Wikipedia:Reference desk/Archives/Science/2018 November 8

= November 8 =

How big does a camera pixel need to be for the small pixel width to visible light λ ratio to have negligible effect?
How do I calculate how many photons per pixel at native ISO and (impossible) 100% efficiency from stuff like luminance of what the pixel's photographing, focal length, pixel and sensor size, and f-stop? Sagittarian Milky Way (talk) 18:12, 8 November 2018 (UTC)

How do you use the photons per pixel and number of pixels to calculate the chance that a hypothetical photo where every pixel would be exactly between 127,127,127 and 128,128,128 in 8-bit per channel RGB if shot noise didn't exist would still not contain numbers besides 127 and 128 even though light is quantized and the time between photons is only an average? Sagittarian Milky Way (talk) 18:55, 8 November 2018 (UTC)


 * Sukhwan's paper was the first search-result! Characterization of Noise in Digital Photographs for Image Processing (HP Labs, 2008).
 * If you're wholly unfamiliar with photoelectric photoconversion, a good place to start reading is our articles on the photoelectric effect and quantum efficiency.
 * Your answer will require quantum mechanical statistics. All we can measure is whether an electron is actually ejected from a pixel's photoactive area; and we have to infer whether that electron was ejected by a "valid" incident photon, or if it was ejected by one of the many other sources of noise that can cause an electron ejection.  We never actually know whether one individual photon was "valid" - we can only describe the aggregate statistical properties of the incident radiation.
 * In practice, we measure the approximate value in a well-calibrated laboratory, and we don't usually spend much time to describe it with an equation - because we already know that the equation to calculate that value would be really complicated, and it would depend on a lot of details we'd have to measure in a lab anyway.
 * Nimur (talk) 21:04, 8 November 2018 (UTC)


 * Cant be answered because Image sensor technology like sensor technology in general is in constant development and takes impressive huge leaps every year. Some companies just released 48 megapixel cameras in Smartphones. Just a few years ago you had to buy a chunky and very expensive digital camera with a huge sensor chip inside to get such resolutions. Today its all in a pin head on a circuit board so tiny, you need calipers to place it correctly in its little spot in these smartphones. In 10 years they can probably place nano cameras with battery and transmitter on dragonflies to follow them "live". I think i read somewhere some secret services actually already use such Insect-drones for high profile infiltration and surveillance. --Kharon (talk) 03:26, 10 November 2018 (UTC)


 * P.S.: Better watch our for them suspicious flies that keep following you :D --Kharon (talk) 03:35, 10 November 2018 (UTC)


 * Two related issues: first, you only know how many pixels your camera has. unless you take it apart you don't know whether the actual image sensor is huge with large pixels or small with tiny pixels. The second issue is the size of the lens. The small diameter, limited depth lens typical of smartphones simply cannot get as good a picture as a digital camera of the same quality with a bigger lens. The laws of physics make it so that super thin smartphone equals poorer-quality camera and tinny sound. --Guy Macon (talk) 04:04, 10 November 2018 (UTC)


 * Well I don't think you have to take it apart. For higher end smart phones from reputable brands, it's not that uncommon the sensor is known e.g. the global and Chinese version of the Xiaomi Redmi Note 5 [//www.gsmarena.com/xiaomi_redmi_note_5_ai_dc-review-1777.php] which for developed work income levels is probably mid range or low mid range comes [//www.gsmarena.com/xiaomi_redmi_note_5_ai_dc-review-1777.php] with the Samsung S5K2L7 [//www.samsung.com/semiconductor/image-sensor/mobile-image-sensor/S5K2L7/] and S5K5E8 [//www.samsung.com/semiconductor/image-sensor/mobile-image-sensor/S5K5E8/] sensors on the rear. While okay Xiaomi can be confusing with various models with different specs but the same name in different locations, generally the info out there. Crappy Chinese brands often publish completely fake info, but the better ones are generally better at that. And of course other non Chinese brands like LG, Samsung etc are likewise often reliable at least for their mid range and higher phones. In many developed countries, publishing such specs carries a legal risk if false, and the precise model of the image sensor is probably possible to check unlike say storage capacity of batteries so for reputable stores and reputable brands there tends to be a reasonable incentive not to lie. As for the image sensor specs, well that's more complicated, still for a reputable brand I think you can resonable trust the image sensor size specs amongst other things because again it's something easy to check. Nil Einne (talk) 12:32, 10 November 2018 (UTC)
 * These are easier to find for camera cameras. 36x24mm for cameras advertised as full frame, smaller like APS-C for cheaper cameras. The lens opening is well advertised (focal length divided by f/number), angle of view has the photo diagonal in degrees for each focal length. Some of the light hits outside the sensor corner though, you can't have vignetting and poor corner image quality in a decent camera. Sagittarian Milky Way (talk) 13:23, 10 November 2018 (UTC)

1) the CCD charge wells hold 1000s of electrons for pixels of reasonable size, so the 8-bit number is the result of A-D conversion. 2) If it's RGB that means you'll lose some photons to the Bayer filter but you might want to ignore that too. 3) Maybe what you're really asking is how to compute the number of photons from the luminance. Answer: look at luminous efficiency to find some parameters to convert lumens to watts.  Then find the photon energy by multiplying the photon frequency (which depends on the color) by Planck's constant.  Make sure to get all the units consistent and it should just be arithmetic from there.  173.228.123.166 (talk) 06:41, 11 November 2018 (UTC)


 * Actually this is a pure statistics problem. You will have to establish the distribution of response to constant illumination, and then see how likely it is that all pixels have the same value. Graeme Bartlett (talk) 07:17, 17 November 2018 (UTC)