Sensor and Performance

The ThundeRobot ML903 NearLink is equipped with the PixArt PAW3395. According to specifications, the 3395 is capable of up to 26,000 CPI, as well as a maximum tracking speed of 650 IPS, which equals 16.51 m/s. Out of the box, four pre-defined CPI steps are available: 800, 1200, 1600, and 3200.

All testing was done on the latest firmware (1.0.39.0/1.39.0.0). As such, results obtained on earlier firmware versions may differ from those presented hereafter.

CPI Accuracy

"CPI" (short for counts per inch) describes the number of counts registered by the mouse if it is moved exactly an inch. There are several factors (firmware, mounting height of the sensor not meeting specifications, mouse feet thickness, mousing surface, among others) which may contribute to nominal CPI not matching actual CPI. It is impossible to always achieve a perfect match, but ideally, nominal and actual CPI should differ as little as possible. In this test, I'm determining whether this is the case or not. However, please keep in mind that said variance will still differ from unit to unit, so your mileage may vary.


I've restricted my testing to the four most common CPI steps, which are 400, 800, 1600, and 3200. As you can see, deviation is consistently positive and very low, which is an excellent result. Despite the sensor supporting increments of 50, CPI adjustment is limited to multiples of 100.

Motion Delay

"Motion delay" encompasses all kinds of sensor lag. Any further sources of input delay will not be recorded in this test. The main thing I'll be looking for in this test is sensor smoothing, which describes an averaging of motion data across several capture frames, in order to reduce jitter at higher CPI values, increasing motion delay along with it. The goal here is to have as little smoothing as possible. As there is no way to accurately measure motion delay absolutely without special equipment, it is done by comparison with a control subject that has been determined to have consistent and low motion delay. In this case, the control subject is a Logitech G403, whose PMW3366 sensor has no visible smoothing across the entire CPI range. Note that the G403 is moved first and thus receives a slight head start.

Testing is restricted to NearLink as Bluetooth is not suitable for non-casual gaming applications.

Wired testing

First, I'm looking at two xCounts plots—generated at 1600 and 26,000 CPI—to quickly gauge whether there is any smoothing, which would be indicated by any visible "kinks." The second plot shows very noticeable kinks, which strongly suggests there being any smoothing. In fact, even the first plot shows several kinks, though these likely aren't related to smoothing but are causally related to polling instability instead. SPI timing is low owing to MotionSync, which is permanently enabled.


In order to determine motion delay, I'm looking at xSum plots generated at 1600 and 26,000 CPI. The line further to the left denotes the sensor with less motion delay. Asserting motion delay is complicated by the fact that the curve is non-linear. At the onset of motion, the first count is delayed by close to 15 ms, which continually decreases throughout the motion until reaching a minimum. As such, giving a plain number is not possible.

Wireless testing

While 1600 CPI appears rather normal, even though the degree of SPI timing jitter is unusual given that MotionSync is permanently enabled, 26,000 CPI shows a malfunction, which is likely due to a 12-bit HID descriptor. As such, if a report contains more than 2047 counts (+/-), a malfunction is the consequence.


Once again, 1600 and 26,000 CPI are tested. Unlike in wired mode, the curve is typically linear in wireless mode. At 1600 CPI, a motion delay differential of roughly 2 ms can be measured. At 26,000 CPI, this increases to 16 ms, which again is due to smoothing.

Speed-related Accuracy Variance (SRAV)

What people typically mean when they talk about "acceleration" is speed-related accuracy variance (or SRAV). It's not about the mouse having a set amount of inherent positive or negative acceleration, but about the cursor not traveling the same distance if the mouse is moved the same physical distance at different speeds. The easiest way to test this is by comparison with a control subject that is known to have very low SRAV, which in this case is the G403. As you can see from the plot, no displacement between the two cursor paths can be observed, which confirms that SRAV is very low.

