- Your digital camera and lens collect photons (light particles) and convert these to photoelectrons in the sensor.
- These photoelectrons are the signal we want to record for each pixel.
- Photoelectrons are stored in each pixel on the sensor, then transferred to the electronic system of the camera where they are amplified (ISO), and then digitized (ISO), and finally stored in an array of integer numbers on the memory card.
- Photons from the scene arrive at random times. There is an inherent error (noise) in counting these photons. This shot noise, or photon noise, is natural (physics) and has nothing to do with your sensor or camera model.
- Signal-to-Noise ratio (SNR) is a good measure of image quality relating to noise or grain. Signal is the number of photoelectrons and the noise is comprised of shot noise, and readout noise from the electronic system of the camera.
- Dynamic range is the ratio of the highest possible photoelectron signal to the noise floor, which contains sensor readout noise, camera processing noise and the dark current shot noise. Dynamic range tells us how well the camera can reproduce the brightest and darkest parts of an image, and the gradations in between.
- When the number of photons per second illuminating the sensor is high (lots of light) the dominant noise is photon noise. In this realm the SNR increases with the square root of the number of photoelectrons. Quadrupling the number of photons doubles the SNR. For example reducing your shutter speed from 1/1000 sec to 1/250 sec will result in twice the SNR.
- When the light available is low the challenge is in collecting enough photoelectrons relative to readout noise electrons (i.e. low SNR). It is in these situations where we really need to understand how the digital camera works and what is the best strategy.
- The ISO setting on your camera does two things (learn more here):
- Amplifies the post sensor electronic signal before the analog-to-digital converter (ADC) that digitizes that analog signal for each pixel.
- It changes the digitization range, the number of photoelectrons to assign to the maximum digital value.
- ISO does not affect the true exposure and it does not affect the sensitivity (electrons per photon).
- There are two important ranges of ISO and noise
- Camera electronics limited region (ISO < 1600 on most cameras )
- In this region the ISO can amplify noise electrons that are not related to the photoelectrons.
- In this region we presumably have a high enough SNR due to high numbers of photons that the main source of noise is from camera electronics.
- Sensor limited region (ISO > 1600 on most cameras)
- In this region we are amplifying the photon noise, since we presumably have less light.
- The photon noise begins to dwarf the read-out noise.
- Camera electronics limited region (ISO < 1600 on most cameras )
Photoelectrons are generated inside each pixel, stored there, then extracted as a current, which is then amplified, and then digitized to an integer signal level that is recorded on your memory card. This is done for each pixel on the sensor. Image from https://www.ptgrey.com/white-paper/id/10912 .
The dynamic range versus ISO is shown here for some representative Canon and Nikon cameras. The plateau on the Canon curves for ISO < 800 shows that the post-sensor electronic noise is the limiting noise source in this low ISO range. The Nikons are doing much better in the low ISO region. Once ISO 1600 is reached the main noise source is photon noise and as we increase the ISO we are amplifying the signal and photon noise, while also reducing the digitization range of the number of photoelectrons — thereby reducing dynamic range (and SNR) as the ISO is increased. This shows that it does not really help to raise the ISO above 1600. In this region it maybe best to take the images at ISO 1600 and then increase exposure, if needed, in post processing using a curves adjustment. From https://luminous-landscape.com/dxomark-camera-sensor/
Field Technique to Mitigate Noise Issues
Confused by all this talk about electrons, photoelectrons, photons and noise?
Understanding a few of these key points on noise goes a long way.
Here are some rules of thumb.
- First set your shutter speed and aperture for the specific situation, balancing motion blur and depth of field with the amount of light available.
- In low light situations ask yourself which is more important, lack of noise or avoiding motion blur? Sometimes it is best to de-noise the image in post-processing.
- If you have to keep a high shutter speed, then open up the aperture, including all the way if needed. Focus on the face of the moving wildlife and let the rest of the animal field blur (low depth of field).
- Is there a away to use a much slower shutter speed and minimize motion blur? (i.e. stabilize the camera on tripod or make shift rest, take lots of images in burst mode, and cherry pick the images where the subject was still, discard all the others.)
