Why Two Mice With the Same Sensor Track Differently

Two mice sit on the same pad, both list the same sensor model, and one of them feels noticeably better. The obvious conclusion is that the feeling is imagination, because the sensor is the part that does the measuring and it is identical. The less obvious conclusion is that the sensor is one component in a chain, and the rest of the chain is where the difference lives.

We logged both mice on the same guided rig and the traces were not the same. The average tracking error was close, but the behaviour at speed and during direction changes differed enough to explain the preference. Sensor name tells you the component; it does not tell you the implementation.

What does a sensor model actually specify?

The sensor is a chip with a published resolution range, a maximum speed, and an acceleration ceiling. Those are the conditions under which it will track correctly, and they are measured on the vendor’s own test setup. They say nothing about the lens, the illumination, or the code that reads the chip.

That last part is the gap. The sensor outputs raw frame data, and something has to turn it into a movement report. Two manufacturers can use the same chip and make different choices about filtering, calibration and reporting, and those choices are what reach your hand.

How much does the lens matter?

A great deal. The lens focuses the surface onto the sensor, and its shape and height decide how sharp that image is. A lens that sits slightly too far from the surface produces a softer image, which increases jitter and shortens the speed at which tracking fails.

We saw this directly when two mice with the same chip produced different jitter figures at rest. The difference was not in the silicon; it was in the plastic in front of it. Lens geometry is rarely discussed, and it belongs near the top of any list of causes.

Why does firmware change the same chip?

Because the vendor supplies a reference implementation and each manufacturer adapts it. One may add filtering to reduce noise, another may leave the raw path clean, and a third may add prediction. Each choice produces a different feel from identical hardware.

Firmware also sets the lift-off threshold and the surface calibration behaviour, both of which affect the path during normal use. Two mice sharing a chip can differ in how far they track off the surface, which becomes obvious the moment you lift to reposition.

What did the guided traces actually show?

At moderate speed the two paths were close enough that a casual comparison would call them identical. At high speed one path cut corners slightly and the other followed the guide. During abrupt stops one arrived promptly and the other overshot by a small but repeatable margin.

That is the shape of the disagreement between the two owners. Both were right about what they felt, because they were describing different parts of the speed range. A single-number comparison would have hidden it completely.

Does weight and balance change tracking?

Indirectly, yes. A heavier mouse decelerates differently for the same hand force, which changes the speed profile the sensor sees. A mouse that is poorly balanced can also rock slightly during a lift, which changes the height the sensor experiences.

Neither of those is a sensor property, but both show up in the reported path. That is why two mice with the same chip can feel different for reasons that have nothing to do with measuring movement at all.

How do the feet fit into this?

The feet set the ride height, and the ride height is part of the optical path. Worn feet lower the body, which moves the lens closer to the surface and changes focus. We measured a small but real change in jitter between fresh and worn feet on the same mouse.

Feet also decide how consistently the mouse sits flat. A mouse that rocks even slightly presents a changing height to the sensor, and that variation is indistinguishable from surface noise in the reported data.

Can cable drag masquerade as a tracking difference?

Yes, and it is one of the most common confusions. A stiff cable resists movement unevenly, which produces a path that feels like sensor lag during fast direction changes. We reproduced this with a heavy cable on a mouse that tracked perfectly once the cable was supported.

Before concluding anything about tracking, lift the cable off the desk or use a bungee and repeat the test. A surprising number of reported sensor problems disappear when the cable stops pulling.

What is the right way to compare two mice?

Use the same pad, the same surface calibration state, the same report rate, and the same movement pattern, and compare paths rather than feelings. Feelings are real but they are not comparable across sessions; paths recorded minutes apart are. Where possible, test both within one sitting.

Record at several speeds, not one. A single-speed test would have declared our two mice equivalent, and the disagreement between their owners would have remained unexplained.

Which differences are worth paying for?

Lens quality, firmware maturity and physical stability are worth paying for, because they determine the path at the speeds you actually use. Raw resolution past a certain point is not, because it stops mattering long before the marketing stops mentioning it.

The most valuable thing a vendor can publish is a description of the processing pipeline. A manufacturer willing to say what it filters has usually thought about the trade-offs, and that thinking shows up in the path.

What should you take from this when shopping?

That the sensor name is a starting point, not a verdict. Two implementations of the same chip can differ in the ways that matter most, and only a recorded path will tell you which one you prefer. Reviews that quote the sensor and stop there have not answered the question.

Does firmware maturity explain some of the difference?

Often. A sensor’s first firmware is usually the reference implementation, and later revisions tune filtering, calibration and power behaviour. Two mice using the same chip at different points in its life can behave differently for exactly that reason.

That also means the comparison can change over time. A mouse that measured worse on arrival can improve after a firmware revision, which is why conclusions drawn at launch do not always hold a year later.

Why do reviewers disagree about the same sensor?

Because they test different implementations at different speeds on different surfaces, and because they usually describe feel rather than recording paths. Both sources can be reporting accurately and still appear to contradict each other when their conditions differ.

Looking for the conditions before the conclusion resolves most of it. A review that states the surface, the report rate and the firmware version can be compared with another; one that does not, cannot.

Is there a way to check claims without equipment?

Yes, with the same drawing tests used for smoothing and prediction. A slow circle, a fast flick with an abrupt stop, and a lifted reposition cover the three filters that matter, and each produces an obvious artefact when present.

Doing those three takes under a minute and tells you more about a mouse than any specification. It also makes it much easier to interpret reviews, because you know which of these behaviours you are sensitive to.

What does this mean for buying on a sensor name?

That the name is a hint about capability and not a description of behaviour. It tells you what the chip can do in principle, and the lens, firmware and physical build tell you what this mouse will do on your desk.

Buying on implementation evidence, where it exists, is more reliable than buying on a component name. Where it does not exist, the drawing tests on your own desk are the next best thing, and they are free.

Is the lens something you can inspect?

Yes, and it is worth doing. Removing the shell shows the lens and its distance from the surface, and a lens that sits noticeably proud of the base usually means a softer image. Comparing two mice side by side makes the design difference obvious.

You cannot change it, but knowing it explains behaviour. A mouse with a poorly placed lens will always be noisier at rest, and no firmware update will fix a geometry problem.

Does the pad change the comparison between two mice?

It can reverse it, because sensors respond differently to different surfaces. A mouse that tracks beautifully on cloth may be ordinary on a hard pad, and the reverse happens just as often. Comparing on one surface describes that combination only.

Testing on the surface you own is the only comparison that predicts your own experience. Reviews that use a different pad are describing a situation you will never be in.

The practical next step is to define what speed range you actually work in, because that decides which part of the path matters to you. Measuring your own movement rather than guessing it is the natural continuation.