Luminous Landscape Forum
Equipment & Techniques => Digital Cameras & Shooting Techniques => Topic started by: JonathanRimmel on December 02, 2011, 01:09:11 pm
-
I have been surfing the web checking out various photography related articles and videos and have come across a few which indicated the ISO's I thought were native to my camera may not be. Perhaps there is already some info pertaining to this on LuLa, if so please point me in the right direction. What my real question is what ISO's are best to use? Should I use 200, 400, 800, 1600, ect or the intermediate ISO's such as 160, 320, 640, 1250, ect. Are there ISO's that give better overall performance and translate to better image quality that others? (obviously higher ISO's usually translate to more noise) Is there perhaps something I am missing here? I am very much interested in this topic as I have to use high ISO's frequently for my work and I would like to get a higher level of image quality than what I am.
-
There is not a clear response.
To obtain the maximum quality in terms of noise, expose as much as possible your RAW file without clipping the highlights (ETTR), using the lowest real ISO that allows you to achieve that. Only real ISOs are worth using, so intermediate ISOs should be avoided. On the other hand ISOs over ISO1600 don't improve noise on any camera, so they can also be avoided (in fact in Sony sensors, it's hardly worth to go beyond base ISO100). I.e. if you still cannot ETTR at ISO1600, it is not worth pushing ISO any more, just stay at ISO1600 underexposed and push exposure in the RAW developer (that will save highlights at no cost).
When shooting JPEG, ETTR is nonsense, and the only rule is expose correctly your JPEG at the lowest possible ISO. So here ALL ISOs are available (intermediate ISOs or any ultra-high ISO value)
But all that story is about image quality in terms of noise. Imagine that to achieve ETTR in the RAW file (or to expose correctly your JPEG) using the lowest possible ISO, you set such a large aperture that produces insufficient DOF, or such a low shutter speed that you loose sharpness because of motion blur. In those situations you are making your final image worse just because of insisting in reducing noise. Lack of sharpness is worse to an image than some additional noise.
Regards
-
What Guillermo said. But to add something: I don't think you need to worry about 160 vs. 200 (on the basis of being an "intermediate" ISO value) except in the case of (some?) Canon SLRs that simulate 200, 400, etc. Instead of amplifying the signal, the camera instead applies an adjustment to the meter, so you're not actually shooting at 200, 400, etc. Instead, if you've got one of those cameras, stick to, as Guillermo said, one of the real ISOs (160, 320, etc.) on your camera.
A reference re: simulated intermediate ISOs: http://www.fredmiranda.com/forum/topic/682945/0
-
Great info gentlemen
Are all cameras "true" ISO's the multiples of 100 or are some multiples of 160?
-
The other aspect of "correct ISO" to consider is dynamic range - the ISO you choose should also provide the greatest dynamic range. For most cameras, the native ISO does this, and is often 200, but that might be something worth testing or at least checking on a reputable camera testing site.
-
Are all cameras "true" ISO's the multiples of 100 or are some multiples of 160?
It all depends on the sensor...Nikon is usually 200, Canon 100. The new Phase One IQ180 backs have an ISO base of 35. So, it all depends...
-
To the best of my knowledge, Canon is the only company for which (some of) their cameras have settable ISOs that are not "real" (w.r.t changing sensor gain), excluding push/pull ISOs at the extrema.
-
Here is a good discussion:
http://www.naturescapes.net/phpBB3/viewtopic.php?f=2&t=139621
Most of the posts (apart from mine, of course ;) ) in the FM thread linked to above are misinformed. What Guillermo wrote above is correct, but only for Canon cameras. Bottom line is that the optimal choice of ISO for a given situation very much depends on the camera model you are using. The point of the ISO adjustment in raw capture is to optimize the DR of the capture device relative to the scene DR. One needs to know the DR characteristics of the camera (which can be obtained eg from DxO measurements available at their website DxOmark.com) and apply that to a given shooting situation.
-
Are all cameras "true" ISO's the multiples of 100 or are some multiples of 160?
I'd say that true ISOs are indicated in the dxomark plots - their way of measuring sensibility according to ISO12232 (and saturation) (http://www.dxomark.com/index.php/About/In-depth-measurements/Measurements/ISO-sensitivity) should effectively defeats the intermediate ISO trick (actually 1/3 over-or underexposing and compensating).
There, you can also see if lowest ISO is actually base ("native") ISO, or overexposed next stop. And you can even see if it's worth to increase ISO (does the DR falls less than 1 EV with 1 stop more ISO?).
