Using ISO and ISO invariance

Discussion in 'Beyond the Basics' started by freixas, Apr 10, 2019.

  1. freixas

    freixas TPF Noob!

    Joined:
    Jun 11, 2010
    Messages:
    112
    Likes Received:
    22
    Location:
    Portland, Oregon
    Can others edit my Photos:
    Photos NOT OK to edit
    OK, I've been a photographer for a long time (as in, I've spent more years with analog than digital cameras). After looking at some really long threads about ISO elsewhere, I'm starting to understand how ISO works in digital cameras. It's not ASA.

    I'm convinced: you lower noise by giving the sensor as much light as possible. For cameras that are not ISO invariant, for any given exposure (shutter/aperture), a higher ISO setting may produce less noise than a lower one. For an ISO invariant camera, the ISO setting won't matter.

    The light meter is always going to factor in the ISO value. So what's the best strategy for setting exposure on ISO variant and ISO invariant cameras?

    Let's pick a specific scenario: I'm going to photograph birds with a hand-held 400mm zoom. I need at least a shutter speed of 1/400, but even faster would be better. I pick a shutter speed, not less than 1/400. For the aperture, I'll live with whatever I can get.

    If the camera is "ISO variant", I'll set the mode to shutter priority and ISO to auto. The hope is that the camera will raise ISO only after the aperture is maxed out. If the aperture is also important, I'll go to manual and also set the largest aperture I can live with. ISO is still auto. Some bird shots may turn out to have a lot of noise, but this will always be the best possible result given the camera and the lighting conditions.

    If the camera is ISO invariant, I could use the same strategy or I could leave the ISO at 100, so that the camera maximizes the amount of light received by the sensor (given the limitations of the shutter and possibly aperture settings). The camera might flash warnings that the scene does not have enough light; that's OK, as long as it allows the photo to be taken.

    An alternate scenario: I'm going to photograph landscapes and I can put the camera on a tripod. Nothing is moving, so I can use a long exposure. The aperture might be the only thing I care to set. What's the best strategy now?

    I'm less sure here. For an ISO variant camera, I want the longest exposure that doesn't blow out the highlights, so I might set the ISO to 100 and adjust the shutter speed by going to manual mode. Or I could rely on the camera's "evaluative metering", but check the histogram. I could then raise the ISO, if needed, until the highlights are no longer clipped. If the dynamic range is extreme, I could take multiple photos to combine as an HDR.

    I'm not sure that an ISO invariant camera would be set any different, although I see articles which seem to suggest that these cameras have an advantage in this scenario. It's not clear; for every camera, you want to maximize the amount of light received by the sensor. The brightest areas of a scene may limit how much light you can allow in before those parts are clipped. This limit affects all parts of the scene, so it's not clear how ISO invariance helps.

    Any tips on how to set exposure in various scenarios or corrections to my guesses above would be welcome. Thanks!


     
  2. petrochemist

    petrochemist TPF junkie!

    Joined:
    Mar 9, 2014
    Messages:
    1,445
    Likes Received:
    426
    Location:
    North Essex UK
    Can others edit my Photos:
    Photos OK to edit
    I think that's generally right, BUT it's not as much as possible just maximised within the sensors range. A 30s exposure at f4, outside at noon will be just pure white as every pixel is clipped to it's maximum.
    With non ISO invariant bodies the degree of amplification prior to digital conversion varies with the set ISO, and this can help reduce noise. Of course there are multiple factors involved so it's not always the case.

    With small birds subject movement is often an issue so you might need shutter speeds much faster than simply the reciprocal of your focal length. Fast long lenses are expensive & heavy so I'm likely to be limited to f/4 or slower. Noise is generally preferable to movement blur IMO.

    I've usually found some movement in my landscapes, usually just blown foliage, but also water, & even stars. The best tripod mounted compromise here depends a lot on the conditions - at night & using too much ISO, simple shot noise can become an issue - at other times diffraction may limit the usable aperture, or an ND filter may be needed to get the DOF/shutter speed required for the desired effect. With a few of my adapted lenses (some are fixed f/2.8 or even f/1.2) a ND filter has been needed to shoot at base ISO outdoors - my camera simply can't go fast enough.

