Change contrast threshold

330 views
Skip to first unread message

C S Darshan Bhandari

unread,
May 11, 2022, 4:55:19 AM5/11/22
to davis...@googlegroups.com
Hello Group,

I have a Davis 346 mono. I was wondering what the temporal contrast threshold is, by default, and whether I can control it using software or is it fixed.

The spec sheet says that the dynamic range is 120 dB and contrast sensitivity is 14.3% (ON) and 22.5% (OFF) with 50% pixels responding. So, an event will be triggered when the brightness difference at a pixel crosses this threshold. Now, is this threshold fixed by hardware? Or do I have the option to change it via software somehow? Also, is there an option to display the current contrast threshold in the DV software?

Thanking you

Regards
Darshan


Luca Longinotti

unread,
May 11, 2022, 6:19:35 AM5/11/22
to davis...@googlegroups.com
The contrast thresholds for events to trigger can be influenced on
DAVIS346 by adjusting its biases. This is a very finicky process. Some
documentation is available here: 

https://gitlab.com/inivation/inivation-docs/-/blob/master/Advanced%20configurations/User_guide_-_Biasing.md#dynamic-vision-sensor-biases

You can change these settings in DV, open the Capture module's advanced
configuration (+ button in right hand menu) and search for the bias
names, you'll be presented with the 'fine value' and 'coarse value'
sliders, which you can use to modify them.

There is no specific display of contrast threshold value in percent.
--
Luca Longinotti (llongi)

Head of Embedded Platforms
iniVation AG - https://inivation.com/
Zurich, Switzerland
Office: +41 44 500 32 14

Tobi Delbruck (UZH-ETH)

unread,
May 16, 2022, 8:10:11 AM5/16/22
to davis...@googlegroups.com

If you run a DAVIS with jAER, you get a display of the estimated contrast threshold computed from the bias currents that are estimated to be generated by the on-chip bias generator bias current ratios, according to Nozaki, Y., and T. Delbruck. 2017. “Temperature and Parasitic Photocurrent Effects in Dynamic Vision Sensors.” IEEE Transactions on Electron Devices 64 (8): 3239–45. https://doi.org/10.1109/TED.2017.2717848. See the jAER users guide section https://docs.google.com/document/d/1fb7VA8tdoxuYqZfrPfT46_wiT1isQZwTHgX8O22dJ0Q/edit#heading=h.egpfaofkar9d (for rather out of date information) or https://www.youtube.com/watch?v=xKv-RZXcCHo&t=317s for a video. Both of these lack the latest estimated threshold computations however. I will try to update the user guide.

However, I recommend to use DV because it is faster and more supported now and writes much better aedat4 data files than jAER. Perhaps you could contribute a pull for a user friendly control panel for DV which would be very useful for the community who do not want to know the detailed pixel circuit.



Reply all
Reply to author
Forward
0 new messages