-
I'd say that true ISOs are indicated in the dxomark plots - their way of measuring sensibility according to ISO12232 (and saturation) (http://www.dxomark.com/index.php/About/In-depth-measurements/Measurements/ISO-sensitivity) should effectively defeats the intermediate ISO trick (actually 1/3 over-or underexposing and compensating).
DxO misuses the term ISO, which refers to what particular light intensity input yields a particular output brightness. Since raw data is unrendered, ISO does not apply. Their measurement determines what particular light intensity input yields a particular raw level in the raw data, which is something different.
-
Hi,
For best image quality you would use the lowest ISO on the camera and expose to the right, avoiding clipping any important highlights.
Reason: you want to collect as many photons as possible.
A photon is a "particle" or "quantum package" of light.
Best regards
Erik
Here is a good discussion:
http://www.naturescapes.net/phpBB3/viewtopic.php?f=2&t=139621
Most of the posts (apart from mine, of course ;) ) in the FM thread linked to above are misinformed. What Guillermo wrote above is correct, but only for Canon cameras. Bottom line is that the optimal choice of ISO for a given situation very much depends on the camera model you are using. The point of the ISO adjustment in raw capture is to optimize the DR of the capture device relative to the scene DR. One needs to know the DR characteristics of the camera (which can be obtained eg from DxO measurements available at their website DxOmark.com) and apply that to a given shooting situation.
-
DxO misuses the term ISO, which refers to what particular light intensity input yields a particular output brightness.
From what I understood, DxO adapted the saturation-based technique, which would be described for rendered jpegs in the ISO norm, to the raw output, which is definitely what interest us raw shooters isn't it?
It would seem to me that this misuse (if we conform to the lettre of the norm) is indeed more faithful to the goal of the norm, which is (or at least should be?) to give photographers an indication of sensitivity with the greatest repeatability (hence, no rendered converter-dependant jpeg output shall be used but raw output only) and practical interest (hence the reference point taken on raw saturation, which should primarily base the exposure, rather than an arbitrary gray level).
But yes, I fully agree that all the above is subjective speculation rather than strict ISO norm.
Edit : Sorry I didn't open your naturescape link before, and so have to reckon that my precedent assumption is wrong : Nikon's sensors should have measurement points at intermediate ISO in their DxOMark graphs, and it seems that DxOMark just doesn't care about it.
-
From what I understood, DxO adapted the saturation-based technique, which would be described for rendered jpegs in the ISO norm, to the raw output, which is definitely what interest us raw shooters isn't it?
Not exactly. Let's differentiate:
- Real ISO: ISO based on an analogue amplification into the camera. They are the only ISOs useful to improve SNR, but just up to a certain value (ISO1600 on most cameras, ISO100 in the recent Sony sensors)
- Effective ISO: ISO based on levels encoded in the RAW file (i.e. the saturation criteria you mention)
DxO measures effective ISOs (the only ones that can be measured from RAW data), not real ISOs (the ones that need to be analysed to find out).
That means Canon 5D's ISO3200 in DxO is double effective ISO than ISO1600. But 5D's ISO3200 (http://www.guillermoluijk.com/article/isos5d/iso3200z.gif) is not a real ISO since it is achieved in that camera by multiplying (digital amplification) an ISO1600 shot (http://www.guillermoluijk.com/article/isos5d/iso1600z.gif). So 5D's ISO3200 is completely useless to the RAW shooter; you gain nothing in shadows SNR but can clip one extra stop in the highlights.
An interesting case is the Fuji X100 ISO range:
(http://www.guillermoluijk.com/article/dxomark/isox100_large.gif)
Unlike it could seem, this is a CLEVER form of ISO (http://www.luminous-landscape.com/forum/index.php?topic=53451.0). When an ISO superior to ISO1600 is set on this camera, the ISO setting is stored as metadata to let the RAW developer know about it, BUT the shot is internally done at ISO1600. So you can save up to one (ISO3200) or two (ISO6400) stops in the highlights, that other cameras (like the Canon 5D) stupidly clip.
Recent Sony sensors (K5, D7000) could almost work this way from base ISO100. Since there is a negligible SNR improvement from pushing ISO, they could always store the ISO setting as metadata, internally shoot at ISO100, and display a JPEG according to the ISO set by the user but keeping in the RAW file non-clipped highlights. This is what Emil has sometimes refered to as an isoless camera.
Regards
-
How does what is being said in this thread about ISO square with this article, http://www.digitalphotopro.com/gear/cameras/the-truth-about-digital-iso.html and the referenced video (link in the article)?
