Shootin' The Shot
13May/1042

Canon HD DSLR “Native ISO”

If you've spent any time researching Canon's HD DSLR cameras, you've probably come across discussions of which of the cameras' ISO settings to use and which to avoid. There seems to be a common misconception out there, held by even some very well-regarded experts (http://www.zacuto.com/shootout, webisode 2, about 3 minutes in) that the cameras' "native ISO" or "true ISO" settings are the multiples of 160 (ISO 160, ISO, 320, ISO 640, etc), and that the rest of the ISO settings are produced digitally. This is not true. In fact the cameras' native ISO settings (that is, the settings that are derived from analog gain rather than digital exposure compensation) are the multiples of 100 (ISO 100, ISO 200, ISO 400, ISO 800, etc). However, that doesn't necessarily make them better.

I believe that much of the confusion comes from mis-analyzing tests such as these:

http://marvelsfilm.wordpress.com/2009/10/10/canon-7d-noise-and-iso-test/

http://www.cinema5d.com/viewtopic.php?f=14&t=4736

that clearly show that the 160-multiple ISO settings exhibit less noise than the 100-multiple ISO settings. Viewing these test results is often surprising to people because it goes against the conventional wisdom that the higher the ISO, the noisier the image will be, and the lower the ISO, the cleaner it will be. In fact, it is very surprising to learn that ISO 1250 can look on par with or even better than ISO 125. The natural assumption when looking at these tests is that the native ISOs must be the cleanest ones. This seems like a logical assumption at first, but it doesn't hold up when you look at it a little bit closer.

It makes sense that noise would increase if you added a digital push to the exposure, but if the lowest native ISO is 160, and all other non-native ISOs exhibit greater noise levels due to a digital exposure push, then how are ISO 100 and ISO 125 derived? Also, what about the Canon Rebel T2i? It does not offer any intermediate ISO settings, but only allows the selection of the 100-multiple ISOs. Given that this camera is so similar to the 7D, it seems much more likely that they would simply eliminate the intermediate (non-native) ISO choices than that the camera's sensor would be somehow fundamentally differently designed from that of the 7D and 5D Mark II.

So, if the 160-multiple ISOs are not the native ones, why are they cleaner, and how are they derived? Well, it is correct that the 125-multiple ISOs are the noisiest because they are derived by a digital exposure push. ISO 125 is actually ISO 100 with a 1/3 stop digital exposure push, ISO 250 is ISO 200 with a 1/3 stop digital exposure push, etc. However, the 160-multiple ISOs are actually the cleanest not because they are "native", but because they are a result of a digital exposure pull. This pull brings down the exposure of the entire image, and hides much of the noise that would be visible at the next higher ISO. ISO 160 is the cleanest because it is the native ISO 200 with a 1/3 stop digital exposure pull, yielding even less noise. ISO 320 is actually ISO 400, with a 1/3 stop exposure pull, etc.

So now that we know this, which ISO settings are best? Just because 160, 320, 640, etc, are not "native" does that mean you shouldn't use them? Well, no, not necessarily. They are less noisy, after all. It's just important to understand that there is a price to pay for this lack of noise. That price is decreased dynamic range.

Because ISO 320 is actually ISO 400 pulled 1/3 of a stop, that means that the highlights are going to clip at exactly the same point as they would at ISO 400. The 1/3 stop pull is just making that point 1/3 stop darker than pure white. The entire image at ISO 320 is 1/3 stop darker (and may be less noisy) than the image at ISO 400, so the blacks lose detail 1/3 stop sooner, but you don't get that 1/3 stop back at the highlight end of the range -- it's still gone. Therefore, at ISO 320 you're losing a net 1/3 stop from the total usable dynamic range that you would have if you were shooting at ISO 400. The tests here confirm that this is the case: http://www.cryptobola.com/PhotoBola/Canon7D_ISO.htm. [Update: these tests no longer seem to be online.]

