Feeler: Mac Portable max-out RAM upgrade

[techknight]

My 5120 doesnt have any expansion cards, and I had noticed decent cards are hard to come by, if not impossible. 

Soo... 

I have been thinking, I was going to design and build a RAM expansion card for the portable, with the potential to max it out at 9MB. 

Question is, would anyone be interested? SRAM isnt exactly cheap, and with the only one remaining in production these days for 1MB, weighs in at around $6.88 a chip. So i would technically need 8 chips. Plus the on-board RAM to max it out. 

So I am curious if anyone would be interested in buying any if I do get one made. Just trying to justify the cause. Spread out the costs among a bunch of them instead of just 1. it would ease the pain on me, but make availability to you. 

Any thoughts? 

 

[Gorgonops]

Isn't the Portable's RAM actually psuedo-static, not static? Just spitballing, but if you could make a board that could take a BGA here's a 4Mx16 70ns Psuedo-RAM that's $5 a head:

http://www.mouser.com/ProductDetail/ISSI/IS66WVE4M16BLL-70BLI/?qs=sGAEpiMZZMt9mBA6nIyysDQy6Jcm%252bVH0YaEEWLnPRRM%3d

Granted the 3.3v supply voltage might be problem? (Portable's 5V logic, right?)

The data sheet *does* say contact the manufacturer about availability as TSOP-I instead of BGA.

 

[trag]

I have been thinking, I was going to design and build a RAM expansion card for the portable, with the potential to max it out at 9MB. 

Question is, would anyone be interested? SRAM isnt exactly cheap, and with the only one remaining in production these days for 1MB, weighs in at around $6.88 a chip. So i would technically need 8 chips. Plus the on-board RAM to max it out.

I don't have a portable, so I'm not interested in a card. However, I might be able to help your SRAM issue. I have 250 TC55VBM416AFTN on hand. These are 1M X 16 SRAM, although they can also operate as 2M X 8. They take a 2.3 - 3.6V supply, but the datasheet claims "Direct TTL compatibility for all inputs and outputs". If that will interface with the Portable, then four of these chips should do the trick; just add a small Voltage Regulator to drop 5V to 3.mumble. I could send you eight or twelve of them to try out, although that would mean that your board design was laid out for their package. And we can arrange a price if you want to buy more later, although the market is probably tiny.

The only problem I see is if Apple does something like dropping the supply voltage to the RAM to preserve contents during power down. The VR probably wouldn't appreciate a drop in input from 5V to 1.5V.

I also have 310 Hitachi HM658512ALFP 512K X 8 Pseudo-Static RAM chips on hand. IIRC, one model of Portable wanted SRAM and the other wanted PSRAM. I don't know if you could use the SRAM in the model that was meant for PSRAM.

Anyway, with Seed Studio available for boards, you ought to be able to run off a few proto-types very affordably, once you've done the layout. PM me if you want to use my chips.

Does the memory expansion board need a special connector? That's probably your only other obstacle.

I've been hanging on to these chips for years with the idea of doing a portable memory expansion card. I am never going to get to it. It would be great if you did it. I'll be happy if I can just get that PEx ROM done...

 

[max1zzz]

i'd deffinatley buy one since i only have a 1mb card for my portable and as the onboard ram is bad there is a good chance the upgrade card might be bad aswell

Even at $6.88 a chip it isn't that bad, it took me long enough to find a portable i could afford so i'd bet i will never come across a 8mb ram card for it (and even if i did it would probably be more expensive)

 

[techknight]

The portable keeps the card hot at all times however there is a standby signal. But I wouldnt use it.

3.3v logic isnt a problem. The question really wasnt about the design. More about if others would be interested in the ram card.

Non backlit uses sram. Backlit uses psram. Apple says you cant interchange the two cards. But I dont see any reason why not. Psram acts as sram on the bus....

Trag ill take you up on that RAM offer.

 

[techknight]

I would rather have the Toshiba SRAMs trag. Why? because they are 16 bit and plus they have both an upper and lower byte access mode. This is a requirement for the 68000 as there is no Address 0. it uses Upper/Lower data strobes to get the upper and lower part of a word, aka byte mode. In that fashon, they would never need an A0, so they excluded it. 

Plus that keeps me down to 1 chip per 2 megabytes. (as your chip is a 16MBit or 2megabyte IC) Instead of 2 even/odd chips. So in theory I would only need 4 chips for an 8 Meg card. 

Less CPLD lines I need the better. I will end up using an XC9572XL CPLD as the memory address chip select decoder. 

The connector is just an ordinary right angle female IDC.

The 1 thing that concerns me, is there is no DTACK line routed to the RAM expansion connector. Without DTACK, the processor cannot know that there is valid data on the bus. RAM of different timing would have wait-states inserted in before DTACK gets raised to let the CPU know data is awaiting. This could present problems with multiple different speed chips on the bus. if the new chips are faster than the old ones, I may just use the RAM expansion and take over decoding for primary on-board RAM. 

In that case, I would disable the on-board RAM and the expansion card would take over all of it. Since technically, all the address/data lines from the entire system bus are routed to the RAM expansion. But again, no DTACK. So not sure how to handle that. 

 

[uniserver]

oh man this is pretty exciting. 

 

[trag]

I would rather have the Toshiba SRAMs trag. Why?

Less CPLD lines I need the better. I will end up using an XC9572XL CPLD as the memory address chip select decoder.

PM me your physical address and I'll post some chips out to you. It sounds like you hunted up the datasheet, so you already know these are 48 pin TSOP.

