Source for SDRAM chips

purchased? yes.

comes to mind. sometimes and or have them in too. sometimes has the chips for sale or auction too.

Are you looking for specific chips, or just any parts in approximately the right density?

I had big trouble trying to source 1Mx16 SDRAMs for an ARM prototype run we did about a year ago. I eventually cannibalized SODIMMs :(

I've been able to source Micron's 256Mbit SDR DRAM from Arrow electronics on one-off quantities for ~$10 ea. (MT48LC16M16A2TG-7E, the

4Mx16-bit version). They seem to have been on allocation forever, but in small qty it hasn't been a problem.

be purchased?

In the UK: Farnell

Wim

Thank you. That was very helpful.

I'm not looking for any particular chips, mainly just anything at about 1M x 8 and up.

and up.

what happened when you did web searches? what did you find with mouser, digikey, and other vendors that you found from your web search?

No stock, or minimum purchase a waffle, for every single part number I could find that was compatible with my design. Mostly it was no stock. I eventually begged some samples out of Micron, I think.

I haven't found anything with Digikey, Mouser, AvNet, Future, Allied, Newark, All Electronics, or BG Micro. So far Arrow, which was suggested by Richard H., seems to be the only major vendor that sells them in small quantities. I've seen them on eBay before, but I'd like to find a source that's a little more reliable.

I'm glad that pointer was useful. IIRC, I went about it the reverse direction - asked the manufacturer who stocks their parts and sells qty

1. In fact, I think I was asking about samples sources. I stopped when I found the first one, but you might try this with other manufacturers.

I got the sense that DRAM prices are too volatile for most distributors to stock...

What quantities are you looking for? A TSSOP tray looks to be 108 pcs (~USD$1000 for 32MB chips) - a bit much for hobby, but maybe not for short-run production. Distributors might be more willing to broker full trays.

Also, you might check with smaller shops that make products with similar parts - they may be willing to sell you loose parts from their inventory.

I think the main problem is that not many people these days use loose DRAM chips, as modules are so widely available - for development quantities, it will often be easier and cheaper to strip chips off a PC memory module where possible.

If I have to do it again, I think I'll design in a SODIMM socket anyway (at least for prototypes). Being able to upgrade the RAM ex post facto is a Good Thing :)

Lewin,

When I last looked into this, I discovered SIMM sockets were very expensive through the Usual Sources - ballpark USD$25. Even if the SDR SIMM was free, the solution was much more expensive than just buying an oversized chip. Have you found a source for SIMM & DIMM sockets that's reasonably priced?

Also, you mention SODIMMs - are you using a DRAM controller? SDR seems to be the end of the road without one (timing stability, clock speed, etc.), but apparently they aren't available as discrete parts anymore - know of any discrete DRAM controllers?

Richard

Hi,

The cheapest I've seen 144-pin SODIMM sockets is on the secondary market at about $1.50 each. In production quantity of course they are cheaper than this. Digi-Key sells them at $5.98 qty 1, going down rapidly with quantity. WM18501-ND is the part#. 64Mb PC66 or PC100 SODIMMs - if you can find them - start at around $10 in surplus quantities.

I was using ARM devices with on-chip SDRAM controllers, e.g. Cirrus

7312, Sharp LH79520. Async DRAM is virtually a dead product line, no? If I had to add SDRAM to a micro that didn't have an on-chip controller, I would either change micros (most likely) or if ABSOLUTELY necessary I would think about doing the controller in an FPGA. opencores has an SDRAM controller.

I was thinking about doing this. Do you happen to know how many gates would be needed for a basic SDRAM controller? I have a project in mind where I'll need to put an SDRAM controller (among a few other things) into a cheap CPLD, but at this point I'm not even sure if that will be possible. I don't need to implement all of the functionality of the SDRAM - just reads and writes of single locations, as well as refreshing when necessary.

Thanks. I'll take a closer look.

The part # I spec'd earlier is sync, but older single data rate stuff (SDR SDRAM). Still viable, but relatively dead technology. I'm not familar with Async DRAM - another name for SDR?

DDR being the next generation, the clock needs to be too regular and the signals too well timed to bit-bang. Yet to be proven if I can even get SDR to work, but there are encouraging projects that show it can be done.

This seems like the logical course of action, since most MCUs that'd be used with SDRAM have integrated DRAM controllers. But this is one of those personal quests to work the impossible. :-)

Thanks! Richard

Not sure that this can be done with a CPLD, at least not without driving it from the MCU. My limited knowledge of CPLD and FPGA leads me to think you'll need an FPGA to make it run autonomously.

I don't think there's much more, except burst modes for higher performance I/O. But refreshing is no small task - it's a huge overhead to bit-bang without a controller.

See if the SDRAM you're looking at supports self-refresh mode. I'm not sure if it's a standard capability, but I know Micron supports an internal self-refresh that's 'set & forget'.

If you're looking at newer SDRAM technologies, you'll probably have to use a DRAM controller of some sort because of the tightness and consistency of the timing specs. What type of SDRAM are you exploring?

Richard

Hi,

Asynchronous DRAM is what we had before SDR. When you say "DRAM" with no leading S, I assume you mean asynchronous DRAM. It's much simpler to use than SDR. I haven't used DDR in a design, I kind of think of it as a "PC component" because the micros I work with don't support it.