- Take some test images and check the histogram. If they are under exposed (histogram bunched on the left), then do not increase the ISO, instead increase the true exposure (lower the shutter speed and/or open up the aperture).
- Always remember that increasing ISO does not increase the amount of light falling on the sensor. If we increase the ISO on an underexposed image we will amplify noise and we will digitize a smaller range of light — resulting in a noisy image.
- For very long exposures (i.e. nightscapes) thermal noise on the sensor becomes problematic. Give the sensor time to cool between frames, keep the ISO at or below ISO 1600 and improve the image in post processing.
- De-noise software works best in uniform areas of the image (i.e. clear sky, water surfaces) that lack detail. It smoothes out noise by averaging over local pixels, which can also reduce detail you want to retain (i.e. fur, feathers, foliage, cloud details, tree bark, grass, or other fine details).
Canon 1DX with EF 600mm f/4L II + 1.4x III, manual exposure 1/250 sec, f/5.6, ISO 4000. Difficult lighting in the pre-dawn light in Alaska.
An Example from the Field
The photo of the brown bear family above is a good example of how one’s excitement over a photo opportunity can interfere with proper technique.
Up early in the field at 4:30AM in Katmai National Park, Alaska I was hoping to photograph a bear backlit by the sunrise. As I waited for sunrise I began to photograph a bear family in the very low light of dawn. I was using a long lens (840mm) and the bears were about 200 meters away. I opened up my aperture all the way since I had enough depth of field at this distance. Normally, with mid-day light, I would be able to use a fast shutter speed of ~ 1/1250 second for a situation like this. In the low light I lowered my shutter speed to 1/250 second. I then increased the ISO to 4000 as I spot metered off the mother bear. The image came out very noisy. Why?
My camera sensor was photon starved. Increasing the ISO to 4000 should have been a warning sign to me. If I instead increased it to ISO 1600, and lowered the shutter speed to ~ 1/1oo second I may have done better. Sure, the bear would be moving some, however I could fire away in burst mode at 14 frames per second and cherry pick the frames where the bear was still (this does work..). The lower shutter speed would let in more than double the photons at 1/250 sec. In that case I would be amplifying more photoelectrons relative to noise electrons (since I have a higher signal to noise due to higher photon statistics, and the number of signal electrons would be higher relative to read noise electrons).
Why not increase the ISO to above 1600? At this level of ISO I have to worry most about amplifying photon noise and less about read out noise. Increasing the ISO above 1600 would mean that I would also be digitizing a smaller range of photoelectrons, resulting in more digitization artifacts (fewer shades of gray resulting in a grainier image). If the image at at ISO 1600 was too dark I could improve that in post-processing (not by changing the exposure or levels adjustments, but instead by keeping the black point and white point of the image fixed and using the curves adjustment to lighten the image).
