A take at explaining the legendary bit rot

[drx]

Who hasn't heard of bitrot? I think we all have. Someone asked me if there is any truth to it, and I decided to check it out. Basically I checked the datasheets and did some calculations.

I did those for Atmel's 27C080 EPROM (8 Megabit - 1 Megabyte), based on their datasheet.

To erase an EPROM, an integrated exposure of the EPROM to a light intensity (irradiance) of 12000 µW/cm^2 for 20 minutes. This equals to about 14.4 Ws/cm^2 of integrated exposure.

Now, let's assume two scenarios:

* you keep the EPROMs uncovered, and they are prone to sunlight. Sunlight's irradiance is 120 W/cm^2. Let's divide 14.4 Ws/cm^2 by 120 W/cm^2. What do we get? 0,12 seconds. That's right, it takes a small fraction of a second of exposure to sunlight to kill an eprom. Pretty depressing, isn't it?

* you keep the window covered with something. I have no idea how much UV does that filter, so I cannot tell.

The general rule of thumb is - if you're storing prototypes, keep them in a dark place, and cover the windows. I've seen 30 years old carts not damaged by bitrot, but then again I've seen 10 years olds completely bitrotted.

Of course, the data I presented applies only to the specific EPROM I have chosen. It may vary with other EPROMs (for the better, or for the worse).

Hope that clears things up.

And if I convinced anyone to backing up his collection of prototypes, let me know, I'll be happy to assist in doing so :)

[ProgrammingAce]

Hold on there, let's check your math...

12000 µW/cm^2 = 0.012000 W/cm^2

According to this site (http://www.racerocks.com/racerock/abiotic/solar/solar.htm), the generally accepted standard is 1020 watts per square meter at sea level.

that would make sunlight's radiance 1020 W/m^2

(1 020 watts) per (square meter) = 0.102 watts per (square centimeter).

I love google calculator. So your math is already off by 3 decimal places.

You're trying to convert W/c^2 into Ws/c^2, that's not mathematically possible. So instead what we're going to do is work within the 20 minute boundry stated by the eprom's spec sheet. If i were to say that my computer screen is 14 inches wide, you can't just go and say that it's 14 inches wide per second. They don't mean the same thing.

Without doing any complicated math, we can see that sunlight has a higher wattage then is required to erase the eeprom.

Sunlight----------Eprom tolerance
0.102 W/c^2 > 0.012000 W/cm^2

It is safe to say that leaving an eprom in the sunlight for 20 minutes is more then enough to erase it. As for how long the eprom can be exposed, that can't be determined by these numbers, none of the values have a calculation for time. If you want to determine a time limit for exposure, you will need to find the watts per minute (centimeter squared) for the sun, which isn't a value i could find in google. We can assume an eprom is erased at 0.012000 W/cm^2 (20 minutes), so we could work from there.

That being said, eproms are like batteries. The UV writing process stores a charge inside the eprom. Like any charge, it can be dissipated through contact with air. An eprom can normally hold a charge between 10 and 20 years.

Here is far more information about them then you could ever want to know:
http://www.answers.com/topic/eprom

[drx]

Oh, I see my mistake. Maybe I could test it empirically.

[Mark30001]

I better push my sneakers over and make way!

[discordance]

ProgrammingAce a Watt Is a joule per second, drx's math was fine the units were muddled but his answer using his value for irradience of sunlight was correct. Using your value for irradience:

12000 µW/cm^2 = 12000 µJ/s cm^2

The tolerance is 1200s x 12000 µJ/s cm^2 = 14,400,000 µJ/cm^2 = 14.4J/cm^2

the energy received from sunlight is
0.102 W/cm^2 = 0.102 J/s cm^2

the time to exceed the tolerance is
14.4 / 0.102 = 141.18s = 2mins 20s

[Mark30001]

Thanks anyway, drk! You really should post this on other forums full of proto collectors. ;)

[DogP]

But you guys gotta remember that bitrot isn't only caused by exposure to sunlight. UV light is one way that erasure can occur (IIRC UV-C is the optimal wavelength, which IIRC is only ~0.5% of sunlight), but EPROMs work with a floating gate, which can naturally lose it's charge (I believe that's what the datasheets generally talk about for data retention length). Of course they're talking about worst case, so many chips will hold their programming for longer, but you can't know how long for sure.

Another thing to note is that you're looking at a datasheet of a relatively new device (8mbit)... many of the older and smaller devices didn't require as much UV exposure to be erased. I can tell you from experience that it does take a long time of sunlight exposure to erase an EPROM though. While I was waiting for my EPROM eraser to arrive a long time ago I wanted to erase a chip so I left it outside in the sun... after a day it still wasn't erased (then the eraser showed up the next day). I'm sure some bits were erased though, which definitely isn't good if you're trying to preserve a game.

DogP