Luminous Landscape Forum

Equipment & Techniques => Digital Cameras & Shooting Techniques => Topic started by: Bill Brooks on August 11, 2011, 10:07:58 am

Title: ETTR: ISO vs shutter speed
Post by: Bill Brooks on August 11, 2011, 10:07:58 am
The Camera to Print & Screen videos demonstrate the benefits of ETTR very clearly.  I have also read Michael's earlier articles on this and I understand the reason why this works.  The videos also point out the obvious fact that lower ISO settings also improve S/N ratios.  Now, by exposing to the right, I am also using a slower shutter speed.  If instead of exposing to the right I used the same slower shutter speed but at a lower ISO setting, I would also benefit from lower noise.  My question is this: which result would give me the better image - the ETTR image at the higher speed, or the "middle"exposure at a lower ISO, using the same shutter speed and f-stop? 
Title: Re: ETTR: ISO vs shutter speed
Post by: Guillermo Luijk on August 11, 2011, 11:08:02 am
If instead of exposing to the right I used the same slower shutter speed but at a lower ISO setting, I would also benefit from lower noise.  My question is this: which result would give me the better image - the ETTR image at the higher speed, or the "middle"exposure at a lower ISO, using the same shutter speed and f-stop?

If you use the same shutter speed and f-stop, noise will be more visible the lower the ISO. If that was not clear enough: ISO100 will display more visible noise than ISO1600.
On some cameras (e.g. Pentax K5, Nikon D7000) the difference will be negligible, in some others (e.g. Canons) will be monstruous.

Canon 350D, 2 shots with the same aperture and shutter speed:
(http://www.guillermoluijk.com/article/iso/versus.jpg)

PS: if you have a digital camera, or you can borrow one from someone, it will take you 5min to take two shots with the same aperture/shutter but different ISO, develop the RAW files so that exposures match, and compare visible noise.
Title: Re: ETTR: ISO vs shutter speed
Post by: Bill Brooks on August 11, 2011, 12:14:43 pm
If the slower ISO is "pushed" during processing I can well understand this.  I'll try to illustrate my original question with a concrete example.  Lets say the camera's lightmeter indicated a "standard" exposure of 1/250th at f8 using ISO400.  The ETTR shot might then be 1/125th at f8, also at ISO400.  In ACR/Lightroom you would then back off the ETTR exposure to give a processed image with lower noise than the "standard" shot.  However, how would this processed ETTR image compare to an alternative "standard" shot of 1/125th at f8, but at ISO200?
Title: Re: ETTR: ISO vs shutter speed
Post by: cybis on August 11, 2011, 12:19:14 pm
If the slower ISO is "pushed" during processing I can well understand this.  I'll try to illustrate my original question with a concrete example.  Lets say the camera's lightmeter indicated a "standard" exposure of 1/250th at f8 using ISO400.  The ETTR shot might then be 1/125th at f8, also at ISO400.  In ACR/Lightroom you would then back off the ETTR exposure to give a processed image with lower noise than the "standard" shot.  However, how would this processed ETTR image compare to an alternative "standard" shot of 1/125th at f8, but at ISO200?

The 1/125th at f8 ISO400 will have LESS noise than the 1/125th at f8 ISO200.
Title: Re: ETTR: ISO vs shutter speed
Post by: Guillermo Luijk on August 11, 2011, 12:27:25 pm
If the slower ISO is "pushed" during processing I can well understand this.  I'll try to illustrate my original question with a concrete example.

Your original question was already answered. Lower ISO = more visible noise.
Title: Re: ETTR: ISO vs shutter speed
Post by: Bart_van_der_Wolf on August 11, 2011, 04:45:15 pm
Your original question was already answered. Lower ISO = more visible noise.

Hi Guillermo,

Your "Lower ISO = more visible noise" answer is incomplete, it should have been followed by "assuming identical exposure parameters, which effectively means underexposing the lower ISO shot much more, compared to a normal exposure for that ISO."

Cheers,
Bart
Title: Re: ETTR: ISO vs shutter speed
Post by: Wayne Fox on August 13, 2011, 02:48:23 am
Hi Guillermo,

Your "Lower ISO = more visible noise" answer is incomplete, it should have been followed by "assuming identical exposure parameters, which effectively means underexposing the lower ISO shot much more, compared to a normal exposure for that ISO."

Cheers,
Bart
So are you interpreting the OP the same as I am, which seems to be the theory that rather than slow down the shutter or open the aperture to move the data to the right to achieve EttR, you would instead increase ISO which would move the data to the right in the histogram?  I'm reading the question as " if the shutter speed is as slow as you want to go, and the aperture as wide as you want to go, and your exposure looks pretty good but isn't "to the right" can using ISO to move it the right help, hurt, or be irrelevant."

Logic says that if the original exposure is a good one (not grossly underexposing) increasing ISO will add noise that may not be offset sufficiently by moving the data up in the histogram.

Guillermo's example seems to show otherwise, but perhaps missing the point a little since in this case both exposures will provide an adequate exposure, the higher ISO one EttR, the other one "normal". This means only a about a .6 to 1.3 stop ISO increase, not a 4 stop one like the example, where the ISO 100 one seems to be very underexposed (I haven't seen that much noise in shadows at ISO 100 in a camera in a long time).

And Guillermo's right - curious enough to try it now.  I've never thought of increasing ISO to move the data to the right because as I mentioned, logically seems counter to the concept.  I'll have to do some tests while out shooting tomorrow.
Title: Re: ETTR: ISO vs shutter speed
Post by: ErikKaffehr on August 13, 2011, 03:14:12 am
Hi,

Some cameras have high read noise. In these cases increasing ISO actually improves signal over read noise as the initial increase in ISO is achieved by pre amplification of sensor signal before raw conversion.

The enclosed figures shows this quite clearly. Nikon and Phase drop linearly with increasing ISO while Canon maintains it's low DR perhaps up 500 ISO before it starts to drop.

Best regards
Erik


So are you interpreting the OP the same as I am, which seems to be the theory that rather than slow down the shutter or open the aperture to move the data to the right to achieve EttR, you would instead increase ISO which would move the data to the right in the histogram?  I'm reading the question as " if the shutter speed is as slow as you want to go, and the aperture as wide as you want to go, and your exposure looks pretty good but isn't "to the right" can using ISO to move it the right help, hurt, or be irrelevant."

Logic says that if the original exposure is a good one (not grossly underexposing) increasing ISO will add noise that may not be offset sufficiently by moving the data up in the histogram.

Guillermo's example seems to show otherwise, but perhaps missing the point a little since in this case both exposures will provide an adequate exposure, the higher ISO one EttR, the other one "normal". This means only a about a .6 to 1.3 stop ISO increase, not a 4 stop one like the example, where the ISO 100 one seems to be very underexposed (I haven't seen that much noise in shadows at ISO 100 in a camera in a long time).

And Guillermo's right - curious enough to try it now.  I've never thought of increasing ISO to move the data to the right because as I mentioned, logically seems counter to the concept.  I'll have to do some tests while out shooting tomorrow.
Title: Re: ETTR: ISO vs shutter speed
Post by: Bart_van_der_Wolf on August 13, 2011, 06:50:38 am
So are you interpreting the OP the same as I am, which seems to be the theory that rather than slow down the shutter or open the aperture to move the data to the right to achieve EttR, you would instead increase ISO which would move the data to the right in the histogram?  I'm reading the question as " if the shutter speed is as slow as you want to go, and the aperture as wide as you want to go, and your exposure looks pretty good but isn't "to the right" can using ISO to move it the right help, hurt, or be irrelevant."

As Guillermo has demonstrated, when we use the same aperture and shutterspeed (because we are somehow restricted to those settings), it can be advantageous to increase ISO because it can reduce the noise (although with a few caveats).

Quote
Logic says that if the original exposure is a good one (not grossly underexposing) increasing ISO will add noise that may not be offset sufficiently by moving the data up in the histogram.

That's correct for a 'good' exposure. Guillermo's example on the other hand compares a grossly underexposed (for ISO 100) image with a much better exposure (for ISO 1600). That scenario is fine, and in sync with the OP's dilemma of a fixed aperture and shutterspeed. But is is misleading to conclude that therefore higher ISO always leads to lower noise. It does not, when the image is exposed properly for the selected ISO setting, but that would require the freedom to increase the exposure for the lower ISO setting and thus improve the photon statistics.

Therefore one needs to consider 2 scenarios:

1. Set the camera to ISO 1600, exposure meter a uniformly lit surface, and set the aperture and shutterspeed as suggested by the metering with the camera in Manual mode. Take an image. The histogram will show a hump in the middle of the range. Now, while using these same aperture and shutterspeed settings in Manual mode, change to ISO 100. Take an image. That second image will show a hump/spike on the left side of the histogram, because we now underexposed it by 4 EV for the ISO. Unsurprisingly, the earlier better exposed (for ISO 1600) image will usually exhibit lower noise, which is a useful thing to know in case of shutterspeed or aperture restrictions.

