Serial Rams
pjv
Posts: 1,903
Hi All;
I've been looking for a nice physically small (8pins?) inexpensive yet reasonably fast (10 Mhz clock, non I2C) serial·RAM chips for storing SX variables, perhaps·1 or 2 KBytes,·and I'm not having much luck. Some costly I2C units I did find had some unnecessary features such as a real time clock·that is of no interest for the extra cost. I think some also had some EERAM, and that does have some interest. But that whole I2C thing is just too slow and ugly. Been there.
Can anyone point me to a good ISP solution?
Many thanks,
Peter (pjv)
·
I've been looking for a nice physically small (8pins?) inexpensive yet reasonably fast (10 Mhz clock, non I2C) serial·RAM chips for storing SX variables, perhaps·1 or 2 KBytes,·and I'm not having much luck. Some costly I2C units I did find had some unnecessary features such as a real time clock·that is of no interest for the extra cost. I think some also had some EERAM, and that does have some interest. But that whole I2C thing is just too slow and ugly. Been there.
Can anyone point me to a good ISP solution?
Many thanks,
Peter (pjv)
·
Comments
Bean.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Check out· the "SX-Video Display Module"
www.sxvm.com
"A problem well defined, is a problem·half solved."
·
http://www.ramtron.com/lib/literature/FM25256ds_r1.1.pdf
25MHz 32Kbyte fram
More serial chips here
http://www.ramtron.com/doc/Products/serial.asp
regards peter
Available from Farnell, so I'd imagine that US Disties carry them too.
The fast ones would seem like good candidates if a faster Stamp was ever planned. Strapped to the side of an SX, there should be very little waiting around [noparse]:)[/noparse]
Steve
I know, I know google is my friend, but I've looked for about a half hour now with no luck...
The only distributor I found is newark and they don't have the FM25256.
Bean.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Check out· the "SX-Video Display Module"
www.sxvm.com
"A problem well defined, is a problem·half solved."
·
http://www.ramtron.com/doc/Products/serial.asp
Just click order samples
Or search their website for a local dealer
regards peter
There is also the Windbond W55206B a 128kx1bit in a 18 pdip (why they use an 18 pin when only 8 pins are used, I cant say).
Theres also the 1K-bit DS1200, but good luck on finding any (Dallas has a buy direct option on thier website). I think it uses I2C though.
Thanks for the pointer, looks to be exactly what I need. Future Electronics carries stock on some of the sizes, at a price of a couple of dollars, depending on quantity.
In looking at the spec sheet, some of the devices list 10 billion writes, some 1 trillion writes, and some (the low power "L" versions) "unlimited" writes. This does concern me a bit for general RAM applications.
One trillion writes, assuming a 1 MHz clock, and on a continuous write basis (unrealistic, but a potential upper limit) one trillion writes would be consumed in 1,000,000,000,000 divided by roughly 100,000 writes per second, is 10,000,000 seconds, or 2777 hours, or 115 days.
Using the device as a general purpose RAM with a clock of 10 Mhz could reduce this to just 2 weeks; hardly a long time.
I wonder if "unlimited" rally means that. Best to call and ask the question.
Cheers,
Peter (pjv)
so must·you not multiply the outcome by the number
of bytes? (I don't think you write the to the same address
all the time)
regards peter
No, I don't think so, I believe it to be a group issue. I will querie the manufacturer on this.
If ithe case is as you suggest, then it is potentially less of a problem unless of course the software writes one particular byte much of the time.
Thanks,
Peter (pjv)
It seems to me you could do this with standard SRAMs, say 65kx8. You'd need·4 pins (two for·latches and two for RAM controls) and a port for byte-wide data.
You basically latch the high address and low address bytes (16 address lines = 65k) in octal flip-flops (or latches). That's two latch lines.
The third line is for a R/-W (Read/-Write) pin on the RAM. This determines·access direction. The SRAM's -OE (Output Enable) control is number four. The RAM's Chip Select is just·forced to the enabling level, high or low, whichever is specified.
All the bytes, address or data, happen at the same port.
To do a read, start with the R/-W and -OE lines high, then:
1. use the data port to put out the high address byte and latch it with one latch line.
2. Same as above for the lower address byte, using the other latch line.
3. Set the port for input.
4. Set the -OE low. (this causes the RAM to·put out·the stored data)
5. Store the data at the port into a variable and process it.
6. Set the -OE high. (tristates the RAM's· output)
Done
To do a write, start with the R/-W and -OE lines high, then:
Do as in lines 1 and 2 above.
