Feeds

back to article Boffins read memory bits with light

Researchers from the US and Singapore have demonstrated a form of ferroelectric RAM that can be read with light instead of electricity – along the way overcoming a problem that has kept the technology locked away in a relatively small niche. FRAM is attractive as a memory technology because, unlike dynamic RAM (DRAM), it only …

COMMENTS

This topic is closed for new posts.
Silver badge

So does the cat exist, or not?

"While bismuth ferrite can be set and read electrically, it also has a useful photo-electric property. When illuminated, it produces a voltage, and the voltage varies depending on the material's polarisation state. Conversely, the polarisation state reflects its electrical condition."

Something fishy here ... Will continue to read up on this ... potentially useful, if true :-)

0
1
Silver badge

Long term reliability problems?

The Ramtron FRAM we're using in somewhat, er, difficult conditions is rated for 10e14 accesses. I'd be curious to see any papers showing long term reliability issues.

2
0
Gold badge
Boffin

Re: Long term reliability problems?

Depends on the clock frequency.

1Mhz That' 100 million seconds or about 3 yrs continuous operation.

1Ghz (reasonably high performance phone) 100 000 secs a bit over 1 day.

Of course if you run slow and switch in between the game changes totally.

0
0
Bronze badge

Re: Long term reliability problems?

Life depends on read/write cycles to a cell, not on the processor clock speed (unless you access every cell every clock cycle) ...if there's only a statistical average of one read per cell every 100 or 1000 cycles life extends greatly. Use it for your windows swap file and it will die!

0
0
Silver badge

Re: Long term reliability problems?

40Mb/s is the maximum clock speed for the serial part I'm using; it requires about forty clocks to do something useful, so in practical terms the access time is something under a MB/s at best. And I'm nowhere near that; I have to bit bang the part and I write each cell only two or three times in it's expected lifetime (hostile conditions).

0
0
Anonymous Coward

Re: Long term reliability problems?

Indeed. We are using a Ramtron chip as well and its expected life is in excess of 10x the PCB life.

0
0
Thumb Up

Re: Long term reliability problems?

I have a couple of those Ramtron devices for some homebrew stuff I'm doing. They're great --- absurdly easy to interface to and utterly reliable. I'm using SPI/I2C parts, and they speak the same protocol that serial flash and SRAM do, except simplified not to need erasure before rewriting. Could not be easier to use.

Weren't expensive, either, although I got the impression that Cypress (who own Ramtron) aren't really set up for people wanting to buy less than a thousand or so. (I got two. In a very large box.) Here's a datasheet:

http://www.cypress.com/?docID=42534

I do notice that they don't seem to have the parts in DIP8 any more, which is a shame is it's vastly more homebrew-friendly.

Also of interest is that you can get MSP430 microcontrollers with FRAM instead of SRAM. Which means, of course, that they can power down completely and restart with all their RAM intact. Again, not DIP, but it's still awesome.

0
0

You wouldn't need to illuminate individual bits, and it would be slow - one row would be fine, with the columns read off the conducting strips in parallel.

0
0
Silver badge

Intersecting grid, magnetic cell at each intersection, destructive readout... this design looks somehow familiar.

Everything old is new again. Just much, much smaller.

4
0

Ferrite-core memory

That's what I was thinking. Ferrite-core (magnetic-core) memory was one of the earliest implementations for random-access memory. And gave us the term core memory.

I seem to remember many years ago, Thompson were investigating reading magnetic tape using a laser. It would have eliminated the rotating head on VCRs but DVD was already looming so it never appeared.

Put the two together...

0
0
Bronze badge

What Danny4 said.

0
0
Silver badge
Coat

Sometimes, destructive read-out is an advantage

Destructive read-out is not necessarily a disadvantage.

It might be desirable in crypto systems, for instance.

Another case where the destructive read-out property is useful: Seeburg juke boxes (from 1955 onwards, anyway; starting with the V200, the industry's first 200-selection machine) used magnetic core memory in the TORMAT memory unit (mounted underneath and behind the record magazine). One bit is used per side of a record, to indicate whether or not it is to be played. One write-in wire through all the A's, another through all the B's ..... one write-in wire through all the 1's, another through all the 2's ..... you get the picture. Write-in pulse supplied (from a capacitor, which is only charged while there is credit) for the instant while both a letter and a number key are depressed, just before both are unlatched.

Read-out pulses come from the record player mechanism as it moves back and forth along the magazine (A-sides are played while travelling left-to-right, B-sides right-to-left). And the act of reading-out from the core implicitly resets the record side not to be played.

Mine's the one with the Select-O-Matic Mechanism manual in the pocket .....

1
0
Facepalm

Wait, so our new computer memory will be made of *oil filters*?!

http://www.fram.com/

0
0
Bronze badge

Wrong article title.

Should be "memory array finally sees the light", surely?

0
0
This topic is closed for new posts.