-
How does what is being said in this thread about ISO square with this article, http://www.digitalphotopro.com/gear/cameras/the-truth-about-digital-iso.html and the referenced video (link in the article)?
The author of the linked article is very much misinformed, for instance 160 is not the 'native' ISO of the 7D, nor is it 'less noisy' than its neighboring ISO's. On the 7D, ISO 160 is derived by pulling ISO 200 via digital multiplication of raw values by 0.8 (160/200). So it is not less noisy, it is the same analog amplification, rather ISO 160 meters 1/3 stop brighter, so leads to 1/3 stop more exposure, so is less noisy because more photons are captured. The read noise test referred to in the link simply shows that if you take ISO 200 read noise and multiply it by 0.8, it looks 1/3 stop dimmer (duh!) but until this is considered in reference to some signal it doesn't mean much -- indeed, if you maintain the same exposure (Tv and Av) both signal and read noise are multiplied by 0.8 and so S/N remains the same.
Rather than giving you 'the truth about digital ISO', the article is instead promulgating a lot of misinformation >:(
-
From what I understood, DxO adapted the saturation-based technique, which would be described for rendered jpegs in the ISO norm, to the raw output, which is definitely what interest us raw shooters isn't it?
It would seem to me that this misuse (if we conform to the lettre of the norm) is indeed more faithful to the goal of the norm, which is (or at least should be?) to give photographers an indication of sensitivity with the greatest repeatability (hence, no rendered converter-dependant jpeg output shall be used but raw output only) and practical interest (hence the reference point taken on raw saturation, which should primarily base the exposure, rather than an arbitrary gray level).
But yes, I fully agree that all the above is subjective speculation rather than strict ISO norm.
There are I think a couple of goals:
1) To know how much headroom you have in the raw data for a given scene illumination (the DxO 'ISO' figure gives you this directly)
2) To know the actual sensitivity of the sensor -- its quantum efficiency, the percentage of incident light that is recorded by the sensor (the DxO 'ISO' figure gives this indirectly, when combined with other of their measurements)
Unfortunately by calling the measurement 'ISO' misleads I think unless one understands what it is for.
-
Most of the posts (apart from mine, of course ;) ) in the FM thread linked to above are misinformed.
Haha! That was actually the thread I was looking for (but couldn't find after a cursory search)! So it turns out you were the author of the post I was actually trying to cite. :)
-
DxO misuses the term ISO, which refers to what particular light intensity input yields a particular output brightness. Since raw data is unrendered, ISO does not apply. Their measurement determines what particular light intensity input yields a particular raw level in the raw data, which is something different.
From what I understood, DxO adapted the saturation-based technique, which would be described for rendered jpegs in the ISO norm, to the raw output, which is definitely what interest us raw shooters isn't it?
It would seem to me that this misuse (if we conform to the lettre of the norm) is indeed more faithful to the goal of the norm, which is (or at least should be?) to give photographers an indication of sensitivity with the greatest repeatability (hence, no rendered converter-dependant jpeg output shall be used but raw output only) and practical interest (hence the reference point taken on raw saturation, which should primarily base the exposure, rather than an arbitrary gray level).
DXO (http://www.dxomark.com/index.php/About/In-depth-measurements/Measurements/ISO-sensitivity) is using the original ISO 12232 saturation standard, where the ISO (S sat) = 78/Hsat. Hsat is the exposure in lux seconds necessary to saturate the sensor. This standard can be applied to exposures with raw files.
The newer ISO 12232:2006 (http://en.wikipedia.org/wiki/Film_speed#Saturation-based_speed) standards also have a saturation standard (SOS) but it applies only to sRGB rendered files and is not of much interest to those of us who shoot raw. The only ISO 12232:2006 standard that can be applied to raw files is the REI (Recommended Exposure Index) which can be anything that the manufacturer wants it to be. My experience with newer Nikon cameras indicates that it is close to the older saturation standard and will give 12.7% saturation when exposed according to the camera light meter. This allows 0.5 EV for highlight headroom.
If one is exposing according to a hand held meter, corrections for light falloff and lens transmission must be made as explained in the DXO post.
Regards,
Bill
-
DXO (http://www.dxomark.com/index.php/About/In-depth-measurements/Measurements/ISO-sensitivity) is using the original ISO 12232 saturation standard, where the ISO (S sat) = 78/Hsat. Hsat is the exposure in lux seconds necessary to saturate the sensor. This standard can be applied to exposures with raw files.
