Probable RFI issue between servos and locos

[mmagliaro]

A related question in my mind is what makes a group of wires "twisted"?

Would a flat cable (with 3 conductors) that is twisted along its path between components be any different than three individual wires that are twisted together along the same path?

I had been told that the purpose of twisting the wires for things like the track power bus was to cut down on the space between the wires along the path by holding then close together.  Wouldn't bonding them into a flat cable do that, maybe even better.

And, if the twist is actually the functional aspect with respect to the issue(s?), why wouldn't twisting a flat cable be as helpful as twisting individual wires?

Let's consider 2 wires carrying DC.  If the wires get hit by some RF, they act like antennas.  There will be an electromagnetic field around each wire, inducing a current that you don't want in that wire - a random, noisy, AC current.   At any instant, the DC current flowing in one wire is opposite the polarity of the current flowing in the other wire.  The magnetic fields induced in each wire are therefore opposite.  You would think that two wires close together, side-by-side, such as in a ribbon cable, would allow the two magnetic fields to cancel each other out.   And they do... somewhat.  But in fact, the two wires are not EXACTLY the same distance from the incoming interference wave, so the induced magnetic fields are slightly out of phase, and they do not exactly cancel each other.

But if you twist the wires around each other, what now?  Well, the two wires are STILL not exactly in the same position relative to the incoming interference, but over their length, they "average" - for one inch, wire A is closer, and for the next inch, wire B is closer.  As it turns out, this achieves a much closer match between the two fields and better cancellation.

Twisting a whole 3-wire ribbon cable around itself would accomplish almost nothing, because the individual conductors wouldn't be each paired with an opposite-polarity mate that is alternating position with it along its length.  The "magic" that makes it work is having a twisted pair such that the average position of each conductor is the same relative to the incoming noise wave.

[alhoop]

Very good explanation on twisted wire pair cancellation effects in reply #15 ,Max
Al

[KevinHunter]

Let's consider 2 wires carrying DC.  If the wires get hit by some RF, they act like antennas.  There will be an electromagnetic field around each wire, inducing a current that you don't want in that wire - a random, noisy, AC current.   At any instant, the DC current flowing in one wire is opposite the polarity of the current flowing in the other wire.  The magnetic fields induced in each wire are therefore opposite.  You would think that two wires close together, side-by-side, such as in a ribbon cable, would allow the two magnetic fields to cancel each other out.   And they do... somewhat.  But in fact, the two wires are not EXACTLY the same distance from the incoming interference wave, so the induced magnetic fields are slightly out of phase, and they do not exactly cancel each other.

But if you twist the wires around each other, what now?  Well, the two wires are STILL not exactly in the same position relative to the incoming interference, but over their length, they "average" - for one inch, wire A is closer, and for the next inch, wire B is closer.  As it turns out, this achieves a much closer match between the two fields and better cancellation.

Twisting a whole 3-wire ribbon cable around itself would accomplish almost nothing, because the individual conductors wouldn't be each paired with an opposite-polarity mate that is alternating position with it along its length.  The "magic" that makes it work is having a twisted pair such that the average position of each conductor is the same relative to the incoming noise wave.

I don't think the "almost nothing" estimation is quite right. Lots of noise suppression can be done without demanding the near-perfect noise control required in something like an analog audio signal.

In most electronics that control a servo the power and ground wire are closely coupled with capacitors large and small throughout the power circuit. Provided that the servo includes a small cap across the power, as is good practice, at higher frequencies the positive and ground should act almost as a single ground in relation to the data cable. Further the twisting will help rejection across the power / ground pair. Although the noise suppression of twisted flat wire is not an ideal "twisted pair" it will still be many times better than an untwisted wire.

In regard to the original problem, the electrical noise created by sparking brushes can be very powerful. The surges of EMF can be much larger than the supply current or voltage, created as magnetic fields are collapsed when the commutator irregularly breaks contact.

Further, low cost servos may have no flitering at all. They were originally designed to operate on batteries and far from external electrical noise. I would be interested to see the behavior with external power filters near the servos, or of any slightly higher quality servo that might include better noise rejection.

Kevin Hunter