Mike's PCjr Page

Alright, now that my head has had time to clear from this, I'm working on making a cleaned-up version of the schematic. Also, since the old schematic was done in DipTrace, and that's a commercial program, I'm re-doing it in the open-source program 'gschem'. I did use the freeware version of DipTrace to make the old schematic, but I just like the idea of releasing a schematic generated with open-source software.

EDIT: I'll post it when I'm done, so hopefully you guys can figure out if I'm doing anything wrong, or if there's an alternative approach that would work better.

I haven't checked it too carefully, but here's my first try, exported as a PNG.

I asked an expert (Andrew Lynch) over at VCForum - he has a few projects under his belt, including the NV8EM computer and XT-IDE adapter. This was his answer:
I've never designed anything big enough to need software, but I'm willing to learn. And we have backup.


Mike

I have KiCAD on my machine as well, but gschem was easier for me to get used to. There's supposed to be a way to make PCB designs using other products in the GEDA suite, but I haven't figured that part out yet.

EDIT: Oh, actually it's pretty easy. You can just import the schematic, place the components, and auto-route. I kind of wish I could figure out how to make it route without vias, though.

EDIT 2: GEDA PCB's auto router is seriously lacking compared to DipTrace. I guess it doesn't matter too much though, since I haven't successfully tested my design yet, anyway.

Alright, I started re-doing my schematic on KiCAD, and discovered an error. A16-A19 are connected in reverse order to the 74154. This is probably also why when testing the upgrade (while it was working without the 74ls245 installed) I only got it to count to 448k.

I guess I like GEDA better than KiCAD. Schematics are about the same in both, but the options for PCB routing are clearer in GEDA.

Hmm, I'm wondering if I might be able to combine all the logic onto a microcontroller or PLD. That would reduce the cost of getting circuit boards printed (should the project advance that far) by quite a bit. I already had a design done where I did the address decoding with a PLD, but I still had the '373 latches and the '245 transceiver in the design.

I went back to DipTrace.

It's just so much quicker than all the other offerings out there. If I were doing this to make money, I would totally consider buying a license for this program.

Anyway, I reconstructed the old design I had around the PLD, and I'm concentrating on squeezing it into the smallest space possible. Luckily, the DipTrace autorouter can work magic, and I can have all the chips almost touching each other. I have the board cost with BatchPCB down to around $16, or $31.10 with shipping and the $10 'handling' charge, assuming they're ordered individually.

I don't want to give anyone the purchase link until I have the design successfully tested on perfboard, though. The PLD code works in a simulator, but since I don't have the device yet, I have no idea if it'll work practically.

EDIT: Got the PCB below $15 now, so that's under $30 with the shipping and $10 handling charge. I think that's a good goal. The board is, of course, unpopulated in that state.

I'm working on a PCB design for the CPU socket version of this upgrade again, using all discreet components, with KiCAD and FreeRouting. I believe that the design I prototyped should have worked, and I'm not sure why it didn't, so I'm hoping using a printed circuit board and new chips will clear up any problems. I'll probably be getting the PCB prototype from BatchPCB, or maybe PCBsingles, if I can make a decent one-sided design.

Alright, I expanded the circuit board to provide a full 736k, and I also discovered a great new option for getting PCBs made (at least if your size is less than 10cm x 10cm).

Check it out:
http://iteadstudio.com/store/index.php? ... cts_id=175

They're in China, looks like air mail to the US is $5, so $33 total for 10 boards! From what I've heard they get boards made up in about 6 days, and shipping takes 1 to 2 weeks. I've read some mixed reviews about them, but apparently you can get good results if you design with a few quirks in mind (I think my design will be fine without any modification).

At that price, I think it's definitely worth a shot.