Re: [NOISE] Nasty-Quibble-Punks

[jim bell <jimbell@pacifier.com>]

At 04:06 PM 3/31/96 -0800, Mike Duvos wrote:
>"Perry E. Metzger" <perry@piermont.com> writes:

>Uh huh.
>
> > Incidently, EEPROMs don't work by simply charging a
> > capacitor or something silly like that. No insulator is
> > perfect, no dielectric is perfect, and charge would
> > eventually leak away were that the case. However, if it
> > were, it would be fairly easy to determine the state of a
> > cell without having to get particularly close to it. Beyond
> > that, there is this insane notion you seem to have that a
> > charged object will lose its charge if the "insulator" is
> > "stripped off" -- I wasn't under the impression a vacuum,
> > for instance, was a particularly good charge carrier.
>
>Uh huh.


Turns out Perry is wrong about this.  I believe that UV EPROMs and probably 
EEPROMs do indeed work by storing charge on a buried, totally-isolated 
capacitor.  The capacitor is charged with a system called "Fowler-Nordheim 
tunneling," which involves placing a relatively high voltage on a nearby 
electrode and causing the thin interface to temporarily conduct.  (It's odd. 
 That's why it's called "tunnelling.")   The charge, surprisingly enough, is 
stable for years, in fact decades, and probably (statistically) centuries at 
room temperature.  The reason the charge stays around so long is that the 
insulator, silicon dioxide, is extremely good.  It has to be.  If the 
capacitor were, say, 1 picofarad, and the resistance was 10E18 ohms (a 
billion gigohms) the resulting time constant would be 1E6 seconds, or about 
12 days.  Since EPROMs obviously hold data far longer than this (well over 
100 times longer, or else our computers wouldn't work!), and since the 
capacitance is probably not nearly 1 pf, that tells you that the effective 
resistance is far above 1E20 ohms.

UVEPROMS are erased, naturally enough, by exposing them to UV light, which 
is usually produced by a mercury vapor lamp.  This UV causes enough 
electrons to be excited into upper electron shells in the insulator to 
temporarily turn it into a slight conductor, and the charge dissipates.  I 
think EEPROMs are erased by, more or less, reversing the voltage on the 
charging electrode.

As for keeping the charge when that insulator is stripped off, that would be 
a problem.  It isn't that a vacuum isn't a good enough insulator; it is, but 
it would be hard to imagine a technique to strip off an SiO2 insulator that 
doesn't also allow a substantial amount of charge to flow.  You could strip 
it off with HF (hydrofluoric acid) but that's electrically conductive.  Even 
a gas-phase process would probably result in enough conductive products to 
discharge the capacitor.  Ion-beam milling would also remove SiO2, but as 
the name implies that's applying a current to the system.

Fortunately, all this is moot:  Since the floating gate is inherently part 
of a transistor, it isn't necessary to expose it to detect its charge state: 
Just activate the transistor in-circuit


BTW, some PLD's have a so-called "security bit" which (when set) is designed 
to prevent reading of the state of the rest of the programmed bits.  Years 
ago it occurred to me that if you knew where this particular bit was stored, 
you could expose this bit location alone to a UV source through a tiny mask 
to discharge it.  Finding that location wouldn't be all that hard:  Just expose 
the chip with a series of exposures, moving a linear mask slightly, and 
eventually the security bit will erase.  Note the location of the mask, and 
rotate the mask 90 degrees and repeat the process.  At that point, you've 
located the bit (this may require a few iterations), so you expose the 
target part through a tiny pinhole (Edmund Scientific sells them in many 
different sizes, exposing only that security bit location.


Jim "Mr. Bell talks about one field" Bell

(Let's see, I covered solid, liquid, and vapor phase chemistry, a bit of 
particle physics (ion-beam milling), semiconductor physics, minor optics, 
electronics, some trivial math, and maybe even some detective work!)

jimbell@pacifier.com