I have historically taken many of these topics on a "rule of thumb" basis. However, I am looking at a new application that may be noisy and over long distances (up to 60m) by nature, so I want to take all the precautions I can.

Low data-rate RS232 would be fine, but I have some questions:

1. I see it recommended to use twisted pair cable for RX/TX. For single ended signaling, I do not see the advantage of doing this. RS232 doesn't filter common mode voltage, so why would there be any advantage? If anything, I would think the second line of relatively high dV/dt would be prone to capacitive coupling with the first.

2. This site has some specific information regarding RS232 over long distances: http://www.marcspages.co.uk/tech/long232.htm. However the suggestion to implement a twisted pair (with ground) for each signal raises the question of what does a twisted pair with a shared ground do to noise attenuation? I am assuming the author is suggesting (to retain full duplex operation) both the TX and RX grounds be tied together at both nodes. So how would return currents be distributed fortunately to promote noise immunity (he mentioned typical RS232 hardware, not galvanically isolated receivers). Or, is it the additional physical separation two twisted pairs provide between RX and TX the noise benefit while utilizing the second conductor in both pairs for significant ground mismatch currents?

3. How is the best way to handle shielding in RS232? I expect on typical PC/Modem applications, it is likely best to tie the shield ground to earth (typically on both ends) while using the PSU ground to connect to Gnd (pin 5) on both DTE and DCE devices. If you don't want to go to earth ground, do you simply shield at one end to PSU ground?

Practically, being forced to use 2 pair cable for extending RS232 begs the question why not go with a differential serial connection like RS422/485. However, if I do this, how do I handle two nodes with floating (isolated) power supplies? I assume this would take a dedicated wire to share the grounds, giving me the same 2 pair cable as using RS232. I am trying not to ground the supplies to earth in this application.

If anyone has answers to these questions, or even a good book on the applied side of this subject, it would be much appreciated.

 

The article is ~20 years old. There are several better solutions today which is why RS232 is an anachronism. Which makes hardware for it that interfaces with today's motherboards... What is it that you are trying to do?

 

@SamR My questions are meant to be more generic than they seem. I realize there are a number of more capable schemes I could utilize, but I wanted to clarify some details of serial cabling that contradict what I understood regarding twisted pairs and shielding.

 

The article was based on 9600 baud which is pretty slow. Back when I was building cables for it in the 80s MicroSoft MS-DOS only supported 9600 baud but you could get drivers for 19.2kbps and 38.4kbps. At that rate, you maybe could go 60' using individually shielded twisted pair data cable. And it was serial, not parallel. Even getting the best cable available only bought you a few extra feet. Slower rate, longer distance. Which in turn meant longer transmission time. Trying to drive a drum plotter for 30x42" sheets took quite a bit of time. So what are you trying to do or is this just out of curiosity?

 

You need to use twisted pair on RS-485.

 

/snip So what are you trying to do or is this just out of curiosity?

It has an end application in mind, but the application prompted me to look at how to properly setup cabling for these situations, and some of the information I have found contradicts my understanding of the methods.

 

You need to use twisted pair on RS-485.

Of course. I am more confused with the seemingly common advocation for using twisted pair for single ended serial comm. If I did go RS-485, I would like to do the shielding correctly, which I believe is how I mentioned above.