So which is more important, shooting images with less noise, or shooting images with more dynamic range? It's obviously subjective, but for me there is no one-size-fits-all answer. In some cases I prioritize one, and in some cases I prioritize the other. I really can't think of any reason to use the digitally pushed ISOs (125, 250, 500, etc). If you need more exposure, you're better off going up to the next exposure-pulled ISO, or the next native ISO, rather than shooting with the increased noise and decreased dynamic range that comes along with the pushed ISOs. In general, if I am shooting a bright scene with a large contrast range, such as a daytime exterior, I will prefer to use the full-stop, native ISOs. In that situation, since I will likely be using the lower ISOs (100 or 200), noise really isn't much of a factor anyway, and I will prioritize dynamic range to prevent the highlights from blowing out as much as possible. If I am shooting a dimly lit scene, I will likely be using the higher ISOs where noise is more of an issue. In that case, keeping the highlights in check isn't usually a problem, so I will prefer to use the 160-multiple ISOs (320, 640, or 1250) so I can get more exposure with less noise.

Comments (42) Trackbacks (11)
  1. Great article – and clears up some very common misconceptions.

  2. Hey, so I appreciate this post. I’m not convinced that 100 increments are native on these sensors. I’ve seen a lot of arguments for 160 and a lot for 100, but besides the test results that are now taken down, do we have anything “official”?

  3. Garrett,
    I can’t find any other tests of the cameras’ dynamic range at various ISO settings to show you online. All the tests I have seen show that the noise levels are lowest at the 160-multiple ISOs and highest at the 125-multiple ISOs and that the dynamic range is highest at the 100-multiple ISOs. I also know that the T2i only offers ISO settings of 100, 200, 400, 800, 1600, 3200, 6400, and 12800. You can draw your own conclusions.
    -Josh

  4. Thanks. Passed this along as most everyone I know goes by the multiples of 160. Had no idea it was taking off a 1/3 stop from the native ISO for each of those but it makes a lot of sense. Thanks for bringing the dynamic range issue in a straightforward manner.
    Joe

  5. Great post, I didn’t know about the highlight information being clipped at the intermediate pull ISOs. This clears up a lot of stuff thanks!

  6. I think I would rather more Dynamic range and accept the noise because I can remove noise later but not add dynamic range.

    What situation would be better to go the other way?

    • In low light, where I would use high a ISO setting which would result in additional noise, highlights clipping would probably not be much of a concern anyway, so I would use one of the “pulled” ISO settings. Less noise and the slight decrease in dynamic range wouldn’t really bother me in that situation.
      -Josh

  7. Brilliant post, cleared up a lot of my wonderings and general rumors going around my college. Quick Q; Does this apply to other cameras such as the EX3, RED, F3? If not why? Or is it completely down to the sensor type?

    • No, this is really only applicable to the Canon DSLR cameras. There may be other cameras with similar characteristics at the various ISO settings (Nikon cameras, perhaps? I am not really too familiar with them), but I wouldn’t necessarily assume that to be the case without testing. I know the EX3, like traditional video cameras, is not setup in terms of ISO. The comparable setting would be gain and is measured in + or – dB. The “native ISO” would be 0 dB gain. I’m guessing the F3 would be the same. The Red’s native ISO is 320 (yes, even with the MX sensor). But with the Red, changing the ISO does not actually alter the image. The Red shoots in a raw format, and the ISO setting only affects metadata attached to the image, not the image itself. And just because it’s natively 320 doesn’t necessarily mean that you always want to rate it at 320. In fact, Red’s recommended rating of 800 for the MX sensor is not a bad choice — it gives you over a stop of extra headroom in terms of protecting the highlights, and the noise level at that ISO is still very acceptable. If you rated it at 320, you might risk exposing it in such a way that your highlights get clipped in exchange for only a minimal reduction in noise levels, which may not be worth the trade-off.

