gd@geovision.uucp (Gord Deinstadt) (06/09/89)
In article <8YWwgEy00WBn01718g@andrew.cmu.edu> zs04+@andrew.cmu.edu (Zachary T. Smith) writes: > >Hi, I have a question about a form of EEPROM/RAM >that I heard about going on one year ago, called some- >thing like Ferromagnetic RAM; I don't remember exactly. Ferroelectric RAM. The usual silicon dioxide insulator in the RAM cell is replaced or augmented with a ceramic such as lead-zirconinium-titanate (PZT). This is a non-linear dielectric with hysteresis; in effect, the crystal structure is altered by the applied electric field (yes it is a piezoelectric material) and it stays altered until an electric field is applied in the other direction. While in the altered condition it emits an electric field, just like a permanent magnet emits a magnetic field. Therefore this memory is NON-VOLATILE! > >What I want to know is: > >Do these parts exist?? What kind of r/w access times >can I expect on them, and where can I get them? As far as I know they exist only in the laboratory right now. Two small companies are pursuing them. One company has licensed the technology to ITT in Europe. ITT wants to use them in digital TVs. I don't think they really care about the non-volatility; they are more interested in the fact that PZT has a much higher dielectric constant than SiO2, giving a higher capacitance in each RAM cell. This makes it much easier to scale down the cell size for higher density. I believe it's a factor of about 6 improvement, meaning you can go right to the next-generation (4 times denser) without doing much else besides just shrinking the transistors. >Are they pin compatible with std. SRAM I.C.s? I imagine they will be. I also expect that in five years the standard DRAM technology *will* be ferroelectric. The major chip companies all ho-hummed at the announcements, but I think once they see some real parts out there they will switch over pretty quickly. ITT reports that it is pretty easy to convert standard production lines to make FRAMS. >Thanks in advance, > >Zach Smith (zs04@andrew.cmu.edu) You're welcome. Gord "I LIKE FRAMS" Deinstadt gdeinstadt@geovision BTW, the name for a thing that has a permanent electrostatic field, without actually having any voltage on it, is an "electret". You may have heard of an "electret microphone".
pcg@aber-cs.UUCP (Piercarlo Grandi) (06/13/89)
In article <661@geovision.UUCP> gd@geovision.UUCP (Gord Deinstadt) writes: In article <8YWwgEy00WBn01718g@andrew.cmu.edu> zs04+@andrew.cmu.edu (Zachary T. Smith) writes: >Hi, I have a question about a form of EEPROM/RAM >that I heard about going on one year ago, called some- >thing like Ferromagnetic RAM; I don't remember exactly. Ferroelectric RAM. The usual silicon dioxide insulator in the RAM cell is replaced or augmented with a ceramic such as lead-zirconinium-titanate (PZT). This is a non-linear dielectric with hysteresis; in effect, the crystal structure is altered by the applied electric field (yes it is a piezoelectric material) and it stays altered until an electric field is applied in the other direction. While in the altered condition it emits an electric field, just like a permanent magnet emits a magnetic field. Therefore this memory is NON-VOLATILE! I had read many years ago (around 1981) an article in Electronics on ferroelectric memory. It is the invention apparently of a single person (an old style inventor?), and was developed in coolaboration between him and some West Coast govnt. lab. They had an argument on it (he alleged they wanted exclusive rights for peanuts or else...), and this set back the entire idea. Ferroelectric is a misnomer; from what I remeber a certain *potassium* compound has a rombohedrical crystalline structure; one of the electrons can be indifferently in either apex of the rombohedron, and can be moved or detected from one to another by electric field. Once moved, it stays there, so it is not volatile. The advantages are: high noise immunity, very low power dissipation (none on standby), which means that multilayer (i.e. 3D) memories are possible, and almost infite scalability (down to the single crystal cell), the limit being the resolution of the electric grid. It is therefore an old technology, and I haave been hoping for the last ten years that it comes of age. Before these, the most recent news were that the technology had been licensed by Australians, in a hope to leapfrog current technology and build an indigenous industry around them. If FERAMs become common, the impact on architectures will be very profound. Just think of 3D layering, and non volatility. Mass storage, even more than core storage, could be swept away... (wishful thinking probably, but who knows...). -- Piercarlo "Peter" Grandi | ARPA: pcg%cs.aber.ac.uk@nsfnet-relay.ac.uk Dept of CS, UCW Aberystwyth | UUCP: ...!mcvax!ukc!aber-cs!pcg Penglais, Aberystwyth SY23 3BZ, UK | INET: pcg@cs.aber.ac.uk
sms@ficc.uu.net (Stanley M. Sutton) (06/14/89)
In article <1012@aber-cs.UUCP>, pcg@aber-cs.UUCP (Piercarlo Grandi) writes: > I had read many years ago (around 1981) an article in Electronics on ferroelectric > memory. It is the invention apparently of a single person (an old style > inventor?), and was developed in coolaboration between him and some West > Coast govnt. lab. They had an argument on it (he alleged they wanted exclusive > rights for peanuts or else...), and this set back the entire idea. > I believe there was an article about a pair of US companies that were marketing FRAMs as military replacements for EEPROMs. The major advantage was that they were much more radiation resistant than standard EEPROMs. Like EEPROMS, the write cycle was *much* longer than the read cycle. Both companies felt that they would eventually solve the problem. If anyone is interested, I think I could find the companies names and the article reference. Please send email, thank you. Also, I believe both companies were going to market the memories at exactly the same prices as EEPROMs, even though FRAMs are much cheaper to produce. -- Stanley M. Sutton, Service & Test, Ferranti International Controls Corp. Work: uunet.uu.net!ficc!sms, sms@ficc.uu.net, +1 713 274 5023 Home: bigtex!texbell!sugar!sms, sms@sugar.hackercorp.com Opinions may not represent the policies of FICC or Service and Test.
