How to Find the Best ISO for Astrophotography: Dynamic Range and Noise

how-to-find-the-best-iso-for-astrophotography

ISO is one of the three major exposure settings in the exposure triangle of a digital camera. Of the three: shutter time, f/number, and ISO, it is ISO that is probably most misunderstood. Even more so than f/number. In fact, it is a common misconception that higher ISO settings will cause images to be noisier. In fact, the opposite is often true. Wait, what?

That’s right, higher ISO settings alone do not increase image noise and higher ISOs can even be beneficial to low-light photography. In this post, I talk about the craziness surrounding ISO settings, how ISO actually affects exposure and how to find the optimal ISO setting on your camera for astrophotography.

Introduction

This article is the first of three articles in the Astrophotography 101 course about optimizing exposure. Futures articles on shutter time and aperture f/number will follow. Learning how to optimize exposure settings is one of the most helpful skills when attempting astrophotography. “What exposure settings should I use?” is probably the most common question I get. For beginners who are new at trying astrophotography with their regular digital camera and lens, I usually recommend starting with my Milky Way Exposure Calculator. That calculator will provide an excellent starting point when making your first attempts at shooting the night sky.

Alabama Hills Photography Workshop

Once you’re comfortable making your first exposures, the next thing I recommend learning about exposure is how to optimize your ISO setting. That’s what this article is all about. To begin, here are a couple of glossary items that will hopefully help:

ISO

In digital photography, ISO is a standard (specifically ISO 12232:2006) for exposure brightness developed by the the International Organization for Standardization (ISO). Different camera sensor models have different sensitivities so we need some way to correlate them so like exposures yield like brightnesses. Some people pronounce each letter (aɪ-es-o) but that’s technically incorrect according to the organization for which it stands. It’s correct (and easier) to just say it like the two syllable word aɪso.

Signal

Signal is the part of the photograph that we want. Light is signal. Signal is the image. Without the signal (without light), we can have no image. The more light that we can gather, the more signal we have. In general, the more signal, the higher the quality of the photo.

Noise

Noise is the part of the photograph that we do not want. Noise is interference appearing as speckled grain that obscures the signal and thus the details of the photograph. Noise is usually generated by heat or imperfections in the behavior of the electronics of our digital cameras. Some noise is random with every shot (shot noise) and some noise is produced consistently by the camera’s sensor (upstream read noise) or produced by the electronics after the sensor’s signal has been amplified (downstream read noise). In general, the more noise, the lower the quality of the photo.

Signal-to-Noise Ratio (SNR)

The ratio of signal to noise in an image. The higher the signal-to-noise ratio, the higher the quality of the image. More light = more signal = good. More noise = bad. Collecting more light is the best way to increase signal-to-noise ratio.

alabama-hills-1400px

Higher signal-to-noise ratio is the best way to improve image quality. Sony a7S, 55mm f/1.8 @ f/2.8, 48x5s, PP7, ISO 12800

Dynamic Range

The full range of light of a scene, from the darkest darks to the brightest brights. A high dynamic range scene has extremely bright highlights (such as the sun) and extremely dark darks (such as a fully shadowed black rock). A low dynamic range scene has relatively uniform light across the scene where the brightest parts of the image are not much brighter than the darkest parts of the image. Cameras only capture a limited dynamic range of light. If the dynamic range of a scene is high enough, anything outside the range of the camera’s sensor will either be blown out to pure white (in the case of very bright areas) or crushed to pure black (in the case of very dark areas). In general, a camera sensor capable of capturing a higher dynamic range of light is more desirable.

A disclaimer: I’m an engineer, but I’m formally and primarily trained in mechanical engineering. I have some relevant experience, but electrical and computer engineering isn’t my main mode of expertise. My intention with this article is to simplify these concepts in a way that hopefully makes sense to a larger, non-technical audience. If you’re familiar with this topic and you see any glaring mistakes in this article, please feel free to let me know.

Also, all of the points made in this article apply to RAW image files. It’s very important to shoot astrophotos in RAW format to preserve the best data collected by the sensor. Don’t start complaining when you try any of the tests in this article on your JPEGs. Also, much of the benefit of optimizing ISO selection applies primarily to low-light shooting (like astrophotography) where we have a relatively small amount of signal competing with the various noise sources that encroach upon our photographs.

ISO is Amplification or Gain

It’s a (very) common misconception that increasing ISO increases the sensitivity of a camera sensor. ISO doesn’t change sensitivity. Increasing ISO simply increases the brightness of a photo by amplifying the sensor signal. In the electronics world, amplification is sometimes called “gain.” Like we can “gain” weight if we increase our eating, we can “gain” brightness if we increase our ISO.

ISO doesn’t change sensitivity.

ISO in no way affects how much signal (light) the camera can collect. If we actually want more sensitivity with a camera, we need to either increase shutter time or aperture size (lower the f/number).

Higher ISOs Don’t Increase Noise

OK, to the main point: Higher ISOs won’t increase the visible noise in a photo.

Read that again, realize that it contradicts what you probably think you know about ISO and then let me elaborate:

All other things being equal, a higher ISO will do the following:

  • A higher ISO will increase the brightness of an image
  • A higher ISO will decrease the total dynamic range of the image
  • And, in many cases (like astrophotography), a higher ISO will actually decrease the visible noise

OK, I know what you’re thinking: “How come when I use a higher ISO, I get more noise?!” Here’s why:

For most imaging situations, photographers will usually use P (Program), A/Av (Aperture Priority/Aperture Variable), or S/Tv (Shutter Priority/Time Variable) modes on their camera. In these exposure modes, using a higher ISO setting will result in an image with more relative noise. What most people don’t realize is that the increase in noise is not because of the increase in ISO. The increase in relative noise when using a higher ISO in an automatic exposure mode (like P, A/Av or S/Tv) is actually due to the reciprocal decrease in shutter time or the decrease in aperture size as a result of using an automatic exposure mode. Most people are misattributing the increase in noise to the ISO when it’s actually caused by lower signal-to-noise ratio due to the shutter or aperture.

When setting a higher ISO on one of these autoexposure modes, the camera tries to achieve a neutral exposure and compensates for the increase in ISO by decreasing the amount of light entering camera. This reduction in light is done automatically by the camera by either decreasing the time the shutter is open (when in A/Av mode) or by using a higher f/number and thus decreasing size of the lens aperture diaphragm and letting in less light at a time (when in S/Tv mode), or by a combination of both (when in P mode).

So a reduction of light by the shutter or the aperture is the reason that the image appears noisier. It’s not noisier because of the higher ISO. This reduction of light is a reduction of signal and a reduction of signal yields an overall lower signal-to-noise ratio and thus a noisier photo.

How Do Shutter, Aperture and ISO Affect Noise?

A simple comparison test can show that relative noise levels are primarily affected by shutter and aperture and not affected nearly as much by ISO. In these tests, all settings are kept identical except the one that we wish to test which is adjusted by two stops. Then, in post processing, the images are equalized in brightness and compared.