The newer ISO 12232:2006 (http://en.wikipedia.org/wiki/Film_speed#Saturation-based_speed) standards also have a saturation standard (SOS) ...
A little warning: all of that's talk of "true ISO" seems related to measuring the _lowest_ exposure index at which there is adequate highlight handling. It is thus utterly different that what we meant by the ISO/ASA speed of film (actually, a film+processing combination), which was about the _ highest_ exposure index that gives adequate shadow handling --- very crudely, noise floor about four stops below metered midtones. The two measures are roughly like the lower and upper limits of pull/push processing.
Frankly, I suspect that most of us are far more interested in the shadow handling, noise related upper limit on usable exposure index, which by the way seems to be at 1600 or beyond for all modern SLR-sized sensors. And even higher if you wish to compare noise levels after normalization to some moderate resolution level, say the 8MP used in some DXO measures.
-
Maybe somebody reading this very interesting thread can help shed some light on a mystery that has been confusing me for some time concerning an ISO bug that pops up when I run .dng images from my old MFDB that is not officially supported by C1 through this program.
The mystery is that differences in MFDB ISO settings (eg 100, 200) are identified and reported by C1 in the metadata of the .dng files but do not show in the visual appearance or histogram of the image when comparing two files of exactly the same subject shot at different iso (that I often get from bracketing the iso in a portrait session), though the differences in ISO is clearly visible when the same files are run through other programs apart from C1 that do support this MFDB. Now what does that mean?
This is a question I have asked before but got no answer. But at the moment I am more concerned with what follows:
Is it that the actual data of raw files shot at different isos is the same, the MFDB only records the iso data as metadata, then the iso data is applied to the raw file automatically when the computer program opens the file - unless as in the case of C1 the program has not been taught how to do that with the metadata from the MFDB?
If this were the case (a big if) then could I stop wasting my time adjusting iso of raw files in an ignorant attempt to catch the ideal combo of maximum data captured with minimal motion blur, and just do this in post, as adjusting iso while shooting would make no difference to the actual data captured by the camera (apart from the camera settings info. stored in the metadata)?
But come to think of it this must be poppycock, since the iso setting does effect the exposure setting (as reflected in the histogram) and surely this must make a real difference to the actual data captured in the raw file as well as the metadata, right.
Maybe this should be posted under beginners questions rather than here?
-
Is it that the actual data of the raw files shot at different isos is the same, the MFDB only records the iso data I set as metadata, then the iso data is applied to the raw file automatically when the computer program opens the file - unless as in the case of C1 the program has not been taught how to do that?
If this were the case (a big if) then could I stop wasting my time by adjustin iso of raw files in an ignorant attempt to bracket my exposures and just do this in post, as adjusting iso would make no difference to the actual data captured by the camera (apart from the camera settings info. stored in the metadata)?
I would answer yes to both questions, but to make sure just open a couple of your DNG files (same aperture/shutter, different ISO) into Rawnalyze. If they have the same histogram, they hold identical RAW data and you are wasting your time with your ISO bracketings.
Regards
-
Are you using the same f-stop and shutter speed, different ISO speed setting and then getting same raw histogram with the raw converter that does not support that back's raw format?
If so, this sounds like the situation I have heard of with some (older?) DMF backs that the ISO setting has no effect on analog amplification and thus on the raw digital output, but just sets a flag in the raw file which tells the raw converter how much to adjust levels in the digital domain. The thinking being AFAIK, that the ADCs used in those DMF backs could handle the whole dynamic range from full well signal down to the noise floor anyway (being about 12 stops) so that variable analogue gain would be pointless, and indeed a source of degeneration of the signal.
-
Thanks Guillermo & BJL... this is quite intriguing. Maybe my Sinarback Emotion75 is a good solid "100iso"-equivalent (or maybe instead 200iso?) MFDB with a few extra buttons for amusement?
As you suggest Guillermo I downloaded the program Raw Image Analysis version 2.10.1.0 and installed it on my Win7 PC to compare 2 .dng files from the MFDB. Unfortunately the program could not fully open the files instead reporting invalid image file directory form (maybe as the Sinarback .dng format was not around when the the program was developed?) so no histogram :-( but the Rawanalyze program did provide the following information:
for image 1:
Sinarback Emotion 75, 2011:09:27 08:27:35, 4992x6668pix
ISO 100, F/5.6, 16666/1000000s, Corr 1/3 EV
for image 2:
Sinarback Emotion 75, 2011:09:27 08:27:43, 4992x6668pix
ISO 200, F/5.6, 16666/1000000s, Corr +2/3 EV
As you can see it reports 2 different values for the EV correction (my emphasis). Could this be a clue to how the .iso works?