beyer@cbnewsh.ATT.COM (jean-david.beyer) (06/14/89)
When I was a summer student at Bell Laboratories, in the late 1950's, a man named J.Reid Andersen (I am no longer sure of the spelling) was making ferroelectric memories in the lab. He was using Barium Titanate, as I recall. The things worked, but since they were so much slower than thin film and magnetic core memories, they seem to have lost interest. He put down 256 bits on a little chip about the size of 1 square centimeter. I assume todays technology could increase the density somewhat, but I do not know enough about it to know what would be required to make them faster. -- Jean-David Beyer AT&T Bell Laboratories Holmdel, New Jersey, 07733 attunix!beyer
gnb@bby.oz (Gregory N. Bond) (06/15/89)
In article <1012@aber-cs.UUCP> pcg@aber-cs.UUCP (Piercarlo Grandi) writes:
[...]
It is therefore an old technology, and I have been hoping for the last ten
years that it comes of age. Before these, the most recent news were that
the technology had been licensed by Australians, in a hope to leapfrog
current technology and build an indigenous industry around them.
In fact there are some serious patents held by an Australian company
working in the area, Ramtron Australia Limited, arising out of
research conducted at RMIT in Melbourne, Australia. There was an
article in the paper a few weeks ago that Ramtron's U.S. subsidiary
had signed a very large contract ($50m?) with, of all people, Alcan to
develop the technology. Apparently there is a very good match between
the manufacturing of ceramic components (which Alcan is very involved in)
and the Fram devices, much closer than traditional Si processing.
Ramtron needed manufacturing technology and Alcan wanted to expand its
hi-tech materials exposure.
According to the hype in the financial press (where Ramtron is
somewhat of a white knight of the hi-tech stocks around here) they
hold fundamental patents on the whole technology and should do
extremely well out of it. However, it is quite unlikely that there
will be an "indigenous industry" as there is virtually ZERO
infrastructure for a components manufacturing facility. And little
chance the Govt. will provide the required support.
Disclaimer: I'm not a Ramtron shareholder! and this is from hazy memory.
Greg.
--
Gregory Bond, Burdett Buckeridge & Young Ltd, Melbourne, Australia
Internet: gnb@melba.bby.oz.au non-MX: gnb%melba.bby.oz@uunet.uu.net
Uucp: {uunet,mnetor,pyramid,ubc-vision,ukc,mcvax,...}!munnari!melba.bby.oz!gnb
steve@basser.oz (Stephen Russell) (06/15/89)
In article <1012@aber-cs.UUCP> pcg@cs.aber.ac.uk (Piercarlo Grandi) writes: >[...] It is the invention apparently of a single person (an old style >inventor?), and was developed in coolaboration between him and some West >Coast govnt. lab. From article in June 1989 "Elcectronics Australia": the "technique for applying ferro-electric materials to semiconductor devices [was] invented by George Rohrer [from] Michigan". Rights to this technology was purchased in 1983 for $7.5M by (what is now) Ramtron Australia, a wholly Aus company. >Ferroelectric is a misnomer; from what I remeber a certain *potassium* >compound has a rombohedrical crystalline structure; It is the titanium atom in a PZT ceramic which can be in one of two stable positions within the lattice. >The advantages are: high noise immunity, very low power dissipation (none on >standby), which means that multilayer (i.e. 3D) memories are possible, and >almost infite scalability (down to the single crystal cell), the limit being >the resolution of the electric grid. Disadvantage of current technology is a "half-life" of around 10**10 state reversals. Current implementations team the FRAM with CMOS SRAM which is initialised/saved at power on/off to reduce FRAM cycling. >It is therefore an old technology, and I haave been hoping for the last ten >years that it comes of age. It may happen yet. Here's some other snippets from the article: o Ramtron International Corp, a subsiduary of Ramtron Aus, is developing chips at their plant in Colorado (no way it could happen in Aus, sigh...) Research was also done with U of Colorado, Colorado Springs, at cost to Ramtron of $13M. o Ramtron has signed licensing agreements with ITT Semis, NMB Semis of Japan, TRW and Seiko Epson. o R&D team at Colorado includes Fred Gnadinger, Sheffield Eaton and Doug Butler, ex Inmos. Richard Horton, ex Honeywell and TI, is President. o Ramtech showed first device, FMx801, at ISSCC Feb 88. This was 256x1. Current production is 4K and 16K devices, with 256K due "early in 1990". Early device used 3 micron rules and a 6 transistor cell, current stuff is 4t, new design is 1t DRAM style cell. o Charge storage density is around 100 times that of SiO2. o Current goal is 10**15 cycle "half-life" for DRAM-style cell, giving "conservative 500,000 hour lifetime for 1T/1C [the C is FRAM `capacitor'] operating at 20MHz". So, should be worth watching. Ramtron seem confident they can scale up to 1M+ chips in next copule of years. If the 1T/1C cell works, it should make a dent in the DRAM market. Also, it appears the PZT will work with other semiconductor materials such as gallium-arsenide.
