Making an XLR cable

Posted by (twitter: @_sorceress)
February 12th, 2017 3:31 pm

I realise this doesn’t need a tutorial, but I enjoy small stuff like this. Documenting it gives me an opportunity to share some background/theory, and a chance to show some of my methods.

Balanced and Unbalanced Signals


Signals can be carried in a variety of ways, be them analogue or digital, high or low frequency, high or low impedance. Unbalanced cables employ one “live” conductor which carries the signal, and one “ground” conductor, which provides a return path, and as it’s name suggests is grounded somewhere.

Such cables are sometimes constructed in a cylindrical fashion, with the ground taking the form of a braid surrounding the live core. The grounded braid acts like a faraday cage, helping to prevent the live core from picking up interference (amongst other things). Because these two conductors play different roles and have different properties, there is an undeniable asymmetry.


Conversely, a balanced cable consists of a matching pair of wires, neither of which is grounded. The signal is the differential voltage between these two conductors. This arrangement maintains a symmetry, which has certain advantages.

Such cable also may be surrounded by a braid, which is grounded, to create a shielding effect. This braiding plays no role in carrying the signal though. The two conductors are sometimes distinguished as ‘hot(+)’, and ‘cold(-)’. Switching them around effectively inverts the polarity of the signal, which may or may not be important.

For examples, coaxial cable is unbalanced. It carries a signal in the live core, while the outer braid is grounded. Computer USB cables are balanced. A differential signal is carried between the D+ and D- wires.

Connector Construction

XLR cables are often used to carry balanced audio signals. They were invented exactly for this purpose in the 1950s for radio work, and stood the test of time through the age of television. Due to a combination of them already being ubiquitous and quite satisfactory, they have persisted into the 21st century.

An XLR patch cable as one would use in balanced audio applications is made of three basic parts: a length of suitable cable and an XLR connector for each end: a standard 3-pin male, and a standard 3-pin female type.

These types of cable are typically used to carry a single channel of balanced audio, such as from a microphone to a preamp.

The three pins of the connectors are numbered and utilised in a standard way. Pin 1 is for ground, and is usually connected to the outer braid of the cable, while pin 2 is for the ‘hot’ wire, and pin 3 for the ‘cold’ wire.

Cable Construction

Cable consists at a minimum of two parallel conductors for the balanced signal. Though there is almost always an outer braid too, for shielding. Shielding is important to protect the conductors from picking up interference, and also to prevent the cable radiating out it’s own electric field.

Copper braid is effective at blocking high frequency fields, like radio waves. But because copper is a low permeability metal, it much less effective at blocking low frequency magnetic fields, like the hum produced by mains transformers. A formidable solution may be to use two layers of braiding, with high and low permeability metals (such as permalloy and copper). This is utilised in extreme cases, like EMP-hardened cables, but it is not really seen in audio applications.

So copper braiding alone may not be enough to reduce the noise level below a desired threshold. Especially important in electromagnetically noisy environments, or with very sensitive signals. We can see this in the following graph.


The horizontal axis is the distance from the cable to a noise source, and the vertical axis measures the noise level induced in the signal.

There are some assumptions here, because the graph for any particular cable will depend on several factors, like it’s length, the separation of the two conductors, the specifics of the noise source, etc. So don’t read too much from it.

Though it should be clear from the shape of the graph that the cable is quite susceptible to induced noise, even from noise sources relatively far from the cable.


The two conductors may additionally be wound around one another to form a “twisted pair”. This twisting “averages out” the side-by-side position of the two conductors, over the length of the cable, making them appear equidistant from external points of reference.

Although were are only talking about a couple of millimeters at most, the twisting can help to remove induced noise significantly. This works because apparent equidistance makes any induced voltage “more equal” in each conductor of a twisted pair, compared with an untwisted pair. And since it is a differential signal in balanced audio, “more equal” means less induced voltage *between* the two conductors.


The graph for this cable shows the noise rejection quite clearly, with a distinct logarithmic slope of 15dB per unit distance this time, well below the corresponding graph from an untwisted pair.

As stated, we shouldn’t read too much detail from the graph due to free variables and assumptions, but at minimum it should be clear that it is “better”!


Another innovation is star-quad cable. Instead of two conductors, there are now four conductors arranged with some precision around a central non-conductive thread. Diagonally opposite pairs are soldered together at the ends of the cable, forming two effective conductors.

This geometry creates a different form of averaging, as each diagonal pair appears electromagnetically “more like” a single conductor in the centre of the cable. The differential signal thus appears to run along the centre of the cable in both the hot and cold pairs, and apparent equidistance from external points of reference is achieved. Star-quad geometry can eliminate induced noise, and it can also be combined with twisting and braiding to achieve a very low noise cable if desired.


Keeping all other assumptions the same, we can generate a graph for this cable, and it turns out to show a steeper noise rejection of 30dB per unit distance. ie, star-quad cable rejects induced noise at twice the dB of a twisted pair.

This is all theoretical of course – actual cables are not constructed to mathematical perfection, so any conclusions about it’s betterness may turn out to be a bit optimistic in practise.


Leave a Reply

You must be logged in to post a comment.

[cache: storing page]