Here’s what one of my complete test image looks like. It’s a RAW shot of Orion from a city suburb, made on a Sony a7S (my review) with the Zeiss 55mm/1.8 (also my review) lens:

Constellation Orion, Sony a7S, 55mm

How Shutter Time Affects Noise

  • 8s, f/2.8, ISO 3200
  • 4s, f/2.8, ISO 3200 (+1 stop in post)
  • 2s, f/2.8, ISO 3200 (+2 stops in post)
Astrophotography shutter time noise comparison

How Shutter Time Affects Noise – Sony a7S, 55mm, f/2.8, ISO 3200

Conclusion: shorter shutter time = less signal-to-noise ratio = noisier photo

How Aperture (f/number) Affects Noise

  • 8s, f/2.8, ISO 3200
  • 8s, f/4.0, ISO 3200 (+1 stop in post)
  • 8s, f/5.6, ISO 3200 (+2 stops in post)
Astrophotography aperture f/number noise comparison

How Aperture (f/number) Affects Noise – Sony a7S, 55mm, 8s, ISO 3200

Conclusion: higher f/number = less signal-to-noise ratio = noisier photo

How ISO Affects Noise

  • 8s, f/2.8, ISO 3200
  • 8s, f/2.8, ISO 6400 (-1 stop in post)
  • 8s, f/2.8, ISO 12800 (-2 stops in post)
Astrophotography ISO Noise Comparison Test

How ISO Affects Noise – Sony a7S, 55mm, f/2.8, 8s

Conclusion: higher ISO ≠ more relative noise

So of the three tests on my Sony a7S, shutter speed and aperture very obviously directly affect the apparent levels of noise in the photograph while ISO has nearly no effect. This is completely contrary to what many people would expect when they think about higher ISO.

In low-light photography, there is one aspect of ISO that can greatly affect the amount of perceived noise for any given ISO setting: downstream electronic noise. Let’s see how different types of cameras can be affected by downstream electronic noise.

ISO-Invariance and Downstream Electronic Noise

There are variations from sensor to sensor and camera model to camera model in how ISO affects low-light images. Understanding how your camera sensor behaves can help you find the optimal ISO setting for astrophotography. There are two fairly common configurations that we see in most modern digital cameras so we can split most cameras into one of two camps, ISO-variant and ISO-invariant.

ISO-Variant Cameras

Cameras use varied levels of analog amplification to adjust ISO. In a simplification of this case, the amplifier boosts the electronic voltage readout from the sensor by 2x for each ISO: 100, 200, 400, 800, 1600 and so on. Higher ISO means more amplification of the sensor output data.

After the sensor data is amplified by the ISO, it’s sent through some (downstream) electronics (such as an analog to digital convertor) to ultimately change our data from voltages into a digital file of numbers that’s readable by a computer. One of the distinct characteristics with ISO-variant cameras is higher contribution of noise from these downstream electronics.

If there is relatively little signal to begin with (e.g. in low-light situations), the lower ISO settings might not apply enough amplification for the voltages of the sensor data to overcome the contribution of electronic noise made by the downstream electronics. That means that in low-light situations like astrophotography, ISO-variant cameras will actually show more noise at low ISO settings and less noise at higher ISO settings. The Canon EOS 6D, still one of my favorite choices for a DSLR for astrophotography, is highly ISO-variant and actually shows its best low-light noise performance at ISO 6400 and higher!

higher-iso-less-noise-canon-eos-6d

The Canon EOS 6D is highly ISO-variant and achieves its best low-light noise performance at ISO 6400 and higher.

Most Canon DSLRs are highly ISO-variant. There are a few exceptions to the Canon lineup that are not as ISO-variant including the new Canon EOS 5D Mark IV and the Canon EOS 80D.

sensor-diagram

ISO-Invariant Cameras

ISO-Invariant cameras have lower downstream read noise such that in low-light shooting conditions, the signal to noise ratio stays more constant as ISO settings change. In a simplification of this case, the sensor data is already amplified above the minimal contribution of downstream read noise sources before being converted to a digital signal. The result is a camera with low ISOs that tend to have less shadow noise and less of a variation between ISO settings. Most of these types of cameras are considered relatively ISOless or ISO-invariant. One camera that shows a great example of ISO-invariance is the Fujifilm X-T1. An example of the X-T1’s ISO-invariance test is available at the end of the article.

sony-a7s-review-6

Modern digital cameras made by Sony and Fujifilm tend to be relatively ISO-invariant.

Notes and Exceptions

Okay, it’s not all black and white: many ISO-variant cameras eventually act like an ISO-invariant camera above a certain high ISO setting. Above some threshold ISO, these cameras fully overcome their noisy downstream electronics and show minimal difference in signal-to-noise ratio with higher ISOs. Most Canon cameras act this way above about ISO 1600. Knowing what that threshold ISO setting is can help us achieve the best low-light performance.

Similarly, many ISO-invariant cameras may have one or two distinct jumps in gain that will affect the overall read noise contribution to the image. In this case, there may still be a threshold ISO above which it is beneficial to shoot in low-light conditions. The Sony a7S acts this way with changes from approximately ISO 100 to 200 and 1600 to 3200. The Sony a7S’s best low-light performance is actually around ISO 3200 and above. Otherwise, the differences between ISO settings in low-light conditions on the a7S is relatively minimal.

Ultimately, both configurations achieve the same goal of brightening the photo to correspond with the particular ISO setting but the end result can be quite different, especially when shooting in low-light scenarios. ISO-invariance is a distinct enough trait in the behavior of a camera that dpreview.com has added an ISO-invariance test to most of their latest camera reviews. I personally think it’s very helpful to know how a camera acts in order to find out where it will perform best in low-light photography.

ISO vs. Dynamic Range

One of the distinct negative aspects of using too high of an ISO is reduced dynamic range. The more that we amplify the data that makes up a digital image, the more that we risk brightening it so much that it blows out the brightest parts of the image to pure white and loses detail in those parts of the image.

In the dynamic range test below, I made exposures of the star Antares at the highest ISO settings of my Sony a7S using the same exposure settings and varying only the ISO. As the ISO increases, the star appears to get larger because it’s being gradually more and more overexposed with each higher ISO. In practice, with the Sony a7S, the reduction in dynamic range doesn’t become too much of an issue until about ISO 51200 and higher but the difference in each stop is still apparent.

As a side-note, notice how similar most of the ISO settings between 1600 and 204800 look to each other in terms of noise, especially relative to the Canon EOS 6D sample above. The Sony a7S is a fairly, although not completely, ISO-invariant camera.

astrophotography-iso-dynamic-range-comparison-sony-a7s-stars

ISO Dynamic Range Test on the Star Antares – Sony a7S, 50mm, f/2.8, 8s

In my experience, except for the brightest stars, blowing out any part of an astrophoto to the point where we’re losing a lot of data is very, very rare. The bigger risk of using too high of an ISO in landscape astrophotography occurs when there is a larger, brighter (usually artificial) light source in view of the shot such as a street lamp, light pollution from a nearby town or your buddy’s headlamp.