2. Now repeat the experiment, but leave the camera on (preferably) Aperture priority mode. take one shot at ISO 1600, and one shot at ISO 100. In that case the spike/hump on both histograms is roughly in the same position (the exposure levels were compatible for the given ISO setting, but with a 16x (4 EV) absolute difference in exposure time. In this case the lower ISO image will usually exhibit lower noise, because of the much improved photon statistics. In this scenario the lower ISO shot has lower noise overall.

To complicate matters, different cameras exhibit different noise characteristics depending on the chosen ISO setting. Therefore, some cameras are better not pushed beyond ISO 800 (e.g. the Canon EOS 1Ds Mark III) even for scenario 2. It can be better to underexpose (as in scenario 1.) at ISO 800 when e.g. a minimum shutterspeed is needed for a given aperture, and push in postprocessing. But that only applies to ISO 800+, otherwise there is usually nothing that beats the noise performance of more photons.

Cheers,
Bart
Title: Re: ETTR: ISO vs shutter speed
Post by: ejmartin on August 13, 2011, 11:26:43 am
If the slower ISO is "pushed" during processing I can well understand this.  I'll try to illustrate my original question with a concrete example.  Lets say the camera's lightmeter indicated a "standard" exposure of 1/250th at f8 using ISO400.  The ETTR shot might then be 1/125th at f8, also at ISO400.  In ACR/Lightroom you would then back off the ETTR exposure to give a processed image with lower noise than the "standard" shot.  However, how would this processed ETTR image compare to an alternative "standard" shot of 1/125th at f8, but at ISO200?

There is no short answer; it depends on the camera.  What camera do you use?

The main point of ETTR is that increasing exposure decreases image noise.  And it is important to note that ISO setting is not part of exposure -- exposure is the light intensity times exposure time of the capture, the total light captured, and so depends only on shutter speed and aperture.  ISO is an amplification of the electronic signal after capture, and therefore has nothing to do with the amount of light that was captured.  There are two main sources of image noise: (1) photon noise, statistical fluctuations in the number of photons in the light arriving at the sensor, and (2) electronic read noise, noise in the camera electronics which processes the signal during and after capture.  The larger the exposure, the more light there is, and as a result the photon noise fluctuations are smaller relative to the amount of light captured; this is the main point of ETTR -- that noise relative to signal is smaller as exposure is increased.  So if the ISO is fixed, expose to the right to improve the image noise characteristics.

Now, how does changing the ISO affect things?  As far as photon noise is concerned, not at all, if exposure is kept fixed and ISO is varied; photon noise depends only on exposure, and ISO is not part of exposure.  But the read noise is a property of the camera electronics, that depends on the camera design details.  The second component of noise, the read noise, is ISO dependent on many cameras.  Since the exposure is about light, which is made up of photons, it is useful to measure the read noise in photon equivalent units.  Since the photons are converted into electrons during the capture process in the pixels, the unit of measure 'electron equivalents' is conventionally used.  Here are the read noise characteristics in 'photo-electrons' (e-) as a function of ISO of a variety of cameras:

(http://theory.uchicago.edu/~ejm/pix/20d/posts/dpr/readnoise_vs_iso.jpg)

The first point is ISO 100, and each successive point is one stop higher ISO.  So, on the 5D2 and D3s, read noise drops steadily until about ISO 1600 or 3200 (there is no true ISO 100 on the D3s, rather it uses ISO 200 metered for ISO 100, which is why the ISO 100 and 200 read noises are the same on this camera).  So for these cameras (and all Canons, and Nikons with CMOS sensors designed by Nikon) the read noise is lowered by raising the ISO, up to about ISO 1600 or so, provided the exposure is kept fixed.  Please note that this is always a secondary consideration to simply increasing the exposure, which will always lower the noise more than keeping the exposure fixed and raising the ISO.  Only if you are exposure limited should you consider ETTR via increasing the ISO for cameras such as the D3/D3s/D700 or any Canon.

A new breed of cameras such as the D7000 use the Sony 'Exmor' CMOS sensor family.  These turn out to have an essentially flat read noise characteristic.  This means that there is no advantage to raising the ISO to reduce noise.  Similarly for any camera with a CCD sensor, such as older generations of Nikon such as the D2x, or any MFDB such as the P65+ (I am ignoring here the pixel binning features of the more recent Phase offerings, which lower image read noise at the cost of reduced resolution).  For all these cameras, ISO is essentially irrelevant; the only thing that controls image noise is the exposure.  In fact, there is little reason to do anything other than peg the camera at or near base ISO, and set the largest exposure you can subject to the constraints of motion blur and needed DOF.  Raising the ISO will do nothing but chop off highlight headroom.  The meter will say you are underexposing in many cases, but a properly designed raw converter should be able to amplify the raw data accurately in the same manner as raising the camera ISO setting (note however that Adobe changes internal settings on its raw conversion based on the ISO indicated in metadata, so one will have to play around a bit to undo that; see also a parallel thread on DPP (http://www.luminous-landscape.com/forum/index.php?topic=56951.0)).  The difference is that one can put a shoulder on the tone curve that rolls off highlights more delicately.  Canon and Nikon already offer a limited version of this with Highlight Tone Priority/Active De-Lighting
Title: Re: ETTR: ISO vs shutter speed
Post by: Wayne Fox on August 16, 2011, 02:12:41 am
Similarly for any camera with a CCD sensor, such as older generations of Nikon such as the D2x, or any MFDB such as the P65+ (I am ignoring here the pixel binning features of the more recent Phase offerings, which lower image read noise at the cost of reduced resolution).  For all these cameras, ISO is essentially irrelevant; the only thing that controls image noise is the exposure.  In fact, there is little reason to do anything other than peg the camera at or near base ISO, and set the largest exposure you can subject to the constraints of motion blur and needed DOF.  Raising the ISO will do nothing but chop off highlight headroom.  

OK,  your blowing my mind here ... you're saying with my IQ180 if I shoot a shot at 1/10th at f/11 at ISO 200, and I shoot the same shot by just going to ISO 50 I won't really see any difference in the final results, even though the second one is way "under exposed"?

really hard to wrap my mind around that .... (and curious enough to test it).
Title: Re: ETTR: ISO vs shutter speed
Post by: sandymc on August 16, 2011, 06:51:16 am
OK,  your blowing my mind here ... you're saying with my IQ180 if I shoot a shot at 1/10th at f/11 at ISO 200, and I shoot the same shot by just going to ISO 50 I won't really see any difference in the final results, even though the second one is way "under exposed"?

That's right. There is a "but" in there however - there are side effects to be aware of - some cameras try to reduce visible noise by raising black points, many raw converters apply adjustments before you can "dial out" underexposure or ETTR, etc, so there might be some small differences - tone curve, hue shifts, etc. But side effects from the raw converter and read noise effects, etc in some cameras aside, all that counts is the number of photons that hit the sensor, aka shutter speed and aperture.

Sandy
Title: Re: ETTR: ISO vs shutter speed
Post by: ejmartin on August 16, 2011, 09:40:57 am
Wayne,

I would have suggested you try one of the converters I know applies exposure compensation correctly and doesn't jigger the controls secretly according to the ISO set in metadata (two that come to mind are RPP and RawTherapee), but unfortunately they use dcraw code to open images, and the IQ180 is not listed among the supported cameras.  It's possible that the raw format is similar enough to another supported Phase One back that it will open, but no guarantees.
Title: Re: ETTR: ISO vs shutter speed
Post by: ErikKaffehr on August 16, 2011, 12:35:13 pm
Wayne,

This is exactly what Guillermo has seen on the Pentax K5. You are actually in position to try out on IQ180.

Now, it is quite possible that a raw converter looks at ISO and applies different processing. That is a reasonable approach. Raw conversion is about actual photography and not about "objective testing".

It's a little bit parallel to some users using vendor software for superior image quality, while Jeff Schewe says that if you know how to drive LR/ACR you can get the very same results.

Best regards
Erik


OK,  your blowing my mind here ... you're saying with my IQ180 if I shoot a shot at 1/10th at f/11 at ISO 200, and I shoot the same shot by just going to ISO 50 I won't really see any difference in the final results, even though the second one is way "under exposed"?

really hard to wrap my mind around that .... (and curious enough to test it).
Title: Re: ETTR: ISO vs shutter speed
Post by: ejmartin on August 16, 2011, 01:39:16 pm

Now, it is quite possible that a raw converter looks at ISO and applies different processing. That is a reasonable approach. Raw conversion is about actual photography and not about "objective testing".



