1
\$\begingroup\$

Some camera use cases (like exoplanet transit photometry) require measuring smaller variations in brightness across an image and over time than the bit depth resolution in 16 bit monochrome cameras can afford (2^16=65536 counts, where the detected variation is just a tiny fraction of that.)

Do (monochrome) camera sensors beyond 16 bits per pixel exist?

\$\endgroup\$
8
  • 2
    \$\begingroup\$ hm stupid question, since I'm not familiar with the sensors for such use cases, but aren't they photon counters, essentially? meaning that more bits doesn't improve the lower end of measurement, just the maximum count per cell? \$\endgroup\$
    – Marcus Müller
    Commented Jun 13, 2023 at 11:52
  • \$\begingroup\$ If a photomultiplier tube can count single photons, quantifying brightness takes on a different perspective. \$\endgroup\$
    – glen_geek
    Commented Jun 13, 2023 at 11:52
  • 2
    \$\begingroup\$ @MarcusMüller Yes they are essentially counting photons and some noise, so more bits means more dynamic range on the high end. High bit cameras are frequently used to measure small signals on top of large offsets (variations in brightness). They are useless (or at least no better) for dim signals that would fit into fewer bits. \$\endgroup\$
    – user1850479
    Commented Jun 13, 2023 at 11:57
  • 1
    \$\begingroup\$ I am not an astronomer but if I wanted to track the brightness of a star over time I would record many sequential frames and then filter the output to generate a plot of photons per second over the interval. In this scheme the (per frame) dynamic range of the ADC would not be a limiting factor. This would have other advantages, including the ability to track the star as the Earth moves and the ability to look at variations at different time scales. \$\endgroup\$
    – user1850479
    Commented Jun 13, 2023 at 12:07
  • 1
    \$\begingroup\$ Astrophotography at that level involves large amounts of post-processing of the data, including sometimes "stacking" dozens or even hundreds of individual exposures in order to increase the overall dynamic range. \$\endgroup\$
    – Dave Tweed
    Commented Jun 13, 2023 at 12:31

1 Answer 1

Reset to default
2
\$\begingroup\$

I looked up the sensors for the PLAnetary Transits and Oscillations of stars (PLATO) space telescope:

https://www.spiedigitallibrary.org/conference-proceedings-of-spie/10564/1056405/The-PLATO-camera/10.1117/12.2309075.full?SSO=1

It uses 32 81x81mm CCD (not CMOS) sensors, each with 4510x4510 18 micron wide pixels. Due to the enormous pixels, the saturation power is nearly 1 million photons, far more than could be counted with a 16 bit sensor.

They split the array up into fast (2.5s) and slow (tens of seconds) exposure sensors, with the former being used for orientation and angular control of the space craft. The latter do long exposure to monitor the brightness of stars.

Interestingly they use 14 bit ADCs for the fast sensors and 16 bit for the long integration time slow sensors. No explanation is given, but presumably they do not require more than 16 bits dynamic range (although they could have chosen more).

Stopping at 16 bit probably makes sense. 96 dB dynamic range is very high for an optical signal given shot noise statistics. Considering shot noise, at saturation the sensor would have a shot noise standard deviation of nearly 1000 photons (sqrt(N)). If the gain of the ADCs was set to 10 photons per DN for an experiment, they could record up to 655350 photons per pixel and still have a quantization step size 100-fold smaller than the shot noise. At the same time, the sensor read noise is more than 20 photons, so their quantization size would also be less than half the minimum sensitivity, so they aren't losing anything at the low end either.

I'm not sure if anyone has ever made a sensor with >16 bits. Probably someone somewhere has, but as you can see it requires truly extreme conditions to make much difference. Often I suspect 16 bit is used because relatively cost effective, not because it will have a significant effect on real images.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.