Internal 512k expansion?

[jmetal88]

Ha! That's great. I actually came up with almost the exact same design totally independently! Well, at least for the 512k portion of the upgrade. I'm trying to eliminate a fair amount of components for the revision and the extra 96k.

I wanted to do the 96k like it's shown in the schematic there, but I've been having a hard time sourcing 128k SRAM chips, at least in a DIP. If I went surface mount, I could do it, it's just surface mount is harder to solder (although I have done it successfully on components where the pins weren't too close together).

EDIT: Oh, nevermind. It was 64k that I couldn't find. I found a 128k from the same company that makes the 512k chip I went for.

[Brutman]

Actually, the idea to use a single SRAM came from that design for the Racore by the person Bob mentioned ... I just passed it on.

Bob - the old forum isn't lost, but it's not in a readable state. I basically have to reverse engineer the database schema that they used, and then write code to repopulate a new database from the existing tables. It's doable - it just takes a lot of time to do it right.

[jmetal88]

Actually, the idea to use a single SRAM came from that design for the Racore by the person Bob mentioned ... I just passed it on.

Yeah, I gathered as much.

I just went back through my CUPL code for the PLD and revised the truth table quite a bit, as well as added the 512k and 96k stuff in the same file. The new code eliminates three traces from my original +96k PCB design in addition to making an upgrade to 736k on one board a lot easier.

It's too bad I actually have homework this weekend, as I'll have to stop this at some point and work on that. As for the board, it might actually be worth using a surface mount SRAM to keep the design simpler.

[jmetal88]

Alright, I just finished another board design. This one uses two SOIC SRAM chips, one 512k and one 128k, and should upgrade a PCjr to 736k. It requires an Atmel ATF16V8B properly programmed with a small address decoding program, which I've written, and which worked as expected under a CUPL simulator.

I'll be able to order a nice looking prototype board for about $38 (shipped) from BatchPCB, albeit without any components attached. The chips and DIP carrier will add a decent amount to the cost, and I'll probably need some new soldering equipment to take care of the surface mount components.

So, I'm on my way, but there's still quite a ways to go.

[jmetal88]

*Whew!* Looks like this thing is gonna cost me over $80 to put together!

I wish I had a cheaper way to put together the circuit board, but I just don't see it happening. If I didn't have to drill the board or do 2 layers I could probably knock a board together using my laser printer, but unfortunately I do have to drill the board, and buying PCB bits and a drill press wouldn't be cheap.

If someone wants my files (got 'em in DipTrace and Gerber formats) to make their own board, I can send them, but I don't see myself making my own prototype for quite a while on this.

EDIT: I managed to save a little cost by switching to all surface mount parts, with the obvious exception of the 8088. It's still kind of costly though. For myself, I might look into making a simpler board for 736k that I could wire into the same sockets I already soldered onto my motherboard.

[jmetal88]

I also just designed a DIP-only version of the board. It costs more component-wise, but I'd save on soldering equipment.

I'd still really like to get this done, I just don't know when it's going to happen.

[southernbob]

If your still having problems finding 128k RAM Futurlec has the 128K & 512K RAMs in DIP as per the drawing I posted.

http://www.futurlec.com/ICRAM.shtml

Mike, what I was referring to on my previous message was the FTP site that Brian had started.
That is just files so may be easier to recover. However, it may not be worth trying to
recover/convert either group as there is not enough action on the forum to warrant it.

Bob

[jmetal88]

Yeah, I did end up finding some. Thanks for listing an alternative store for me, though!

I just had a decent idea though - I might try to order one of those chips and hook it up to my PCjr by piggybacking it onto the 512k chip. That way it's not too much extra wiring to squeeze in, and I don't have to get my prototype circuit out of the way just yet.

I still want to see if I can refine my PCB design a little before I commit to having one produced. And of course triple-check it for errors.

EDIT: Ooh, I can get a better deal from Futurlec, and it looks like there's no minimum order either!

[southernbob]

I have been looking at Chuck's schematic, for the 512K & A segment memory project, for several
days trying to figure out how the 512K module is filled completely. At first I thought
the first 128K is not used and only 384K is used. But once I layed out the address lines
(A19,A18,A17,A16) on a piece of paper and put binary values underneath each address
line it became apparent how it worked. When addresses 80000 and 90000 are on the bus
A19 is active but the memory chip only has up to A18 so only the lower 3 bits of the address
nibble are used which becomes 000 and 001. so the data for address 80000 & 90000
are stored as follows on the chip starting at location 00000: 80000,90000,20000,30000,
40000,50000,60000,70000 which would use all 512K of the chip. At least that is what I think not having built the circuit.
Bob

[jmetal88]

Yup, that's the same math that led to my circuit design, which is why mine is so similar to Chuck's I guess.