Since we lose a little bit of highlight data with each higher ISO, choosing the optimal ISO for astrophotography is a little bit of balancing act between using a higher ISO for better noise performance (especially in the case of an ISO-variant sensor) or a lower ISO for better dynamic range.

Finding the Optimal ISO for Astrophotography: The ISO-Invariance Test

Stand back, we’re going to try science! In order to find the best ISO to use for astrophotography, I recommend doing an ISO-invariance test. Most of the samples shown in this article up to this point were made with an ISO-invariance test. It’s a super easy test to run: all we need to do is to take about 7-10 RAW photographs, one at each whole-stop ISO and then we match the exposure brightnesses in post processing. This test is easier to perform in a low-light scenario so I recommend doing this test outdoors at night or in a dimly lit room. Maybe make it an astrophotography trip.

If you’re performing this test while shooting the dark night sky, use my Milky Exposure Calculator to determine the shutter time and aperture setting. If doing the test in a dimly lit room, first use your camera’s (P) Program exposure mode at ISO 3200 to determine your shutter time and f/number.

Example: Canon EOS 700D

For my example, I’ll be testing out the Canon EOS 700D/T5i. Here’s a summary of the test:

  • Shoot in dark conditions: a dimly lit room or outdoors at night
  • Shoot in RAW file format!
  • Use (M) manual exposure mode
  • Set “daylight” white balance (just so it doesn’t drift)
  • Disable all forms of noise reduction (Long Exposure NR, High ISO NR)
  • Shoot one exposure at each whole stop ISO (100, 200, 400, 800, etc.)
  • Keep all other settings the same, change only ISO
  • Match exposures in post processing and compare

For my test on the T5i, here’s what the complete images looked like with the crop of the test area highlighted. I cropped the results of the test to a small area that included some midtones and some shadows.

iso-invariance-test-frame-canon-700D-t5i

Straight out of the camera, the crops of the RAWs looked like this:

ISO Comparison – Canon EOS T5i / 700D, 18mm, f/3.5, 25s

In terms of noise, this comparison is deceiving because the brightnesses don’t match between exposures. In order to level the playing field, we need to match the brightnesses. To do so, I used Exposure adjustment slider in Adobe Lightroom to match all of the exposure brightnesses to the ISO 3200 exposure. The ISO 100 image was pushed all the way to the max +5EV setting on the Exposure slider, the ISO 200 +4EV, the ISO400, +3EV and so on…

+5ev-push-exposure-slider-adobe-lightroom
Here’s the complete summary of how we match all the exposure brightnesses in Adobe Lightroom.

  • ISO 100 gets pushed +5EV
  • ISO 200 gets pushed +4EV
  • ISO 400 gets pushed +3EV
  • ISO 800 gets pushed +2EV
  • ISO 1600 gets pushed +1EV slider
  • ISO 3200 has no adjustments made
  • ISO 6400 gets pulled -1EV

Another way to do this in Adobe Lightroom is to select all of the exposures, then highlight the ISO 3200 exposure and select Photo > Develop Settings > Match Total Exposures or press Command+Option+Shift+M (Ctrl+Alt+Shift+M).

Once equalized, here’s what the exposures look like:

iso-invariance-test-iso-comparison-canon-700D-t5i

ISO-Invariance Test – Canon EOS  700D / T5i

Upon comparison of the exposures, it’s immediately apparent that the Canon EOS 700D/T5i is not completely ISO-invariant. It appears as if that the camera reaches its best low-light performance at ISO 1600 and higher. ISO 1600, 3200 and 6400 look almost identical meaning that the 700D might be ISO-invariant from ISO 1600 upwards. Below ISO 1600 is a different story: As the ISO lowers, image quality degrades until the point of being nearly unusable at ISO 100. In order to preserve some dynamic range, but still get the best low-light performance on the 700D, it’s clear from the results of the test that ISO 1600 is the optimal setting.

Example: Fujifilm X-T1

Just for comparison, I ran a separate ISO-invariance test on my Fujifilm X-T1, this time at 30 seconds and an aperture of f/2.8. The results are distinctly different from the Canon.

ISO-Invariance Test – Fujifilm X-T1

The difference is that there is no difference… between the ISO 200 setting (the lowest it goes on the X-T1) and the ISO 6400 setting, noise levels are identical. This means that the Fujifilm X-T1 is completely ISO-invariant. The noise levels across the ISO range don’t change in the slightest. This means that it doesn’t really matter which ISO you use on the Fujifilm X-T1 and the optimal setting might even be ISO 200 in order to preserve dynamic range.

That said, there’s also a little bit of impracticality if attempting to shoot astrophoto at ISO 200 as the image preview on the back of the camera would be very dark and evaluation of other important factors like focus and composition would be difficult at ISO 200. Luckily, we’re usually not risking too much dynamic range by bumping ISO up to a moderately high level, assuming there are no bright artificial light sources in the photo. So using ISO a slightly higher ISO might be the more practical choice, keeping in mind our tolerance for reduced dynamic range.

Conclusion

Contrary to popular belief, higher ISOs don’t create more noise and using a higher ISO can actually be beneficial when shooting in low-light scenarios, especially on cameras with ISO-variant sensors. Run an ISO-invariance test on your camera to determine the best ISO setting to use when shooting astrophotography. ISO behavior varies from camera model to camera model and testing out each ISO setting can help determine the best ISO to use for the best noise performance in your astrophotography.

It’s important to understand that ISO-variance or invariance doesn’t necessarily make a camera better or worse at low-light, it’s just different. Knowing how a camera behaves is an important step to achieving the best image quality.

More and more cameras manufacturers tend to be making their cameras more and more ISO-invariant, as they develop sensor technology with reduced downstream read noise and improved dynamic range at low ISO settings.

Do you know which ISO on your camera gives the best low-light performance? Do a test to find out!

Disclosure

We are a participant in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for us to earn fees by linking to Amazon.com and affiliated sites. We are also a participant in the B&H Affiliate Program which also allows us to earn fees by linking to bandhphotovideo.com.

Learn Astrophotography

Astrophotography 101 is completely free for everyone. All of the lessons are available on the Lonely Speck Astrophotography 101 page for you to access at any time. Enter your email and whenever we post a new lesson you’ll receive it in your inbox. We won’t spam you and your email will stay secure. Furthermore, updates will be sent out only periodically, usually less than once per week.

Join 10 236 other subscribers.

Help us help you!

Believe it or not, Lonely Speck is my full-time job. It’s been an amazing experience for us to see a community develop around learning astrophotography and we’re so happy to be a small part of it. I have learned that amazing things happen when you ask for help so remember that we are always here for you. If you have any questions about photography or just want to share a story, contact us! If you find the articles here helpful, consider helping us out with a donation.

Donate

Thanks so much for being a part of our astrophotography adventure.

-Ian

*This article was edited on March 20, 2017. A previous version of this article incorrectly characterized certain aspects of ISO-invariance. This point has been corrected. 