-
Since your back outputs dng files, you have a couple of options. Among them are
1) Download the command line tool dcraw and execute the command
dcraw -4 -T -D sinarfile.dng
for two images shot at different ISO. This will process the raw file without demosaic or any other processing to a tiff file. Open the resulting sinarfile.tif's and see if they have the same brightness/histogram; if they do then ISO is purely metadata on your camera.
2) RawTherapee 4.0 opens .dng files; it may be able to open yours. There is an option in the histogram palette to display the raw histogram, so you should be able to see whether the histograms for two different images shot at different ISO.
Guillermo's histogrammar program may be a third option.
-
Aliens 4 - Humans 0 >:(
* Reansalyse cannot open the Sinarback eMotion75 Raw .dng file
* DCRAW just flashes in a black box on my screen and disappears forever into the ether
* DCRAW X (for Mac) cannot open the eMotion75 raw .dng files
* Another nice raw interface program MEGUI that uses DCRAW can open raw .dng files from Canon 5D that I have on my computer but crashes when I feed it a raw .dng file from the Sinarback
So maybe my MFDB is the alien???
Meanwhile, I have been informed by a friend who knows about such things that it is true, the .iso switch on the Sinarback eMotion75 is mainly decorative, though have not been able to verify this myself yet.
If this is true, which increasingly seems to be the case, then what are the implications:
a) perhaps things do not always work as we are lead to believe they do
b) maybe this applies to your MFDB too?
c) perhaps more recent MFDBs work differently in this respect?
d) what is the point of relying on the histogram on the MFDB for such entertainments as ETTR etc if it is just hooked up to an iso switch that is mainly decorative….
e) am I the only one who didn't know this already???
Fortunately none of this need prevent me from continuing as I have always done opening the .dng files in C1 and boosting the exposure to the desired level. In my opinion This still produces a "better" 16-bit TIFF file than importing and exporting the .dng file from Sinar's own Captureshop or Adobe Raw (even though both of the latter two programs do recognise and automatically adjust the image and histogram to the iso setting in the metadata of the file) so life could be a lot worse.
-
First off, thanks to everyone for your responses. It would seem once again Canon deviates from everyone else in how they handle ISOs. I use a Canon 7D at work and a Nikon at home. I am still a bit uncertain here though, what ISOs are considered "native" and additionally, do I understand correctly that it would be better to underexpose (using a fast enough shutter speed to prevent blur) rather than use ISOs above 1600 on any DSLR and then brighten up the image in ACR / Lightroom?
-
First off, thanks to everyone for your responses. It would seem once again Canon deviates from everyone else in how they handle ISOs. I use a Canon 7D at work and a Nikon at home. I am still a bit uncertain here though, what ISOs are considered "native" and additionally, do I understand correctly that it would be better to underexpose (using a fast enough shutter speed to prevent blur) rather than use ISOs above 1600 on any DSLR and then brighten up the image in ACR / Lightroom?
The ISO's that are purely analog gain on the 7D are 100-200-400-800-1600-3200-6400; all others are obtained by digital multiplication from the nearest of these (eg 160 is 200 pulled 1/3 stop, 250 is 200 pushed 1/3 stop).
As for underexposure at high ISO, yes there is not much advantage in going above 1600 as far as the quality of the raw data is concerned. Though ACR/LR is not the most accurate processor if you need large amounts of exposure compensation; note also that ACR changes its internal defaults as a function of ISO, so the result out of the converter will differ from a file taken at a higher ISO (for instance more NR is applied to the higher ISO image). You can be confident however that the raw files are equivalent and equivalent results can be had if the converter is up to the task.
-
d) what is the point of relying on the histogram on the MFDB for such entertainments as ETTR etc if it is just hooked up to an iso switch that is mainly decorative…
ETTR based on the histogram in a camera where ISO is pure metadata, is a guarantee to be underexposing by at least as many stops as the difference between camera's base ISO and the ISO set by the user. In other words, a ruin.
-
Apologies Jonathan for deviating so far into the bushes from the main fairway of this thread that is digital camera's native iso. To avoid this I have posted a new thread on MFDB iso & ETTR here (http://www.luminous-landscape.com/forum/index.php?topic=60009.0). Hope this is ok.
