Sign in with
Sign up | Sign in
Your question

Thinking of assembling my own modules!

Last response: in Memory
Share
a b } Memory
August 24, 2002 3:27:08 AM

Well, the only company that even tried PC175 never released it to the public I believe. And PC166 is almost impossible to find, a lot of places selling PC150 Cas2 as PC166 Cas3. Thinking of taking an old PC100 module and removing the chips/replacing them with 6ns parts, or better yet 5.5ns parts, but I don't even know where to find them. Any advice?

<font color=blue>You're posting in a forum with class. It may be third class, but it's still class!</font color=blue>
August 24, 2002 5:38:35 AM

Have you ever done sm? It's possible, not fun though. From experience of designing and making our own boards there is a lot of issues to be dealt with and even with the faster IC's it may not be stable at that speed. But I'd love to see it done! Let me ask a professor what he thinks, Yale Patt will now if it can be done. Maybe were to get them.
a b } Memory
August 24, 2002 5:44:15 AM

In fact, the only work of this type I've done is to reattach several legs of a controller chip to an overdrive processor's controller board.

<font color=blue>You're posting in a forum with class. It may be third class, but it's still class!</font color=blue>
Related resources
August 24, 2002 7:35:09 AM

Depending on how close the pins are they can be done with an iron. If they are ball grid or really close organic solder is best. A clean joint is a must! Before you try it I suggest getting some cheap memory and practice taking off the IC's and putting them back on, then see if it still works. My soldering skills are ok, but when it comes to sm I usually let one of two people in our robotics group do it. Turns out better! It may be a while till I see Yale Patt, but when I do I'll see what he says.
Anonymous
a b } Memory
August 24, 2002 8:25:29 AM

Hey crashman, use PCB's from a good company as layout does matter.

You want to identify the chips on the board. This will allow you to get a datasheet and find out many clues you will need to find faster chips that will work. First step in identifying them is obviously the chip markings. Hopefully you see a logo you recognise, and go straight to the manufacturers site, but more likely it'll take a bit of research. Once you know the part number and manufacturer you can get the datasheet off their website.

The datasheet will help you narrow down what package you are dealing with as they will have the mechanical specs as well or a reference to where they can be found. A set of calipers can get you a few key measurements of IC dimensions. Other than package there are a few other gotchas to watch out for, pinout, voltage requirements, , and obviously speed ratings will all be found in the manufacturer data sheet.

Once you have fully identified what you have, you need to go out and find an electrically compatible but faster part to replace it with. Don't forget to consider all manufacturers as they often have crosses for each other. Once you start pouring through manufacturer data sheets hopefully it will become clear if you are dealing with some popular pinout/package that newer faster parts will be available in. A bit of advice: print out all the datasheets of the original memory chip and any real candidates, and double-triple check everything before you buy.

Also once you find a memory you think'll work, since your not going to be bying all that many, you'll most likely have to go through a distributor. I would suggest someone like Allied, Digikey, FAI etc. A huge warning here is: be sure to get the exact part number for whatever you want to buy before you call, as the people running the phones are probably stupid and won't know anything other than pricing and availability. These people can be useful for doing crossing dirty work as long as you fully check out their suggestions, and take whatever they say with a huge grain of salt.

Sounds interesting. Post back how it goes.
August 24, 2002 8:27:31 AM

After some research it looks like the pinouts are different. You could get 183mhz dram, to put on the pc100, but ddr won't work. Now if you had a ddr board 6.0ns could work, but you can buy this. Ddr2 will be FBGA so it won't fit on a ddr PCB. Besides only a madman would try doing FBGA with out the cool tools needed! I will wait to see what Patt's response is, he might know a trick. If the above is what you want its possible to convince manufactures to give you engineering samples.
August 24, 2002 8:47:28 AM