      In any case, this is not really such a big issue with the Red, since it shoots raw, or to some extent the Sony cameras you mentioned, because they were all designed specifically for shooting video and have much better compression than the DSLRs. There’s more exposure latitude, and there’s much more you can do with the image in post. The reason this is so important with the DSLRs is that the video image is so highly compressed that there’s just not a whole lot you can do with it in post. If you try adjusting the levels too much, the image quality just falls apart. The ISO setting is one of the adjustments you can set in camera, BEFORE the compression happens, so knowing how the ISO is derived and what that setting is doing to your image is a very valuable tool to have at the stage when it’s really going to matter.
      -Josh

  8. This is the perfect blog for anyone who wants to know about this topic. You know so much its almost hard to argue with you (not that I really would want…HaHa). You definitely put a new spin on a subject thats been written about for years. Great stuff, just great!

  9. Great article. I have tried to explain this to people on forums many times, now I will simply point them to this article. The dynamic range is never measured or mentioned when people present these lens-cap-on noise tests.

    “I think I would rather more Dynamic range and accept the noise because I can remove noise later but not add dynamic range.”

    If the scene you are shooting has less dynamic range than the camera’s sensor can accomodate, there is no need to worry about losing clipping/crushing detail.

    As the author says, one is no better than the other. It’s horses for courses.

    “The Red shoots in a raw format, and the ISO setting only affects metadata attached to the image, not the image itself, much like shooting in raw mode with a DSLR for stills.”

    This doesn’t sound right to me. I shoot RAW for stills, and I’m pretty sure it is not meta data, but the analog gain on the sensor that ISO settings affect. Maybe I read you wrong.

    • Yes, you are correct about the ISO setting on stills. I was actually thinking of the metadata adjustment options in the Red software, where ISO is just a setting that can be changed after it’s shot like white balance. The ISO settings on the Canon cameras DO actually effect the exposure in raw mode.
      -Josh

  10. BRILLIANT!!

  11. Great article. Intelligent and illuminating. Now you have my undersized brain churning away at the following question: is changing the exposure of a shot in post the equivalent of a digital push or a pull, or is it more akin to an analog adjustment? I’m curious, e.g., whether it is better (all things being equal) to expose a scene correctly at a higher ISO (say 200), or to shoot at a lower ISO in RAW format (say 100), underexpose by one stop, and lighten in post. And does the answer to that question depend on whether you are seeking to optimize DR or low noise? Argh, my head hurts!
    -Wayne

    • Wayne,
      It’s always better to shoot in RAW, but there is no RAW mode for video, which is what the article is about. When you’re shooting video, it’s basically the motion equivalent of shooting jpegs. The image processing is happening in camera. That is why noise is such a major factor to consider. The ISO setting is one of the key choices you can make that will affect the look of your video. By choosing an ISO with less noise, you are making sure that less noise is baked into the image prior to the compression that happens when writing the video to the CF card in the h.264 codec. Of course you can make an exposure adjustment in post, but you will be making that adjustment to already highly compressed video, which will give you much less information to work with, and will still exhibit the noise of the ISO setting you are using. In other words, video shot at ISO 400 and then brought down 1/3 of a stop in post will likely be noisier than video shot at ISO 320 with no exposure adjustment. When I’m shooting video with these cameras, I get the best results by exposing it as closely as possible to how I want the final image to look, or in some cases over exposing and bringing the brightness down in the color correction process. I would never recommend under exposing and then trying to increase the brightness in post, as I have found there’s simply not enough information in the compressed video files to work with when trying to lighten a scene.
      -Josh

  12. Um, I’m pretty sure everything in this blog post is incorrect. I’ve been told personally (as have many other people) by Canon EOLs and even Canon technical reps that the 160 multiples *are in fact* the “native” ISOs (I was told this most recently last night, May 11, 2011).

    That said, you may still be correct that the best dynamic range is achieved at other ISOs… I have not discussed DR specifically with the Canon people as it relates to ISO, but given that the 160 multiples are the native ISOs (and I can say definitively that you are incorrect that 160s are a “pull” exposure from the 100 multiples), it seems very likely that that is incorrect as well.