I think the relevant question is why any raw converter would look only at the ISO in metadata in order to set its default processing 'look'.  The ISO (International Standards Org) specification of ISO refers to output density, not to the amount of signal amplification at some intermediate stage of processing.  If a raw converter is going to change its internals according to the actual ISO (which I agree can be a reasonable approach), it should use the metadata ISO together with the setting on the exposure compensation control of the converter in order to set those defaults, since that is what sets the ballpark output density (apart from sigmoidal tone curves etc).  Using only the metadata makes it harder for the user to get good results using more advanced shooting techniques that take into account the camera's noise characteristics to optimize the capture.
Title: Re: ETTR: ISO vs shutter speed
Post by: ErikKaffehr on August 16, 2011, 02:41:59 pm
Hi,

My suggestion would be that it would be reasonable to increase default noise reduction for mages that are shot at high ISO. That could also explain differences in appearance between low and high ISO shoots.

In the other hand, it could also use the histogram data to decide optimal noise reduction. Both approaches would deceive any user who is not peeking into internal workings of converter to believe that setting higher ISO has real benefits.

Please note that I don't know if this is the case, I just suggest it could explain differences between observations and theory.

Best regards
Erik



I think the relevant question is why any raw converter would look only at the ISO in metadata in order to set its default processing 'look'.  The ISO (International Standards Org) specification of ISO refers to output density, not to the amount of signal amplification at some intermediate stage of processing.  If a raw converter is going to change its internals according to the actual ISO (which I agree can be a reasonable approach), it should use the metadata ISO together with the setting on the exposure compensation control of the converter in order to set those defaults, since that is what sets the ballpark output density (apart from sigmoidal tone curves etc).  Using only the metadata makes it harder for the user to get good results using more advanced shooting techniques that take into account the camera's noise characteristics to optimize the capture.
Title: Re: ETTR: ISO vs shutter speed
Post by: mouse on August 17, 2011, 07:39:48 pm
Wow!  My head is spinning; or at least it was spinning when I first came across this thread which, at that time, ended with Reply #8.  Then ejmartin came along and restored my faith in what I thought I had previously (and painfully) learned about noise and ettr.

My initial perplexity was percipitated by Mr. Luijk's terse (for him) response and his subsequent failure to elaborate upon it.  I should state that I have been a fan of Guillermo's "internet escapades" for several years now, and have always found his contributions to be very high in nutritional value, though not always within the capability of my digestive system (my shortcoming, not his).   Thus my trust in GL's knowledge caused me to question my own.

Now that I have had the opportunity to read all of this thread to date, it becomes clear that my confusion (and perhaps that of others) stems from the rather vague (or at least incomplete) manner in which the OP posed his question, and which was made even more perplexing (at least to me) by his addendum in Reply #2.

With apologies to Mr. Brooks, allow me to rephrase the original question and supply my answers, so that others my confirm or critique my understanding of the problem:

Q1.   My in-camera meter suggests an exposure of S sec at aperture A, with my chosen ISO setting.  I know from experience that this reading allows at least 1.5 - 2.0 stops of headroom before there is any danger of serious highlight clipping.  I wish to expose to the right in order to minimize noise in the image.  In order to ettr (push the histogram to the right) I have the option to increase the exposure (by altering either S or A) or, alternatively, to increase the ISO setting.  Which of these alternatives will produce an image with the least noise?

A1.  Increasing the exposure by changing either shutter speed or aperture will produce the cleaner image.  Increasing the ISO will add noise to the image.

Q2.  I have set my shutter speed and aperture to values dictated by the shot and which I prefer not to alter.  I have set the ISO to a value such that the in-camera meter shows a "proper" exposure.  My only alternative to achieve ettr (push the histogram to the right) is to increase the ISO.  Will doing so improve, degrade or have no effect on the noise in the image?

A2.  If you are using a camera whose read noise diminishes with increasing ISO then increasing ISO up to 1600 will lessen noise.  If you are not using such a camera, there is nothing to be gained by increasing ISO.  In either case, be aware  that increasing the ISO comes at a cost of decreasing highlight headroom and may cause some clipping.

And finally the question which (I believe) was never asked, but nevertheless answered by GL:

Q3.  I know that increasing the ISO of my camera, especially above X, results in very noisy images.  However the exposure parameters I have chosen require an ISO of X or more to produce a properly exposed image.  Can I reduce the noise in the image by shooting at an ISO less than X and compensating for the underexposure in post-processing?

A3.  See Reply #1  ;)
Title: Re: ETTR: ISO vs shutter speed
Post by: ejmartin on August 17, 2011, 10:15:11 pm

Q1.   My in-camera meter suggests an exposure of S sec at aperture A, with my chosen ISO setting.  I know from experience that this reading allows at least 1.5 - 2.0 stops of headroom before there is any danger of serious highlight clipping.  I wish to expose to the right in order to minimize noise in the image.  In order to ettr (push the histogram to the right) I have the option to increase the exposure (by altering either S or A) or, alternatively, to increase the ISO setting.  Which of these alternatives will produce an image with the least noise?

A1.  Increasing the exposure by changing either shutter speed or aperture will produce the cleaner image.  Increasing the ISO will add noise to the image.

The first part -- increasing the exposure results in a cleaner image -- is correct.  Increasing the ISO for fixed exposure will not add noise to the image, as may be seen in any of the examples I gave, increasing ISO at worst does not change the noise, and in many examples results in less noise.

Quote
Q2.  I have set my shutter speed and aperture to values dictated by the shot and which I prefer not to alter.  I have set the ISO to a value such that the in-camera meter shows a "proper" exposure.  My only alternative to achieve ettr (push the histogram to the right) is to increase the ISO.  Will doing so improve, degrade or have no effect on the noise in the image?

A2.  If you are using a camera whose read noise diminishes with increasing ISO then increasing ISO up to 1600 will lessen noise.  If you are not using such a camera, there is nothing to be gained by increasing ISO.  In either case, be aware  that increasing the ISO comes at a cost of decreasing highlight headroom and may cause some clipping.

Do you see that this answer is in contradiction with the your first one?

Quote
Q3.  I know that increasing the ISO of my camera, especially above X, results in very noisy images.  However the exposure parameters I have chosen require an ISO of X or more to produce a properly exposed image.  Can I reduce the noise in the image by shooting at an ISO less than X and compensating for the underexposure in post-processing?

A3.  See Reply #1  ;)

The answer is no, you cannot reduce the noise through use of a lower in-camera ISO.  The noise at fixed exposure will at best be about the same.  The advantage of a lower ISO, if present, is not in the arena of noise but rather in added highlight headroom.
Title: Re: ETTR: ISO vs shutter speed
Post by: mouse on August 18, 2011, 01:13:28 am
Many thanks for your prompt reply.

The first part -- increasing the exposure results in a cleaner image -- is correct.  Increasing the ISO for fixed exposure will not add noise to the image, as may be seen in any of the examples I gave, increasing ISO at worst does not change the noise, and in many examples results in less noise.

Yes, that is clear.  I meant simply to emphasize that, if one wishes to push the histogram to the right, whenever possible increasing exposure will always produce the best result.


Quote
Do you see that this answer is in contradiction with the your first one?

Yes.  In A1 I recognised that increased exposure reduces noise but  jumped to the conclusion that increasing ISO adds to noise; I knew better.

Quote
The answer is no, you cannot reduce the noise through use of a lower in-camera ISO.  The noise at fixed exposure will at best be about the same.  The advantage of a lower ISO, if present, is not in the arena of noise but rather in added highlight headroom.

Yes, Guillermo's answer made that clear.  Again many thanks.

P.S.  Are you perhaps the same person as one Emil Martinec at the U. of Chicago (my alma mater, class of 1952).  His (your?) paper on "Noise, Dynamic Range and Bit Depth" was most enlightening and a pleasure to read; even if partly over my head.
Title: Re: ETTR: ISO vs shutter speed
Post by: ejmartin on August 18, 2011, 01:35:51 am
P.S.  Are you perhaps the same person as one Emil Martinec at the U. of Chicago (my alma mater, class of 1952).  His (your?) paper on "Noise, Dynamic Range and Bit Depth" was most enlightening and a pleasure to read; even if partly over my head.

Perhaps I am  ;D
Title: Re: ETTR: ISO vs shutter speed
Post by: Bart_van_der_Wolf on August 18, 2011, 04:25:06 am
[...]
The answer is no, you cannot reduce the noise through use of a lower in-camera ISO.  The noise at fixed exposure will at best be about the same.  The advantage of a lower ISO, if present, is not in the arena of noise but rather in added highlight headroom.

Hi Emil,

I beg to differ (slightly), based on the following experiment (don't we love empirical evidence...) with my 1Ds3.