The pinouts of a certain memory type are pretty much standard. PC100/133, ddr, Rambus, ddr2, etc. So are the voltage requirements, etc. What is important is the logical layout of the IC's 4mbx32, 8mbx16, etc. You only have to be careful that you don't get an IC that wasn't designed for applications other then the pc, but same interface. These would have different voltages, etc. If your really crazy grab eagle, design a pcb for Samsung's 400mhz (yes 800ddr) FGBA IC's, for about a $100 there are companies that will make the boards for you and make friends with a small company in your area and get them to put it on for free. A lot of research hardware isn't done for a price it’s given in engineering samples and cheap labor. Explain the glory and show the skills, ECE's are egotistical and will do it for fun. We're going to build a control board for an autonomous aerial robot and we convinced someone to put BGA's on for free, it can be done (maybe even the PCB's)! As a last resort pay for it.
Anonymous
a b } Memory
August 24, 2002 9:02:30 AM

that is true, it never hurts to ask.
August 24, 2002 9:31:23 AM

knewton,

I didn't mean to make what you said unimportant, but reliance on specs only can be dangerous. I've seen to many projects release the "magic smoke" that shouldn't have by specs! Most hardware companies use specs as a guideline, then they test it for themselves. The guy who designed the IC doesn't write the manual sometimes, that kind of stuff.
a b } Memory
August 24, 2002 9:47:29 AM

Damn dude, you're scaring me!

<font color=blue>You're posting in a forum with class. It may be third class, but it's still class!</font color=blue>
a b } Memory
August 24, 2002 9:51:31 AM

Dude, I wouldn't happen to be thinking about DDR anyway, won't work with my chipset. And DDR chipsets for my processor suck. I have a PIII 1000EB (Coppermine) running at 1307MHz/174FSB, and am trying to come up with some memory that will work at 166MHz Cas2, so I can raise my memory ratio back up to 1:1 and drop my bus speed to 166 if necessary.

<font color=blue>You're posting in a forum with class. It may be third class, but it's still class!</font color=blue>
a b } Memory
August 24, 2002 9:53:31 AM

Er, I know that the stuff (SDR, 6ns) exist because it's used on low end video cards.

<font color=blue>You're posting in a forum with class. It may be third class, but it's still class!</font color=blue>
August 24, 2002 9:59:23 AM

Memory shouldn't be to bad though. As its mostly just traces from the IC. The part to worry about is making good connections for the 54 pins on a Sdram, or 66 for ddr. Maybe its because the only good time to do this stuff is between classes, ie 10pm - 7am with little sleep. Ever seen an IC walk away from you?
Anonymous
a b } Memory
August 24, 2002 10:07:14 AM

Of course, but isn't the point helping him find something to try? That's what I was doing in that huge post! I'm confused now. How else would you propose that crashman got going on this?

As far as I know every succesful board level design starts with someone checking out some datasheets. Sure theres testing, but that comes later. It helps to make intelligent decisions about what to try right off the bat. It saves time and money. Who wants to rework some fine pitch sm parts over and over cause you're to cool to look at data sheets. Succesful designers and part crossers spend a good portion of time doublechecking datasheet details. Face that fact now and get fired less later.
August 24, 2002 10:17:22 AM

Try this:
<http://www.samsungelectronics.com/semiconductors/DRAM/S...;
Its runs at 183 mhz standard 54 TSOP pinout, so there should be no problems with pins. Still waiting on the prop, though it could take several days. CAS is 2 or 3 as stated below. And the data sheets are there to. Before you get anything always ask for engineering samples! They might give you some. National Semiconductor gives us half a dozen or so LMD18201 H-bridges ($20 each) for free, no reasons given, just we want them. I know I keep saying that, but it can save you a lot of money for a little effort!
August 24, 2002 10:29:01 AM

Of course you use data sheets for design. Just don't think because the data sheet said it should work that it will. There can be a lot of problems that must be traced with oscope before the design is done. Maybe I'm being harsh, but I dislike telling someone that it will be this easy and it turns out that its not. I'd rather say what could happen and help them. So that if it works right away, great. Remember Scotty? "You don't actually tell the captain how long it will really take?" Meaning that if you give estimates that have no error buffer in them things will go wrong thanks to Murphy's Law.
Anonymous
a b } Memory
August 24, 2002 10:36:07 AM

Do you have a module picked out that you want to disect?

