RS-485 Cable impedence mismatch

Instead of trying to guess go and get an oscilloscope and see if the data looks reasonably square on the end of the cable without significant ringing, see if the voltage reaching the end of the cable end of the cable is above the minimum specified in the receiver chip datasheet.

What is the DC resistance of the cable? Go and short one end and measure it if necessary.

It would also be a fairly trivial exercise to get a couple of laptops, wire up a couple of veroboards with a max232 and your 485 chip and actually measure the bit error rate on the link.

A discussion of transmission line effects may result in lots of arguing here. As I understand it 38Kbit/sec is low enough that the cable characteristic impedance will look complex rarther than being the value specified by the manufacturer. The results can be counterintuitive, such as inserting inductors in long telephone lines can improve transmission of low frequency audio signals. In some situation a compensation capacitor on the end of the line may acutally improve the risetime of logic edges, it's not linear, too small a capacitor may be worse than no capacitor.

Assuming a velocity factor of 0.66 a level change at the transmitter takes 1.2uS to propagate to the end of the cable, a small fraction of a bit period, a small fraction of your 26uS bit period.

Bob

It is also worth looking at crosstalk and common mode voltages.

Hopefully the cable is a twisted pair with a screen.

Connecting the screen to earth at the ends may help.

Personally for an 800foot data link I would allways go for a system that uses transformer isolation or fibre optic cable unless if at all possible.

Bob

You're edging into a cable length where that impedance mismatch could cause problems, but your system would have to be clinging to the operable region by it's fingernails for the mismatch to be the culprit. 800 feet in free air would be 800ns one way, perhaps 1 to 1.2us with reasonable velocity factors. So in one bit time your signal has a chance to bounce around 12 times -- that doesn't leave much of a prior bit to interfere with a current bit.

Like the other responder said, take a look with an oscilloscope, or at least an ohmmeter. There are so many different things that could possibly be wrong that any theorizing done before adequate facts is just so much gas. Figure out how to check for one wire grounded, for one wire open, for noise being injected into the system, if it's a really severe problem check for polarity, think of all the things I haven't thought of and check for those, and then some more.

What does the signal look like at the far end? _YEARS_ ago I had to resort to a home-brew driver and pre-emphasis to get reliable signaling over such a length.

...Jim Thompson

Wish I knew... all equipment that would tell me is 1K miles away - after all, I was *assured* it was a software problem ...

ANYTIME, ANYTHING goes wrong, in my world it's an analog problem... digital guys never admit mistakes ;-)

...Jim Thompson

the

it

ohm

or

My gut feeling is if the impedance is that far off, the balance isn't very good either.

Tektronix makes various adapters for twisted pair to 50 ohm. I have no first hand experience with them, but it might be something to investigate if you try to look at the waveform.

As mentioned earlier, at that data rate the mismatches look theoretical not practical. 800' of (distributed) cable resistance is a great helper in absorbing cable reflections. If easy, then drop the baud rate to say 1200, get the mismatch idea out of the way. Main problems to contend with are voltage loss due to bad connection/ trapped wires and excessive pickup of mains hum somewhere down the line, either induced or galvanic. But you've then got to have the scope. As a first step I'd short the far end pair and confirm the DC resistance of the cabling is somewhere near.

They learned that from the people making crap computer hardware.

--

aioe.org, Goggle Groups, and Web TV users must request to be white listed, or I will not see your messages.

If you have broadband, your ISP may have a NNTP news server included in your account:

There are two kinds of people on this earth: The crazy, and the insane. The first sign of insanity is denying that you're crazy.

We fired it by bonding the conduit to the girders supporting the building. We got the differential to less than .1 V. This was measured by disconnecting the shield at one end, and measuring between it and the safety ground.

If the equipment is intended to be run over an 800foot unisolated data link then not doing a checksum on the data, not resending if data loss is a critical problem or not giving some end user indication if there is lots of packet loss is a software design problem.

With a signaling system that can be upset by a comon mode offset that is running over a non-isolated data link you are guaranteed to get a few bit errors now and then, even if is is just once a year when there is a nearby lightening strike.

Bob

As often is the case, it was not a case of one thing being wrong, but several. In addition to being the wrong impendence, for about 450ft of an underground run it was in the same conduit as the electrical - which, by the way, was also about a 750ft total run. Of 220v on 14/3 NOMEX. Running lights, a 1500w space heater and a 1/2 hp gate opener. And they wondered why the lights were dim... Anyway, replaced both low and high voltage cables, fixed a really nasty ground loop and what do you know - hardly any transmission retries anymore. The software is still an absolute POS though

- and the hardware appears to have been designed by cutting and pasting from cookbooks - "A" has an output, "B" has in input. They automatically should work, right?

Thanks to everyone for their help and suggestions; now it's time to go appease an Marine Major who has a limited understanding of the word "patience."

I would say it was just the ground loop....

matching the termination at both ends to the cable should help if your drivers can handle the lower impedance.

A terminated 90 ohm line has impedance of 45 ohms (measured at any point) and that's less than the 50 and 54 mentioned in the datasheet.