    • Chris,
      It sounds to me like it’s more a question of semantics than anything else (which is why in my post I use the term “native ISO” in quotes – as I don’t believe “native ISO” is a technical term at all, and have only heard it discussed with regards to these cameras). When most people ask what the native ISO is, what they want to know has nothing to do with the technical details of the way the sensor is designed, they just want to know what ISO they should shoot at to obtain the best results. It is much easier to give the simple answer: multiples of 160, than to explain that there could be a minor loss of dynamic range compared to some of the noisier ISO settings. The noise is lowest at the 160 multiple ISOs and the dynamic range is highest at the 100 multiple ISOs. Explaining that is the point of the post, and it doesn’t matter to me whether some ISOs are referred to as “native” and others are not, I just want to know how each setting is going to affect the image before I select it. It seems odd to me that the ISOs that are called “native” by the Canon reps are not even available as options on the T2i without hacking the firmware, but hey, I don’t design cameras so what do I know?
      -Josh

  13. Next maybe you could tell me why “they” say to only shoot at shutter speeds of 50 if shooting 24 fps, or 60 if shooting 30fps.

    • This is in order to get a look closest to what you are used to seeing in projects shot on film — that “cinematic” motion you are trying to emulate by shooting 24fps. Film cameras traditionally have 180 degree rotating shutters. The shutter must be open while the camera holds each frame in place, and closed while it advances the film to the next frame. The shutter is open half the time and closed the other half of the time, so if you are shooting 24fps, that gives you an exposure time of 1/48th of a second (very close to 1/50th of a second).

      If you shoot with a shutter speed of say 1/24th of a second while shooting 24fps, it would be like shooting with a theoretical film camera with a shutter that is open all the time. This is impossible with a film camera because if you remove the shutter, the film would be exposed during the pull down part of the cycle as well, leading to a smeared, blurry image, but it is possible with a digital camera. However, you would be getting motion unlike that of the 24fps film camera. Because each frame is exposed for longer, there would be more motion blur and it would not look as close to the cinematic motion you are used to seeing in movies or projects originating on film.

      On the other hand, if you shoot with a faster shutter speed, say 1/200th of a second, this would be like shooting on a film camera with a 45 degree shutter. There is much less motion blur than you are used to seeing in a film. You can do this with a film camera, and I’m sure you’ve seen many examples of this technique used to obtain a specific motion effect — think of the battle scenes in “Saving Private Ryan” or “Gladiator”. The motion looks strangely crisp. This is how this effect is achieved. You may want this effect, and in that case shooting with a fast shutter speed is the way to go, but assuming you are just looking for normal “cinematic” motion, 1/50th is the proper shutter speed for 24fps, and 1/60th for 30fps. This is why people use ND filters to achieve the f-stop they want to shoot at rather than just compensating with shutter speed. I use Lightcraft Workshop’s “Fader ND” filter rather than carrying around a whole set of NDs. It’s pretty convenient for these cameras.

      That being said, I’m not super strict about always shooting at exactly 1/50th when shooting 24fps. I find you can adjust it slightly in either direction without any noticeable effects. I will aim for 1/50th, but then vary between 1/40th and 1/60th to make micro adjustments to the exposure if it’s easier to do it that way than with aperture or ND.

      -Josh

  14. I’m taking a guess here, but maybe using the new Technicolor CineStyle Picture Style profile for Canon DSLR cameras, with a 160, 320, 640, 1250 ISO would be the perfect combo, as the Cinestyle profile is said to increase dynamic range… yes?

    • You can never have too much dynamic range. You can always lose whatever you don’t want in the color correction process, but you can never bring it back. The Technicolor people say they recommend using the 160 ISO multiples for best results, but I haven’t done enough testing with the picture profile myself to know if there’s something inherent about CineStyle that makes those ISOs work better with it, or if it’s just a general recommendation because starting with the cleanest image will generally produce the best looking picture. If I had to guess, I would say it’s probably the latter, and might consider sticking with the 100 multiple ISOs up to 200 or 400, before switching to the 160 multiple ISOs around 320 or 640, but without doing some serious testing, I can’t really say for sure.
      -Josh

    • Don’t really think that CineStyle actually increases dynamic range like increasing the headroom. What it does really is allowing more information to be captured during conversion, so you get to keep more detail after color correction. There are differences in dynamic range within the Canon DSLR range. The 5D mark II has a higher dynamic range than the 550D for instance

      • More dynamic range is achieved using flatter picture style settings (such as CineStyle) than by using the more contrasty picture styles that come with the camera (standard, neutral, etc).

