The idea behind "twisted pair" is:

1. To use a pair of cables, which carry the signal as symmetrical signal (the actual signal and an inverted copy of this signal), instead of having one signal ground wire, where several other signals are also referenced to. The actual signal is only referenced between this pair of cables. No other cable is intended to carry neither a portion of this signal current nor any return current. Everything "plays" between these paired cables.

2. To twist this pair of cables. Of course, the loop area between these paired cables must be as small as possible, just to reduce emission and absorption of magnetic and electric fields. But in demanding applications this might not be enough. Only by evenly (!) twisting these paired cables you can achieve, that the field vector of magnetic interference hits the loop areas (from twisting node to twisting node) twice, but with opposite sign, so that the induced voltages cancel another, the more the more evenly the pair is twisted. Similar for electric fields: The unavoidable noise injection into both cables of the pair is also better balanced when the cables are twisted.

Another advantage of this twisting comes into play, when other signal cables are running within the same bundle. Then, twisting makes, that all interfering cables come close to both cables of the pair at a ratio of 50% to 50%. And, as the twisted pair carries a symmetrical (differential!) signal, the interference injected into both cables of the pair (evenly as 50% to 50%!) will finally vanish in the receiver!

The use of high grade twisted pair cables can yield outstanding results!!! A good microphone cable in recording studios, for instance, can transmit signals in the 100 microvolt range at a signal to noise ratio beyond 80dB.

To your question, Per. If your cable also transmits other signals, omitting the use of a twisted pair for your RS485 can end in a disaster. If, on the other hand, the RS485 is the only thing to be transmitted by the cable, then use at least a slew rate limited RS485 driver, like the LT1785 for instance, because the lack of twisting often results in a wide tolerance of characteristic impedance of cable, especially after years of bending and maltreating the cable!


Kai


PS.: Remember: Running symmetrical signals even in combination with a poor cable still yields decades better results than running signals asymmetrically, even with a state-of-the-art-cable.