53 Responses

  1. Dave November 2, 2017 / 10:07 am

    Wow really this article is really useful!! I want to try your “experiment”.
    But I’m a little bit confused, I’ve got a Canon EOS 80D and you said this camera (and 5D MarkIV) is almost ISO-invariant.
    Fine, which the translation for a dummy like me? What I have to take care of when shooting night landscapes? I’ve only to “play” with shutter speed and f aperture?

    • Dave November 3, 2017 / 4:09 pm

      I’ve done some ISO-invariant test with my Canon 80D (with a Tamron 17-50 f2.8), you wrote it isn’t so ISO-variant, but if I’m not wrong there is a evident jump at 1600.
      Lower than 1600 it seems almost horrible, 1600 ISO seems to be the best, higher than 1600 not very useful, 3200 still usable but the others…meh…

  2. Scott October 18, 2017 / 6:16 am

    I’m shooting a Canon 6D with a Rokinon 14mm f2.8 wide open for 20 seconds in suburban settings with artificial light sources in the frame. Can someone tell me what ISO to use? My best guess from reading the article is 1600-6400…

    • Scott October 18, 2017 / 6:36 am

      Also thank you so much to Lonely Speck you have inspired me!

  3. Rob October 3, 2017 / 5:41 am

    ISO note.

    Thank you for this article and all the others you have contributed. You mentioned you were an engineer and welcomed corrections – one to bring up is the question of ‘eye-soh’ vs ‘eye-ess-oh’ pronunciations of ISO. It seems almost everyone in the photography community pronounces it according to the the second variation suggesting it’s an acronym for the organization that makes the standard. In actuality it looks like that’s incorrect. It’s not an acronym at all and never has been. It’s a short form and should be pronounced as ‘eye-soh’.

    The International Standards Organization doesn’t actually exist and neither does the acronym. It’s the International Organization for Standardization and the short form ISO is actually derived from the Greek word isos meaning equal.

    See this detailed explanation from Wikipedia under the ‘Name and Abbreviations’ section:
    https://en.wikipedia.org/wiki/International_Organization_for_Standardization

    And this shorter explanation from ISO themselves (if one doesn’t trust Wikipedia):
    https://www.iso.org/about-us.html

    • Ian Norman October 10, 2017 / 6:50 am

      Yes, this is one that I’ve mentioned in several articles, thanks for the detailed references!

  4. Dmitry August 9, 2017 / 12:34 am

    Re:”OK, to the main point: Higher ISOs won’t increase the visible noise in a photo.”
    The higher ISO – the more signal amplified. The more signal amplified – the more noise in a final picture, law of physics, very simple!

  5. Yves August 4, 2017 / 12:37 pm

    Hm, this all sounds plausible, albeit a bit misleading. If you change brightness in post, you introduce a fourth parameter besides shutter speed, aperture and ISO. (Actually this serves the same purpose as ISO, just with an alternative implementation.) And then it just comes down to this simple equation:

    Analogue ISO amplification directly in the sensor vs. digital post-processing amplification on the digitised file.

    Can you imagine after this short comparison which will introduce more noise? Of course the second. So if you expose correctly for you don’t need to use the crappy post-processing amplification, then higher ISO will increase noise. And if you just increase ISO because your lens can’t open wide enough, getting another lens so that you can decrease ISO will decrease noise. Everything else (like described in this article) is just confusing number twisting.

  6. Brad Grove June 24, 2017 / 9:01 pm

    I don’t want to appear too critical here because I can see that the basics of your article are founded on reproducible results and I can see that you know what you are doing. However these types of blogs are meant to instruct and inform people and therefore require good communication technique.

    My problem with what you have written here is that some of the statements you have made are going to be very misleading to non-technical people and inexperienced photographers (I can see that by way of some of the comments already made by others here).

    You simply can’t make statements like the following without being properly qualified…

    You wrote: “In fact, it is a common misconception that higher ISO settings will cause images to be noisier. In fact, the opposite is often true. Wait, what?”

    By itself… what you have said (above), is completely incorrect. I know that the rest of the article was meant to be qualifying and instructive but it was, in my view, slightly irresponsible. It would have been much better and more accurate to indicate that…

    “Astrophotography specifically places some rather challenging boundaries/restrictions on exposure in that we have very defined limits on the length of exposure (so that we don’t get unacceptable star trails appear in the image). Because under-exposure is also a significant contributor to noise, it becomes very important to find the ISO setting which gives us the best balance between the noise resulting from higher voltages being applied to the image sensor (ISO setting) and the noise resulting from this underexposure.”

    I don’t disagree with your explanation of “ISO-Variant” and “ISO-Invariant” cameras but readers need to understand that applying 5 stops of exposure in post processing to an image taken with an ISO of 100 on an ISO Variant camera body (like most Canon bodies) will produce the results you have illustrated (lots of noise). But if that same image had of been exposed properly the noise levels would be far less that the image which was correctly exposed at all of the higher ISOs.

    In other words, we all know that noise is adversely affected by under-exposure and it is undisputed fact that noise levels will increase when voltages are increased (especially on ISO Variant sensor implementations), so this article is more about balancing those two dynamics under the umbrella of “Astrophotography” and the restrictions which specifically apply here.

    I hope this helps others.

    • Kaveh September 6, 2017 / 10:23 pm

      Thanks Brad. I am so glad I came across your comment here as I was going to ask the same question that you clearly and accurately answered. Thanks for clarifying it.

      I would also like to thank Ian for this great post that cleared up a lot of things for me.

  7. Guido Da Re June 17, 2017 / 6:18 am

    Hi Ian, awesome and interesting article! I have a Fuji xt10, which shares the same sensor and processor of her ‘sister’ xt1, so it should have the same results in terms of noise and image quality. My question is, I develop my raw files in lightroom but I’ve read that even with the lastest updates the demosaicization of the fuji’s raf is still not perfect and introduces artefacts, do you use some oder softwares to do demosaicization or do you do everything in lightroom?
    Waiting for your response! 🙂

  8. Sriram Murali April 7, 2017 / 9:51 am

    Wow. Great article, breaking the myth about ISO. Say I’m shooting under moonlight, I find that f2.0, 10 secs, ISO 3200 to be a good exposure on my Canon 6D, are you saying it is actually better to make it f2.0, 5 secs, ISO 6400? I’ve always thought reducing exposure time and increasing ISO increases noise.

    • Ferdigrafie April 11, 2017 / 1:43 pm

      No. It may be actually better to choose f2.0, 10 secs, ISO6400 and pull back the exposure in post by -1EV. But as you´re speaking of a moonlight scene, you may have be big bright lights in your image, so be careful when overexposing your image (dynamic range!).

    • Dan Bridges April 12, 2017 / 5:34 pm

      ” I’ve always thought reducing exposure time and increasing ISO increases noise.”

      The reason for the increased noise in the image is the weaker exposure. You’re capturing less photons. The image is “dimmer”, (actually lower digital values in the raw data), requiring more “brightening” (increasing the digital values representing a pixel’s luminance). This brightening can be done in either the analogue or digital domains. Doing it in the analogue domain is used for the low-to-medium-ISO range, while digital amplification/multiplication is used for high ISOs.

