The new sensor has 1MP resolution and depth accuracy of 10cm at the distances from 10m to 100m. This is an improvement over the previous generation Panasonic APD sensor announced in June 2018 that had an accuracy of 1.5m.
Wednesday, February 19, 2020
Panasonic Develops 1MP APD-based ToF Sensor with 250m Range
Panasonic has developed a TOF image sensor that uses APD pixels and is capable of acquiring highly accurate 3D information at distances up to 250 m. The applications of the new sensor include automotive range imaging and wide-area surveillance.
The new sensor has 1MP resolution and depth accuracy of 10cm at the distances from 10m to 100m. This is an improvement over the previous generation Panasonic APD sensor announced in June 2018 that had an accuracy of 1.5m.
The new sensor has 1MP resolution and depth accuracy of 10cm at the distances from 10m to 100m. This is an improvement over the previous generation Panasonic APD sensor announced in June 2018 that had an accuracy of 1.5m.
1.2 kW (kilo-watt!) pulse and performance only given at night.
ReplyDeleteThis is not that much for a flash LiDAR with 250m range and 1MP resolution. It's about 1.2mW per pixel peak power. The average power is much lower, I'd guess.
DeleteWith 10K gain, this APD acts almost like a SPAD, though it does not go out of order the moment first false photon comes. So, still analog ways of summation, averaging, etc. can be done to improve the overall SNR. Still, 1.2kW is very high for any civilian application at this wavelength.
ReplyDeleteWhat is the frame rate of this, please?
-AA
I'm pretty sure this operates in Geiger mode, so this IS a SPAD. The nomenclature is a bit confusing giving the standard in the community. This is a VSPAD (Vertical SPAD)
DeleteYes indeed, it is a SPAD.
DeleteAny information on the pixel size?
ReplyDelete6 um.
ReplyDelete@laser power comment. To reduce the power of the pulsed light by one order of magnitude I would recommend Panasonic to design another sensor for real LIDAR automotive applications with only 320x160 resolution (just an example but comparable with others). Binning may be considered but making it simple and cheap would be preferable. This would be 20x less power requirement and 20x increase in frame rate reading. If the reading noise would be less than 5 e- rms and a QE of 70% at 905nm would be achievable, that would be a winner among ALL LIDAR sensors.
ReplyDelete@at night comment: if the integration is not happening during 1.67us (that is the duration of the light to bounce back from the 250m distance) but with Reset/TG window of 5-25ns, the ambient light noise would be reduced by a factor of 1670ns/5ns=360 ! Hopefully this is enough to make it work in plain light conditions. Of course, optical filtering (BP) at 905nm would be required, no question about it.
Comparing with other LIDARs: no mechanical or MEMS mirrors, low pulsed light power, rejection of ambient noise; the technique to integrate (analog) several pulsed light sessions in the same bin is excellent for long distances comparing with Espros for example, which cannot do this technique because the shift register should move another position every 4-5ns.