    In any of the more extreme situations bracketing your shots is the best answer, at least if you have time - my wife won't wait for such matters!
     
    • Like Like x 1
  3. Ysarex

    Ysarex Been spending a lot of time on here!

    Joined:
    Nov 27, 2011
    Messages:
    6,266
    Likes Received:
    2,887
    Location:
    St. Louis
    Can others edit my Photos:
    Photos OK to edit


    Raising the ISO in the scenario you're describing in this paragraph would further clip the highlights not reduce clipping.



    Yes! Overall image IQ apart from the shutter/motion and aperture/DOF concerns is a function of exposure -- more exposure is better. Exposure is primarily what determines noise in the image. This is the same for both ISO invariant and non-invariant sensors. And so the same usage strategy applies to both.

    If ISO invariance is no more than an academic curiosity it's not worth a lot of attention. You're looking in this post at the correlation between ISO and noise and that's the common view of what ISO does given the Exposure Triangle model of confusion. ISO reduces sensor DR. With non-invariant systems we had to make a trade-off choice between loss of DR and increased read noise (in a non-invariant system ISO amplification suppressed read noise at a level that mattered). It was typically worth the trade to lose a couple stops of DR for less noise in the shadows. An ISO invariant system gives us the option to retain full sensor DR without the read noise penalty. And that's where it can become a strategy in pragmatic use; in very high lighting contrast where the DR loss would matter.

    iso_invariant3200.jpg

    In the above image the exposure was calculated for the basket/chair with the ISO set to 3200. Exposure was 1/8 sec, f/6.3. At ISO 3200 the sensor DR for the camera I used was 4.5 stops. The window is nuked into oblivion. But the sensor DR for that camera at base ISO is 9 stops. So I used the same exposure of 1/8 sec, f/6.3 but set the ISO to base 125 and took the photo again. The SOOC JPEG shows the basket/chair as nearly black. But because the sensor is ISO invariant I pay no penalty to lift the shadows in post process (same exposure means I record the same shadow data) and because I retained the sensor's full DR capacity I have a much different rendering of window.

    iso_invariant125.jpg

    Joe

     
    • Like Like x 1
  4. freixas

    freixas TPF Noob!

    Joined:
    Jun 11, 2010
    Messages:
    112
    Likes Received:
    22
    Location:
    Portland, Oregon
    Can others edit my Photos:
    Photos NOT OK to edit
    Sorry, because of the way the quoting came out, your reference was hard to track. I found it and yes, I said the opposite of what I should have: if the highlights are getting clipped, one show lower ISO, not raise it.

    That makes sense. The sensor receives whatever light it gets. The higher the ISO, the more likely the pixels receiving the most light are likely to get clipped.

    You go on to give an example of two shots, one with a high ISO where the highlights are blown; the second with a low ISO but same exact shutter/aperture--the highlights are not blown and the shadows can be raised to the same level is in the image with the high ISO and with no worse noise penalty.

    But if the sensor is ISO invariant, then it seems there should be absolutely no difference in the two shots. The sensor records the same exact data. You should be able to take the first image and reduce the highlights, just as you raised the shadows in the second image. That is, unless the RAW file scales the pixels based on ISO so that the highlights are lost beyond any recovery. So maybe things aren't quite as "invariant" is I assumed. I was assuming that an ISO invariant camera would deliver the exact same RAW data regardless of the ISO setting, with only some meta data (the ISO) to determine the default way the data would be viewed.

    If you get the same data regardless of ISO ("true" ISO invariance), then you should be able to shoot any scene with a low or high ISO and not worry about it.

    If you do not get the same data (the DR drops as ISO is raised and the clipped pixels are permanently clipped), then you might as well always shoot with the lowest ISO. It won't hurt the shadows and will prevent highlight clipping. You can adjust whatever you want in post. Other than saving some post processing time, I would see no advantage to ever using anything except the lowest ISO.

    For an ISO variant camera, you can choose to accept more noise in the shadows (low ISO, raise the shadows), accept blown-out hightlights (high ISO and somewhat less noise in the shadows) or take two shots and combine them as an HDR. In that sense, I now get the value of an ISO invariant camera in the scenario I was asking about.

