Jump to content

Fabbing 30 pin SIMMs


Recommended Posts

I was wondering about the feasibility of making new 30-pin SIMMs and wanted to see what thoughts members here might have. Finding 16MB modules can be bit tough if you have space constraints like within the SE/30 chassis. Could making our own be feasible without being too expensive?

 

trag pointed out the simplicity of the circuit here viewtopic.php?f=7&t=18527 So perhaps it would be feasible to make 30-pin SIMMs by fabbing them at SeedStudio and finding the appropriate chips if they are still being manufactured. I tried finding a source for the IC used in my 3-chip SIMMs (KM44c16100BK-6) but could not readily find any in stock. I would think a 2-chip design would be optimal to reduce complexity/cost and power consumption assuming it is compatible with most machines.

Link to post
Share on other sites

Good point, but I bought some from them a few months ago through their website and had to send them back since they were the taller type that cannot fit in all the banks of an SE/30 due to clearance issues. At the time they said they had no shorter SIMMs, maybe they have some "new" inventory now, but a lot of times the photos posted are not representative. OWC could not confirm the height of their SIMMs, but someone that ordered from them recently said they received the taller type.

Link to post
Share on other sites

I don't know what they stock to be honest, or how "NEW" those parts are. You need SIMMs for compacts, not sure what the other buyers would need them for and if the premium of you making SIMMs would be worth it to them (there seems to be no clearance issues with old PCs or Macs like the 950 for example). I stuffed 256MB into a 950 just for the hell of it, hard to justify the expense when you don't realy need that much RAM. I did have 128MB in a SE/30 but 32MB seems to be good enough and I removed it.

 

The only time making small batches of RAM is warrented is when it just cannot be found anymore and the market is big enough to make it worth the effort. Very few sets of 16MB IIfx SIMMs seem to have been sold (and the maker quit when a set of 4 would only bring in $60 if they did sell) and those are unobtanium.

 

I know somebody made some GVP 64 pin SIMMs for Amiga users with GVP hardware, but those people pay crazy prices for that kind of stuff and it is not available anywhere else.

Link to post
Share on other sites

8's are rare and some macs have issues with them, same with 2's. macs in general since they take non parity and parity 30 pin SIMMs are not that picky about RAM (except the II, IIx models that need composite SIMM).

 

I don't understand what you mean by marketability/profitability of 16MB SIMMs, if money is no object and you don't care about selling them then ANYTHING is possible to make, they were commidity RAM and the specs are out there. All I was trying to say is you can find them cheaper then it would cost to make them so there isn't a market for them.

Link to post
Share on other sites

2 & 8 MB SIMMs do not work on the SE/30 -- just 256K, 1, 4 & 16 M :(

 

The IIsi (and I presume the IIci too) seems to take almost anything 30 pin, parity or not, composite or not, including 2 and 8 MB sizes.

 

But the clearance issues aren't unique to the SE/30: the Quadra 700, for instance, has a tighter spacing between sockets, thus some SIMMs with chips on both sides won't fit.

Link to post
Share on other sites
I don't understand what you mean by marketability/profitability of 16MB SIMMs, if money is no object and you don't care about selling them then ANYTHING is possible to make, they were commidity RAM and the specs are out there. All I was trying to say is you can find them cheaper then it would cost to make them so there isn't a market for them.

I know it doesn't totally make sense, part of this is driven by wanting to make something "new". When I looked for SIMMs that met my specs, the cost was getting in the realm of maybe just looking into making it myself and learning something new along the way. My thinking is if the SIMM takes 3 main components and each component is $3-4 then it is probably feasible cost-wise. I'm just not sure what parts would be best to use, if they are still manufactured, or more contemporary/available equivalents could be used.

Link to post
Share on other sites

The memory chips are available, if you are willing to do some extra work.

 

Get 128 MB 72 pin SIMMs or 64 MB 72 pin SIMMs. Desolder the eight (or 16) 16M X 4 memory chips. Put two each on your home designed 30 pin SIMMs.

 

One $16 128MB 72 pin SIMM will supply enough memory chips for an entire SE/30, provided you don't destroy any in the desoldering/soldering process.

http://www.ebay.com/itm/KMM53632000BK-6-128MB-32x36-60NS-FPM-SIMM-/110677960034?pt=US_Memory_RAM_&hash=item19c4eba562

 

A two-layer SIMM should work fine. I used a two-layer design for my IIfx memory, and Doug is using a two-layer design for his ROMs.