I did a test by determining the standard deviation (luminosity) of the 6 grayscale patches of a MacBeth Mini Colorchecker, by Raw converting in Capture One, and using Photoshop for the statistical readout of exactly the same cropped pixels between conversions. I made sure that the different exposures gave almost the exact same RGB output levels for the corresponding patches after Raw conversion. The push was performed in postprocessing only, and I made sure they resulted in similar RGB values, so I didn't rely on exposure slider values only but also used a little WB where needed (to keep the luminosity standard deviation stable). All noise reduction settings were set to zero. Caveat, we do not know if the ISO setting influences the Raw conversion in other ways, but Capture One seems pretty well behaved.

Noise standard deviation of a 50x49 pixel area of each patch.
(http://www.xs4all.nl/~bvdwolf/temp/OPF/HighISO+Push.png)

The patches named 1 through 6 in my table are formally called patches 19 (white) to 24 (Black) in CC speak.

As can be seen, the ISO 400 ('unity') gain has lower noise than the ISO 800 group, and the ISO 800 group has lower noise than the ISO 1600 group. Each group has a 'correct' exposure at the main ISO setting, and the lower ISOs in that group were underexposed by one or two EV, effectively resulting in the same amount of photons in each exposure for that group.

Within the ISO 800 group, there is little difference between the normal exposure and the 1 EV underexposed plus 1 EV pushed (in Raw conversion) setting, but the 'pushed' settings will have more highlight clipping latitude (a stop headroom).

Within the ISO 1600 group there is also little difference, although the ISO 800 pushed 1 stop is slightly better than the rest, and again it has 1 stop overexposure headroom. Even ISO 400 pushed 2 stops is a bit better than the ISO 1600 gain setting. The differences within each group are actually very difficult to see, but they are measurable.

IMHO the 1Ds3 doesn't perform very well above ISO 1600 (for my type of use), so I didn't test higher ISO settings, but I expect e.g. the 5D2 to show similar performance at ISO 3200.

It all boils down to the amplification of readnoise by increasing the gain for these camera models. When there are less and less photons for the actual exposure at higher and higher ISO gain settings, the readnoise amplification generates a higher contribution in the mix of noise sources. So for my specific camera model, I do not set the ISO above 800 if I can underexpose by one or two stops and push in Raw conversion.

Cheers,
Bart
Title: Re: ETTR: ISO vs shutter speed
Post by: ejmartin on August 18, 2011, 10:57:15 am
Hi Bart,

What happens if you measure the raw data directly, rather than filtering it through the converter (eg by sending it through dcraw -T -4 -D -v)?  I am a bit suspicious that either (a) the number of photons is not the same between shots; or more likely (b) the converter is doing something without telling you.  the first three squares (1-3 in your counting, 19-21 on the CC chart) should be entirely dominated by photon noise, and yet your results show consistent differences (I think we need to regard 400+2EV, square number two in your chart, as an anomaly).
Title: Re: ETTR: ISO vs shutter speed
Post by: Bart_van_der_Wolf on August 18, 2011, 02:26:49 pm
Hi Bart,

What happens if you measure the raw data directly, rather than filtering it through the converter (eg by sending it through dcraw -T -4 -D -v)?

It's an older experiment, so I'll have to search for the files when I have some spare time. I'm not too fond of doing noise analysis on single frames though, I prefer to really cancel camera specific influences by subtracting 2 frames, which I didn't take for this complete workflow test.

Quote
I am a bit suspicious that either (a) the number of photons is not the same between shots; or more likely (b) the converter is doing something without telling you.

If anything not already in the Raw data, it would be the Raw converter (which would be a bit of a surprise with Capture One). I made exposures of the CC by bracketing in 1/3rd stops, and matching the closest exposures, in order to reduce the potential effect of shutter inaccuracies which would directly influence the photon shot noise. So the exposures are within 1/3th (probably even 1/6th) of a stop accurate.
The Raw converter (using a linear tonecurve) was in this case specifically included in the equation because I was testing a workflow solution for someone with Capture One when faced with poor (stage performance with dance movement) lighting conditions. No matter how nice the sensor statistics, it's the final result after demosaicing that counts in practice. Final noise reduction was not a part of the workflow test (although an option afterwards), and thus set to zero in the Raw converter.

Quote
the first three squares (1-3 in your counting, 19-21 on the CC chart) should be entirely dominated by photon noise, and yet your results show consistent differences (I think we need to regard 400+2EV, square number two in your chart, as an anomaly).

Outliers, however unlikely, are still possible in a probability distribution, apparently. I have no explanation for it.

The experiment is easy to repeat though (for someone with enough time to do it, I'm a bit swamped now), even for those without Raw analysis tools. So I invite others to try it as well.

Cheers,
Bart
Title: Re: ETTR: ISO vs shutter speed
Post by: torger on August 31, 2011, 07:09:18 am
The Camera to Print & Screen videos demonstrate the benefits of ETTR very clearly.  I have also read Michael's earlier articles on this and I understand the reason why this works.  The videos also point out the obvious fact that lower ISO settings also improve S/N ratios.  Now, by exposing to the right, I am also using a slower shutter speed.  If instead of exposing to the right I used the same slower shutter speed but at a lower ISO setting, I would also benefit from lower noise.  My question is this: which result would give me the better image - the ETTR image at the higher speed, or the "middle"exposure at a lower ISO, using the same shutter speed and f-stop? 


There's been many very good and very technical responses. Here's one a bit less technical.

ETTR is only relevant when your shutter speed does not need to be short. The purpose of ETTR is to gather as much light as possible, and thus have the shutter open as long as possible without over-saturating the sensor. The sensor can gather most photons at base ISO so that is the setting that should be used.

However, often you end up with too long shutter speeds if doing ETTR at base ISO. The workflow is then to use as long shutter speed as the scene situation allows and then use a low ISO as possible to get a good "middle" exposure (an exposure that gives properly exposed in-camera JPEGs). The logic here is that it is not much difference between ISO800 ETTR and ISO400 in the middle when same shutter speed f/stop is used since you're capturing the same number of photons.

A perfect sensor would actually not need ISO setting at all, it would just deliver the exact number of photons captured at each sensel, but today's sensor need analog amplification of the sensel signals, higher ISO higher amplification (and more noise). However, it differs a bit between cameras and sensor how it behaves at high ISO, so there is no safe answer if ISO800 ETTR or ISO400 in the middle will be better image quality, you would have to test for your camera. However, the difference is likely small, so the rule to not use shorter shutter speeds than necessary and combine that with a low ISO setting but not necessary press it to ETTR is a good one I think. If you are the optimizing kind of guy you may want to do your own tests with your equipment though to find out the best strategy.
Title: Re: ETTR: ISO vs shutter speed
Post by: mouse on September 01, 2011, 12:28:05 am

The experiment is easy to repeat though (for someone with enough time to do it, I'm a bit swamped now), even for those without Raw analysis tools. So I invite others to try it as well.

Cheers,
Bart

I have recently become interested in this problem and, although I am certainly not in the same league as you and Dr. Martin. I hope you don't mind if I accept your invitation and join the game.

I started by making a few preliminary analyses on a single exposure using different methods.  The results are somewhat puzzling to me and I hope you can give me some suggestions.  The exposure was of a daylight lit gray card filling the field of a 105mm lens on a D700.  Exposure was F/8 @ 1/1000, which was about 2 stops below the metered value.

1.  Raw file opened in ACR using default values and then opened in PSCS5.  The histogram shows a mean value of 85.7 with a S.D. = 1.6

2.  The raw file was converted with dcraw (-v -d -r 1 1 1 1 -T -4).  When the resulting tiff file was opened in PSCS5 the histogram shows a mean = 12.36 and a S.D. = 2.9

3.  Using IRIS software, the raw file was converted to a .pic version of a CFA file.  The color components were separated with the command CFA2RGB and the following data were extracted.  Looking only at the green channel: mean = 959; S.D. = 46.4

4.  Using the IRIS software I attempted to calculate a S.D. employing the difference method.  Not having a duplicate exposure, I constructed two pic files by taking a 300x300 pixel window from 2 different areas of the green channel frame.  I then subtracted one window from the other.  Results:
window #1  mean = 966  S.D. = 15.2
window #2  mean = 984  S.D. = 14.6
win2 - win1  S.D. = 21.2   (corrected (??) 21.2/1.4 = 15.0)

5.  Opening the raw file in Rawnalyse software, the following stats are obtained, Green channel only mean = 975  S.D. = 18.5

Allowing for the fact that the scale of ADU in PS is 0 - 255 and that in the IRIS files and Rawnalyse is 0 - 16383, I am still confused by the large differences in magnitude between the statistics read in PSCS5 and the other programs.  I am heartened by the reasonably good agreement between iris and Rawnalyse statistics.  I am not surprised by the larger SD read from the IRIS green channel alone when compared with the SD computed from the difference between 2 windows.  I do wonder how Rawnalyse arrived at a similar SD.