      Brightening after the Capture phase in a raw converter, i.e in the Rendering phase, is also performed digitally. In a camera shooting JPEG, capturing and then rendering are performed immediately, whereas, in raw shooting, these two phases are separated, allowing more flexibility.

      There are two basic noise types (ignoring long-exposure sensor-heat-generated noise) : Electronic Read Noise (sensor, ADC) and Image/Photonic noise, aka “shot noise” , the SNR of which decreases with the square-root of the number of photons captured.) At low ISOs ADC noise dominates. At high ISO, sensor noise dominates. In weak exposures the shot noise SNR will be lower than for a strong exposure.

      You generally use high ISO to compensate for the lower digital values in a weak exposure. Shot noise will be in the whole image. Read noise will be in the dark parts.

      The optimum ISO setting is to use the lowest ISO where the sensor noise dominates the ADC noise in the Total RN (a quadrature combination of the uncorrelated/random electronic noise sources). Then the image is said to be “shot-noise limited”, and the perceived noiseyness of the image comes down solely to the number of photons captured by the sensor. Using any higher ISO than this when capturing the image is just wasting headroom, which is important for capturing the very brightest parts of the image, e.g. specular reflections off metal, even if the rest of the image is low-level.

    • Dan Bridges April 13, 2017 / 12:29 am

      ” Image/Photonic noise, aka “shot noise” , the SNR of which ***decreases*** with the square-root of the number of photons captured.)”

      I meant ” the SNR of which INCREASES with the square-root of the number of photons captured.)” . For example:

      10,000 photons captured in a single pixel. SNR due to shot nosie is 100:1
      40,000 photons captured in a single pixel. SNR due to shot noise is 200:1

  9. Jordan March 31, 2017 / 9:58 am

    Ian – If i can ask one more question in this topic (which is so fascinating!):
    – Do noise characteristics change in predictable ways at higher ISO/noise levels(i.e., at a consistent EV and ISO 800 vs 3200 vs 12800 etc), or in certain light situations? By this I mean, are there situations that need higher color noise correction but not as much luminance noise correction, or vice versa?

    I’m asking because I would like to know if there is more advanced strategy with dealing with noise in post processing. Currently I trial & error the levels of luminance vs. color noise reduction in Lightroom, but I would be interested to know if there is a more sophisticated way to think about it. For example, maybe shooting at high ISO in medium light (to get high shutter speed) has a different noise signature than high iso in the dark dark (for long exposures), and therefore we should concentrate on color noise in one situation and luminance in the other.
    I hope that makes a little sense!

  10. Mostafa Refai March 31, 2017 / 5:16 am

    Great article Ian, very informative!

    I have a question regarding ISO-invariant.

    From what I understand, ISO-invariant cameras would let me shoot at any ISO and then I can edit the exposure later. Can I make advantage of this fact for HDR photography?

    Instead of taking multiple photos at different ISOs, I can shoot only one photo then create exposure compensated versions on Lightroom. Btw, I use luminosity masks to create HDR photos, I don’t stack all the photos together.

    Again, great work. Please keep it up 🙂

  11. leisure travel March 27, 2017 / 2:36 am

    It is in reality a great and helpful piece of info.
    I’m satisfied that you just shared this useful information with us.
    Please stay us informed like this. Thank you for sharing.

  12. Anthony R. March 24, 2017 / 10:04 am

    Ian,
    Thank you very much for the response and the recommendations. I think one of the cheaper options will suffice, I especially like that the first option can be used as a monopod. I knew I should probably consider an L-bracket, thanks for pointing that out, I just didn’t realize how expensive those could be. The Q66 plus the L-bracket will do just fine.
    Thanks again and keep up the great work,
    Anthony

  13. Brad Vietje March 23, 2017 / 3:13 pm

    Excellent information well presented — Thank You!

    I wonder if a lot of confusion about ISO settings comes from the good-ole film days, when ASA and ISO were often used interchangeably, and higher speed film really did have larger emulsion grains, thus a grainier image. We’d use Tri-X (ASA 400) for dim light, and either Pan-X (ASA 32) or 2415 Technical Pan for very fine grain imaging. Then exploiting either push or pull processing, we’d change the whole scene altogether.

    I don’t currently use a DSLR for night-time imaging, but I often use a CCD camera on a permanently mounted observatory telescope. We tend to deal more with binning, pixel size, and local seeing conditions for deep sky imaging, as well as Nyquist sampling for planetary photography.

    Is there a DSLR equivalent of binning? I need to learn more about how DSLR cameras work.

    Thank you kindly,

    Brad Vietje
    Newbury, VT

    • Ian Norman March 25, 2017 / 3:11 pm

      The closest think to binning that I would know of is just downsampling. Not a technique that I typically use. Might be useful on very high resolution cameras.

  14. Larry Olsen March 23, 2017 / 3:07 pm

    Ian, great article. I learned a lot. In discussion with friends we have kicked around what magnification is appropriate when evaluating noise (and focus, for that matter). Available data is confusing and contradictory. 1:1, 2:1, 11:1? It seems impractical to examine photos at such huge levels of magnification. 11:1 produces squares so that’s useless. “Fit” doesn’t provide enough detail. I’ve heard talk of 1:1 but then sensor resolution and monitor resolution give different coverage for same magnification. Any comments?
    Larry O

    • Ian Norman March 25, 2017 / 3:08 pm

      I personally use 1:2 most of the time for evaluation of small details but it really is dependent on the situation.

  15. Anthony R. March 20, 2017 / 4:44 pm

    Ian,
    Fantastic article. I have a question that is a little off topic, but I need an expert opinion. I’m getting ready to start shooting astro for this year on my D7200 with the Sigma 18-35 f1.8 Art. This setup is HEAVY and my tripod can’t hold the kit for vertical shots safely and without it loosening the screw mount. Any recommendations on a tripod that can hold a monster setup without breaking the bank?
    Thanks,
    Anthony

    • Ian Norman March 20, 2017 / 5:36 pm

      Thanks Anthony! I have a couple recommendations for tripods that should be great at supporting that kit: In order of least to most expensive.

      CAMETV Q66: https://bhpho.to/2nXQsWy (Cheap, lots of features, shortest of the bunch at 60″, lightest at 2.9lb)
      Davis and Sanford TR654C-36: https://bhpho.to/2nu3Sx8 (also cheap, taller at 65″, 12lb load capacity, less features (no monopod option), 3.4lb)
      Oben CT-3581: https://bhpho.to/2nu0y5f (tallest of the bunch at 67.9″, 26lb load capacity, 3.9lb)
      MeFoto GlobeTrotter: https://bhpho.to/2nXQVrM (most well-reviewed, medium 64.2″ height, 26lb load capacity, lots of features, heaviest at 4.6lb)

      I also highly recommend adding an L-Bracket ( bhpho.to/2ntSzVu) to your D7200 for use with any of these Arca-type plate compatible tripods. It will increase your options for mounting (vertical orientation) to the tripod and increase the security of your gear to the tripod.