 

As others wrote, it may not be worth the time/money, but it would certainly be fun. I've often considered running off a batch myself.

Link to post
Share on other sites
well how much does it cost to have a company fab up some blank simms and send them back to you?

i would assume were talking about made in china…

 

http://www.seeedstudio.com/depot/fusion-pcb-service-p-835.html?cPath=185

 

For the Seed Studio 2-layer boards, it's $2.50 a board. I think that's their price for up to 5cm. X 10 cm. (2" X 4"). It's $10 + $15. It puzzled me at first, because I kept reading it as being $15 for 5 cm X 10 cm.

 

So, a set of eight of them (128 MB of RAM) would cost $25 (PCBs) + $16 (chips) + ~$6 shipping + supplies (solder and the like). With two circuit boards left over.

Link to post
Share on other sites

Thanks trag, that's a clever idea.... I was hoping you would chime-in since you have a bit of experience in this area. :)

 

The 3-chip SIMMs I have have 2 DRAM chips and the other I thought was maybe parity, or is it some kind of controller? Without looking at P/Ns It looks like the 128MB 72pin SIMM also has them.

Link to post
Share on other sites

Yes, the extra chip is a X 1 chip for parity. If you scavenge from a 128MB SIMM you get a bunch of 16M X 1 chips as a by-product. That's why I built both two-chip and eight-chip IIfx SIMMs. The X4s are a lot easier to come by, but they require me adding a couple of ICs to the SIMM to handle the IIfx's special write buffering scheme. That's not an issue on any other Mac, so one can just use the X 4 chips by themselves.

Link to post
Share on other sites

tt, in another thread; you or another member were wishing for a way to kick-start development for boards that require 4 layers in order to make all the connections.

 

Suggestion:

 

Almost every single sided PCB that I've seen in PSU and KBD matrix applications, wires on the component side jump across across traces on the copper side, effectively implementing a two sided layout on a single sided board.

 

Doing a two sided "Seed board," with vias designed in for patch wiring these same impossible connections would effectively convert a two sided proto-board to four sided function, sans the additional ground fill planes, of course.

Link to post
Share on other sites

I was thinking the PDS connector necessitated at least 3 layers since there are 3 rows of pins close together, but the Asante card sneaks two lines between each pin, so if that clearance, tolerance and minimum feature size scheme is allowed by the fab-house, then I could see it done with a 2-layer board. There may be electrical reasons for needing more than two layers though, but that determination is out of my design space.

 

I don't see any reason why a 30-pin SIMM cannot be done with two layers without digging into details, but I think dougg3's ROM design shows it can be done.

Link to post
Share on other sites

That sounds like a fun project, I'm not sure why it never occurred to me to try it. I've used Seeed Studio many times for PCBs and have been very impressed. The quality is quite good, delivery can take close to a month but the price is a fraction that of the next lowest cost fab house I've found. The base price is $9.90, with larger sized boards adding on top of the base price. The dimensions they list are the maximum dimensions, smaller boards are fine. I recently had a batch of 10cmx10cm PCBs made and for 10 boards it worked out to only $2.50 per board. Shipping is free if you spend $50 so I get 2-3 designs ready to go and order them all at once.

 

Somebody mentioned making some IIfx RAM SIMMs, do you have the PCB layout available for those? It's something I'd be interested in doing if I ever get my (badly corroded) IIfx resurrected.

Link to post
Share on other sites

I'll throw a knuckleball into the works at this point, this one has been on back burner for quite some time, but I just remembered a recently downloaded pic of a similar adapter.

 

Instead of moving just the 72 pin DRAM ICs back a form factor generation. Move the entire 72 pin SIMM back a generation by converting a 72 pin socket (or more?) to the older 30 pin SIMM form factor.

 

 

3_Slot_PCI_Riser.JPG.1733a1da327f9e5dd3e72624391e4645.JPG

 

 

I've always wanted to do a "reversed-funnel" SIMMspender™ hack of ye 'ole SIMMsaver Hack. Three bare minimum profile RAS-CAS line address transfer SIMMs would be wired up to one larger SIMM that lays a 72 PIN SIMM flip-flopped right back across the top of them.