it shouldn't be too hard. Especially because of the type of soldering that will most likely be necessary, and your soldering chops it's a moderately risky operation. It all depends on how much you stand to lose if you damage the PCB, or it didn't work because of microscopic solder bridges for instance. It might even be illegal I wouldn't know.
August 24, 2002 10:41:47 AM

I agree with that. At first I thought he was thinking of ddr or something packaged in a BGA, these are harder. Sdram can be done with an iron. Organic is better for not having bridges. No heat required, just put it on like a paste.
Anonymous
a b } Memory
August 24, 2002 11:21:26 AM

Look at Crashmans posts he's rightly nervous. I've seen lot's of his posts, and realize he's the do it yourselfer type. A meticulous skilled person might even do this in one shot, assuming thay made good choices in the beginning and the proper components are actually available.

BUT your absolutely right,assuming there are reasonable parts available, the soldering could be a real bear, especially for someone inexperienced with the small stuff. I probably wouldn't touch it without a nice scope and some good soldering equipment, and definitely a nice strong magnifying glass. There is also a very real possibility that the whole project isn't even possible using presently purchaseable components under the restrictions we've been given.

IC's made by reputable memory manufacturers of SDR type memories would be very refined and well documented. I would say that it is extremely likely that if these parts exist it wouldn't be the undocumented characteristics in data sheets that gets you, but more likely poor choice in PCB, or soldering gotchas.

I'm not trying to say that there aren't undocumented device characteristics, which are not in manufacturer data sheets. Or am I saying it's good to ship a new product when it looks good on paper, without an appropriate design/build/testing phase. No, I'm just saying that in my experience in many cases I have seen designs work well on a first design pass.
Anonymous
a b } Memory
August 24, 2002 11:35:59 AM

oops sorry didn't see this until after I made my post.

We agree!
August 24, 2002 11:58:05 AM

Crashman,
Sorry if I scared you. From what Knewton says I think you can do it. Sdram should have no problems. Especially if you get a good pcb like Knewton says.

Knewton,
I agree with you 100%. I didn't know he wanted to do something this simple. If you look I changed my mind when I found out that he just wanted to do sdram. At first I thought he wanted to put a high end video memory like Samsung's ddr that I mentioned. Note that these modules aren't available till Nov as engineering samples, so there are bound to be a lot of "mishaps" till the design is out for awhile. This standard may not be approved last I heard, so things often blow up with on the edge hardware. If I offended you in any way I'm sorry. As for IC spec, after a team in Canada (part of a contest be AUVSI) who's sensor quit working at 4C but was certified for much less. And a tour by Sclumberger who's engineers assured us of how much testing they do on IC's, because of odd specs. These are things I've seen in the last month with school out and not much being done. But you are correct that most IC's are close enough. Particularly those that Crashman is considering.
a b } Memory
August 24, 2002 6:26:29 PM

Not an IC, but I did have one of those paper thin foil data cables walk away from me. BTW, those things are a bugger to strip.

<font color=blue>You're posting in a forum with class. It may be third class, but it's still class!</font color=blue>
a b } Memory
August 24, 2002 6:28:23 PM

The address didn't work for me


<font color=blue>You're posting in a forum with class. It may be third class, but it's still class!</font color=blue>
a b } Memory
August 24, 2002 6:35:35 PM

I have a nice Micron 8-chip single sided PC100 module that's identicle to my last PC133 part, but it's 64MB, would the computer detect it at 128MB if I used 8 128Mbit parts? Hoping to locate a quality 16-chip PCB so I could get 256MB using 16 128Mbit parts (I've meantioned the 128Mbit/chip limitation in the FAQ as 16MB/chip)

<font color=blue>You're posting in a forum with class. It may be third class, but it's still class!</font color=blue>
a b } Memory
August 24, 2002 8:00:05 PM

Nope, still don't work, can't even go to DRAM from their homepage.

<font color=blue>You're posting in a forum with class. It may be third class, but it's still class!</font color=blue>
August 24, 2002 8:09:25 PM

Odd? I just cut and pasted it in to a new window and it came right up? What error do you get?
a b } Memory
August 24, 2002 8:36:15 PM

A Cannot Find Page error, in Japanese.

