Questions about pixel behavior in DAVIS346

[Shota Sotome]

Dear all,

I'm working on research on visible light communication using an event camera (DAVIS346).  I’m trying to capture a blinking LED with DAVIS346 and use the obtained event data to estimate the timing when the LED turned on or off.

Fig. 1 shows some of the actual event data. Only events detected in a single pixel were extracted. The horizontal axis is the timestamp (subtracted to start at 0).

The experimental conditions are as follows:

• The LED was driven by a square wave with a duty ratio of 50% at 1 kHz.
• Distance between the LED and DAVIS346 is 1 m.
• Regarding biases of DAVIS346, the ON and OFF thresholds were adjusted so that ON and OFF events are detected equally.
• Event rate was about 30 kEPS.
• I’m using DV and Export_csv module to output the data. The OS is Windows 11.

As can be seen from Fig. 1, ON and OFF events alternate at intervals of approximately 500 us, capturing the blinking of an LED.

Here, I have questions about the behavior of DAVIS346. First, sometimes events are not detected. Fig. 2 shows another portion of the same data as Fig. 1. There are periods when no events are detected. Why does this happen?

Second question. I reduced the LED frequency to 100 Hz and captured the LED. All other conditions are the same as in Fig. 1. The result is shown in Fig. 3. Since the LED frequency is 100 Hz and the duty ratio is 50%, I expected that ON and OFF events would alternate at 5000 us intervals. However, this is not the case, and we can see that there are more ON events and they are detected for a longer period. I don't understand why it works like this.

Regards,

Shota