Sensor

As mentioned earlier, the name of the sensor is "Owl-Eye optical sensor", which is Roccat's name for the PixArt 3361, a custom version of the 3360.


Here's a close-up of the sensor and the microcontroller.

The sensor behaves like you would expect a top-end sensor to perform: perfectly. The tracking feels very smooth and direct and adjusts to small movements easily - even very fast flick shots can't make it malfunction.

Performance

To showcase the performance, I tested the mouse at various CPI steps in MouseTester and tried to move the mouse as fast as possible across the area of my mousepad. The result is excellent, showing no signs of tracking loss even at 5.5 m/s. Roccat claims the sensor to be capable of tracking at up to 6.35 m/s and probably even faster on a good surface. So it's fair to say that it's fast enough for any human to never worry about exceeding tracking speeds.

CPI accuracy

The CPI steps are also very close to the proclaimed value set in the software. There usually is a slight variance to CPI with all mice (i.e. you set 400 CPI, but it really is something between 390-410); however, the more accurate the better. I perform this test because in the past, certain mice had a very high divergence at their higher settings, where 6000 CPI would really be 4900, for example. So the CPI accuracy you see below is actually very good and most of all consistent.


As you can see, the CPI steps are all a bit above their proclaimed values; however, they are very consistent in that you get used to it quickly anyway. So here, everything looks perfectly fine.

Speed related accuracy variance

With this test, I show how accurately the sensor reports movement at different speeds. You can see me move the mouse at varying speeds from a starting position and back again at a different speed. In the top-right corner is the showpos command in the Source engine. The second row shows the viewing angle, of which my point of reference is mainly the second one, the x-axis angle. With a script, I set my viewing angle to 0 0 0 at the start. After every swipe back to my starting position, I record the deviation from the starting position.

There is obviously some degree of error here due to me performing this test manually because I lack a test bench that is able to limit movement to the x-axis only. I have done this same test with a variety of different mice and sensors, so it's possible to compare. In this test, the angular displacements were 0.53, -0.53, -0.95, -1.45, 2.55, -2.51, -0.11, -1.68, -1.52, -0.27, which is actually a very good result and as expected for a 3360 sensor. The biggest outliers are in the test where I move to the right slowly and back again quickly; my fast return movement is really inaccurate because I can push away the keyboard I use as a stop if done with enough speed.

Lift off distance

The lift-off distance (LOD) can be set to either "high" or "low" in the software. The classic way to repeatably measure LOD is to see at how many stacked CDs a mouse still tracks. In general, a low LOD is preferred as the sensor really only tracks accurately if the surface it tracks is at the optimal distance. However, some people tend to hover their mice at the end of swipes and feel uncomfortable with a lower setting.

The Kone Pure Owl-Eye has less than one disc of LOD in its low setting. With the higher option for LOD, it seems to be at just below one CD since I get tiny jittery cursor reactions with two CDs. So people from both worlds should be happy with these LOD options. The standard setting out of the box is low LOD.