<font color=blue>You're posting in a forum with class. It may be third class, but it's still class!</font color=blue>
a b } Memory
August 25, 2002 2:26:08 AM

Yes, but can't get through the link Products>semiconductors>DRAM.

<font color=blue>You're posting in a forum with class. It may be third class, but it's still class!</font color=blue>
Anonymous
a b } Memory
August 25, 2002 3:43:25 PM

That's a good question. Alot of manufacturers will pull tricks to make things like that possible in order to save money. To answer that you need to find out what exactly the PCB has in it. Ie are the traces for the higher order address lines present on the PCB? There may also be some other traces that are needed, mayba a DRAM expert would know a little more.

It may very well work, but probably only if the PCB was designed for that purpose. If you had the pinouts for the larger memory IC's and knew the pinouts standards for SDRAM modules you might be able to ohm it out to make that determination.

Another thing to look for: if it is a single sided module, then often there are empty part patterns on the other side. If this is the case, obviously you would be able to use those too. I would think the safest bet would be to fill up the front and back with chips which are crosses to the manufacturers oiriginal choice, but with the speed improvements, however I certainly understand why you may want to end up with more than 64-128MB per module when you are done.

A couple things to consider. Oh BTW, I think Micron PCB is a good choice. I have always liked Micron memory.
August 26, 2002 7:02:55 PM

It works...

Crash, try the link above.

<b>"If I melt dry ice in a bathtub, can I take a bath without getting wet?" - Steven Wright</b>
a b } Memory
August 27, 2002 12:24:40 AM

d00d, the link worked!
OK, I'm looking at my old memory and discovering there are a lot more than just RAM chips to this stuff, and I really don't want to start trying to track down the small parts, so if you know anyone with burned high quality double sided SDRAM could you tell me?

<font color=blue>You're posting in a forum with class. It may be third class, but it's still class!</font color=blue>
August 27, 2002 6:50:15 AM

not too long ago i saw some PC166 at a swapmeet.
only 128mb sticks though... quality unverified.

have you considered getting a DDR board for you P3T? then at least you could run nice PC3200.

<b>Due to Customer Complaints, this sig has been witdrawn from public use. Thankyou. :lol:  </b>
a b } Memory
August 27, 2002 12:43:13 PM

Can't find a DDR chipset that offers the same performance level (being that DDR offers no improvements in performance by itself for the SDR bus P3). BTW, I don't have a P3T, I have a Coppermine, it's the 1.75v version, check the stepping and so forth at <A HREF="http://www.crashman.dns2go.com" target="_new">My Site</A>.

Anyway, the only advantage I can see in running DDR is that I could use PC3200, BUT at PC1600 performance levels because of the SDR CPU bus. It would at least solve my memory speed issues. Do any of these boards/chipsets offer a 1/5 PCI divider as well?

<font color=blue>You're posting in a forum with class. It may be third class, but it's still class!</font color=blue>
August 27, 2002 2:25:10 PM

Lame might be a good word for it but it is 10x better than Fatburger's and 1000x better than my cheesy Geocities page.

Maybe when I set up my real web page I might "really" call your's lame. Until then I will just have to admire yours.

BTW: Cute way of getting all of that info one one screen. No typing needed..."take a screenshot and post that on my website."

Sometimes simplicity is bliss.

<b>"If I melt dry ice in a bathtub, can I take a bath without getting wet?" - Steven Wright</b>
a b } Memory
August 27, 2002 4:44:42 PM

Yes, but all my PCI devices running at 42MHz, and anything more they give up! In fact, I had a hard drive that could't even cope with the overclocked IDE controller, so I sold it. A 1/5 PCI divider would be nice!

Heck, I wouldn't even need PC3200! I could get by with PC2100 running at 2/3 bus speed, considering the CPU bus is the bottleneck when using DDR. So a board that supports that would also be nice.

<font color=blue>You're posting in a forum with class. It may be third class, but it's still class!</font color=blue>
!