Erasing a powered EPROM with UV?

[Artlav]

Hello.

I was curious on how the content of an UV-eraseable EPROM chip degrade when it's being erased.
So, i put a picture into a chip from a TV test signal generator (M27C512), powered it up, and put an UV erasing lamp next to it, expecting pixels to go pop... pop... pop pop popopopopop or something.

Instead, nothing happens.
The lamp normally erases this chip in 10 seconds, yet in as much as 30 seconds i waited not a single pixel disappeared.
Granted, the distance is a bit higher that usual (things in the way to put it flush against the chip), so intensity will be less, but can it be so much less than not a single pixel disappears?

Does the power being applied neutralize the erasing somehow?
If i pull the chip out and put it under the lamp for a second, it shows lines of half-erased pixels.

Also, is it harmful in any way to the surrounding electronics - like, would there be any voltage spikes from photons charging up the bits that can propagate outside, or erased bits latching up in current draining states, or etc?

 

[Floid]

Interesting question and I don't have an answer. It could very well be that the presence of power is strong enough that it prevents the UV light from dissipating the programming charge and thus erasing the chip. I don't know enough about the low level construction of EPROMs to know if that would be the case. Anytime I have ever erased an EPROM it was always without power applied.

But to your second question, the light will not hurt any other electronics.

 

[Averagesupernova]

Interesting. I have never tried erasing one while powered. It has been years since I have had anything to do with EPROMs. I recall the erasers we used would take care of one in about 20 minutes. Had never heard of anything that would erase in less than a couple of minutes. I know that in a window with direct sunlight during the day it would take a few weeks.
-
A side note, I recall an EPROM that had sine tables loaded into it for audio test tones had upper level address lines that were not used. They were left floating accidentally. When light shined on the window one or more of those address lines floated one way and when the window was dark the address lines floated the other way. This was in the development stage for a piece of test equipment so the windows were not covered.

 

[davenn]

firstly, what is the EPROM feeding its data into to display that test pattern ?
something that has video RAM or something ?
if so you may not see any thing until you rebooted the system and it tried to upload the data from the EPROM again

just my thought :)

but as others have said... I have never tried erasing an EPROM whilst its powered

Dave

 

[Artlav]

A few more experiments show that power does not matter.
Looks like it's just less intensity due to the square of the distance and maybe related to the lamp reflector geometry.

The chip does erase, but it takes over a minute.
And it erases differently than i expected.

At first nothing happens - a minute or so.
Then the pixels on one side of the screen start to flicker.
The flicker form vertical lines.
Lines become whiter and whiter.
Then the second side start to flicker.

If you turn off the UV, the flickering remains - as if the bits are stuck in a half-charged state.

As far as i understand it, there is some charge level that is defined as 1, and below that - 0, but the programmed bit level is much higher than that threshold.
As the UV light shines, it erodes the charge, but there is a long way from fully programmed to the threshold, so a minute pass before anything start to happen.

The flickering bits are also interesting - it suggests that the chip left in the sun or staying for decades in the dark would develop intermittents, probably in a form of weird bugs in the code, not just go out quickly.

Here is the video of the interesting part (better to watch in HD):
https://www.youtube.com/watch?v=eafhFs-ZOgs

I guess the half of the screen at a time is due to the chip being made out of two slightly different banks? Then, the address space is interleaved between them at 128 byte intervals.

All in all, you live and learn. :)

Had never heard of anything that would erase in less than a couple of minutes.

Soviet vintage mercury vapour induction lamp. 10W of pure 230-280nm UV, no light, no heat, no safety features. When it is placed face-down on the table no light leak out and it fits over a chip perfectly.

firstly, what is the EPROM feeding its data into to display that test pattern ?
something that has video RAM or something ?

No, just a CPLD that reads the pattern directly from it, and makes the video signal.

 

[the_emi_guy]

Very cool post!
Thanks for sharing it.

 

[Averagesupernova]

Very interesting.