[HN Gopher] SoDaCam: Software-Defined Cameras via Single-Photon ...
___________________________________________________________________
SoDaCam: Software-Defined Cameras via Single-Photon Imaging
Author : PaulHoule
Score : 41 points
Date : 2023-09-10 14:58 UTC (8 hours ago)
(HTM) web link (arxiv.org)
(TXT) w3m dump (arxiv.org)
| rgovostes wrote:
| Better page: https://wisionlab.com/project/sodacam/
| ooterness wrote:
| This project is amazing. There's a video about halfway down the
| project that gives a good summary.
|
| They demo dynamic adjustment of frame rate, i.e., by choosing
| the time internal for summation of each output pixel. But for
| me, the really interesting part is when they start summation
| over more complex trajectories. For example, they know a
| falling object in the video has a certain initial velocity and
| acceleration, so they can completely remove motion blur.
|
| It would not surprise me if next-gen ultra-high-speed cameras
| start using this tech.
| wcerfgba wrote:
| My understanding is, CCDs accumulate charge in response to
| photons, whereas these diodes don't, is that correct?
|
| If so, what are the current limits of these diodes which stop us
| from manufacturing cameras with these diode arrays instead of
| CCDs?
| DoctorOetker wrote:
| What is your real question?
|
| CCD is a technology, a strategy or approach to a specific
| problem if you wish.
|
| diodes are semiconductors, with a positive and negatively doped
| region (optionally an intrinsic region in between).
|
| A CCD also relies on a diode semiconductor bandgap to liberate
| charge carriers when a photon exceeding the bandgap energy
| strikes.
|
| What is your actual question?
|
| Is it "Why don't we use SPAD's in our cell phones and web
| cams?" ? Thats like "I don't know what it is, but I want it
| too".
|
| You can't use these in brightly lit scenes, the lighting needs
| to dim enough in order to register electrical pulses due to
| individual photons. Instead of a single electron-hole pair,
| SPAD's are reverse biased beyond "breakdown voltage" such that
| the released charge carrier will in turn release others (the
| "A"valanche in SPAD). The breakdown voltage and reverse bias
| (together with a dark current) means the sensor will heat up.
| So it would need significant cooling and energy consumption
| (higher currents per photon at higher breakdown voltage) at
| normal daytime light intensities. The sheer rate of photons
| would not allow individual photon pulses to be discriminated as
| their pulses would overlap... The reason it's not in your phone
| is because it would be expensive, power hungry (decreasing
| battery life) and require cooling.
| mnode wrote:
| This explanation isn't consistent with my understanding. The
| key reasons are that SPADs have lower quantum efficiency than
| conventional CCDs and typical SPAD arrays have fewer pixels.
| Both are improving, but still not competitive for most use
| cases. Heating is an issue for data transfer, but as the
| video shows this can be addressed by on-chip processing.
| Y_Y wrote:
| What is this answer? I'm very familiar with the devices in
| question but I can't understand why you gave this response.
|
| CCDs count photons (in large numbers), SPADs work hard to
| register single photons. You can make a camera based on
| either one, depending on your requirements.
___________________________________________________________________
(page generated 2023-09-10 23:01 UTC)