-
ETTR based on the histogram in a camera where ISO is pure metadata, is a guarantee to be underexposing by at least as many stops as the difference between camera's base ISO and the ISO set by the user. In other words, a ruin.
This is true of any camera. To raise ISO is to gather less light (normally...if you're just increasing ISO to push the histogram then photometric exposure remains the same.) It's just a question of when the exposure compensation gets applied...either immediately by the sensor's amplifiers, or later in post.
-
Hmm. Perhaps I am just dense, but I am not entirely sure I understand this just yet. (although I might be getting closer)
It sounds to me like it would be more beneficial for me to set the ISO to the base ISO (100?) and use manual exposure mode to get the shutter-speed and aperture I want even if it underexposes and just brighten up and increase exposure later on in post.
But if the above was the case, wouldn't this go against the ETTR line of thought? Wouldn't noise be just as bad as increasing the ISO?
I just want to get the best IQ I can in very low light situations. It wouldn't be so bad if most of my low light photography didn't involve people, in which case I could just use a slower shutter speed.
-
To raise ISO is to gather less light (normally...if you're just increasing ISO to push the histogram then photometric exposure remains the same.).
This is not the case we are speaking here. We mean pushing ISO because maximum aperture/slowest shutter allowed do not permit ETTR. In that case pushing ISO to get ETTR improves noise in most cameras; not in the back discussed here (since ISO is just metadata on it).
-
But if the above was the case, wouldn't this go against the ETTR line of thought? Wouldn't noise be just as bad as increasing the ISO?
When RAW exposure is still not maximum, ETTR can be achieved in 3 ways:
- Use slower shutter speed
- Use wider aperture
- Pushing ISO
In some cameras (the MFDB discussed here, cameras with the Sony sensor: K5, D7000,...) the third option doesn't work, and ETTR must be achieved only through maximizing the amount of collected photons. In other cameras (most of them) pushing ISO is a good idea to achieve ETTR is a good idea if we cannot get it by optical means.
Regards
-
When RAW exposure is still not maximum, ETTR can be achieved in 3 ways:
- Use slower shutter speed
- Use wider aperture
- Pushing ISO
In some cameras (the MFDB discussed here, cameras with the Sony sensor: K5, D7000,...) the third option doesn't work, and ETTR must be achieved only through maximizing the amount of collected photons. In other cameras (most of them) pushing ISO is a good idea to achieve ETTR is a good idea if we cannot get it by optical means.
So it looks as if I still want to use ETTR even in low light situations even if this means pumping up the ISO. I often find myself in situations where this is the only choice anyway. The fastest lens I have at my disposal at work is an F/1.8 50mm but this often times is not the right focal length for the job so I have to use one of the two other lenses here which are usually set at 5.6. This means a slow shutter speed. But with people in the scene I try best to stay above 1/80. This usually means an ISO of 1600 - 3200. This isn't as much of a problem with my own setup at home, but at work...
-
This is not the case we are speaking here. We mean pushing ISO because maximum aperture/slowest shutter allowed do not permit ETTR. In that case pushing ISO to get ETTR improves noise in most cameras; not in the back discussed here (since ISO is just metadata on it).
Okay. It's just that I personally don't consider that to be ETTR simply because ETTR is about increasing the light collected. If you're increasing ISO just to take advantage of some read-noise quirk of the camera then that's a different technique altogether (and one I feel is of dubious value...but to each, his own.)
-
Okay. It's just that I personally don't consider that to be ETTR simply because ETTR is about increasing the light collected. If you're increasing ISO just to take advantage of some read-noise quirk of the camera then that's a different technique altogether (and one I feel is of dubious value...but to each, his own.)
I extend your definition of ETTR to RAW exposure. After all ETTR stands for 'Expose [the histogram] to the right', does ISO affect the histogram?.
If you feel ISO is of dubious value to reduce noise, just compare what happens in a Canon camera when ISO is pushed (same aperture/shutter on both shots):
Canon 350D:
(http://www.guillermoluijk.com/article/iso/bodegon.jpg)
(http://www.guillermoluijk.com/article/iso/versus.jpg)
Another example of marginal noise improvement thanks to pushing ISO (same aperture/shutter on both shots):
Canon 5D (image on the right):
(http://img826.imageshack.us/img826/415/16799979.jpg)
(http://img268.imageshack.us/img268/8320/isouf.jpg)
In some other cameras ISO is (nearly) useless to reduce noise (the K5 on the left would have hardly got any benefit from pushing ISO).
Regards
PS: the fantastic still life picture of mine is copyrighted, please don't make money out of it without written permission.