 

SIMMspender00.thumb.jpg.ab83f35fccf3154144e51c0a9b34ccfe.jpg

 

Maybe it's more of a Knucklehead throwing a Screwball into the works here. All the signals would need to be reversed to match the banks of the "backwards" 72 pin SIMM socket.

 

Installation of the adapter SIMM and the 72 pin SIMM will probably need to be done simultaneously to keep the adaptation low profile. hold the two ready to click into their slots then and push them down together like two legs of a parallelogram for near simultaneous clicks.

 

Removal should be fun too! }:)

 

The original concept involved a single adapter PCB and patch wiring the additional connections. This way no MoBo modding would be required. It'd be a lot more straightforward for most to do.

 

These are the less attractive configs available in terms of height.

 

SIMMspender01.thumb.jpg.948593ea6bca831509769899737e9f05.jpg

 

The top version should be easier to install, but the overhang may not fit in the SE/30.

 

The middle version uses a more difficult to source flat SIMM Slot Connector and the screwball wiring setup. Probably low enough for most Macs.

 

The bottom version uses it as well, but could be silly looking and use the angled connectors for a straight thru, but very high config.

 

The lower two, with jumper settings to re-arrange the RAS/CAS line feed, might be staggered to convert adjacent banks.

 

Three of the four might work in tandem banks.

 

My first choice, the most difficult to install of course, along with the top, of the alternate choices, might work in a 4x4 bank config if right and left handed versions were to be made available.

 

. . . or it's a silly idea all together due to SIMM type incompatibility. :-/

Link to post
Share on other sites
recently downloaded pic of a similar adapter

 

Just for edification, the dinguses hanging off that PCI slot adapter are there to take a couple of the unique select/arbitration lines from two of the motherboard's PCI slots and connect them to the appropriate pins on two of the adapter's slots. PCI isn't a *completely* parallel bus like ISA/S-100/whatever; each slot has a few dedicated lines running from it to the PCI controller to handle selection and arbitration. (In other words, a PCI controller has to know *which* slot a card is in, unlike something like ISA where in theory you could extend the bus infinitely via a passive slot extender, subject to electrical loading and propagation time limitations, and the system will have no clue where on the bus a card is located.) If one or both of them aren't plugged in the associated slots on the riser won't work.

 

(In theory you *can* add more slots to a PCI slot by using an "active" riser that itself has a PCI-PCI bridge on it. But from a software standpoint that's not completely transparent.)

 

A 72 pin to 30 pin adapter would need at least 8 connections running from each of the "satellite" SIMM slots to the "main" slot the 72 pin socket's plugged into. Any cables would have to be kept *very* short and shielded, so the thing would be utterly monstrous to plug in. There may also be issues with faster machines experiencing errors because of the almost unavoidable timing "skew" you're going to end up with because of the different total wire lengths for each of the four 8 bit banks. It might work, but it would be an even grosser hack than SIMMsavers were in the first place.

 

(I never got the point of SIMMsavers. Every time I remember upgrading motherboards to one using a newer memory form factor the price of memory had dropped enough it made more financial sense to just buy more new memory than I had in the first place than try to save the old stuff. I can see some edge cases where if you'd *really* broken the bank maxing the memory on the last machine they might save you a few bucks, temporarily, but... factor in the headaches and it's still a stupid idea.)

Link to post
Share on other sites
(I never got the point of SIMMsavers. Every time I remember upgrading motherboards to one using a newer memory form factor the price of memory had dropped enough it made more financial sense to just buy more new memory than I had in the first place than try to save the old stuff. I can see some edge cases where if you'd *really* broken the bank maxing the memory on the last machine they might save you a few bucks, temporarily, but... factor in the headaches and it's still a stupid idea.)

 

The one place I've found such a thing useful is the SimmXpander in a Q605. This converts two 72 pin SIMMs to one 72 pin socket.

 

I've been able to get 260 MB of RAM in the Q605 using one expander and two 128 MB SIMMs. However, thinking about the addressing and architecture of the SIMMs, I can only think of one way it can possibly be working. The Q605 must have four independently operable RAS lines. I thought they were usually paired.

Link to post
Share on other sites

I've owned one SIMMsaver since back in the day for tinkering purposes, but I've actually never used it, IIRC.

 

On a less silly note: the DATA/Addressing lines aren't made all THAT much longer to make timing an issue for the SIMMspender hack.

 

In the straight thru iterations, it should be duck soup to keep all the traces/patches equal in length.