  15. Josh, this is a great article and it really took my mind for a spin. I found that every single shot I took at ISO 100 had a higher bitrate than the ones shot at 160, which I thought could mean two things: The ISO 100 shots have a higher data rate due to the higher dynamic range OR the the ISO 160 shots have significantly less noise and therefor less data picked up in the conversion. What do you think? It’s a shame the tests you posted aren’t available anymore

  16. Very Interesting. My Canon GL-1 has Gain settings of 0, +3, +6, +9, +12 db (ISO 125, 160, 200, 250, 320… as measured/interpreted with a gray card and a Minolta Flash Meter IV ). The +3 setting (160) gives very clean results. Vaguely familiar yet interestingly strange!!!

  17. I think you’ve got the “pull” part wrong. ISO 160 uses native ISO 100 with a faster shutter speed. The less exposure then gets ‘push’ed.

    I’m still researching. The native ISO are probably different depending on sensor. More modern, denser sensors may have native ISO in multiples of 160, while older sensors such as in Canon 5D classic have native ISO in multiples of 100.

    Sensor chips have operating regions (voltages) where the response is linear. Voltages immediately above and below these regions operate the chip in non-linear response regions, which is why intermediate voltages are not used for intermediate ISO settings. The Canon 5Dc achieves ISO 50 by operating the chip at half the voltage of ISO 100, in a region that is slightly non-linear.

  18. I really appreciate you sharing this information. Just wondering how much this applies to photographs or does it all just apply to video?

  19. nice blog, very helpfull i my search on iso standards etcetera

  20. Hey everyone….

    Great article!

    I worked as a studio photographer for a number of years photographing everything from fashion to wine bottles and in that time, intuitively, I always felt queasy when I needed to begin adjusting ISO on the multiples of 125 and so on. I never seemed to have a problem with 160 multiples, but for some odd reason the 125’s was an issue that my eyes could see but my mind couldn’t wrap itself around.

    So, I began asking the other photographers if they perceived any issues with certain shots at certain ISO’s (below 800) . Sad to say, they didn’t notice and honestly, seemed bewildered that I was finding faults with our brand new 5D MK II.

    I realize this is a issue that .jpg photogs will be hit with (if they ever realize it) and NOT raw shooters, as the majority of noise can be minimized quite effectively, however, as professionals will argue which format to shoot in, my rule of thumb is – shoot raw when needed and shoot .jpg if raw is not needed. Oh! and nail your exposure as best as you can EVERYTIME. This will save you hours and hours of cleanup work.

    Sorry to go off on a tangent.

    My eyes instinctively knew there was something wrong with the 125’s x’s but it wasn’t until I came across this article and another one put forth by Josh, then it clicked.

    WOW!

    What’s funny, after reading this, I have gone back and looked at my exif data from some of my most beautiful shots…and sure enough when I needed a full dynamic range ( if I wasn’t shooting in raw) I was on the 100’s x’s and during low light, I was in the 160’s x’s.

    In the end, does this matter at all? It’s like those people (me included) who pixel peep, for one reason or another.

    The answer is: Of course it matters! Well, that is if you have made photography your life…bread and butter…and have spent years improving your technique, your craft.

    I leave it here. Thank you to all who have commented towards this post.

    One last thing, sometimes you need to trust your instinct!

  21. Very informative, would be nice to see some sample footage to compare has anyone found a reputable source that has done this?

  22. Thanks for the post, very interesting. I’m about to do a series of tests on the 5d2 and 5d3 using the standard, cinestyle, and a flat picture profile and look at different ISO settings under controlled conditions.