-
If you feel ISO is of dubious value to reduce noise, just compare what happens in a Canon camera when ISO is pushed (same aperture/shutter on both shots):
But when would you ETTR by 4 stops? If every element in your scene is either 12.7% gray or darker, the most you can overexpose by is 3 stops. I find that saturated reds and blues really limit ETTR. Taking a pic of a red fire truck? You'll get 2/3rds of a stop before you clip the red channel.
I was wondering what improvements you'd get with only 2 stops of ETTR. I don't have a Canon 5DMII, but I figured I could fake it. I downloaded the ISO 100 and ISO 400 test images from Imaging Resource, increased exposure on the ISO 100 image, and then compared the three darkest gray patches from the ISO 100 image to the gray patches that were the closes match in the ISO 400 image. Then I did the reverse (though that didn't quite work as well.) The results are in the attached image...
As we already know, applying EC to a very dark tone results in more noise than having set a higher ISO. That's visible in the first comparison. But with an increase of two stops, the difference isn't much.
However, when ETTR is used, exposure isn't increased in post...it's decreased. So the second comparison shows the ISO 400 image decreased (I should have played with EC more to better match the patches...sorry.) Even if you zoom in to 300%, there's no real discernable difference in the noise levels.
This is the basis of my feeling that boosting ISO is of dubious value...you'll rarely boost more than two stops, and when you reduce exposure in post the apparent noise appears the same. Just wanted to clear that up.
PS: the fantastic still life picture of mine is copyrighted, please don't make money out of it without written permission.
I'll try to fight the urge ;D
-
I was wondering what improvements you'd get with only 2 stops of ETTR.
That strongly depends on the sensor technology and how far from saturation we are exposing (i.e. how deep the shadows are). For Canon cameras, in the deep shadows going from ISO100 to ISO200 nearly halves the visible noise (6dB of improvement), and pushing to ISO400 improves by some extra 4dB. So the 2 stops from ISO100 to ISO400 mean an improvement of up to 10dB, i.e. 3 times less noise in the deep shadows for 2 extra ISO stops.
Following the thick black curves you get the SNR improvement for these 3 cameras:
(http://www.guillermoluijk.com/article/snr/curvassnr_iso.gif)
As can be seen, most of the improvement is obtained in the first ISO pushes (going from ISO100 to ISO200 improves more than going from ISO200 to ISO400, and so forth). Improvement in Nikon sensors is more reduced than in Canons. Improvement in the K5 (Sony isoless sensor) is negligible.
Regards
-
As can be seen, most of the improvement is obtained in the first ISO pushes (going from ISO100 to ISO200 improves more than going from ISO200 to ISO400, and so forth).
Yes, I understand the numbers...I do believe that the read-noise issue is real. I simply question whether it's visible (in a pixel-peeping sort of way, mind you) in actual photographic works.
-
An obvious reason why, with many cameras, it makes sense to increase ISO above the minimum (which I insist is not the ”native" ISO, and has nothing to do with a film's ISO speed) when you cannot get enough light to fill the photosites is to reduce the effect of noise sources that enter after the amplification but before conversion to digital is completed. One such source is quantization noise in the ADC, when its dynamic range (true bit depth) is less than that of the sensor's signal. Another is read noise entering during signal transmission after the amplification. The latter is a factor in CMOS sensors that do the amplification in charge transfer from photosite to the edge of the sensor, before transmission along the sensor edge and so on. At least some Canon sensors described in research papers do the variable gain this early; I do not know for sure about any sensors actually in production but the graphs above suggest that Canon does this more than others.
As to the debate about ETTR, I think of this strategy as exposing the ADC to the right, ensuring that the most significant bits in the output are used.
P.S. If using the alternative is using a lower ISO setting and then doing extra work pushing levels up in post, and the results from that extra effort are at best no better and more likely at least a little bit worse, what reason is there for not using the higher ISO setting that gets the midtones at the normal, convenient level?
-
G-
Love your graphs. But they would be much more helpful (especially to the uninitiated) if the variables were more clearly defined. For example, in the above does the abscissa (EV) include ISO as well as actual exposure? Does the heavy black line then imply response at constant exposure? (EJ Martin's definition: exposure does not include ISO). And if so, why label the axis as "EV"? Sorry if I'm being a little dense.
-
...exposure does not include ISO
Photometric exposure doesn't include ISO. Photographic exposure does include ISO.
Photometric exposure is about light.
Photographic exposure is about the response to light. You can't have a photograph without something responding to light. ISO defines that response.