So my questions to the experts:  Do my methods make any sense?  Are my results in the proper ballpark?  Any suggestions appreciated. 
Title: Re: ETTR: ISO vs shutter speed
Post by: ejmartin on September 01, 2011, 10:34:07 am
I started by making a few preliminary analyses on a single exposure using different methods.  The results are somewhat puzzling to me and I hope you can give me some suggestions.  The exposure was of a daylight lit gray card filling the field of a 105mm lens on a D700.  Exposure was F/8 @ 1/1000, which was about 2 stops below the metered value.

1.  Raw file opened in ACR using default values and then opened in PSCS5.  The histogram shows a mean value of 85.7 with a S.D. = 1.6

ACR conversion opened in PS will have gamma applied, as well as tone curve etc unless you have zero'd out the controls in ACR.  With sRGB gamma of 2.4 that brings 985/16383 up to about 79/255, close to what you are observing.  Tone curve could account for the rest.  Similarly, SNR has been increased by a factor of about 2.5 relative to IRIS/Rawanalyze values, presumably also because gamma raises the value of S and compresses the N.

Quote
2.  The raw file was converted with dcraw (-v -d -r 1 1 1 1 -T -4).  When the resulting tiff file was opened in PSCS5 the histogram shows a mean = 12.36 and a S.D. = 2.9

dcraw -d will give the raw data, without separating the color planes the std dev is not meaningful.

Quote
3.  Using IRIS software, the raw file was converted to a .pic version of a CFA file.  The color components were separated with the command CFA2RGB and the following data were extracted.  Looking only at the green channel: mean = 959; S.D. = 46.4

If you look at the entire color plane then there is a chance that vignetting affects the result; best to choose reasonable size patches, say 100x100 or so.

Quote
4.  Using the IRIS software I attempted to calculate a S.D. employing the difference method.  Not having a duplicate exposure, I constructed two pic files by taking a 300x300 pixel window from 2 different areas of the green channel frame.  I then subtracted one window from the other.  Results:
window #1  mean = 966  S.D. = 15.2
window #2  mean = 984  S.D. = 14.6
win2 - win1  S.D. = 21.2   (corrected (??) 21.2/1.4 = 15.0)

You can have IRIS separate the channels, then subtract one green channel from the other.  Many effects such as vignetting and other sources of signal variation such as uneven lighting cancel out. 
Title: Re: ETTR: ISO vs shutter speed
Post by: hjulenissen on September 01, 2011, 11:13:37 am
If you look at the entire color plane then there is a chance that vignetting affects the result; best to choose reasonable size patches, say 100x100 or so.
Or perhaps highpass filter the entire image with a really low cutoff?

-h
Title: Re: ETTR: ISO vs shutter speed
Post by: mouse on September 01, 2011, 05:47:51 pm
ACR conversion opened in PS will have gamma applied, as well as tone curve etc unless you have zero'd out the controls in ACR.  With sRGB gamma of 2.4 that brings 985/16383 up to about 79/255, close to what you are observing.  Tone curve could account for the rest.  Similarly, SNR has been increased by a factor of about 2.5 relative to IRIS/Rawanalyze values, presumably also because gamma raises the value of S and compresses the N.

Yes, I had a hunch that gamma conversion was responsible for most of the problem.

Quote
You can have IRIS separate the channels, then subtract one green channel from the other.  Many effects such as vignetting and other sources of signal variation such as uneven lighting cancel out. 

I recall reading one of your posts, some time ago, that you did just that.  Can you tell me where to find that message?  I was unable to find the proper instruction in IRIS to obtain the two green channels as separate files.  Can you enlighten me?

Many thanks for your comments.  I shall continue my experiments and bore you all with the results.
Title: Re: ETTR: ISO vs shutter speed
Post by: ejmartin on September 01, 2011, 09:09:05 pm
I recall reading one of your posts, some time ago, that you did just that.  Can you tell me where to find that message?  I was unable to find the proper instruction in IRIS to obtain the two green channels as separate files.  Can you enlighten me?

The IRIS command

split_cfa c1 c2 c3 c4

will split the raw data array into four separate color channels, assuming a Bayer type pattern.  You can find them as c1.fit etc in whatever directory is the default for IRIS (which you can find by pulling up the preferences dialog).  You can load each color plane in turn via

load c1

and so on.  Find the two channels that are closest in value; these will be the two green planes.  Suppose it is c1 and c3.  Load c1 and then execute the command

sub c3 1000

(here 1000 is a fixed amount that is added so that the average is not zero; make it whatever you want).  IRIS can then compute average and std dev of a patch -- just drag the mouse with left click to select a window, and right click to select 'Statistics' to get the mean and std dev for the selection.
Title: Re: ETTR: ISO vs shutter speed
Post by: mouse on September 01, 2011, 11:49:32 pm
The IRIS command

split_cfa c1 c2 c3 c4

will split the raw data array into four separate color channels, assuming a Bayer type pattern.  You can find them as c1.fit etc in whatever directory is the default for IRIS (which you can find by pulling up the preferences dialog).  You can load each color plane in turn via

load c1

and so on.  Find the two channels that are closest in value; these will be the two green planes.  Suppose it is c1 and c3.  Load c1 and then execute the command

sub c3 1000

(here 1000 is a fixed amount that is added so that the average is not zero; make it whatever you want).  IRIS can then compute average and std dev of a patch -- just drag the mouse with left click to select a window, and right click to select 'Statistics' to get the mean and std dev for the selection.

Thanks.  Worked like a charm.  Results: S.D = 22.7  corrected x 0.7 = 16.2
Agrees well with result of subtracting two windows from green channel as described above.

I am still puzzled that the results from Rawnalyse come so close without the benefits obtained from subtracting two images.

More to follow.  Hope I don't wear out your patience.
Title: Re: ETTR: ISO vs shutter speed
Post by: Bart_van_der_Wolf on September 02, 2011, 05:59:19 am
dcraw -d will give the raw data, without separating the color planes the std dev is not meaningful.

That's the reason I don't use DCRaw for ths type of analysis. Besides, I wonder if the -d is perhaps better replaced by -D which does no scaling at all. Likewise, it might be useful to add the -k 0 parameter as a safeguard against losing the blackpoint offset (maybe it defaults to zero, but it wouldn't hurt to make sure).

Quote
If you look at the entire color plane then there is a chance that vignetting affects the result; best to choose reasonable size patches, say 100x100 or so.

It indeed helps to take a smallish area from the center of the image and use an aperture of f/5.6 or f/8 as that minimizes the vignetting influence. To avoid image detail from interfering, a slight defocus can be used when shooting test images. I also try to select an area that has no hot sensels, or dust bunnies (which tend to pile up in corners if the sensor is not cleaned to perfection).

Cheers,
Bart
Title: Re: ETTR: ISO vs shutter speed
Post by: Bart_van_der_Wolf on September 02, 2011, 06:16:07 am
Find the two channels that are closest in value; these will be the two green planes.  Suppose it is c1 and c3.  Load c1 and then execute the command

sub c3 1000

(here 1000 is a fixed amount that is added so that the average is not zero; make it whatever you want).

I use a minimum amount of 1024 to accommodate for the blackpoint offset of Canon cameras when doing a Blackframe (read noise) analysis. When doing S/N analysis at higher ISOs I use 4000. In any case, I follow the subtraction command by a stat command and check for minimum>0 or maximum<clipping level, just to make sure that the sigma/standard deviation as reported is not based on clipped noise.

My workflow is based on pre-cropped image segments, so the stat command is adequate for obtaining the standard deviation at the same time as the boundary checking with the stat command.

Cheers,
Bart
Title: Re: ETTR: ISO vs shutter speed
Post by: Guillermo Luijk on September 02, 2011, 04:58:36 pm
2.  The raw file was converted with dcraw (-v -d -r 1 1 1 1 -T -4).  When the resulting tiff file was opened in PSCS5 the histogram shows a mean = 12.36 and a S.D. = 2.9

As Emil points, dcraw's -d doesn't separate the colour cells from the Bayer pattern, but there is a very easy and simple way to achieve that in PS: apply nearest neighbour resize to 50% (that will pick one pixel from each 2x2 Bayer cell):

(http://www.guillermoluijk.com/article/rawnoise/extraccion.gif)

(http://www.guillermoluijk.com/article/rawnoise/extraccion2.gif)

Now we are free to calculate mean and StDev over that individual RAW channel (not sure about PS's precision on this though, I prefer to use Rawnalyze).

To obtain the four RGGB components just add one pixel line up/down and/or left/right before resizing.