  16. Gernot March 19, 2017 / 1:50 pm

    Hi Ian! I’ve tested my Canon 5d MarkIV and i located the “sweet spot” at ISO 6400. Do you have experience with this camera? Can you confirm my result? Greets from austria (not australia 😉 )

    • Ian Norman March 20, 2017 / 12:10 am

      Hey Gernot, I’ve personally not used the Canon 5D Mark IV but I’ve seen results from some of my friends that look great. It has much improved low ISO noise capabilities versus older Canon models

  17. Chris March 19, 2017 / 10:40 am

    Hi Ian, great article and very interesting to know.
    One question though, just openly thinking: as far as I followed, higher ISO means higher voltages. Now, could I assume that higher voltages generate more heat in the image taking process? And if so, would a series of high ISO shots generate a significant grade of heat, that might cause sensor reading irritations – aka worse SNR – itself?
    In other words, do you think it is possible that even relatively ISO-invariant cameras could output noisier images from shooting many* high ISO frames rather than the same amount of images at lower ISO?
    * With many being a significant number. I don’t know the exact number, but it kept me thinking that some astro shooters even use active cooling systems for their cameras. And cameras would not be the first electronic systems in which heat changes to on paper designs…

    Love to get your perspective on that.

    • Ian Norman March 20, 2017 / 12:08 am

      The biggest detriment to image quality that I’ve personally experienced at very high ISOs is a noticeable reduction in dynamic range… or… overly strong noise reduction. Some cameras like the Sony a7S and other offerings by Sony switch on a form of noise reduction in certain modes such as Bulb, or for exposures made as ISOs higher than 51200. This noise reduction is good at reducing hot pixels that are present in long exposures at such high ISOs but it’s also detrimental to shooting stars as they end up being eaten by the noise reduction algorithm. For this reason, the a7S isn’t really suited to traditional astrophotography that requires exposures longer than 30″. As far as high vs. low ISO for stacking, it ultimately comes down to total light collected. If the lower ISO exposures are tracked such that the exposure length can be increased to collect more light, the results can certainly be better than the high ISO stack. The more light, the better. As far as cooling… I’ve experienced significantly noisier images on hot summer nights versus cool fall evenings… heat is certainly a problem that if mitigated can yield improved results. But ultimately, it’s a matter of balancing the complexity of gear, expense, and careful post processing. I personally would start employing other methods such as stacking dark frames before investing in a cooled system. But that’s just me personally.

    • Chris March 23, 2017 / 10:32 pm

      Thanks Iam, for your time taken to answer.
      The cooling was just some sort of support for my question. I try one more 🙂
      When taking multiple pictures, do you think that there is heat generated? And if so, would you think that there is more heat generated with multiple high iso images than multiple low iso images? (Because more voltage is used?!)
      So in a nutshell, might – in the case of multiple exposures – high iso have an indirect effect on noise? Via heat generated, I mean.

  18. JohnL March 19, 2017 / 5:06 am

    BTW a quicker way to pick an ISO is perhaps to look at the sensorgen read noise measurements, e.g.
    http://sensorgen.info/CanonEOS-700D.html
    The shot noise is fixed by how much light you collect (i.e. aperture and shutter speed) , so you just want to keep the read noise down without overflowing the pixel’s capacity (saturation) for the bright stuff (except specular highlights where that’s happening regardless).

    I didn’t understand ISO invariance as being a fixed analogue amplification, as then you’d get gaps in the data as it was adjusted digitally which I don’t recall seeing. Specifically I thought most of the difference (invariant/not) was due to the analogue to digital converters (ADCs) being on the sensor die and so you not getting noise added by going off-sensor, as Canon used to do. The only digital gain adjustments I recall are to make intermediate ISOs when the analogue amplifier only supports the main ones (so 160 and 80 use the same analogue gain as 100 and are then stretched/squeezed to suit). This can be seen as the Dynamic Range bounces around as the digital gain is applied, e.g.
    http://www.photonstophotos.net/Charts/PDR.htm#Canon%20EOS%205D%20Mark%20III
    While a more recent camera supports all the possible amplification levels:
    http://www.photonstophotos.net/Charts/PDR.htm#Canon%20EOS%205DS%20R

    My view of this is the Canon sensors are ISO invariant, just not at low ISOs, so you need to get to 800-ish first. This is due to the extra off-sensor noise, the effect of which falls rapidly as ISO rises. The mechanism is that the ISO amplifier boosts the signal and the noise level stays the same, so if you had 100 electrons captured and 10 electrons of off-sensor noise you’re going to notice it, if you amplify the 100 electrons 64x before it leaves the chip (the ISO amplifier is always on-chip) then you’re combining 10 noise with 6400 signal and you won’t see them.

    The Nikon/Sony/Fuji cameras with on-chip ADCs don’t have so much noise between the amplifier and the ADC but the noise will still fall a little at Higher ISOs s the gain increases. If the gain was fixed I would expect this to behave differently.

    • Ian Norman March 19, 2017 / 11:58 pm

      John, thanks for sharing these great resources. Yes, as I mentioned, many cameras do eventually act ISO-invariant above a certain threshold. I think performing a test is a great way to find out where that threshold is. I also learned about photonstophotons.net from another reader via email: what spectacular resource!

      I’ve noticed especially that most of the Sony/Nikon/Fuji cameras still seem to have some mild gain changes, especially between base ISO and the next step up. The data on photonstophotons.net seems to support this. Other than that, their curves of read noise vs. ISO do tend to be quite flat.

    • JohnL March 21, 2017 / 5:32 am

      Dual-gain sensors usually do one gain change at fairly low ISO, but cameras can also throw in some extra noise reduction in the Raw files at high ISOs which looks similar. For example it’s been strongly suggested to me that the Nikon D810 DR step at ISO 12,800 here is due to noise reduction being applied to the Raw file in-camera (so not being all that Raw):
      http://www.photonstophotos.net/Charts/PDR.htm#Nikon%20D810
      (BTW it’s flat at low ISO as they aren’t real ISOs, but just ISO 64 amplification and longer exposures then adjusted back in the processing, so the highlight headroom gets nuked.)
      You can see both (dual-gain and high-ISO Raw NR) combined here:
      http://www.photonstophotos.net/Charts/PDR.htm#Sony%20ILCE-6300

  19. Ioan March 18, 2017 / 8:15 am

    Hello Ian,
    Interesting article, i saw some situation on my camera [Canon 6D]it is better to have high iso than a long exposure, like i see is better an 10s with ISO 3200 than an 20s wish ISO 1600 , even if it is the same exposure the iso high is better than more seconds
    maybe i am wrong, maybe something was different [like an external light] i don’t know.
    even you said, one of your favorites camera is canon 6d , you didn’t said what you conclude about this camera “sweet spot” of iso for astro?

    • Ian Norman March 18, 2017 / 5:09 pm

      I personally think the sweet spot for low-light on the 6D is ISO 6400-12800. That seems to be where it has the lowest amount of visible noise in an astrophoto.

