RS422 over Cat5e: length speed product?

[technix]

This somehow popped into my head: how long and how fast can I pump 5V RS422 down standard Cat5e or Cat6 cables? Is there a “length baudrate product”? How can I estimate it?

Or more generally, how to estimate the maximum length of twisted pair cable (usually Cat5e or Cat6 UTP) can I pump a differential signal at a given voltage and speed down? Is there any general formula that I can look up values from spec sheets and plug numbers into?

[Rerouter]

Look at your driver chip, for differential drivers the max capacitance is limited by the driver output current, e.g. if your drivers can only swing 10mA, your bandwidth is limited by the impedance of the line capacitance.

[mikeselectricstuff]

Off the top of my head :
For low baudrates, up to about 1Mbaud the first limit you hit is cable resistance, at about 200-300m
You may be able to extend this by paralleling up the pairs.
6MBaud is good to at least 50M, 4Mbaud 100M

If you have multiple buses on one cable, then for >50m you should definitely be using cat6 for lower crosstalk

[rstofer]

RS485 is probably a better standard than RS422.  In any event, the chip manufacturers have datasheets and app notes that discuss this very thing because, were it not for distance, these standards might not even exist.

RS485 tops out at around 4000 feet at 800 kbps for one of Maxim's drivers.  Beyond that, there is a tradeoff between distance and speed.  Cable resistance isn't allowed to drop the signal more than -6dB so cable specs will matter.

https://www.maximintegrated.com/en/app-notes/index.mvp/id/3884

[David Hess]

RS485 is probably a better standard than RS422.  In any event, the chip manufacturers have datasheets and app notes that discuss this very thing because, were it not for distance, these standards might not even exist.

Exactly, the datasheets for the drivers often include a data rate versus length graph.

[chickenHeadKnob]

Many decades ago I worked on a system that was routinely deployed in industrial campus type settings and we ran at 19.2K baud over 1km distance on cruddy cheap twisted pair. If going inter building I think that isolated drivers with a well engineered spark gap at either end should be a de facto requirement for safety reasons even if it isn't in the electrical codes of your nation. We used the old school sn75175/ '174 pairings on an optically isolated mezzanine board.  Was very reliable.

[mikeselectricstuff]

Many decades ago I worked on a system that was routinely deployed in industrial campus type settings and we ran at 19.2K baud over 1km distance on cruddy cheap twisted pair.

Though cheap, was that heavier gauge cores than Cat5/6? - with Cat5/6 I tend to find DCR limits to about 300-500m whne properly terminated. Most non cat5/6 twisted pair cables tend to have bigger core size.

[chickenHeadKnob]

Many decades ago I worked on a system that was routinely deployed in industrial campus type settings and we ran at 19.2K baud over 1km distance on cruddy cheap twisted pair.

Though cheap, was that heavier gauge cores than Cat5/6? - with Cat5/6 I tend to find DCR limits to about 300-500m whne properly terminated. Most non cat5/6 twisted pair cables tend to have bigger core size.

M..man you are really putting my memory to a test. I was/am a software guy and never paid the kind of attention to the hardware details that the engineer selecting the wire did. Yes certainly a larger gauge than cat5 maybe 18 gauge? completely unshielded. can't  remember anymore sorry. Back then copper was cheap, now a-days the copper by weight value alone would make more expensive than cat5.

[ogden]

RS485 is probably a better standard than RS422.

I would like to disagree here. From signalling point of view they do not differ. Main difference: rs485 is half-duplex, rs422 - full duplex. Even drivers happen to be universal for both, like this one:

http://www.ti.com/lit/ds/symlink/sn65hvd179.pdf

[T3sl4co1l]

I would suspect, with a little signal conditioning (high frequency peaking) and maybe extra amplification, you could get a bit more length out of it.  (Gain might not even be that big a deal, the receivers aren't terrible comparators as they are.)  And of course -- boosters or retransmitters are always a good idea, if you need really long lengths.  Or just use a better long-distance method (like, uh, DOCSIS or DSL?), or a few watts of licensed radio.

Tim

[David Hess]

I would suspect, with a little signal conditioning (high frequency peaking) and maybe extra amplification, you could get a bit more length out of it.  (Gain might not even be that big a deal, the receivers aren't terrible comparators as they are.)  And of course -- boosters or retransmitters are always a good idea, if you need really long lengths.  Or just use a better long-distance method (like, uh, DOCSIS or DSL?), or a few watts of licensed radio.

You can get much higher data rates.  Use high compliance current drive with termination at each end just like with LVDS and roundoff due to high frequency losses will become the limitation and that can be improved with some high frequency emphasis just like when sending video.  Do it right and existing RS-422/RS-485 integrated receivers can be used.

I designed and fabricated a discrete RS-232 driver like this using 2N3904s and 2N3906s which produced perfect results into unterminated cables resulting in greatly improved data rates over long cables.  The waveform produced could have been used for oscilloscope transient response calibration except of course that it was a little slow.  I was just being lazy though because I did not have any RS-232 level shifters handy and needed a non-inverting solution.