheap that even the solder resist isn’t applied consistently. It seems to be a design from 27th October 2008, so it’s been on the market for a while.


The bad switch was dissected (not pictured) and showed that the switching element had fallen out of place to the side thus blocking the actuation of the switch. As I couldn’t dissect it without damaging the switch, I opted to replace it with a slightly better quality enclosed dual pole switch (utilizing only one pole). As it’s a standard size, it fits in the original cut-out, but the AU$5 I got back from the seller definitely doesn’t cover the cost of parts and labour.


A commonly asked question is how long an EPROM should be erased for? A common answer is to run tests to find out. In order to test the effectiveness of the eraser, I decided to:

  • Erase the EPROM for 10 minutes, verify that it is all 0xFF.
  • Program all cells with 0x00 (as the unit erases to 0xFF).
  • Erase for 30s
  • Check all cells for errors
  • Repeat erase cycle until all cells show 0xFF.

At each 30 second cycle, the cells were checked for errors. A count of number of byte errors (i.e. at least one bit of the byte was corrupted) was recorded, as well as a device clean result (i.e. all device bits are 1’s).


Two independent runs showed slightly different results, possibly due to the difference in positioning and warm-up of the lamp. In both cases, device clean was achieved after 330s of exposure, with the first corruptions happening at the 150-180s mark.

Ultimately, it seems that 330s is just enough to reliably erase this particular chip provided it is placed roughly in the middle and the window is very clean. A safety margin should be added in regular usage, with some people recommending anywhere from 2x to 5x the amount of time that it takes for the device to register all 1’s. This means that in-practice erase times should range from 11 to 27.5 minutes. It is possible to damage EPROMs by over erasing, so exposing them to more than half-hour of radiation under the eraser is probably unwise unless there is good reason for it.

How does the erasure actually happen across the cells? I wanted to find out, so on the second run, I dumped the EPROM data as well. The data was then put through a small C program that took each bit, and created a whole byte from it – if the bit was 1, then it would output 0xFF, otherwise 0x00. This allowed for Photoshop to import it as a RAW image made of 8-bits greyscale mode of 512×512 pixels, each pixel representing a bit in the 256kbit device. From that, the following artful animated GIF was produced.


There are recurring patterns which likely represent different localized regions of the device, which may be closer to the edge or under a thicker piece of quartz. The erasure is not uniform, as cells are not always uniform in manufacture. I found this relatively intriguing, so I will probably repeat this with some other chips when they arrive.


The unit is cheap, and it does work, but its design is pretty basic and it’s very much a “timebomb” and safety hazard in many ways. For occasional light-duty use, it’s better than nothing, but I wouldn’t rely on one of these in any high or medium-volume circumstance, as it’s literally killing itself while in operation.

Appendix: Shoddy C-program to Convert BIN dumps into RAW for Photoshop

#include <stdio.h>

int main (void) {
 int tempchar=getchar();
 int mask=0x80;
 while(tempchar!=EOF) {
  while(mask) {
    if(tempchar&mask) {
    } else {

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, Lighting and tagged , , , , , , . Bookmark the permalink.

4 Responses to Teardown, Repair & Test: AY/Generic UV EPROM Eraser

  1. Mark says:

    That gunk inside the tube is probably mercury or a mercury amalgam used to generate the UV light.

    • lui_gough says:

      Ah yes, quite true and indeed possible, although I was more expecting to see it in one blob rather than dispersed at a cold state. At least the whole eraser (AU$18) is not much more expensive than the unbranded replacement tube (AU$12), so ultimately, it seems to be a bare minimum cost design.

      – Gough

  2. sparcie says:

    I’d be interested to see the same sort of experiment (with longer time frames) for a chip sat in full sun to see what happens there! You might get a nicer animation.

    I’m yet to get an eraser for this type of chips, it’s good to know which kind of one to avoid. Although the question remains if any other erasers are any better.


    • lui_gough says:

      Sadly, the only real “quality” units seem to be vintage ones, which you really don’t want to try since replacement tubes can be a pain to obtain (or not available at all). Instead, I suspect it’s still best to buy one of these “cheap and nasty” units and use them with caution with an expectation you might need to replace the unit after a while.

      As for erasure under the sun – it seems some guys had experimented with that and concluded that it takes weeks to months to achieve an inconsistent erasure. The reason is that the sunlight doesn’t contain enough high energy UV that would work reliably to erase the chip. I could probably run this experiment with the sun, but a peek outside at the weather today tells me I’m not going to get anywhere :). Besides, it’s also heading towards winter as well, which would only lengthen the time required. I will definitely consider it, should I have the time, as I will be getting a few batches of “new old stock” arriving with multiple chips of the same type, which I can afford to “sacrifice” (e.g. should it get blown away, stolen by the neighbour’s cat, etc.)

      Thanks for the comments as usual :).

      – Gough

Error: Comment is Missing!