  20. Matt March 18, 2017 / 5:47 am

    I had always wondered why my a6000 performed so poorly at ISO100.

    I got a star tracker thinking I could do a nice long exposure while having the ISO set very low to overcome the noise issue. I was mistaken! Thank you for explaining why this happens!

    • Ian Norman March 18, 2017 / 5:08 pm

      Yeah, it should definitely yield some better results if you bump it up. I really like ISO 1600 on the a6000.

  21. Dan Bridges March 17, 2017 / 3:19 pm

    “It’s a (very) common misconception that increasing ISO increases the sensitivity of a camera sensor. ISO doesn’t change sensitivity. Increasing ISO simply increases the brightness of a photo by amplifying the sensor signal. ”

    While generally true., this is no longer the case with DR-Pix technology sensors. Sony gained access to this tech in an IP swap with Aptina (now ON Semicondor) and uses it in a number of their recent models, including the A7S.

    DR-Pix sensors have switchable sensitivity. This allows them to change the Sensitivity aka Conversion Gain for the the output voltage, measured in “µV”, produced by the conversion of each photo-electron, “e-“, e.g. 8µV/e-. The sensor does a charge (e-) to voltage conversion internally. The capacitance of the Floating Diffusion inversely determines the CG, and is a compromise. A small CG i.e. larger FD capacitance means a larger FWC and a better DR. While a large CG, a smaller FD capacitance, means better high ISO SNR. (The Source-Follower FET involved in the charge-to-voltage conversion has its own noise. A higher CG reduces the relative contribution of the SF noise to the sensor’s input-referenced RN.)

    With a DR-Pix sensor, there are two CG modes: LCG and HCG. In Low Light/High ISO situations, HCG is used and this uses a small FD capacitance. In Well Lit/Low ISO situations, LCG is used and this switches in an extra capacitance to increase the capacitance of the FD.

    http://www.photonstophotos.net/Aptina/DR-Pix_WhitePaper.pdf

    Looking at the Sony A7S sensor, you can see in the graph link below, the change in the Read Noise curve above ISO 1600 aka 9 EV. This gives a 2 stops reduction in the input-referenced RN at high ISO compared to a standard sensor where the CG would be fixed.

    http://www.photonstophotos.net/GeneralTopics/Sensors_&_Raw/Sony_A7S_DR-Pix_Read_Noise.htm

    • Ian Norman March 17, 2017 / 4:48 pm

      Very interesting information Dan, thanks.

    • Ian Norman March 17, 2017 / 5:17 pm

      Is this conversion gain actually allowing the sensor to convert more photons to electrons or is it just increasing the voltage of the sensor output? Basically, I’m wondering if you can actually call it an increase in sensitivity if it’s not really increasing the quantum efficiency of the sensor. I’m very familiar with the increase in SNR at ISOs above 1600 on the a7S. I’m just wondering if the change is simply the increased voltage amplifying the data above the noise floor generated by the downstream electronics.

    • Dan Bridges March 17, 2017 / 5:44 pm

      “Is this conversion gain actually allowing the sensor to convert more photons to electrons or is it just increasing the voltage of the sensor output? Basically, I’m wondering if you can actually call it an increase in sensitivity if it’s not really increasing the quantum efficiency of the sensor. I’m very familiar with the increase in SNR at ISOs above 1600 on the a7S. I’m just wondering if the change is simply the increased voltage amplifying the data above the noise floor generated by the downstream electronics.”

      This is not a change in QE. It’s a change in Output Sensitivity. For example, look at the Output Sensitivity spec for the Kodak KAF-3900: 26 μV/e-

      http://www.kodak.com/ek/uploadedFiles/Content/Small_Business/Images_Sensor_Solutions/Datasheets(pdfs)/KAF-39000LongSpec.pdf

      Wiht analogue gain change with ISO increase you vary the gain between the sensor and the ADC, by inserting a PGA (Programmable Gain Amp) between them.

      With Analogue gain chain via a DR-Pix step, you instead increase the conversion gain of photo-electrons to μV inside the sensor, so less or no gain chage in the PGA stage is required). The advantages doing this inside the sensor, rather than in the PGA, is better tradeoffs between DR, FWC, and Sensor SNR, when changing from low-light to high-light situations.

      Add in QE and you get the full conversion chain:

      Photons -> Photo-Electrons (accumulating as a decremented charge in the FD – contrary to common knowledge, the FD is filled to the brim with electrons when the sensor is reset – the accumulating photo-electrons during the exposure then reduce the charge on the FD – it’s the change in charge, not its sign, that’s used ) -> Output Voltage.

    • Ian Norman March 18, 2017 / 5:07 pm

      Thanks Dan, that makes sense. very good do know and it coincides predictably with the behavior of the a7S

  22. Matthew McLaughlin March 17, 2017 / 12:49 pm

    Ian, on a Sony a6000 and a6300′ what is the optimum ISO that you’ve found? Just curious, thx for this well written article! – Matthew

    • Ian Norman March 17, 2017 / 1:01 pm

      I have not personally tested the a6300 but the a6000 should be used at ISO 1600.

  23. Jordan Chapell March 17, 2017 / 11:49 am

    What an awesome article on this topic Ian! Good blend of wonky and understandable.
    I shoot with a Sony A7rII that is also relatively ISO-invariant in stages like your a7S. This is an awesome feature, but I wonder if you have a thought on this: Though pushing the EV in post can generally lead to noise levels similar to boosting ISO, I find that there are other consequences: To my eye, the photo’s natural contrast and color information are negatively impacted when pushing EV in post. I wonder if you have any thoughts on this based on how ISO works (i.e., in my mind I’ve pictured something analogous to cones in our eye that struggle with color in low light, even while the rods work well). Perhaps the sensor struggles to capture accurate color in low light with low ISO, so the color information is not complete when pushing EV in post?
    I haven’t done any scientific testing with this, but would love to hear your thoughts.

    • Ian Norman March 17, 2017 / 12:41 pm

      Jordan,

      I have noticed, on Sony cameras in particular, that color balance does tend to be compromised when shooting at very low ISOs in very dark conditions. One of my original hypotheses was that the camera was attempting to apply white balance based on the content of the jpeg preview generated. Since the image is so dark, it incorrectly computes the white balance and throws off the image. That said, it doesn’t seem to correlate with testing because it still can happen when a manual white balance is set. So, yes, I think my experience with Sony cameras is similar to yours. It’s mostly ISO-invariant but there’s still a sweet spot above which we should shoot in low-light conditions.

      Ultimately, there are other factors that might not be immediately obvious with these cameras (or with our post-processing software). We don’t know exactly how Sony is cooking the data from the sensor (e.g. compression on the 1st gen. a7/S/R, etc.). The only real way to find out is to perform a controlled test.

      As a side note, Sony cameras also get a little weird when using picture profile modes (like PP7) when shooting stills. I’ve noticed on my a7S that the “sweet spot” jumps up to ISO 12800 if PP7 is enabled. It’s ISO 3200 normally. If anything lower than 12800 is used with PP7 enabled, noise is worse and colors are not the best.

