some spare parts laying around.. so I built a flash cart

tomaitheous · Post by **tomaitheous** » Sun Jan 17, 2010 6:17 am

Had some spare parts and some free time...

http://www.pcedev.net/md_dev/cart/

Haven't tested it yet, but double checked all the wires. Should be good to go. Gonna test it out tomorrow

Chilly Willy · Post by **Chilly Willy** » Sun Jan 17, 2010 7:27 am

Wow! That looks like a lot of work.

Snake · Post by **Snake** » Mon Jan 18, 2010 1:19 am

What was the cart before you 'converted' it?

TmEE co.(TM) · Post by **TmEE co.(TM)** » Mon Jan 18, 2010 3:43 am

holy shit... !

you should have used thinner wires and PNG is not very good for photos

tomaitheous · Post by **tomaitheous** » Mon Jan 18, 2010 6:47 am

Wow! That looks like a lot of work.

It was ;>_> Especially since I originally had them sideways. It got too tight, so I desoldered everything and rearranged them vertically.

Snake wrote:What was the cart before you 'converted' it?

Bill Walsh College Football, iirc. I just needed something real quick so I could do tests on the real system and such.

I left the decoder and the sram setup intact on the board... for whatever reason I might need save ram or such.

TmEE: I plan to make a more serious flash cart setup like yours in the future. Thanks for showing me that exe proggy that lets apps directly access the parallel ports

It just a matter of me of sitting down and figuring out the logic for flashing the chips (well one large chip hopefully).

I was afraid of the wire length being too long, so I use phone line wire (the solid core). I've heard it's not good to have too long of a distance with wire without a buffer chip. Not sure what "too long" is relative to though.

dtech · Post by **dtech** » Tue Jan 19, 2010 5:00 pm

Ooof that looks horrible

I hope it works even for 8MHz bus, but it's very likely to at least glitch a lot

There will be a lot of ringing, crosstalk and ground bouncing going on due to long wires, no buffers and tree-like structure...

You need to keep all "traces" as short as possible, as matched length as possible, and never make long branches. In this design, last hope to make it work somehow would be to add few bus buffers made of any noninverting 5v cmos logic available. buffer up all address lines, as well as buffer all data lines with buffers that have Output Enable (say 74125 or anything similar) and then keep ROMs allways output-enabled.

Anyways, give a note when you've tried it out!

I think I will put on the web etching/milling files for a high quality flashable board kit (programming adapter and cart itself) someday soon, but it will require soldering SMD components.

Question for everyone: what flashing tool on what port do you use or would prefer to use? Given the fact that LPT ports are phasing out of any modern computer hardware and ROM programmers oftenly are too expensive for their abilites. I am just curious what is popular here

This is the hardest point, as it's pain in the *ss (at least for me and people I know) to create USB drivers and software for a custom device. Especially that are stable on many machines and OS versions... There's only one feasable version for USB: Using FTDI232 or FTDI245, emulating fast serial port. Anyways - is somebody interested in writing programming software with GUI?

tomaitheous · Post by **tomaitheous** » Fri Jan 22, 2010 2:28 am

Ooof that looks horrible

Haha - Not as horrible as it was to make. But still - I kinda like the way it looks

I'll probably try it out this weekend. I do have double sided bare copper boards for etching, just haven't got over to someones house to use their laser printer yet (nor have I bought the etching solution yet). I have some 8line buffers, but not sure that I really want add any more time into this design. I have some other ideas >_>

As far as a programmer, I already an limited 40pin dip for USB. But I'm currently designing/building my own custom parallel port similar to what TmEE did. Of course, that'll only work as long as my MB has parallel port :/

You mentioned USB devices. What about the USB to parallel or such device (DLP-2232M-G, not that cheap though)?

Jorge Nuno · Post by **Jorge Nuno** » Wed Jan 27, 2010 1:32 pm

At 8MHz, don't even bother with length equalizing, the lambda is more than 10 meters, so only worry if you have a wire length of 1 meter or more

freezedream · Post by **freezedream** » Thu Jan 28, 2010 12:10 am

dtech wrote:There's only one feasable version for USB: Using FTDI232 or FTDI245, emulating fast serial port.

If someone designs a USB flash cartridge programmer, capable of programming, say, dual 8-bit flash cartridges, I will be very excited.

dtech wrote:Anyways - is somebody interested in writing programming software with GUI?

Sadly, I'm no programmer, so I guess this may need to be a team effort.

Chilly Willy · Post by **Chilly Willy** » Thu Jan 28, 2010 12:50 am

Jorge Nuno wrote:At 8MHz, don't even bother with length equalizing, the lambda is more than 10 meters, so only worry if you have a wire length of 1 meter or more

The distance will be shorter than that - don't forget you need a half-way decent square wave, so you need at least two harmonics, and square wave harmonics are odd, so you need to consider the wavelength at five times the clock frequency, or almost 40MHz. So instead of a yard, you're looking at a couple feet. Still long enough that you don't need to worry about wires a couple inches long.

HardWareMan · Post by **HardWareMan** » Thu Jan 28, 2010 9:14 am

Just use twisted pair on each signal or use ribbon wire, wich even wires is signals and odd wires is on GND. And you will be happy.

dtech · Post by **dtech** » Thu Jan 28, 2010 2:50 pm

Excellent advice, HardWareMan. Ribbon wire with every odd/even wires as GND will not only fight crosstalk but also an even more tragical issue related to return current surges. If there are, for example, 20 address wires and one gnd wire, the capacity and leakage in chip inputs to gnd will cause gnd to become extremely drifty and shifty. Same about all other pins... There must be a clear current surge return ground path for every line. In case of PCB it's done using large ground area. Genny cart slot is not a good example, as it already has too few ground pins. While 8MHz doesn't seem too much, in case of wide parallel addr/data/control bus (genny is the case) group delay and line behaviour consistency kicks in. I don't say that long wires doesn't work - it's just that it is a hell to fix it if it doesn't and oscilogramms will look quite repulsive (a matter of tase, though) as well as the cart might stop (or start) working when external things are connected to some wires (for example a scope probe's ground, etc).

It's just risky and better if handled with caution.