Regards

Title: Re: ETTR: ISO vs shutter speed
Post by: mouse on September 03, 2011, 04:20:14 am
It indeed helps to take a smallish area from the center of the image and use an aperture of f/5.6 or f/8 as that minimizes the vignetting influence. To avoid image detail from interfering, a slight defocus can be used when shooting test images. I also try to select an area that has no hot sensels, or dust bunnies (which tend to pile up in corners if the sensor is not cleaned to perfection).

Cheers,
Bart

I assume you are referring to the method of using two identical exposures.  Would you also recommend taking a window from the center of the image when one is taking the difference of the two green channels from the same exposure?

And thanks to all for tolerating the questions of a tyro!
Title: Re: ETTR: ISO vs shutter speed
Post by: Bart_van_der_Wolf on September 03, 2011, 04:56:29 am
I assume you are referring to the method of using two identical exposures.  Would you also recommend taking a window from the center of the image when one is taking the difference of the two green channels from the same exposure?

It applies to any noise measurement based on some area average. When there is a slope in average brightness, even if there is no noise, the standard deviation is going to increase. So for a truely acurate noise measurment, one strives for measuring of a uniform surface that's as evenly lit as possible. The larger the area is that's being analysed, the larger the chance is that a brightness slope becomes significant enough to influence the outcome.

Basing the analysis on subtracting an aligned exposure pair will remove any residual slope (pattern noise) from the equation, and leave us with random noise. Subtracting the G1 and G2 green filtered sensel values of a single image from eachother is a close enough substitute to use instead of two separate exposures.

Cheers,
Bart
Title: Re: ETTR: ISO vs shutter speed
Post by: madmanchan on September 03, 2011, 07:20:14 am
You have to be careful with comparing G1 and G2 (and I would not recommend it) because they can diverge due to flare, crosstalk, and other issues. Some cameras correct for this in-camera before writing out raw data, and others do not. For consistency I advise comparing like-colors only, or averaging all greens and comparing two separate images.
Title: Re: ETTR: ISO vs shutter speed
Post by: bjanes on September 03, 2011, 09:25:22 am
I assume you are referring to the method of using two identical exposures.  Would you also recommend taking a window from the center of the image when one is taking the difference of the two green channels from the same exposure?

And thanks to all for tolerating the questions of a tyro!

Subtracting two identical images removes PRNU (http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/#prnu) (pixel response nonuniformity) noise as well as variations in the target, dust on the sensor and other fixed pattern noise. If you measure the standard deviation of a uniform flat field, the result will be considerably higher than the shot noise due to PRNU. PRNU increases in direct proportion to luminosity, while shot noise increases as the square root of the luminosity. With most cameras, PRNU is the most prominent source of noise in the highlights, but the signal to noise in the highlights is such that noise is not perceived in the highlights.

The difference of two flat fields at high luminosity is almost entirely shot noise. One can determine the signal to noise ratio (SNR), and the SNR squared equals the number of electrons collected. Such measurements enable calculation of the full well capacity of the sensor. When performing such measurements, one must avoid clipping of the highlights, since this decreases the noise (a fully clipped image has a standard deviation of zero).

Regards,

Bill
Title: Re: ETTR: ISO vs shutter speed
Post by: mouse on September 04, 2011, 08:41:33 pm
I use a minimum amount of 1024 to accommodate for the blackpoint offset of Canon cameras when doing a Blackframe (read noise) analysis. When doing S/N analysis at higher ISOs I use 4000. In any case, I follow the subtraction command by a stat command and check for minimum>0 or maximum<clipping level, just to make sure that the sigma/standard deviation as reported is not based on clipped noise.

My workflow is based on pre-cropped image segments, so the stat command is adequate for obtaining the standard deviation at the same time as the boundary checking with the stat command.

Cheers,
Bart

First, I fail to understand the procedure for doing a Blackframe (read noise) analysis.  If the Canon camera adds an offset of 1024 to all data, I would think that one needs to subtract 1024 from the data; particularly the mean since this is the significant datum in regard to read noise, and the S.D. is not of any importance.  From what I have read (and perhaps misunderstood),  in doing a black-frame (offset) analysis, one takes several images and computes the average of the means, or alternatively, the median value.  Further it seems that one would want to include pixels whose value was 1024 (or 0 after adjustment) when calculating the mean noise.

When doing a S/N analysis, I can understand that one wishes to exclude pixels outside the range of 0 to clipping level from influencing the S.D.  However, since the S.D. (and the other statistics) are computed first and the offset added after the fact, the addition of an offset should not affect the S.D.  It also eludes me how one recognizes the inclusion of clipped noise by examination of the stats.  For example, if the image used as the subtrahend contains (clipped) pixels at level 0 while the minimun pixel in the minuend are at level 20,  the minimum level returned by the stat command on the difference image will be 20 (or 4020 if one uses an offset of 4000). 

Clearly I must be missing something here.  Would much appreciate your help.

Cheers/Mike
Title: Re: ETTR: ISO vs shutter speed
Post by: Bart_van_der_Wolf on September 05, 2011, 05:55:10 am
First, I fail to understand the procedure for doing a Blackframe (read noise) analysis.

Hi Mike,

No problem, I'll explain. A Blackframe is supposed to be black because it received no exposure. However, when we analyse it there is noise. The noise is produced by the camera electronics. When we take precautions to eliminate as many noise sources (e.g. thermal noise doubles for approx. each 6 degrees Celsius rise) as possible, we could assume that the remaining noise is unavoidable and linked to the action of reading out the sensor data, hence coined "Read noise".

A Blackframe is typically produced by setting the camera to it's shortest possible exposure time (to counteract thermal noise build-up), using a body cap instead of a lens (to avoid light leaks, electronic noise from the lens, and camera gain adjustments at certain apertures), and covering the eyepiece of the viewfinder (to avoid light leaking into the mirrorbox though the back).

The signal that is still recorded is the lowest signal possible and is usually random with a Gaussian distribution. It changes with the ISO (gain) setting. It is not the same as a Darkframe, which is produced with a much longer (>1 sec. typically) exposure time, as used for Darkframe subtraction. By comparing a Darkframe and a Blackframe one can quantify the (mostly thermal) contribution.

Quote
If the Canon camera adds an offset of 1024 to all data, I would think that one needs to subtract 1024 from the data; particularly the mean since this is the significant datum in regard to read noise, and the S.D. is not of any importance.

The offset in most Canons cameras is part of the ADC quantization, so it is not added afterwards. That's why the noise has a Gaussian distribution centered at (usually) ADU 1024. There are also values below 1024 because of the Readnoise.

Quote
From what I have read (and perhaps misunderstood),  in doing a black-frame (offset) analysis, one takes several images and computes the average of the means, or alternatively, the median value.

What you describe is a Darkframe (not Blackframe) noise reduction technique, commonly used in astrophotography where long exposure times are needed to collect enough photons to record faint signals. This is also why Canon cameras are often used in astrophotography, because the Readnoise improves predictably with averaging multiple frames and may reveal faint signals.

Quote
When doing a S/N analysis, I can understand that one wishes to exclude pixels outside the range of 0 to clipping level from influencing the S.D.  However, since the S.D. (and the other statistics) are computed first and the offset added after the fact, the addition of an offset should not affect the S.D.  It also eludes me how one recognizes the inclusion of clipped noise by examination of the stats.

You probably figured it out after the above explanation, but to make sure... When you subtract 2 noisy data sets with a mean value of e.g. 1024, then there is a 50% chance that an image has a value of 1024 or less. There is a equal chance of it being 1024 or higher. When we subtract an image with a higher data value from one with a lower data value we would get a negative number, which cannot be encoded in an integer number calculation, and thus result in a clipped noise distribution.

Therefore we add an offset to both datasets, which only changes the mean value but not the SD around that mean, and the result of the subtraction can be statistically evaluated. My choice of 1024 is not a must, one can use any number that doesn't add to the risk of integer value clipping, although it could also indicate an ADC problem. That's why I use the IRIS stat command after the subtraction, to check that there are no values that resulted in (probably clipped) zero despite the offset. If it would have a minimum of zero, then I redo the subtraction with a higher offset (for light exposure frames), but for Blackframes this is usually not needed (especially for lower ISO gain settings).

Hope that helped to understand the chosen procedure.

Cheers,
Bart
Title: Re: ETTR: ISO vs shutter speed
Post by: PierreVandevenne on September 05, 2011, 05:19:43 pm
No problem, I'll explain. A Blackframe is supposed to be black because it received no exposure. However, when we analyse it there is noise. The

Wouldn't that be a BIAS frame in standard terms? Minor issue, because it doesn't invalidate what you say, but "Blackframe" is too close to "Darkframe" imho and will lead to confusion

http://en.wikipedia.org/wiki/Bias_frame

http://www.cyanogen.com/help/maximdl/Bias_Frame_Calibration.htm

http://www.cfht.hawaii.edu/Instruments/Imaging/FOCAM/calib.html
Title: Re: ETTR: ISO vs shutter speed
Post by: Bart_van_der_Wolf on September 05, 2011, 07:13:25 pm
Wouldn't that be a BIAS frame in standard terms?

