> At 10:47 PM 5/29/97 -0624, you wrote:
> >it would need to record one or two hundred bytes per second for several
> >minutes, ideally quite a bit more. So something like 32K to 256K of memory
> >would need to be written and then later read back. There is almost no
> >requirement to randomly access this memory, just need to write it as data
> >comes in and read it for upload to the PC.
>
> RAMTRON http://www.ramtron.com makes what they call FERAMs.  These are pin
> and software compatible with I2C EEPROMs, but the 10mS delay we are all
> used to using is _!NOT!_ required.  Also, you can write the entire chip in
> sequence after a single addressing command!

In some applications I've done using FRAM devices, there sometimes seems
to be a memory-loss problem when the system is powered down and back up.
I don't know whether this is due to the system's CPU going bonkers and
writing something it shouldn't as it loses VDD, but I'd suggest anyone
using the FRAM devices take care to ensure that their system shuts down
nicely when it loses power.

Also, another important thing to note about FRAMs: they perform a store
operation as soon as they get the last data bit of a byte, even if the
write is not terminated properly.  This is in contrast to most EEPROMs
which will abort a page-write cycle without writing anything if, e.g.,
they receive a stop in the middle of a byte.