Perfect Control Speed

Perfect Control Speed (or PCS for short) is the maximum speed up to which the mouse and its sensor can be moved without the sensor malfunctioning in any way. I've only managed to hit a measly 5 m/s, which is within the proclaimed PCS range and causes no observable sensor malfunction. That said, if a report contains more than 2047 counts (+/-), a malfunction is the consequence due to wireless operation being limited to a 12-bit HID descriptor.

Polling Rate Stability

Considering the ML903 NearLink is usable as a wired mouse as well, I'll be testing polling rate stability for both wired and wireless use.

Wired testing

Only 1000 Hz is available, which shows periodic outliers.

Wireless testing
For wired mice, polling rate stability merely concerns the wired connection between the mouse (SPI communication) and USB. For wireless mice, another device that needs to be kept in sync between the first two is added to the mix: the wireless dongle/wireless receiver. I'm unable to measure all stages of the entire end-to-end signal chain individually, so testing polling-rate stability at the endpoint (the USB) has to suffice here.


First, I'm testing whether SPI, wireless, and USB communication are synchronized. Any of these being out of sync would be indicated by at least one 2 ms report, which would be the result of any desynchronization drift accumulated over time. Several 2 ms polls are visible, which may point towards a desynchronization drift.


Second, I'm testing the general polling-rate stability of the individual polling rates in wireless mode. Running the ML903 NearLink at a lower polling rate can have the benefit of extending battery life. Whereas 500 Hz looks and performs fine, 1000 Hz shows periodic off-period polls.

Paint Test

This test is used to indicate any potential issues with angle snapping (non-native straightening of linear motion) and jitter, along with any sensor lens rattle. As you can see, no issues with angle snapping can be observed. There is no jitter visible at 1600 CPI. Even at 10,000 and 26,000 CPI jitter is minor, which is expected considering the massive amount of smoothing applied at those steps. Lastly, there is no lens movement.

Lift-off Distance

The ML903 NearLink offers two pre-defined LOD levels. At the "1 mm" setting, the sensor does not track at a height of one DVD (<1.2 mm). Using the "2 mm" setting, the sensor does track at a height of one DVD (1.2 mm<x<2.4 mm, with x being LOD height), but not at a height of two DVDs. Keep in mind that LOD may vary slightly depending on the mousing surface (pad) it is being used on.

Click Latency

In most computer mice, debouncing is required to avoid double clicks, slam-clicks, or other unintended effects of switch bouncing. Debouncing typically adds a delay, which, along with any potential processing delay, shall be referred to as click latency. In order to measure click latency, the mouse has been interfaced with an NVIDIA LDAT (Latency Display Analysis Tool). Many thanks go to NVIDIA for providing an LDAT device. More specifically, the LDAT measures the time between the electrical activation of the left main button and the OS receiving the button-down message. Unless noted otherwise, the values presented in the graph refer to the lowest click latency possible on the mouse in question. If a comparison mouse is capable of both wired and wireless operation, only the result for wireless (2.4 GHz) operation will be listed. All the listed values have been gathered without any sensor motion. If latency differs between no motion and motion, it will be noted as such.

In wired mode and at a debounce time of 2 ms, click latency has been measured to be 2.6 ms, with standard deviation being 0.06 ms. In wired mode and at a debounce time of 4 ms, click latency has been measured to be 4.7 ms, with standard deviation being 0.06 ms. In wired mode and at a debounce time of 8 ms, click latency has been measured to be 9.0 ms, with standard deviation being 0.07 ms.

In wireless mode, at a polling rate of 1000 Hz, and at a debounce time of 2 ms, click latency has been measured to be 4.6 ms, with standard deviation being 0.27 ms. In wireless mode, at a polling rate of 2000 Hz, and at a debounce time of 2 ms, click latency has been measured to be 4.1 ms, with standard deviation being 0.16 ms. In wireless mode, at a polling rate of 4000 Hz, and at a debounce time of 2 ms, click latency has been measured to be 3.4 ms, with standard deviation being 0.10 ms.

The main button switches were measured to be running at 1.80 to 1.83 V. I'm not aware of the voltage specifications of the used Kailh switches, but this voltage seems rather low to me.