Hi Pierre,

Yes, that term is often used. It doesn't really describe what it represents, or what causes it though, so I use Black frame when there is essentially no (or as short as can possibly be set on the camera) exposure involved.

Cheers,
Bart
Title: Re: ETTR: ISO vs shutter speed
Post by: mouse on September 06, 2011, 01:51:32 am
Quote
Hi Mike,

No problem, I'll explain. A Blackframe is supposed to be black because it received no exposure. However, when we analyse it there is noise. The noise is produced by the camera electronics. When we take precautions to eliminate as many noise sources (e.g. thermal noise doubles for approx. each 6 degrees Celsius rise) as possible, we could assume that the remaining noise is unavoidable and linked to the action of reading out the sensor data, hence coined "Read noise".

A Blackframe is typically produced by setting the camera to it's shortest possible exposure time (to counteract thermal noise build-up), using a body cap instead of a lens (to avoid light leaks, electronic noise from the lens, and camera gain adjustments at certain apertures), and covering the eyepiece of the viewfinder (to avoid light leaking into the mirrorbox though the back).

The signal that is still recorded is the lowest signal possible and is usually random with a Gaussian distribution. It changes with the ISO (gain) setting. It is not the same as a Darkframe, which is produced with a much longer (>1 sec. typically) exposure time, as used for Darkframe subtraction. By comparing a Darkframe and a Blackframe one can quantify the (mostly thermal) contribution.


Thank you.  I was aware of the difference between a Black-frame (called an Offset by the Astrophotophotographer, I believe) and a Darkframe.  I am interested in the Black-frame because I wish to use it to estimate the influence of ISO on the read noise of my camera.

Quote
The offset in most Canons cameras is part of the ADC quantization, so it is not added afterwards. That's why the noise has a Gaussian distribution centered at (usually) ADU 1024. There are also values below 1024 because of the Readnoise.

Still scratching my head here.  By afterwards I meant after (or even simultaneous with) conversion of the (amplified) signal from the sensor into ADU.  Thus 1024 will be added to whatever read noise is associated with a given pixel.  And I assume that such read noise cannot take negative values.   So I fail to see how one can get values less than 1024, nor how the mean of the read noise would be at 1024 (if you are talking about values obtained from a single image).

Quote
What you describe is a Darkframe (not Blackframe) noise reduction technique, commonly used in astrophotography where long exposure times are needed to collect enough photons to record faint signals. This is also why Canon cameras are often used in astrophotography, because the Readnoise improves predictably with averaging multiple frames and may reveal faint signals.

From my brief reading in the AstroP sites, the same method (i.e. average mean or median of several images) is recommended for determining a Offset (or what we are calling a Blackframe).  How do you recommend obtaining Blackframe data?  I assume it does not involve taking the difference between two images, as one would do for S/N analysis.

Quote
... When you subtract 2 noisy data sets with a mean value of e.g. 1024, then there is a 50% chance that an image has a value of 1024 or less. There is a equal chance of it being 1024 or higher. When we subtract an image with a higher data value from one with a lower data value we would get a negative number, which cannot be encoded in an integer number calculation, and thus result in a clipped noise distribution.

Therefore we add an offset to both datasets, which only changes the mean value but not the SD around that mean, and the result of the subtraction can be statistically evaluated. My choice of 1024 is not a must, one can use any number that doesn't add to the risk of integer value clipping, although it could also indicate an ADC problem. That's why I use the IRIS stat command after the subtraction, to check that there are no values that resulted in (probably clipped) zero despite the offset. If it would have a minimum of zero, then I redo the subtraction with a higher offset (for light exposure frames), but for Blackframes this is usually not needed (especially for lower ISO gain settings).

I am aware that when one subtracts two data sets, regardless of their respective means, there is an equal chance that any given pixel in image #1 will have a value greater than or less than the same pixel in image #2.  Thus the possiblity of negative values.  However I was under the impression that the Iris software can deal with negative numbers when calculating a mean and SD from individual pixel differences.  Furthermore I believe that the offset is added after the calculation of these statistics is performed.  I have tried this with two images, both with and without adding an offset. First subtracting image #2 from image #1 and then the reverse.  The mean, median and SD (after subtracting the offset) were identical in value but opposite in sign. 

Quote
Hope that helped to understand the chosen procedure.

Cheers,
Bart

It has helped, but still a ways to go

Cheers/Mike
Title: Re: ETTR: ISO vs shutter speed
Post by: mouse on September 06, 2011, 03:06:23 am
OK.  After rereading Dr. Martinec's paper I have managed to correct some of my misconceptions:

Black-frame (or read noise) voltages can (and probably do) exhibit both positive and negative values prior to A/D conversion.  The negative values will be clipped to zero unless a voltage bias is applied prior to conversion.  Further, it is these voltage fluctuations  which constitute the read noise, estimated by the SD of these fluctuation (around a mean of zero plus whatever offset is added by the camera).  Thus the SD is the pertinent statistic in estimating read noise, not the mean of the black-frame images, as I earlier presumed.

With this new found enlightenment, I still believe that (with Canon cameras) one should use an offset of negative 1024 with a black-frame image if one is looking for minimum values < 0 as a flag for clipped pixels.  Alternatively one could use no offsetl and look for minimum values < 1024.  What am I missing?

So next question: How does one estimate read noise based on black-frame images from a Nikon D700, which apparently does not apply a bias voltage.

Cheers/Mike
Title: Re: ETTR: ISO vs shutter speed
Post by: bjanes on September 06, 2011, 08:45:05 am

So next question: How does one estimate read noise based on black-frame images from a Nikon D700, which apparently does not apply a bias voltage.

Cheers/Mike

For the Nikon D3 and other cameras not using an offset, one must measure the read noise by extrapolation as shown by Peter Facey (http://www.brisk.org.uk/photog/d3readn.html). He also gives a link to similar study by Emil Martinec.

Regards,

Bill
Title: Re: ETTR: ISO vs shutter speed
Post by: mouse on September 09, 2011, 08:32:42 pm
For the Nikon D3 and other cameras not using an offset, one must measure the read noise by extrapolation as shown by Peter Facey (http://www.brisk.org.uk/photog/d3readn.html). He also gives a link to similar study by Emil Martinec.

Regards,

Bill

Many thanks for this citation.  It is very thorough and interesting.  I have work to do.  Thanks again.

Cheers/Mike
Title: Re: ETTR: ISO vs shutter speed
Post by: ejmartin on September 09, 2011, 11:15:27 pm

So next question: How does one estimate read noise based on black-frame images from a Nikon D700, which apparently does not apply a bias voltage.

Cheers/Mike

Another option: Nikon has an area of 'masked pixels' on the sensor that are shielded from light, and for which a bias voltage is applied; apparently these pixel values are written to the raw file.  The package 'libraw' can output these pixel values:

http://www.libraw.org/docs/API-datastruct-eng.html#libraw_masked_t
Title: Re: ETTR: ISO vs shutter speed
Post by: mouse on September 10, 2011, 04:26:35 am
Another option: Nikon has an area of 'masked pixels' on the sensor that are shielded from light, and for which a bias voltage is applied; apparently these pixel values are written to the raw file.  The package 'libraw' can output these pixel values:

http://www.libraw.org/docs/API-datastruct-eng.html#libraw_masked_t

Thanks emil.  In my time I have written thousands of lines of C code, a skill now much diminished by many years of neglect.  Even at my best I found it difficult to adapt others code to my needs, so I think I will pass on this challenge. :)

I think I will give Facey's method a try.  It sounds like a fun project.
Title: Re: ETTR: ISO vs shutter speed
Post by: digitaldog on September 11, 2011, 01:00:42 pm
The first part -- increasing the exposure results in a cleaner image -- is correct.  Increasing the ISO for fixed exposure will not add noise to the image, as may be seen in any of the examples I gave, increasing ISO at worst does not change the noise, and in many examples results in less noise.

Indeed! I did the test too, with a 5DMII. Used the meter in-camera to average an exposure for ISO 100. Shot the image then left everything the same and upped ISO to 800. On LCD and in Lightroom, as one would expect, this 2nd capture is much brighter appearing. Normalize using Exposure setting to match the ISO 100, examine both, ISO 800 is clearly less noisy. Fascinating!