Hopefully next time in a situation like this I will remember to take some test images, open up the true exposure to move the histogram off the left axis, and then refrain from increasing the ISO above 1600. If the shutter speed required to increase the true exposure is simply way too slow (e.g. < 1/60 second) then I would set the shutter speed to the lowest I am comfortable with avoiding motion blur (in this case ~ 1 / 100 sec) and take some images at ISO 1600 (hoping to fix them in post processing) and also some images at higher ISO (hoping to use reduce the noise in post processing). I applied noise reduction in post-processing to the noisy bear family image. See the image slider below to compare before and after noise reduction software is applied. [/av_textblock] [av_textblock textblock_styling_align='' textblock_styling='' textblock_styling_gap='' textblock_styling_mobile='' size='' av-medium-font-size='' av-small-font-size='' av-mini-font-size='' font_color='' color='' id='' custom_class='' template_class='' av_uid='av-96xd3o' sc_version='1.0' admin_preview_bg=''] [bafg id="20106"] [/av_textblock] [av_textblock size='' font_color='' color='' av-medium-font-size='' av-small-font-size='' av-mini-font-size='' admin_preview_bg='' av_uid='av-83lolg'] Canon 1DX, EF 600mm f/4L USM II + 1.4x III at 1/200 sec, f/5.6, ISO 4000. Noisy image due to high ISO and low signal level (underexposed). The high ISO here is amplifying the noise from the sensor and a small signal from the sensor, and it is also setting the range of the ADC to a low signal level (i.e. the level of photoelectrons is small compared to pixel's full well capacity). This image was taken in very dark dawn conditions where it was photon starved. Note how the noise reduction software does a good job in the areas where the detail is less important, however the bear's fur is softened (not as sharp in the image) after noise-reduction software is applied. The noise also degrades sharpness in the fur before any noise reduction is done in post processing. Moral of the story for this image is that the true exposure was way too low. Since the aperture was fully open, the better thing would have been to ask the bear to sit still, use a tripod and set the shutter speed to ~ 1/100 second, and the ISO to ~ 1600. Next time... [/av_textblock] [/av_cell_one_full][/av_layout_row][av_layout_row border='' min_height_percent='' min_height='0' color='main_color' mobile='av-flex-cells' id='' av_element_hidden_in_editor='0' mobile_breaking='' av-desktop-hide='' av-medium-hide='' av-small-hide='' av-mini-hide='' av_uid='av-n66sk'] [av_cell_one_full vertical_align='top' padding='30px' background_color='' src='' background_attachment='scroll' background_position='top left' background_repeat='no-repeat' mobile_display='' av_uid='av-417u38'] [av_textblock size='' font_color='' color='' av-medium-font-size='' av-small-font-size='' av-mini-font-size='' admin_preview_bg='' av_uid='av-2wgpqc']
Further Reading – Keep Learning
I encourage you to read more about digital cameras and sensors. There is a lot of misinformation on the internet and common misunderstandings. It is easy to make generalizations, looking for “simple answers” without really understanding the physics of photography and sensors.
Some of the common mistakes I frequently read and hear include:
- ISO relates to sensitivity (it seems like 90% of what I see written in books or on the web get this wrong…).
- Always expose to the right (learn more here in the section “ETTR”).
- Bigger pixels have less noise since they collect more photons. (learn more here…)
Online photography forums are notorious for people spreading opinions. Be careful what you read, and even more careful what you believe based only on sound bites and opinions. For some reason photography is over populated with opinions, and actual science based knowledge is rare.
We are indeed living in the golden age of photography. High quality cameras are everywhere now (smart phones), almost everyone is taking photos on a daily basis and sharing their images with huge audiences on the Internet. Camera technology is changing at alarming rates. You have to actively learn new knowledge to stay on top of things (the fact that a photographer used to have their own darkroom for film does not really impress me today). I have a PhD in optics and used to work a lot in darkrooms, and yet today I feel I must keep learning new material and technique to understand modern photography technique and equipment. My photography heroes today tend to be very young, entrepreneurs in spirit, and the ones pushing the envelope. I have learned a lot of new knowledge from many of these younger, and some older, photographers!
It is also very easy to get wrapped up in the latest – greatest equipment, always thinking “if I only had this new camera body…”. Be careful with what you read on the internet along these lines. Many of the authors are paid by manufacturers for link clicks on their website to sellers of the new camera equipment. Often times the level of minutia being discussed is really not that relevant, perhaps one can improve their photography more by going out and taking photos, learning about lighting, composition, and post-processing.
The more you can research the science of photography, the better your images will be in the long run, and you may actually save money by not buying expensive and unnecessary equipment.
Here are some good references and websites to follow:
- The best source I have found for the science of photography is Roger Clark’s website, ClarkVision.com Roger is an imaging scientist and a talented photographer. Many times I have refined, and even changed my view, on how cameras and sensors work based on the information on Roger’s website. Check it out, especially when you realize you do not understand how things are really working.
- For nightscape photography I always learn from Ian Norman’s website, Lonely Speck.
- The Digital Negative: Raw Image Processing in Lightroom, Camera Raw and Photoshop by Jeff Schewe