Can't Remember How To Drive Ribbon Cable

Got a project where I have to interface to a third party board via two wide 0.025" pitch ribbon cables. One of the cables goes G S S G S S G..., and the other is even more sparsely grounded (and seemingly at random. Thanks guys, great design job).

I'm going to be sending digital data at around 1 Mbit/s per digital line down probably about a foot of this stuff and would love for my edges not to ring. Hence I'd like either ballpark controlled impedance on the PCB or a slew rate limiting driver feeding the cable.

Option A) So the specs I'm getting on the half pitch ribbon is 90 ohms SE, 130 ohms differential. But I can't remember what assumptions go into those numbers, and Google is being less than helpful. All of my signals are single ended, but IIRC that 90 ohms is assuming a pleasant S G S G pattern on the cable. Any idea what my impedance might be?

Option B) I've got something on the order of 80 signal lines to deal with, into about 25 pF from the cable, reciever, and assorted strays. Anyone have a recommendation on a good wide (16 bits or better) buffer chip that can drive that with a rise/fall time of about 8 ns? I've got

3.3V and 5V available.

I'd prefer option A to B since I'm not sure how happy the driven chips will be about those slow edges, but I'm not sure that A actually works. Likewise if anyone's got an option C (dual Schottky rail clamps maybe?) I'm open to it. Thoughts?

If it is only a foot of cable, and only 1 MHz, it will probably just work fine. If you can, put series resistors (or resistor packs) near the drivers, and put caps near the load (if you have that option). You can probably use 0 Ohm jumpers for the resistors and leave the capacitors unpopulated, but if you do need to slow down the edges, you will be able to.

I doubt you will need to do controlled impedance of the board with such slow signals and such a short cable. If you do, 90 Ohms will be close enough.

I wouldn't consider 8 ns to be all that slow. You mention 1 MHz, which means your period is 1000 ns. At 8 ns, Your rise time is less than 1% of your period. You could probably afford to go even slower. Slow edges will work out better for both signal integrity, and for radiated emissions. (By the way, if this setup has to pass radiated emissions tests, you could have some problems.)

If the driven chips have a minimum rise-time specification, it would be good to know what it is. Do they?

Also, if this is a one way data transfer, it might be possible to use one or a few serializers instead of a parallel bus. It could really reduce the wire count in the cable. There would have to be room for a board at the receiving connector, though, and there might not be.

Just a thought.

--Mac