So here is where I’m unclear about what’s going on under the hood. The exposure is identical in terms of aperture and shutter. ISO is higher. We expect doing so would make it appear brighter. What I would like explained further (for the non scientist) is how and why? The same amount of light (photons) strike the sensor. What is the ISO doing here to provide a better S/N ratio reducing the noise?
Title: Re: ETTR: ISO vs shutter speed
Post by: hjulenissen on September 11, 2011, 01:15:37 pm
So here is where I’m unclear about what’s going on under the hood. The exposure is identical in terms of aperture and shutter. ISO is higher. We expect doing so would make it appear brighter. What I would like explained further (for the non scientist) is how and why? The same amount of light (photons) strike the sensor. What is the ISO doing here to provide a better S/N ratio reducing the noise?
Perhaps a significant source of noise is located after the analog amplifier, meaning that boosting the signal early on will not affect the additive noise that is added later on, resulting in an improved SNR?

-h
Title: Re: ETTR: ISO vs shutter speed
Post by: Guillermo Luijk on September 11, 2011, 01:18:17 pm
So here is where I’m unclear about what’s going on under the hood. The exposure is identical in terms of aperture and shutter. ISO is higher. We expect doing so would make it appear brighter. What I would like explained further (for the non scientist) is how and why? The same amount of light (photons) strike the sensor. What is the ISO doing here to provide a better S/N ratio reducing the noise?

Andrew look at this simple model of your camera's capture pipeline:

(http://www.guillermoluijk.com/article/iso/modelo.gif)


Think of 2 sources of noise added to the useful signal (S): Npre and Npost. Npre is added before ISO amplification, Npost is added after ISO amplification. ISO amplification by itself doesn't alter the SNR, so the ISO amplification does't improve output SNR with respect to Npre. However with Npost the story changes, and the higher the ISO amplification is, we have higher useful signal vs Npost, so we are improving final SNR with respect to Npost.

In the real world, Npre would be basically the photon noise (inherent to light capture) plus the read noise (electronic noise) produced in the early stages, prior to ISO amplification. Npost would be the read noise produced after the ISO amplification, i.e. in the AD converter.

Hence on those cameras where a big portion of the total noise is added after the ISO amplification (Canons), the higher the ISO the higher the ouput SNR is for a given amount of photons reaching the sensor. Cameras with a very low read noise (and this consequently means very low Npost) hardly benefit from pushing ISO (Pentax K5, Nikon D7000). Rest of Nikons tend to be in the middle point, but definitively still worth pushing ISO on them like in the Canons.

Canon 350D (both captures at the same shutter/aperture, final exposure matched in pp):
(http://www.guillermoluijk.com/article/iso/versus.jpg)

(this image must have been linked in LL more than 10 times  ;D ).

Regards

Title: Re: ETTR: ISO vs shutter speed
Post by: digitaldog on September 11, 2011, 01:21:53 pm
ISO amplification by itself doesn't alter the SNR, so the ISO amplification does't improve output SNR with respect to Npre. However with Npost the story changes, and the higher the ISO amplification is, we have higher useful signal vs Npost, so we are improving final SNR with respect to Npost.

Excellent, thanks! It explains perfectly why it was mentioned that differing cameras may or may not show the results of the test images you provided (and I saw as well on my Canon).
Title: Re: ETTR: ISO vs shutter speed
Post by: Guillermo Luijk on September 11, 2011, 01:28:40 pm
Excellent, thanks! It explains perfectly why it was mentioned that differing cameras may or may not show the results of the test images you provided (and I saw as well on my Canon).

Yes, see this Pentax K5 test where ISO1600 nearly didn't improve noise vs ISO100 at the same aperture/shutter, so it was best to stay at ISO100 which can be useful to prevent highlight clipping:

Pentax K5 (both captures at the same shutter/aperture, final exposure matched in pp):
(http://www.guillermoluijk.com/article/perfect/comp.jpg)

These are Emil Martinec's ISOless cameras
Title: Re: ETTR: ISO vs shutter speed
Post by: hjulenissen on September 11, 2011, 01:58:58 pm
Yes, see this Pentax K5 test where ISO1600 nearly didn't improve noise vs ISO100 at the same aperture/shutter, so it was best to stay at ISO100 which can be useful to prevent highlight clipping:
But does the AE modes optimally? If you set such a camera in P/Tv/Av, stick ISO to 100 and take an image, will it do e.g. 1/100 sec and underexpose badly (something that, according to your image can be easily fixed in pp), or will it bump the exposure time to 1/10 sec, resulting in a blurred image that cannot be salvaged?

Further, if you set it to manual but keep the auto ISO, will it embed a tag saying "please multiply this image by 10 in raw development", or will it multiply the image digitally in-camera, resulting in any highlights being unnecessarily being blown out?

It seems to me that the new Sony sensor is far better than Nikon/Pentax are at exploting its capabilities?

-h
Title: Re: ETTR: ISO vs shutter speed
Post by: Waeshael on September 11, 2011, 04:46:59 pm
Yes you are essentially correct in what you said. Photon flux noise is independent of camera settings. But apparent noise is dependent on sensor type, filtration method, lens light loss, light scatter within the lens, and of course electronic noise from the amplifiers, and thermal noise as the sensor heats up. Some sensors such as CMYG collect twice the spectrum of RGB and so twice the amount of light - this reduces photon noise of course, and it is observable in the images. Some lenses are much more efficient and so more light reaches the sensor. A larger aperture lens naturally transfers more light etc. Small cameras heat up the sensor very quickly due to battery heat, ambient temperatures and hand heat. When the camera internally reaches 104 degrees F, it no longer meets specs. I carry around small cameras in an ice cooled lunch box if I am forced to shoot in low light - this makes a huge difference in mid-tone noise levels for some cameras.

Considering the basic camera design: For example I have two cameras ( I have many cameras) that produce the same mid-tone image brightness at the same exposure time when one is set to ISO 100 and the other ISO 800 and this is because of a difference in sensor type, and a difference in lens characteristics aperture and efficiency.

Another factor is the way that the photodiodes are coupled together at lower "resolution" to increase s/n, and this is proprietary information, and is a significant factor in image noise in the mid-tones. In the deep shadows (Zone 1) twice nothing is still nothing.

Increase in ISO doesn't necessarily mean a higher noise level in the image - as you said. The lower dynamic range as ISO goes up, pulls up the shadow areas - that is: what was the shadow noise at zone 2 of the scene at ISO 100, becomes the shadow noise at zone 5 of the scene in the high ISO image - the zone 2 detail is lost at high ISO and the dynamic range is compressed which masks what is happening.

I have noticed that the older Tessar lens (Leitz) produces images with small dynamic range i.e the shadows always have detail in them even at ISO 3200 using the NEX-5 APS-C sensor, so that the highlights are kept and the shadow detail is kept - this is a function of the lens design. Even the non Aspheric 35mm Summicron has this quality. But the more modern design Elmarit 28mm does not.

So apparent noise is the result of many design factors.

Anyway thanks for the lucid explanation.

Waeshael
Title: Re: ETTR: ISO vs shutter speed
Post by: Guillermo Luijk on September 11, 2011, 04:50:16 pm
But does the AE modes optimally? If you set such a camera in P/Tv/Av, stick ISO to 100 and take an image, will it do e.g. 1/100 sec and underexpose badly (something that, according to your image can be easily fixed in pp), or will it bump the exposure time to 1/10 sec, resulting in a blurred image that cannot be salvaged?

Further, if you set it to manual but keep the auto ISO, will it embed a tag saying "please multiply this image by 10 in raw development", or will it multiply the image digitally in-camera, resulting in any highlights being unnecessarily being blown out?

Are you really asking a camera manufacturer to think of advanced RAW shooters who want to fully exploit their cameras' potential? you must be joking, we don't even have a f****** RAW histogram  ;D ;D
Title: Re: ETTR: ISO vs shutter speed
Post by: mouse on September 14, 2011, 11:55:02 pm

In the real world, Npre would be basically the photon noise (inherent to light capture) plus the read noise (electronic noise) produced in the early stages, prior to ISO amplification. Npost would be the read noise produced after the ISO amplification, i.e. in the AD converter.

 Cameras with a very low read noise (and this consequently means very low Npost) hardly benefit from pushing ISO
(this image must have been linked in LL more than 10 times  ;D ).

Regards


Guillermo-
These two statements seem somewhat contradictory, unless the second sentence refers only to the read noise produced post amplification.  What am I missing?

Best regards,
Mike
Title: Re: ETTR: ISO vs shutter speed
Post by: Guillermo Luijk on September 15, 2011, 02:48:50 am
These two statements seem somewhat contradictory, unless the second sentence refers only to the read noise produced post amplification.  What am I missing?

Npost is never photon noise, just read noise. If total read noise is low (and it is in Pentax K5 and D7000), any read noise contributor considered individually must be low.
In any case this is just a simple model that tries to explain the behaviour of a real sensor; I'm pretty sure a real sensor is something more complex.