UV EPROM amnesia?

[eccerr0r]

Has anyone seen a UV EPROM (with sticker covering window) start losing data over the years? How many years? What device?

Curious how reliable these things are, or whether they need to be backed up as well.

Seems that EAROM/EEPROM I've seen decay over time but mostly due to cell wear and that reduces storage lifetime. EPROM usually doesn't get many erase/reprogram cycles so potentially they could hold longer? The dielectric must be different too...

[stj]

Re: UV EPROM amnesia?

yes - often.

retention on really old eproms like 2716 & 2732 is about 30years.
2764 upto about 27c010 is 20years+
larger roms seem worse - smaller cells = less charge to slowly disipate.

i always backup eproms.

let me explain how they work, an eprom is like a matrix of capacitors, charge one and it's a "0".
they *very slowly* self discharge.
or you can blow the electrons out of them wih UV light ionising the cells.

a flash chip is similar but has an extra transistor connected to each cell that can short it to ground to discharge the cell.
that increases leakage, so flash retention is usually stated as 10years - in reality they are often much worse - probaby because of temperature changes.


the smaller the cells, the worse the retention gets - got a big SSD? back it up atleast once a month.
some of the newer MLC flash stuff can only retain for months - not years.

[eccerr0r]

Re: UV EPROM amnesia?

I was thinking that the dielectric would be better than regular capacitors. Some of those low leakage film caps store charge for a very long time, nothing like aluminum electrolytic.

Just wondering about my Tek scopes, they have 27128s and 27512s in them, both are over 20 years old now and apparently they still checksum OK (at least for now...)

Also I wonder about those bipolar silicide fuse proms, wonder how they survive over time. I have no replacement for these... Need fast 40ns 4Kx8 (Am27S43A) - I don't think standard EPROMs go that fast.

[japlytic]

Re: UV EPROM amnesia?

i always backup eproms.

I always do this, especially for test equipment - the sticker covering the window of the EPROM should be of the foil type (I got these foil stickers from a local newsagent), and needs to be applied immediately once the EPROM is confirmed blank (after proper UV erasure) before programming.

[ozzy214]

Re: UV EPROM amnesia?

yes - often.

retention on really old eproms like 2716 & 2732 is about 30years.
2764 upto about 27c010 is 20years+
larger roms seem worse - smaller cells = less charge to slowly disipate.

i always backup eproms.

let me explain how they work, an eprom is like a matrix of capacitors, charge one and it's a "0".
they *very slowly* self discharge.
or you can blow the electrons out of them wih UV light ionising the cells.

a flash chip is similar but has an extra transistor connected to each cell that can short it to ground to discharge the cell.
that increases leakage, so flash retention is usually stated as 10years - in reality they are often much worse - probaby because of temperature changes.


the smaller the cells, the worse the retention gets - got a big SSD? back it up atleast once a month.
some of the newer MLC flash stuff can only retain for months - not years.

WOW!! Your serious? So SSD and any kid of flash memory data can degrade>???

So whats the best long term? Magnetic hard drives? This sparked my curiousty as I been thinking about getting a few ssds for long term storage.]

My magnetic drives been holding data for 10 years. I have old photos and memories still stored on old hard drives. Occasionally fire them up to insure drive is still good or back the data to another drive!! Some of my oldest data is 10 years on old hard drives!!

[eccerr0r]

Re: UV EPROM amnesia?

Nothing is forever, but some are more forever than others...

[Per Hansson]

Re: UV EPROM amnesia?

WOW!! Your serious? So SSD and any kid of flash memory data can degrade>???

So whats the best long term? Magnetic hard drives? This sparked my curiousty as I been thinking about getting a few ssds for long term storage.]

All magnetic and flash mediums degrades over time.
Allot of magic ECC work takes place in modern mechanical harddrives and SSD's to make the data readable.
SSD's are rated such that when all read/erase cycles have been used up the data must be readable for 12 months, 3 months for enterprise (server) SSD's.

Using up all write cycles of an SSD does not cause dataloss.
The standard mandates that the data must be readable 12 months from when this occurs on client drives.
For server drives it is 3 months.
The SSD will simply become "read only"

Oh, and don't take this as a reason to not perform backups, they are of course still necessary (theft, fire, lightning, flooding, data corruption, viruses & "error 40" just to name a few reasons)

[stj]

Re: UV EPROM amnesia?

old harddrives are probably best for longterm storage, the bits on the platter are physically bigger than modern drives.
i have 40meg and 80meg drives over 20years old that still read o.k.
of course you can get other forms of failure.

[stj]

Re: UV EPROM amnesia?