mslater@cup.portal.com (Michael Z Slater) (06/16/89)
I know of two companies that are sampling, or will soon sample, commercial ferroelectric chips: Kyrsalis in San Jose, and Ramtron in Colorado Springs. Krysalis is making nonvolatile octal latches, pin-compatible with standard TTL chips such as the LS273. Ramtron has (or will have) a series of small, byte-wide RAMs pin-compatible with standard SRAMs, from 1K to 16K bits. The parts are reasonably fast (100 ns) but have one potentially serious problem -- they have a limited endurance, specified as 10^^10 read/write cycles per bit by Ramtron. While this is much better than the 2^^4 or 2^^6 specified by EEPROMs, its not nearly enough to treat them like regular RAMs. I don't know what the failure mechanism is. I'm just starting to research an article on this subject; anyone who is especially interested can send me a note. Michael Slater, Microprocessor Report mslater@cup.portal.com
mdeale@mira.acs.calpoly.edu (Myron Deale) (06/20/89)
Hello, FRAM's show much promise, we shall see if this is fulfilled. Two good references that I've run accross are: Horton, Richard and Castelberry, Randel. "Nonvolatile FRAMs fit standard memory applications." Electronic Products, 6/1/88, pp. 41-47. and Evans, Joe. "Ferroelectric Memories." ESD: Aug. '88, pp. 29-35. The article by Horton covers the physics of FRAMs quite well in my estimation. Some notes: the key ingredient seems to be lead-zirconate titanate (PZT), which is used to form ferroelectric memory cells. Horton works for Ramtron and their goal is an endurance of 10**15 memory cycles, currently having achieved 10**11 cycles. This is for the prototype 256 bit (2**8) part. May have scaling problems? Evans works for Krysalis. Their K28C16 2Kx8 UniRAM nonvolative SRAM with 250ns cycle time "is designed to allow 10**9 power cycles and have one-year retention of state powered-down." The ferroelectric memory is only accessed during power-up or power-down. Evans notes "10**12 accesses in 10 years requires 3000 accesses per second constantly on a single byte." In addition, FRAM density should be superior to EEPROM. These companies currently need to work on manufacturing processes. I think I read some- where that these companies have signed on a few heavy-weights to help develop the products and bring them to market (eg. ATT?). Myron // disclaimer: I have nothing to do with either Ramtron or Krysalis except I like what they make. In addition, the above info is a year old. I expect Michael Slater to have more current info (whom I have nothing to do with except I like uReport). "And futhurmore, I have nothing to ..." // live free or die, Open Letter, Living Color.
mslater@cup.portal.com (Michael Z Slater) (06/21/89)
> Evans works for Krysalis. Their K28C16 2Kx8 UniRAM nonvolative SRAM >with 250ns cycle time "is designed to allow 10**9 power cycles and have >one-year retention of state powered-down." The ferroelectric memory is >only accessed during power-up or power-down. Evans notes "10**12 accesses >in 10 years requires 3000 accesses per second constantly on a single byte." Krysalis is not going into production with their K28C16. They had a number of problems with it, and have decided to target smaller devices first. >In addition, FRAM density should be superior to EEPROM. These companies >currently need to work on manufacturing processes. I think I read some- >where that these companies have signed on a few heavy-weights to help >develop the products and bring them to market (eg. ATT?). That's ITT, a West German company, not ATT. FRAMs have tremendous long-term potential. Unfortunately, the short term goals seem to keep slipping. Michael Slater, Microprocessor Report mslater@cup.portal.com