Question about crosstalk

[Tian JIN]

Hello,

I’m using a DVXplorer Micro to image fluorescent liquid (excitation 488 nm, emission 540 nm) flowing through a 1 mm-diameter tube. Ideally, the camera should register events only from the liquid inside the tube, but I’m seeing very strong ON events in pixels outside the tube as well. Could this be due to pixel crosstalk? 

I accumulated all ON and OFF events within a 2-ms window to form a single frame; the plotted curve is the total ON-event count per 2-ms frame. The FOV is 2.0 × 2.6 mm at 480 × 640 pixels.

Thanks for your support!

Tian

[Tobi Delbruck (INI)]

Is the lighting constant? Have you played with biases?  If the tube
outside is dark, there will be a lot of shot noise (ON+OFF events). But
it doesn't sound like that.

It is critical to understand the effect of biases on threshold,
bandwidth and noise.

[Tian JIN]

Thanks!

The excitation light is constant. The fluorescent liquid only flows inside the tube, so the tube outside should be dark. I also saw the shot noise while the excitation was off, but this strong signal outside tube was stronger than that. 

Regarding the biases, I tried to adjust "contrast threshold off" and "contrast threshold on" to both 9 or both 15, but there is no obvious difference. The strong signal from tube outside still exists. 

I attached a GIF for your reference. The "contrast threshold off" and "contrast threshold on" were 15. 

Thank you again!

[Tobi Delbruck (INI)]

That video is very revealing, thanks.

Either the particles are emitting or reflecting light on the dark
surroundings and cause brightness changes there (the constrast could be
very high since background is dim to start with), or there is some
electrical coupling in the array that causes pixels that are exposed to
dark illumination and are in a high impedance state to be very sensitive
to array electrical coupling.

Can you change the background to be lighter? Do you have a photo of the
setup under actual operating conditions? It is impossible the see the DC
level in DVS recording except by inferring it from dynamics or noise.

[Tobi Delbruck (INI)]

you can turn on frames to at least capture an image, albeit not very
great one.

[Tian JIN]

Hi Tobi, thanks!

Figure 1 is the photo of the set up. The work distance between the last lens and the sample is ~4 mm. We used a fiber to emit 488 nm laser to the sample, and detect the fluorescent light after a filter. Figure 2 and 3 are the lenses we used and the simulation. 

Theoretically, only fluorescence light can pass through the filter and reach the sensor. During the experiment, we will turn off the room lights to ensure that no ambient light enters the camera.

Do you suggest we give a bit room light to make background lighter?

Thank you! 

[Tobi Delbruck (INI)]

It looks like a nice setup, quite practical and with cool optics modeling.

Can you put a piece of flat black paper, or better still, black felt mat behind the tube? That will absorb the scattered light quite well. At least then you can see if it is scattered light from your particles that is creating those events.

It might be an example where a future CSDVS might help, since you care about spatiotemporal response, not just temporal -- you want to see the moving edges, not the diffuse changes of lighting.