Since the function of ISO and the function of software based Exposure Compensation is similar, it's also appropriate to speak of software based EC as adjusting photographic exposure.
-
Photometric exposure doesn't include ISO. Photographic exposure does include ISO.
Photometric exposure is about light.
Photographic exposure is about the response to light. You can't have a photograph without something responding to light. ISO defines that response.
Since the function of ISO and the function of software based Exposure Compensation is similar, it's also appropriate to speak of software based EC as adjusting photographic exposure.
Exposure is measured in lux seconds. Digital sensors can be rated by the ISO 12232 saturation standard, where the ISO (S sat) = 78/Hsat. Hsat is the exposure in lux seconds necessary to saturate the sensor. If you change the "exposure" in the raw converter, you really have not changed the number of lux seconds reaching the sensor. I don't see your point.
Regards,
Bill
-
Exposure is measured in lux seconds. Digital sensors can be rated by the ISO 12232 saturation standard, where the ISO (S sat) = 78/Hsat. Hsat is the exposure in lux seconds necessary to saturate the sensor. If you change the "exposure" in the raw converter, you really have not changed the number of lux seconds reaching the sensor. I don't see your point.
My point is described clearly in my post.
-
Photometric exposure doesn't include ISO. Photographic exposure does include ISO.
A rose by any other name. Good grief, I don't feel even a bit qualified to participate in this discussion. All I wished to know is if the abscissa of Guillermo's graphs are "photometric" or "photographic" exposure.
-
Yes, I understand the numbers...I do believe that the read-noise issue is real. I simply question whether it's visible (in a pixel-peeping sort of way, mind you) in actual photographic works.
Of course that will depend on the level of demand of your application. It is not the same to use an image to build a 50x50 icon for the web (where even the worst compact camera will suffice), than to print a large copy. The extra cleaneness because of cleverly using ISO will be there for you in case you need it.
what reason is there for not using the higher ISO setting that gets the midtones at the normal, convenient level?
If there are absolutely no highlights that could get clipped, then I agree with you. There is no reason not to use higher ISO settings and I would use them. But as soon as there is some highlight source that could get clipped, its surrounding area will be better represented (softer transitions) the lower the ISO setting.
This is just a simulation, but represents the kind of unpleasant colours due to partial channel saturation thay may occur because of RAW clipping:
(http://www.guillermoluijk.com/misc/saturacionparcial.gif)
Another example (this one is real) in reply #3 of this thread (http://www.luminous-landscape.com/forum/index.php?topic=60031.0).
A rose by any other name. Good grief, I don't feel even a bit qualified to participate in this discussion. All I wished to know is if the abscissa of Guillermo's graphs are "photometric" or "photographic" exposure.
The abscissa is the RAW exposure measured with respect to sensor saturation (0EV), so it includes the effect of ISO (what you call photographic exposure). That is why there is a plot for every ISO. The black thick lines represent how SNR improves if you keep aperture/shutter and start to push ISO.
-
Thanks Guillermo :)
That's very clear.
-
In reply to me, Guillermo quite rightly said:
If there are absolutely no highlights that could get clipped, then I agree with you. There is no reason not to use higher ISO settings and I would use them. But as soon as there is some highlight source that could get clipped, its surrounding area will be better represented (softer transitions) the lower the ISO setting.
Yes, and that suggests a strategy for dealing with the strange asymmetry of digital camera response that does not exist with film: that midtones are typically placed only about three stops from the top, while often being far more than three stops above the noise floor. It might be nice for higher exposure index ("ISO") settings to only amplify part way, enough to control read noise well, leaving increasing highlight headroom between mid-tones and maximum level, and noting this in the meta-data. That last step would negate my complaint about extra post-processing.
-
Some cameras do similar things, e.g., underexpose by 1 stop and indicate so in metadata, so raw conversion software and "push" digitally afterwards. Canon's Highlight Tone Priority, Pentax's Dynamic Range Expansion, Samsung's Smart Range, and Fuji's Dynamic Range Priority (some cameras, some modes) are all examples of this strategy.
-
Some cameras do similar things, e.g., underexpose by 1 stop and indicate so in metadata, so raw conversion software and "push" afterwards.
Thanks; that is good to know And one extra stop (so a bit over four stops of headroom?) is probably enough to deal with all but some rather exteme and unusual cases, where the metered light level is below about 5% of the brightest diffuse highlights. I would guess that those exteme cases need special attention anyway, like metering on the highlights and then choosing an ISO speed setting low enough to keep them in bounds.