3. Latch the data to be written out the data port.
4. 'Strobe' (a quick "pull high/drop low") the RAM's R/-W pin. (This latches in the data)
Done
It is a fairly straight forward hardware layout. No quirky address schemes. No page boundaries to monitor and/or deal with. No block erase/write cycles. Unlimited reads and writes. Takes only (only?!) 12 pins.
Can somebody tell me how to·import drawings to these pages?·I'd like to submit a simple SRAM circuit for the SX. If anyone's interested.
Later!
kenjj
The write pulse is a quick "pull low/ go high" pulse.
kenjj
I'm aware this can be done with hardware, but I have a REAL aversion to adding stuff external to the SX. If there is a simple serial answer, I prefer to do it with software, clocking data in the background. For serious memory intensive operations, the parallel approach is perhaps more reasonable.
Thanks for your suggestion Ken.
Cheers,
Peter (pjv)
So in checking with the manufacturer, I'm assured that the "unlimited" write capability on the low voltage parts ("L"), is beyond 10^15, so that's plenty.
Furthermore, writes occur a byte-at-a-time as I had surmised, but accessing is in 8x8 byte matrix groups, and non accessed groups are not deemed to be written. So, statistically, the endurance is even longer.
Will need to get some samples and try this out. It truly appears to be a general purpose serial RAM that doesn't forget when the power is turned off.
Cheers,
Peter (pjv)
·
If you want a stroll down memory lane take a look at the 711/107 class of the US Patent classification system which is ferrite core memories, there are only 13 cases in there that were filled after 1980 and all but two of those are misclassified (thanks to a coworker who didn't understand what the technology was when he first started, so he placed hard drive patents in there). Another interesing class is 711/116 which is bubble memory, another type that went the way of the dinosaurs. If I ever do allow an FRAM case, I'll be putting it in the ferrite core class, because our system of classification only changes once every 20 years or so, so placing cutting edge technologies into technical catagories of yesteryear becomes its own challenge.
Post Edited (Paul Baker) : 3/24/2005 6:55:14 PM GMT
For those who care, I have now received and tested my FerroRAM chips form Ramtron. These things are awesome AND inexpensive. Reading and writing them (serially) with an SX I get approximately one byte per microsecond once the transfer is initiated, and that takes a couple of microseconds first.
Unlimited writes (virtually), and it doesn't loose its mind when the power turns off. Use it like a regular RAM, albeit serial for the model I have. Parallel is available, but of course at the expense of extra pins, but at a significant increase in speed.
Cheers,
Peter (pjv)
What model did you end up getting ?
Bean.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"SX-Video·Module" Now available from Parallax for only $28.95
http://www.parallax.com/detail.asp?product_id=30012
Product web site: www.sxvm.com
"One experiment is worth a thousand theories"
·
The original one is an FM25C160, serial SPI that is 16 Kbit arranged as 2K by 8 at 5Volt.
Since then we have also gotten and designed in an 8K by 8 at 3.3 volts.
Nice and fast, and they don't forget on loss of power.
Cheers,
Peter (pjv)
Have you found any place that sells them in low qty ?
Bean.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"SX-Video·Module" Now available from Parallax for only $28.95
http://www.parallax.com/detail.asp?product_id=30012
Product web site: www.sxvm.com
"One experiment is worth a thousand theories"
·
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
·1+1=10
I went through our local FUTURE rep, and he looked after getting the first samples. It did take a little while. Once we'd tested them and confirmed our satisfaction, then I believe we ordered the second ones through normal procedures. Not positive from where, but presumably the same supplier.
Cheers,
Peter (pjv)
Someone here used them in a project a while back and won't stop raving about them whenever it comes up. Kinda like me and the SX I think.
-dave
p.s. - Yes, the reps I know are amennable to helping engineers with personal/pet projects. As Ken known, the way to get products designed in is to get the engineers to love them first.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
This is not a sig. This is a duck. Quack.
Parallax people, pay attention!
Could Parallax maybe make a mass order and sell these at their web site? I know Parallax stands by their 8-pin serial EEPROMS for the Basic Stamp - Maybe these could be the equivalent for the high-speed SX? These would be great for Basic Stamp memory, as somebody else said, and (Parallax people, listen) you could sell them well to people using SXs. They would like not having to write as long of delay routines to access these.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
·1+1=10
The funny thing was I had known about these, and identified them for a project, and then found this thread...·· ...Glad to see someone else did their homework on expected write cycles.
-Dan
·