    In your example of shooting in a pulled non-native ISO like 320, you talk about the highlights being clipped at a lower value than when shooting a ‘native’ ISO. I don’t understand this though, you will have to adjust your exposure for any ISO change, so I don’t see how a white (for example) would have a different value at different ISO’s if you would adjust your aperture or shutter speed or lighting to compensate for the ISO change. If I shoot a white card at 400 ISO, f4, 1/60 then change to 320 ISO, I would have to change either shutter or aperture to compensate, and wouldn’t the subject (the white card) have the same RGB values as a result?

    • Jules, yes you would change your aperture to compensate, and maintain the same exposure, but if you are using 320, you would just have slightly less highlight headroom than you would if you were shooting at a multiple of 100. Same exposure, less dynamic range. Probably less noise as well.
      -Josh

  23. hi josh,

    … a lot of people are testing many things… would you please show us the test files or screenshots of histograms that prove a loss in dynamic range while using the multiples of 160?

    dynamic range to the top end is generally lost just by raising the ISO. it has nothing to do with the 160x. so, using the ISO 200 without an exposure pull (like you say) will also cut dynamic range. but well, this is anyway not an issue in situations where you usually use HIGH ISO like 800 and above that.

    but, even if your assumptions on whatever observations you made would be correct and it would be true that the multiples of 160 are generated by pulling 1/3 of a stop of the next higher what you call “native iso” while meaning the next higher 100x, how would this explain that ISO 640 has obviously less noise than ISO 100?

    you say: “ISO 160 is the cleanest because it is the native ISO 200 with a 1/3 stop digital exposure pull, yielding even less noise. ISO 320 is actually ISO 400, with a 1/3 stop exposure pull, etc.”

    now look at this comparison and explain how these SHOWN test results fit in your logic. ISO 640 is according to your explanation ISO 800 with pulled exposure… so far so good…. BUT, do you really think, that it therefore has less noise than ISO 100…?

    all kinds of tests show ISO 640 cleaner than ISO 100!

    ISO 800 as a native 100x-ISO (your definition) is 3 stops higher than ISO 100!

    additionally, please have a look how different the amount of noise is between 800 and 640.

    http://www.photographybay.com/2011/05/01/proof-that-multiples-of-iso-160-work-best-on-canon-hdslrs/

    there is also a video:

    http://vimeo.com/23082874

    i shoot since almost 8 years RAW now… on several 1D and also a the 5Dmark2. i do this for a living. i have improved my RAW shooting technique so massively since i started ignoring what my meter shows and started to expose my RAW files like they and the individual scenes deserve it. especially under artificial light.

    HIGH ISO is no issue at all for me. lens cap tests are just as ridiculous as unproven theory talk anyway…

    by the way, did you know that you improve your signal/loss ratio by using a higher ISO? its all about the correct exposure. that makes the magic happen.

    it makes me smile how unexperienced photogrpahers are “scared” of using high ISO settings… grinding always at much too low ISO settings underexposing the heck out of their scenes… with the correct exposure HIGH ISO is no problem at all. all this noise discussion is just meaningless. i dont hesitate to also use ISO6400 and on the 5D also 12800… the 1DMk3 is a killer regarding noise… unfortunatly it operates only up to ISO 6400.

    i shoot a lot of commercial photography for tourism, but also event and documentation shoots for large companies and also high priced weddings. ( a lot of dim lit situations there).

    i easily shoot ISO 12800 on my 5D MARK II without even applying noise reduction and i get cleaner files than many guys who shoot ISO 1600 or ISO 3200 while relying totally on what their meters show them: especially in dim lit situations with artificial light a crucial error leading to massively underexposed images.

    i would really be very much interested in seeing on what kind of files or photographed scenes you base your assumptions.

    especially regarding the cut of dynamic range you are mentioning… how did you measure that? did you compare histograms?

    curious… with kind regards

    luke

  24. Probably the best explanation of this ISO phenomena I have read, having heard numerous people say a similar thing. Thanks for postng this.

  25. damn it

    just use real film


Leave a comment


*