Twisted pair wiring question

Using a twisted pair-wiring for sensors or other data that can be susceptible to noise is something I commonly see recommended, but whenever I try to do additional reading to understand the trade-offs, benefits, limitations, appropriate-use-cases, etc, I just end up more confused than before.

Some places seem to say its not actually helpful, other places say its helpful in some cases but not others, other places seem to suggest that its only helpful if one of the pairs is grounded, others say unless you're looking at a balanced signal its not helpful.

I have an intuitive sense of how I think it works, but I've been wrong before, so I'd like to clear it up if I can. To allay this confusion in practice, I've just kinda taken to doing as many things as I can as twisted pair, if for no other reason than it makes the wiring much tidier in projects.

So I'm left with a few (probably silly) questions of when to use, when not, and how to use.

  1. Do you use twisted pair for DC power-supply lines? In that case would it be VCC and GND twisted
    around each other, or two VCC lines twisted around each other plus two GND lines twisted around each
    other?

  2. When using twisted pair for a sensor does it matter whether the data line is tied high or low, and does
    that affect whether the second wire in the twisted pair is GND?

  3. Would using a twisted pair, or twisted bundle be appropriate for multi-channel lines like RGB LED wire?
    If so, how would that look? Would each data line need to be its own twisted pair with one of the wires
    being the data-line, and the other wire being ground (resulting in a total of 4 twisted pairs for the 3
    channels plus VCC)?

  4. Is there ever a benefit (other than tidiness) to twist-up more than 2 wires together?

  5. Are there instances where you would NOT want to use a twisted-pair arrangement for functional
    purposes (like transmitting certain types of data or data-speeds)?

I doubt I can give you a complete set of guidelines but I can provide some insight.

There are a number of reasons to twist wires and all have to do with operation at higher frequencies or the susceptibility of picking up higher frequency noise.

Why twist?
In theory, if a twisted pair is exposed to an electric or magnetic field, both wires will pickup the same noise but will cancel each other out resulting in little to no noise reaching the input connected to the twisted wires.

Twist also increases the capacitance between the two conductors which can be detrimental to the propagation of high frequency signals. Usually the compromise here it to have a "mild" twist. Cut open an piece of Cat 5 and see the twist that is used.

For DC twist is for neatness only.
Twisting 3 wires is likely not an effective for cancelling high frequency pickup but with low impedance circuits is find for neatness. Here again look at the Cat 5 you will notice the wires are twisted signal pairs.

Hope this helps.

John

Always pair a signal with its ground, and a supply wire with its return (typically ground again). Then the
current returns in the same tightly confined bundle so that the net current is zero.

This means no magnetic field coupling once outside the cable a short distance. Its all about magnetic field leaking
out and leaking in and inducing voltages around the circuit. The twists mean that on average the fields cancel
out, and tight twists are better for this.

With a sensor whose supply wire is decoupled at each end of the cable you can treat the supply wire
as a signal ground (its only ac that matters) - a three wire sensor can simply use a twisted triple of course.

Adding shielding will protect against electric fields, but this is generally a smaller problem with low impedance
circuits which are robust to stray capacitive coupling. Magnetic coupling will affect the voltage in a circuit of
any impedance just the same, capacitive coupling affects higher impedances more.

Magnetic coupling injects a direct error voltage, capacitive coupling injects an error current (which a shield can
drain to ground, or a low impedance circuit converts to a miniscule voltage).

High current equipment like motors can emit RFI directly from their wiring and this is minimized by using
twisted pair for the motor wiring. The low voltages mean little capacitive coupling is present.

The exact opposite scenario might be a spark plug lead - high voltages, low currents, and here a shielded
cable is the correct approach to reduing RFI as capacitive coupling goes up with voltage.

The ideal signal for a twisted pair is a differential signal, such that the average voltage of the two wires
is constant and the net current is zero. This is why things like ethernet use twisted pair with differential
signals - even in close proximity 4 twisted pairs coexist nicely (the amount of twist is different for each
pair to cancel any coupling out).

And the final benefit of twisted pairs and cables is neatness - and they are easy to make using a cordless
drill from whatever wires you need.

This figure illustrates the idea behind twisted pair. A time varying magnetic field passing inside the loop will induce an electrical current on the loop wires. In a twisted pair the direction of the induced current alternates depending on whether the green wire is "on top" or the red wire is "on top". Thus the induced current in each half twist are in the opposite direction and nominally cancel.

Key to the idea is that the wire form a closed circuit in the sense that the same current is going in opposite directions along the paired wires.

  1. It's good practice to twist power supply lines. This reduces the potential for noise pickup by the power lines as a "receive antenna". Likewise it reduces the magnetic fields that might be radiated by the wires as a "transmit antenna".

  2. I'm not clear what you're asking here. The key is that the wires are some signal and the return for that signal. The return may or may not be ground referenced.

  3. The currents in the paired wire should be equal and opposite. For the LED case this means one pair for power and return for each LED. Ideally the grounded wires should be connected to each other at one node of the network only so as to avoid having multiple ground returns for any individual signal.

  4. Because the currents should be equal and opposite the idea of a "twisted triplet" (or more) generally wouldn't be helpful from an electrical standpoint. It is common to find wire with multiple twisted pairs. Ethernet cable, for instance, has 4 pairs. There exist trunk telephone cable with a very large number of pairs.

  5. Practically it is rarely a bad idea to have twisted pair. As JohnRob notes the capacitance will be slightly higher as there is a little more wire between the endpoints.

All very helpful replies, thank you! They mostly-reinforced what I thought I understood, although provided far more details than I understood before. :slight_smile:

So now let me ask specifically about the RGB case. If one is using a few meters or more of RGB wire that is carrying a number of amps among the channels (let's say 10 since that's a nice round number, and a realistic figure), is this something that would functionally benefit from twisted-pair wiring?

And again using the RGB LED example (assuming you do want to do twisted pair), since you want to create a current-return-loop (for lack of a better term), and since the PWM-out from an RGB driver is the negative side, then you'd want to pair the PWM signal with the VCC of the LED strip, correct?

If I understand correctly, with high-speed data, it is the capacitance which can cause problems, correct? Are there certain types of data-transmission (with arduinos) that you would always want to avoid (or always want to use) twisted pair (SPI, I2C, serial)?

silly_cone:
So now let me ask specifically about the RGB case. If one is using a few meters or more of RGB wire that is carrying a number of amps among the channels (let's say 10 since that's a nice round number, and a realistic figure), is this something that would functionally benefit from twisted-pair wiring?

Generally if you're switching many amps on a wire, the concern is that your wire could be emitting a signal that might interfere with something else in your system (or somebody else's). So yes it would help from the perspective of reducing electromagnetic interference emitted from your system.

And again using the RGB LED example (assuming you do want to do twisted pair), since you want to create a current-return-loop (for lack of a better term), and since the PWM-out from an RGB driver is the negative side, then you'd want to pair the PWM signal with the VCC of the LED strip, correct?

Correct.

If I understand correctly, with high-speed data, it is the capacitance which can cause problems, correct? Are there certain types of data-transmission (with arduinos) that you would always want to avoid (or always want to use) twisted pair (SPI, I2C, serial)?

The added capacitance added by twisting wires is relatively small and hence not usually a driving concern. If it matters your probably operating too close to the limits of your protocol anyway. The specific transmission protocol tends to be driven by the distance and throughput required. SPI, I2C, and TTL serial are short range intra-system protocols. Single ended (RS232) and differential Ethernet, RS422 are used for longer distance inter-system communications.

thanks a bunch! :slight_smile: