RS485: when should I consider it a "transmission line"

Look at speed of light in your twisted pair vs. the baud rate -- that'll give you a characteristic bit length.

The length of one bit at 115200 baud on twisted pair is about 1800 meters. I used to work on a system that used RS-485 at that baud rate, with never more than about 100 meters end-to-end, and it benefited tremendously from having terminators. It was in a very noisy electrical environment, but -- it still needed termination.

--

Tim Wescott 
Wescott Design Services 
http://www.wescottdesign.com

There are several considerations for signal transmission in cables: - resistive loss - frequency dependence of cable loss - transition time For the latter, the usual rule of thumb is that a discontinuity that is less that 1/6 the transition time can be neglected, so for a signal with a 1ns risetime, anything less than 1" (speed of light is ~6"/ns) is ignorable. But I doubt you'll be getting that with an RS485 signal!

Consider doing a spice simulation of your circuit.

You're about a factor of two off there. Speed of light is almost exactly 300 Mm/s, which roughly equals 1 foot/ns. Signal speed in cables tends to be about 2/3 times that, so 200 Mm/s, or 8 "/ns.

You have the right propagation speed for polyethylene cables. My 6"/ns is a back-of-the-envelope value for PCBs.

However, in either case they are just rules of thumb, and if the system being considered has characteristics that are vaguely in the same ballpark, then it would be prudent to model it in some depth.

ly 300 Mm/s, which roughly equals 1 foot/ns. Signal speed in cables tends to be about 2/3 times that, so 200 Mm/s, or 8 "/ns.

I went through the same thought process some time ago, calculating the maxi mum baud rate versus cable length.

Then I did some tests, layed out 900 meters of cable in the lab, and to my surprise communicated fine at 115.2kbaud (at 900meters....).

The combination of a 3V drive level from the master and a 200mV detection l evel of the slave meant that even though the waveform was not in any way pr etty, the signal came through fine, just with a time delay (no bit errors o bserved)

Cable used was: ALPHA WIRE - 5472C SL001 - CABLE 24AWG_429588 (~40pF/m)

Cheers

Klaus

And will it work under all (expected) conditions?

Other experience:

115k2, length aprox. 20m. So it was said (by a company we worked with on that project) termination was not required.

In idle, communication was fine, but as soon as the machine was started (high power electrics, signals through slip rings and such fine things), communication reliability went down dramatically. Adding termination improved the situation a lot.

Since that experience (and also before on our own projects) we always use termination whenever we use 485.

So I would only consider leaving out termination if it had a real benefit and only in friendly environments.

--
Stef    (remove caps, dashes and .invalid from e-mail address to reply by mail) 

The life which is unexamined is not worth living. 
		-- Plato