  24. Ahmed March 17, 2017 / 7:18 am

    I can’t say anything but I was really surprised and amazed by the quality and information in this article.
    I would like to raise a point, I think that being iso-variant-camera or iso-invariant-camera is a technical specification of the camera itself. Am I right? If yes, how can I know which type of camera I have?
    I assume that some cameras will have a hybrid system, which may mean that the Analog amplifier will start to operate at certain ISO setting in parallel with the processor.
    Let’s get back to the main point, Does it mean that theoretically if the camera is 100% iso-invariant I can shoot astro at ISO 100 or 200 or whatever gives me the better dynamic range ?
    Sorry for my long comment and questions ! 😀
    Waiting for your feedback ! 😀

    • Ian Norman March 17, 2017 / 9:30 am

      Ahmed,

      Yes, whether or not the ISO is variable depends on the camera model. What camera do you have?

      Some cameras, like my Sony a7S, have small changes in amplification at discreet steps along the ISO range. For example, there is one change from ISO 100 to 200 and another from 1600 to 3200, as if there is a “high, medium, low” approach to amplification. In this case, it’s still beneficial for me to shoot at ISO 3200 and above with the Sony a7S when photographing in low-light astrophotography.

      And yes, a camera like the X-T1 that is basically 100% iso-invariant can use the lowest ISO to shoot astro. The preview of the resulting image, however, will probably appear too dark so I’m not sure how practical that is. I think it highly depends on the situation. If there was an ultra bright element in the photo, it might be highly beneficial to shoot at the lowest ISO.

    • Ahmed March 18, 2017 / 6:35 am

      Thanks Ian for your reply,
      I am planning now to upgrade to Nikon D750
      Do you know is it iso-invariant or not ?

    • Dan Bridges March 18, 2017 / 3:38 pm

      “I am planning now to upgrade to Nikon D750
      Do you know is it iso-invariant or not ?”

      Look at the DxOMark DR curve at the low-ISO end.

      A typical digital imaging system has 3 stages (these days, usually situated along the edges of the sensor chip):

      Sensor -> PGA (Programmable Gain Amp) -> ADC.

      All 3 stages produce noise. PGA input-stage noise can be included as part of the Sensor’s RN (read noise) and the PGA output-stage noise can be lumped in with the ADC noise. So we have:

      Variable Gain
      Sensor -> ADC

      The PGA gain is analogue and is applied up to about ISO 1600 (depending on model). Above this further gain is digital, (just mathematically increasing the raw values for each pixel as, at high ISO, sensor RN predominates.

      But at some mid-ISO, used for stronger exposures, due to the fact the the sensor’s output needs less amplification to get a decent signal out of the ADC, both sensor and ADC noise are approx. the same level. And at low ISO, used for the strongest exposure (i.e. the greatest amount of photons captured during the exposure), the least amount of amplification is applied to the sensor’s output so the ADC noise usually predominates.

      Once Sensor RN clearly predominates the Total RN, no further analogue gain is useful for overcoming ADC noise, and you could just as well shoot at this ISO (and no higher) in raw and later digitally (mathematically) boost the image’s rendered brightness in PP. So the imaging system becomes ISO-Invariant. (Shooting in raw at a higher ISO than this is just needlessly throwing away headroom during the capture phase, since each ISO increase is pushing the amplified highlights closer to the imaging system’s clipping level. )

      The DR curve should be a straight line if the Total RN is low and changes little. Up until a few years ago, ADC noise tended to predominate, and the DR curves would flatten at low ISO. At high ISO, where Sensor RN predominates, the curve is usually ideal (1 EV fall in DR for each doubling of ISO). But, if any NR is being applied in-sensor to the raw data, you can find offset steps in the high ISO DR curves. (Also, the signal is weaker at high ISO, so measurements can be less accurate and some kinks can appear.)

      Theses days most cameras are using column-parallel ADCs. There can be a ADC for every column, so there can be thousands of ADCs. This architecture leads to high bandwidth and low noise. Sony sensors have used this for a long time:(https://en.wikipedia.org/wiki/Exmor).
      https://pro.sony.com/bbsccms/assets/files/cat/camsec/solutions/E_CMOS_Sensor_WP_110427.pdf

      A rule-of-thumb is that the straighter the DR curve and lack of lowe-ISO flattening, the lower the ADC noise. (In the very early days, some P&S cameras could have noisy sensors which only used digital gain. This would also produce a straight DR curve, but the max. DR at low ISO was quite low.)

      The DR750 curve is quite straight. The Canon 5DS and have obvious low ISO flattening, indicating that the ADC performance is still relatively noisy. So the cameras takes longer to become ISO-invariant (The ISO from which Sensor RN predominates).

      But the “Print” (normalised) DR of the 1D X II at ISO1600 is quite high, indicating that the Sensor RN is probably lower than in the D750.

      https://www.dxomark.com/Cameras/Nikon/D750
      https://www.dxomark.com/Cameras/Canon/EOS-5DS
      https://www.dxomark.com/Cameras/Canon/EOS-1D-X-Mark-II

      The Sony IMX-071 sensor used in 2010 in the Pentax K-5, Nikon D7000, Sony A580 and other models was the first CMOS DSLR sensor chip to exhibit very low ADC noise.

      https://www.dxomark.com/Cameras/Nikon/D7000

      The max. normalised (“Print”) DR, when compared in same sensor format (e.g. APS-C), has tended to improve over the years with improvements in ADC noise and FWC.

    • Ian Norman March 18, 2017 / 5:10 pm

      As far as I know, the D750 is also ISO-invariant but some users have reported slight color variation between ISOs. Overall noise levels don’t seem to change between ISOs on the D750.

    • JohnL March 19, 2017 / 8:24 am

      No cameras are 100% ISO invarient, but the D750 is about as close as they get. Look at the read noise numbers here:
      http://sensorgen.info/NikonD750.html
      and you’ll see they fall very slowly with ISO, this means there is very little noise added between the ISO amplifier and the on-chip ADC. It also means that:

      (1) You can shoot at a lower ISO (to give more highlight headroom) then brighten the Raw image on a computer and get about the same noise as if you had shot at a higher ISO. (This is usually what ISO invariance is used for.)

      (2) You can shoot at a higher ISO when you are Aperture and Shutter speed limited (e.g. for astro work by lens choice and star rate of movement) so you can better see what you are shooting. However you won’t get much less noise and you lose highlight headroom (see the saturation column – basically as the signal is amplified you run into the ADC maximum input voltage much sooner).

      For Astro use the Dark Current in the sensor is also a consideration and I have no idea how good the D750 is for this. (I started writing “an important consideration” but sensors are much better at this these days. It is temperature dependant.) The really technical can work it out from the section “Which Cameras Have On-Sensor Dark Current Suppression Technology” here: http://www.clarkvision.com/articles//dark-current-suppression-technology/ – but I wouldn’t worry about it.

      Regards

Leave a Reply

Your email address will not be published. Required fields are marked *