 

Since many of the signals are exposed on the top and bottom layers on the MoBo and on the SIMMs, why is shielding for the CAS/RAS lines above board so important? It's not an insurmountable obstacle to use tiny co-ax, by any means, but it's not like crosstalk or interference should be an issue with chunk combos being strobed sequentially.

 

Then again . . . I could easily be misunderstanding that last bit . . . that's usually the word on my assumptions. :-/

 

p.s. thanks for the clarification on the purpose of the really dumb boards in the dumb PCI Adapter pic, G. I was wondering about their function, addressing made a lot more sense to me, but the color coding of the wires said AC POWER COLOR CODING which is another issue in DC for PCI slot adapters. Looked like a stupid way of doing power, but it'd be waaayyyy better filtered coming from the other slots than from a Molex Y Cable.

Link to post
Share on other sites
On a less silly note: the DATA/Addressing lines aren't made all THAT much longer to make timing an issue for the SIMMspender hack.

 

I don't know, you might be surprised. Once you get into the low-mid double-digit Mhz speed ranges is where you start seeing motherboard designers doing things like waving traces around to make sure that all the wires leading to a memory device are the same length. My guess is you'd *probably* get away with an inch or two of slop, but it's not kosher.

 

Mechanically the thing would be something of a nightmare in any case. You can't really make assumptions about how the SIMM slots are laid out on the host motherboard so the whole thing would have to be flexible. (30 pin SIMM sockets come in both vertical and slanted varieties, the socket-to-socket spacings vary, etc.)

 

Since many of the signals are exposed on the top and bottom layers on the MoBo and on the SIMMs, why is shielding for the CAS/RAS lines above board so important?

 

Someone might correct me, but I don't think you'd need the CAS/RAS lines from more than one socket, or at least the full set of them. You will need the 8 data lines, and you *might* need the strobes. However... taking a brief look at the pinouts of the two types of SIMMs and an operational description of how the two work, something tells me that there are gotchyas to this that might require putting some active components on the adapter for refresh to work correctly, at least if the 72 pin SIMMs used in such an adapter were made using components denser than those used on a 30 pin SIMM 1/4 of the size. IE, if a machine like a Quadra 950 supports 16 MB chips with only 2 [or 3, ignore the third] chips that's 8 chips for 64MB. If you were to install a 64MB SIMM using *less* than 8 chips in your SIMM-adapting nightmare adapter the logical layout of the chips might require a different refresh cycle than the memory controller in the host machine can supply. In particular I suspect this problem would crop up with 72 pin SIMMs that use 8 or 16 bit wide DRAMs.

 

Dual-rank 72 pin SIMMs would also be difficult to use without help. So an adapter like this would have significant operational gotchyas.

 

(There's an old thread on this board where the question is brought up why some machines will only work with the 8/9 pin versions of a given size SIMM, not the 2/3 chip version. The answer is that the wider DRAMs sometimes require more refresh cycles than the memory controller on a given machine can supply. Coincidentally this was brought to my attention recently when researching ways to upgrade an old TRS-80 Color Computer 3 from 128k to 512k. The obvious way, replacing the four 4464 DRAM chips with four 44256s, doesn't work because the denser chips require a 512 cycle refresh while the counter built into the GIME DRAM controller only counts to 256. So in order to get 512k you have to add a satellite board made up of 41256 chips that accept the 256 cycle refresh.)

Link to post
Share on other sites

Yup! Forgot about the data lines for the wider path, my bad! :beige:

 

It's lookin' like four small SIMM Cards connected to one loose, mother goose, card by flex-cables would be the only do-able way for this notion to work with timing integrity preserved in the majority of 30 pin macs.

 

Next! :p

 

p.s. hrmmm . . . where does short run flex cable design and prototyping stand?

Link to post
Share on other sites

I like the idea of putting newer chips on a 30 pin PCB a lot better than some contraption that plugs into the old SIMM sockets and provides "modern" 72 pin sockets. For one thing the PCB is gonna be a lot larger (= more expensive) and a lot more complex. It's pretty trivial to pop all the chips off a SIMM with a heat gun, hotplate or hot air reflow station and then reflow them onto new boards. I use a temp controlled hotplate I built but I've seen it done with various low tech solutions including a thrift store electric skillet. Anyway the end result has a tidy stock appearance that will fit easily in any machine, not a messy hack with a multitude of wires snaking around.

Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
×
×
  • Create New...