    Sadly, Canon seems to be behind the curve on this and I have a bunch of Canon lenses. Supposedly, the Canon 5D Mark IV is "invariant" above ISO 400. The Nikon D810 is looking awfully attractive; a little cheaper and with better IQ. *Sigh*
     
  5. Derrel

    Derrel Mr. Rain Cloud

    Joined:
    Jul 23, 2009
    Messages:
    44,558
    Likes Received:
    16,944
    Location:
    USA
    Can others edit my Photos:
    Photos OK to edit
    I agree with this: raising the ISO typically REDUCES usable dynamic range. Yes.
     
  6. Overread

    Overread has a hat around here somewhere Staff Member Supporting Member

    Joined:
    May 1, 2008
    Messages:
    24,716
    Likes Received:
    4,479
    Location:
    UK - England
    Can others edit my Photos:
    Photos OK to edit
    Sony threw insane money at sensor development because they wanted a serious slice of the camera market. This was money that neither Canon nor Nikon could invest in the same quantity to the point where Nikon just outright bought sony made sensors for a while. I believe that they are now using their own and that Canon is also playing a steady catchup to the new tech.

    In general the leader always shifts around; at one time Canon utterly blasted Nikon for ISO and sensor performance. So sometimes its easier to hold onto your gear and ride out the minor shifts in the market. Remembering that often most of us can't afford to upgrade every time a new body comes out nad that if you change brands - esp if you're heavily invested into a brand already - it will cost you a lot more to sell and replace like for like. Of cousre if you've only got cheaper end gear and are upgrading to much higher levels that can be a time where it might be practical to consider a brand switch.
     
    • Agree Agree x 1
  7. Ysarex

    Ysarex Been spending a lot of time on here!

    Joined:
    Nov 27, 2011
    Messages:
    6,266
    Likes Received:
    2,887
    Location:
    St. Louis
    Can others edit my Photos:
    Photos OK to edit
    It does! And that's a huge point of revelation; same exposure (shutter speed + f/stop) and you get the same data. The sensor delivers an analog signal of electrons -- voltage. But it doesn't end there. That analog voltage signal must be converted into numbers.

    Nope, that data no longer exists. That's how ISO reduces DR. ISO's job is to maintain a standard brightness level in the output JPEG. And so for each stop of ISO increase some method of brightening the data coming from the sensor is applied. Assume that in the exposure used at the base ISO shot the highlights out the window were placed right at the sensor's clipping threshold (they were). Then in the ISO 3200 shot at the same exposure ISO's job was the brighten the image by 4 2/3 stops. It does that by taking the sensor signal and boosting it up in total 4 2/3 stops and any data that was 4 2/3 stops below the sensor clipping threshold gets pushed to clipping. We call the process ISO clipping.

    ISO clipping occurs in the process of creating the raw file. It becomes permanent in the raw file and the loss can't be undone.

    That's an option -- Sigma cameras work that way but most of our cameras either use an analog amplifier to gain the sensor signal before the raw file is created or they multiply the digital conversion values during the creation of the raw file so that the required ISO brightness level is reached in the creation of the raw file.

    That's not what's typically meant by ISO invariance. ISO invariance means that the read noise in the system is so low that raising ISO by the use of an analog amplifier to gain the sensor signal (which would improve read noise) provides no real benefit and so digitally scaling the data in a raw converter works just as well.

    Some pretty fringe nut-jobs take that position but it's pretty extreme behavior. There's no harm done raising the ISO if circumstances require it and you don't have a high DR condition that would suffer the DR loss. The benefits include a better preview JPEG, a better usable EVF image for mirrorless shooters, what if you want to share images immediately....

    Those example photos were taken with a Canon compact G7xmkii and the sensor is surprisingly ISO invariant. Although it's becoming more common I don't see ISO invariance as a big selling feature for a camera. My Fuji XE-2 was one of the most ISO invariant cameras available and when I upgraded to an XT-2 I lost that as the XT-2 is less ISO invariant. That didn't really factor into my decision to upgrade.

