Cable Routing That Changed a Latency Measurement

Two measurements taken minutes apart on the same equipment can disagree, and the usual assumption is that the method is not repeatable. Sometimes that is true, but there is a more interesting possibility: something in the environment changed, and the most common candidate is the cable running past everything else on the desk.

We found exactly this while chasing an inconsistent result, and rerouting one cable removed a discrepancy that had been affecting several measurements. It is a useful reminder that the desk is part of the test rig whether you intended it to be or not.

How can a cable change a measurement?

Through three mechanisms: electrical noise coupled into a signal, physical drag on a device, and radio blockage near a wireless receiver. All three are affected by where cables sit, and all three produce the same symptom, which is results that will not repeat.

The first is the one people underestimate. A data cable running alongside a power cable for a metre can pick up enough interference to cause retries, and retries are irregular delays. The cable still works, which is why the fault is invisible.

None of these mechanisms requires a fault. They are ordinary physical effects that become measurable when you are trying to resolve small differences, and they are why a careful rig separates signal from power as a matter of routine.

What did rerouting actually change?

In our case it removed a cluster of late reports that had been appearing irregularly. The average barely moved, which is why the problem had survived casual checking, but the worst cases improved substantially and the distribution tightened.

That pattern is worth remembering as a diagnostic. Interference does not usually make everything slower, it makes occasional events much slower. Any measurement whose tail improves after a physical change has found a coupling problem.

The size of the improvement surprised us, and it was larger than several hardware changes we had tested in the same session. Physical arrangement is cheap, and it is not a small effect.

Which separations matter most?

Data away from power is the first rule, and it matters more than any other single separation. Power cables carry current that changes with load, and the fields around them couple into nearby signal conductors. A few centimetres is often enough.

Display cables away from audio is the second, because display signalling runs at high rates and can couple into poorly shielded audio wiring. This is the classic cause of a hum that appears only when something is connected.

Wireless receivers away from everything is the third. A dongle surrounded by cables is both electrically noisier and radio-blocked, and moving it into clear space helps on both counts at once.

Does bundling cables make it worse?

Yes, and neatly bundled desks are often worse than messy ones for this reason. Running several cables tightly together for a long distance maximises the coupling between them, which is exactly what you do not want.

Bundling is fine for cables doing similar jobs, such as several power leads, and it is a bad idea for mixed signal types. Separating by type and only tying like with like gives most of the tidiness with none of the coupling.

Where cables must cross, crossing at right angles minimises the length over which they run parallel. That is the standard practice in installations for a reason, and it costs nothing on a desk.

How much does cable quality matter?

It matters mainly through shielding. A well-shielded cable can run near a power lead without picking up much, while an unshielded one cannot. The difference is real and it is visible in a measurement.

Shielding is not always present, and it is not always stated. A cable that works reliably in one arrangement may not in another, which is why arrangement is the more dependable control. Good routing makes even mediocre cables behave.

Connectors matter as much as the cable. A loose or poorly shielded connector breaks the shield at exactly the point where it matters, and a cable with good shielding and a bad connector performs like an unshielded one.

Can physical drag affect a mouse measurement?

Yes, and it is the most underestimated factor in mouse testing. A cable resting on the desk exerts a force that changes as the mouse moves, and that variable force alters the movement itself. The sensor then reports a path that genuinely happened.

This is not a measurement error, which is the point. The mouse really did move differently, because the cable pulled it. Any comparison made with different cable arrangements is comparing two different physical situations.

Supporting the cable removes the variable entirely, and it should be standard practice in any mouse test. A bungee, a hook, or even a loop of tape achieves the same thing, and it makes results repeatable.

Does desk material matter?

It can, through both radio and vibration. A metal desk surface reflects radio and can block the path between a dongle and a device, which is why some setups improve dramatically when the receiver is lifted.

Vibration is the other channel. A desk that transmits machine noise into a mouse pad produces sensor jitter, and a pad on a hollow surface behaves differently from one on a solid one. That shows up in resting jitter measurements.

Neither effect is large on its own, but both are easy to control and both contribute to the kind of variability that makes small measurements unreliable. Standardising the surface is worth doing.

How do you test whether routing is a problem?

Take a baseline measurement, then move one cable and repeat it. If the distribution changes, particularly the tail, routing matters in your setup. Do this with one cable at a time, or you will not know which one was responsible.

The most informative version is to measure a deliberately bad arrangement first. Bundle everything together, measure, then separate it and measure again. The difference gives you an upper bound on how much routing can matter on your desk.

Once you know that bound, you know how much care your measurements need. If the spread is small relative to the effects you study, routing can be ignored, and if it is not, it has to be controlled.

Should a test rig be arranged differently from a desk?

Yes, in one specific way: it should be boring. Fixed cable positions, fixed device positions, no changes between sessions, and nothing resting on the surface that moves. Consistency matters more than any particular arrangement being optimal.

That is the difference between a rig and a desk. A desk optimises for comfort and changes constantly; a rig optimises for repeatability and changes never. Measurements taken on a desk inherit whatever arrangement it happens to be in.

Documenting the arrangement is part of this. A note describing where everything sits makes a measurement reproducible months later, which is the whole point of taking it.

What is the practical advice for a normal desk?

Keep data and power separated where practical, lift wireless receivers into clear space, support the mouse cable, and avoid tight mixed bundles. Those four things take a few minutes and they improve both reliability and measurements.

None of them requires buying anything, which is the appealing part. Cable management is usually sold as tidiness, and the real benefit is that it removes intermittent faults that are otherwise very hard to trace.

Doing it once properly is better than repeatedly moving things. A stable arrangement makes any future problem much easier to isolate, because fewer variables are in play.

What should you conclude?

That the physical arrangement is part of the system being measured, and that inconsistency in results is often a routing problem rather than a method problem. Checking it early saves a great deal of confusion later.

The effect is large enough that it deserves a place in any checklist of things to standardise, alongside port choice and system load. It is cheap, and it removes a category of noise from every measurement you take afterwards.

Where should this lead?

Does a surge protector or extension lead matter?

It can, because everything plugged into one strip shares a path and noise can travel between devices that way. Separating audio and computer equipment across different strips sometimes helps.

It is a low-priority change, worth trying only after the obvious ones. Where it does help, the improvement is usually clear rather than marginal.

Should test cables be permanently separated?

Yes, if you measure regularly. A rig where the routing is fixed and documented produces comparable results months apart, and that is the entire value of keeping one.

It is also easier than re-checking each time. Establish the arrangement once, mark it, and leave it alone.

Is it worth labelling cables?

Yes, because a labelled desk can be restored exactly after anything is moved. That matters more than it sounds, since an undocumented change to the arrangement invalidates any comparison with earlier measurements.

A few labels take minutes and they remove a whole class of mystery from later testing.

Once the desk is stable, the remaining unexplained variability usually belongs to the host system. Measuring with a defined background load is the natural next step, and it isolates the last large source of inconsistency.