Also I wonder about those bipolar silicide fuse proms, wonder how they survive over time. I have no replacement for these... Need fast 40ns 4Kx8 (Am27S43A) - I don't think standard EPROMs go that fast.

they can fail - probably corrosion up the legs or internal matrix-logic failures.
i backup bproms too.

if yours fail i cant help, my largest spares (yes, i have NOS bproms!) is 512x8.
you may get away wih this:
http://uk.farnell.com/atmel/at27c256...-28/dp/2345637

i'm sure u.s. suppliers have them.

[ozzy214]

Re: UV EPROM amnesia?

Hmm thats interestign info. But okay riddle me this. If this supposed manadate is in place where the ssd cant write anymore, but data is still retrievable for up to 12 months right?

Ok why did I have two enterprise drives I bought off of ebay literally go poof on a power cycle? I mean ltieralyl smart was good, trim enabled. Was running fine. Checked with hd sentinel. Shut down system. Next day go to restart and the bios is reporting a sata port error, but still says drive is fine via smart. But I cant access under windows and linux?

This happen with 2 120gb ssd I bought of tab systems within a week!

But even more interesting is I have a ocz ssd I got off ebay dirt cheap in my lappie. Hd Sentinel reports one bad block, but that drive is runniong flawlessly for weeks now?

I think this goes against the mandate you speak of where data can be retrieved even if write cycles is up. I mean literally running perfect, next power cycle....dead. Well kinda, still could be seen. But giving a sata port error and cant be accessed...

[stj]

Re: UV EPROM amnesia?

could be the microcontroller failed that bridges between the SATA and the flash chips.

[ozzy214]

Re: UV EPROM amnesia?

Twice in a row with a week? Literalyl the one I bought last 3 days and the replacement less than 12 hours?>

[stj]

Re: UV EPROM amnesia?

firmware bug is possible - i know a lot of ssd's got firmware ugrades put online.

or a bad psu spiking them?
they arent loaded with caps like a motherboard to reduce that crap.

[fzabkar]

Re: UV EPROM amnesia?

AIUI, a typical SSD failure involves corruption in the Flash Translation Layer (FTL). The FTL maintains a table that maps logical LBAs to physical NAND memory locations. These corruptions don't seem to happen on the fly. Instead they show up at the next power cycle.

Current SSDs use very high density NAND chips (TLC) which are prone to bit errors and require much stronger ECC. I've seen someone at HDD Guru make the comment that Samsung are reverting to an earlier 40nm process to improve reliability.

Here is an interesting article which suggests that lowering the NAND supply voltage reduces "bit noise" and improves the bit error rate.

http://blog.acelaboratory.com/pc-300...ontrol-p3.html

FYI, Intel released a firmware update for their earlier SSDs that addressed a problem where the SSD would die after a power cycle.

[fzabkar]

Re: UV EPROM amnesia?

ISTR many TI 2716 or 2516 EPROMs dropping bits within a few years. It was more than just a bad batch, AFAIR.

[stj]

Re: UV EPROM amnesia?

probably wrongly programmed,
the TI chips needed 25v where as most others needed 21v.

TI 2716 are actually pin compatable wih 2708 and need 3 seperate voltages too.
they have very little in common with other 27' series.

ironicly the 2516 *is* compatable with other company's 27' series chips.

it's all very confusing for people who dont virtualy eat & sleep with early memory devices!

[eccerr0r]

Re: UV EPROM amnesia?

Then there's also the 64Kbit device in 24 pins... there was only one non-jedec that made them IIRC but don't remember. I don't know where all my 2708/2716/2732 devices are, I used to have a pile of them, but all I have now are my 2764-27512 and the 010, 040, 080, 160. I don't know what to do with the 42-pin devices as I don't have a programmer that will handle them...

There are a lot of chips with the "A" suffix that have smaller die, you can see a huge difference between a 2764 and 2764A... suspecting the 2764A has shorter data longevity?

I guess most of my equipment stay within chassis so whether they have a label or not probably doesn't matter. I do worry about accidentally damaging them while unsocketing to read them...

[stj]

Re: UV EPROM amnesia?

Then there's also the 64Kbit device in 24 pins.

2564 mask roms.

same pinout as 2532 but one extra address line instead of VPP
pretty rare - only seen them used by commodore.

[stj]

Re: UV EPROM amnesia?

I don't know what to do with the 42-pin devices as I don't have a programmer that will handle them....

you donate them to me to go with my stocks of 27c322

[eccerr0r]

Re: UV EPROM amnesia?

The 64Kbit 24 pin EPROMs must be pretty well sought after because that's the only thing that can replace the 64Kbit mask ROMs a lot of companies used, if they want to hack their firmware. Still have not come across any yet, at one point wanted to hack my TRS80 firmware but 24 pin 8KB ROM = no can do, at least without external decoding and two 4KB EPROMs; or a 24 to 28 pin converter, which probably would be easier...