    Joe
     
    Last edited: Apr 10, 2019
    • Like Like x 1
  8. Overread

    Overread has a hat around here somewhere Staff Member Supporting Member

    Joined:
    May 1, 2008
    Messages:
    24,716
    Likes Received:
    4,479
    Location:
    UK - England
    Can others edit my Photos:
    Photos OK to edit
    I think the value for it varies depending on the photographer and situation - I know I've been at the highest ISO my 7D can use and still had underexposure. An ISO Invariant sensor would have been invaluable as the light grew more challenging to work with. Meanwhile a studio photographer might never have or rarely have need of such ISO performance because they've always got more studio and flash lights to adjust the light as they need it.
     
  9. freixas

    freixas TPF Noob!

    Joined:
    Jun 11, 2010
    Messages:
    112
    Likes Received:
    22
    Location:
    Portland, Oregon
    Can others edit my Photos:
    Photos NOT OK to edit
    This sounds insane! I'm sure there's some obscure technical reason for it, but still. The clipping process gains you nothing—it only throws information away. What a waste!

    Got it. Again, if the RAW file didn't do ISO clipping, but retained the ISO number, you could have preview JPGs, EVF images, etc. that would look just as they would if you did the clipping. It might require a bit more processing power; performance limitations are about the only reason I can think of for ISO clipping the RAW data.

    Thanks. There are a bunch of factors which make any upgrade path difficult. I currently have two Canon cameras with APS-C sensors and two EFS (not full-frame) lenses. Two Nikon D810s with a 400mm telephoto plus a zoom and wide angle lens would fit my workflow just great, but my budget not at all. However, even going to just one Canon 5D Mark IV requires replacing the two EFS lenses and is still pretty pricey. On the plus side, I might be able to use my 1.4x extender with auto-focus, something I can't do with my EOS 70D.

    Anyway, thanks for the info! The whole ISO thing has been an eye-opener and I started from one of your posts on another thread,.
     
  10. freixas

    freixas TPF Noob!

    Joined:
    Jun 11, 2010
    Messages:
    112
    Likes Received:
    22
    Location:
    Portland, Oregon
    Can others edit my Photos:
    Photos NOT OK to edit
    Thanks for the perspective. Of course, waiting out market shifts is easier at 35 than at 65. The older one gets, the more important instant gratification becomes. :)
     
  11. Ysarex

    Ysarex Been spending a lot of time on here!

    Joined:
    Nov 27, 2011
    Messages:
    6,266
    Likes Received:
    2,887
    Location:
    St. Louis
    Can others edit my Photos:
    Photos OK to edit
    The reason for it is that the design focus of the camera is the SOOC JPEG. The assumption is that when the ISO is raised you're going to lower exposure to place diffuse highlights where they should be after the ISO brightening is applied. A JPEG shooter would approach a shot like the one above and insist that the exposure had to be set for the subject and non-important highlights like the window had to be let go. Remember that in the ISO 125 shot the basket and chair were rendered in the camera JPEG as nearly black.

    Again, Sigma cameras take that approach. The ISO value is simply stored in metadata and then applied to the raw file as needed.

    Joe

     
    Last edited: Apr 10, 2019
  12. Derrel

    Derrel Mr. Rain Cloud

    Joined:
    Jul 23, 2009
    Messages:
    44,558
    Likes Received:
    16,944
    Location:
    USA
    Can others edit my Photos:
    Photos OK to edit


    I am not in agreement with that statement.. not since the Canon 5D was my primary camera, back in 2006. With "some", older d-slr models, yes, lowest-ISO meant BEST quality, for sure, but since ISO-invariant sensors were developed, and since SONY developed the Exmor generation of sensors, the old admonition to "use the lowest ISO" has been rendered an old-fashioned idea.

    Just yesterday, I was cleaning up my iMac's HD and I found a side-by-side comparison done by TPF's own "Tirediron", comparing the Nikon D800 against the Nikon D850. Same test scene, but substantially BETTER detail from the D850. THE ISO used? ISO 6,400. Lemme see... 100, 200, 400, 800, 1,600, 3,2003, 6,400.. SIX full stops more shutter speed above ISO 100...

    When I need shutter SPEED, or more DOF, or farther "throw" with a flash, or faster flash recycle times,elevating the ISO is often my FIRST move.
    This is 2019, and sensor capability has changed significantly... we are now free to "use" different ISO setting and ISO invariance in multiple ways..
     
    Last edited: Apr 10, 2019

Share This Page

Search tags for this page

iso invariance table