Transmit through a ribbon via LVDS

Yes, I use those chips, and a few other LVDS things, over ribbon cable, pinned as you describe. It works fine for short runs, but the cable will be lossy at high frequencies so you'll get ISI and eventually bad amplitude losses for long runs at high speed.

You might google for some data on ribbon cable attenuation.

How fast and far do you need to go?

That apnote used twisted-pair ribbon cable, which will be very different from flat cable. They don't say, but probably had no grounds between pairs, which can lead to crosstalk and jitter.

The soldering note is interesting. He shorted things and didn't bother to fix it, just kept going and published the bad data.

tc

you are right I didn't notice

the ribbon cable with 40 signals and 40 grounds interleaved was good enough for 66MBps ATA so regular ribbon cable should work.

yeh it does seem like something an intern made to pass time...

Those are fine, yeah. If your environment is noisy, do mind common-mode noise at the receiver, yes, even with the interleaved grounds (which is the right way to go on plain ribbon).

Even better is twisted pair ribbon cable, which is quite available. In that case, you can only do sig1+, sig1-, sig2+, sig2-, etc. Maybe some grounds on the end pairs would help nail things down, otherwise, use metal tape to shield the cable and provide a ground plane.

Tim

-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website:

Ok. i just figured out now that pcb routing needs to implement some serpentine track near the LVDS chip transmitter ... those tracks "sig-,+" are microstrips.

Yeah, i found this quite old but useful anyway.

May be a FFC/FPC and ZiF connectors are more suitable ... i realize than many LCB flat panel manufacturer are using this. I'll see it may be.

Cool and Clean.

rise and fall times are about 5ns (3.3v logic) Freq=25Mhz. ribbon cable is 10cm long maxi and the EMI environment could becomes noisy

good point. I will focuse on.

mmhh ... let me see ... yeah a cat6/5 ethernet cable ! I also consider this but too much signals (8 single ended --> 16 diff pair). Good thinking anyway.

No, ribbon with twist. This stuff:

Tim

Oh i see ! as the antic hp64000 in-situ debug probe. Please consider you mention "use metal tape to shield the cable and provide a ground plane." This makes me though about cat6 cable ...

Habib

This kind of stuff:

Basically makes an N-conductor shielded cable on the cheap!

Tim

Ok thx Tim :-)

Those numbers are easy. Expect no problems unless you have a lot (more than 1 volt) of ground-loop offset or noise between the grounds on each end. I guess more than 1 volt would melt the ground wires in the ribbon cable!

I did one 100 MHz link that expected some ground loops. I drove the cable at 5 volts p-p on each differential wire, and used voltage dividers ahead of the LVDS receivers on the other end.

Cat5 or 6 has four pairs. One flavor is FFTP (each pair foil shielded, with an overall foil shield) or SFTP (overall braid, foil shielded pairs), either of which gives you four pairs, no crosstalk, and lots of ground returns. You can get cables with metal-shell (grounding) connectors on each end.

The Cat5 without pair shields (UTP variants) varies the twists to reduce crosstalk, so the prop delay is different on each pair.

I know more about Cat5/6 than I really wanted to.

Well, not at 10MHz+.

Congratulations, you invented high speed RS-422! :^)

(RS-422 is basically a differential receiver, as you'd get within an LVDS receiver, but with a resistor divider which extends the CM range. I suppose it /was/ the LVDS of its day, in terms of speed with respect to the attached logic family, which would've been mainly TTL. Likewise, the interface chips were all bipolar.)

Tim

Yeah right I saw this a long time ago ! the insulation around the gnd wire was melted !! from that time I always put a Y capacitor between system ground and signal ground at each cable termination (e.g. RS485 ...) Unfortunately few designers have this in mind. Good point John ! Hope this will help beginners. Beside it is the rule in the implementation of Ethernet Jack (integrated capacitor)

BTW LVDS transmissions work actually like a charm ! No outrageous behaviour on ringing or time skew between sig+,- hourra !

I will publish some screenshots if i've time for.

Thx John, Habib.

A twisted-pair ribbon cable will perform a bit better that straight ribbon cable. But, the twisted cable has defined pairs, so you need to keep the differential signals on the twisted pairs. Chabin used to make some transmission line cable for Tektronix logic analyzers and IBM mainframe computers that had single ground wires between closely-spaced differential pairs.

Jon

Maybe, but would probably have more crosstalk. And would cost more.

Why? Twisting mostly reduces cross-talk, by making lots of little segments that cross-talk in opposite directions.

Not necessarily. Cost of cable is more about the volume produced than the copper content, and twisted pair doesn't use up that much more copper anyway.

If you can find a cable that's manufactured in high volume, it's likely to be cheaper than something with more limited appeal.