I bet you could do the address decode with a handful of quad NAND gates although the CPLD might be more compact. Is it any more complicated than converting a few upper address lines from binary encoded to a set of one-hot signals which drive the Output Enables?

Does the logic board RAM have provision to drive DTACK? Logically, there isn't really any way for asynchronous SRAM to drive a DTACK signal, that I know of. There's no handshake output signal. In fact, the only output signals are the bidirectional data pins. The data just appears when it does and the designer must use fast enough RAM to make sure that it arrives soon enough. It might be worthwhile to give TGTTMFH and DC&DFTM another look. One of both of those had some detailed info about Portable expansion. They might comment on the RAM timing question.

What speed SRAM does the Portable use? Will 55ns be fast enough? Oh, it looks like access through OE_ is 30 ns.

 

[techknight]

The portable and backlit portable both use 100ns ram.

This is my first time making a ram expansion so if I have the logic wrong its a simple matter of modding the VHDL and flashing. If its 74 series logic. Thats physical cutting and rework.

Also the TC55VBM416AFTN55 from Toshiba is nice, but I fear once we run out of those by making enough cards, I will have to redesign it as those chips are long since out of production. 

However, the demand might be small enough that everyone will be satisfied by the time your stock runs out. 

So it could be a win-win. 

 

[techknight]

The Toshiba has a weird little setup which is A-1, and its shared with I/O16. It appears I/O16 becomes A-1 in byte mode. But A-1 is not A1. And it doesnt really explain what A-1 actually is.

Weird setup. 

 

[trag]

The portable and backlit portable both use 100ns ram.

This is my first time making a ram expansion so if I have the logic wrong its a simple matter of modding the VHDL and flashing. If its 74 series logic. Thats physical cutting and rework.

That makes good sense.

Also the TC55VBM416AFTN55 from Toshiba is nice, but I fear once we run out of those by making enough cards, I will have to redesign it as those chips are long since out of production. 

However, the demand might be small enough that everyone will be satisfied by the time your stock runs out. 

So it could be a win-win.

That's the dream. Run out of chips. Clean out a tiny portion of my attic. They are cool chips. A tiny piece of me squeals, "No, hang on to them. You might need them some day." But that's just the ugly hoarder in me...

250 chips divided by four per board is 62 upgrades. I guess if you put them out on Ebay you might sell that many, eventually. There could be a lot more Portables out there than I think.

 

[trag]

The Toshiba has a weird little setup which is A-1, and its shared with I/O16. It appears I/O16 becomes A-1 in byte mode. But A-1 is not A1. And it doesnt really explain what A-1 actually is.

Weird setup.

It looks like A-1 is just another address signal. When the SRAM is in byte mode, it has twice as many addresses, so it needs one more address signal.

But it is a weird nomenclature. Usually they would label it A20 in such a case. Since it is SRAM, and there is no multiplexing of addresses, nor any burst mode available, the ordering of the address lines does not matter, as long as it is consistent. So, you can put A-1 (different from A1) anywhere you like, I guess, if you're going to use the byte-wise mode.

 

[wilykat]

That makes good sense.

That's the dream. Run out of chips. Clean out a tiny portion of my attic. They are cool chips. A tiny piece of me squeals, "No, hang on to them. You might need them some day." But that's just the ugly hoarder in me...

250 chips divided by four per board is 62 upgrades. I guess if you put them out on Ebay you might sell that many, eventually. There could be a lot more Portables out there than I think.

How many active working Portables though?  Take out a few that is already maxed out, take out some who are not interested in upgrading anyway, and take out some people who are puritan and will prefer genuine Apple memory only.   100 upgrade kit would be a good estimate.

 
I'd jump but I never had a Portable and it's rare enough you probably don't need another competitor on eBay and other used computer marketplaces.  I'll stick to collecting basic Centris and Duo stuff.

 

[techknight]

Thats just it, I dont know. Which is why I am debating whether its worth the effort of designing a card or not. 

I know I want 2 of them for myself. But the rest, is up to demand. 

I already ordered an XC9572 breakout development board, along wtih a Xilinx jtag. 

 

[techknight]

I do want to make the card hold at least 8 Megs. But you dont have to populate all 8 megs. if you want just 1 meg, thats fine. put 1 chip on there, etc etc... 

 

[Paralel]

Can't imagine someone wanting a card that isn't maxed out. That would just be kinda nuts

 

[Macdrone]

I know my backlit ram did not work in my non backlit then was reminded that they were different so I did not interchange at all after that. Also the max ram on each is different.

So beyond that this is still an interesting thread.

 

[techknight]

The question is.. why cant you interchange? The bus is the same...

The max ram also should not matter. The core chips are the same.

Gttmfh only discussed one slot. And they dont refer to any particular model. Maybe I need to run continuity checks on the slot between the two models.

The only thing I see is the backlits PSRAM has a refresh input. But it can be tied high at which point the chip refreshes itself.

 

[techknight]

ok, I answered my own questions. I found this note: 

http://cdn.preterhuman.net/texts/computing/apple_hardware_devnotes/Portable%20(BackLit).pdf

this basically explains the difference. the portable backlit model switches around 2 lines. They removed address strobe and replaced it with chip select. And the old chip select locaiton was replaced with refresh. 

Neither of which will affect my new card, making it compatible with both models. HOWEVER.... the backlit model only generates the DTACK signal, and asserts the chip select signal for the first 5 megabytes of address space. After that, no more. So that means the ceiling on the backlit is at 5 MB. 

but, I can get around that by simply adding a post that says DTACK on the card, and you could in theory clip a wire between that card, and into the PDS. using a pogo-pin style wire. 

that would expand it to the full 8MB.