Ad
Ad
Ad
Pages: [1] 2 »   Bottom of Page
Print
Author Topic: Always Expose to the Right?  (Read 2173 times)
mackster
Newbie
*
Offline Offline

Posts: 2


« on: November 16, 2012, 10:54:38 AM »
ReplyReply

Hi,

In the book "Photoshop CS5 for Nature Photographers: A Workshop in a Book", there's a section on exposure titled "Expose Right", written by Michael Reichmann. In it he says to "...bias your exposures so that the histogram is snuggled up to the right but not to the point that the highlights are blown." His logic for saying this is because the right fifth of the histogram contains 2048 of the available 4096 tonal levels, assuming a 12-bit raw image. This makes perfect sense, but should one always do this regardless of the degree of brightness in the image you're trying to capture? For example, I took some pictures last summer of some mountain goats in fading light. I had to turn the ISO up to 800 on my D80 to establish a fast shutter. Anyway, the resulting histograms had a range from all the way to the left (but not clipped) to just short of the middle; the images definitely were not pushed all the way to the right, as Mr. Reichmann suggests. Yes, the images were dim, but it was dim out when I took the photos. Should I have set my exposure to push the histogram to the right and then "...use the available sliders to change the brightness level and contrast so that the data is spreadout appropriately and the image looks right", as Mr. Reichmann says to do in raw-processing software to maximize the signal-to-noise ratio?

Thanks,

Mac
Logged
kikashi
Sr. Member
****
Offline Offline

Posts: 3669



« Reply #1 on: November 16, 2012, 11:24:51 AM »
ReplyReply

Hi,

In the book "Photoshop CS5 for Nature Photographers: A Workshop in a Book", there's a section on exposure titled "Expose Right", written by Michael Reichmann. In it he says to "...bias your exposures so that the histogram is snuggled up to the right but not to the point that the highlights are blown." His logic for saying this is because the right fifth of the histogram contains 2048 of the available 4096 tonal levels, assuming a 12-bit raw image. This makes perfect sense, but should one always do this regardless of the degree of brightness in the image you're trying to capture? For example, I took some pictures last summer of some mountain goats in fading light. I had to turn the ISO up to 800 on my D80 to establish a fast shutter. Anyway, the resulting histograms had a range from all the way to the left (but not clipped) to just short of the middle; the images definitely were not pushed all the way to the right, as Mr. Reichmann suggests. Yes, the images were dim, but it was dim out when I took the photos. Should I have set my exposure to push the histogram to the right and then "...use the available sliders to change the brightness level and contrast so that the data is spreadout appropriately and the image looks right", as Mr. Reichmann says to do in raw-processing software to maximize the signal-to-noise ratio?

Thanks,

Mac

Briefly, yes; unless perhaps in so doing you create unacceptable problems with shutter speed or aperture. Whether it's a good idea to gain ETTR by increasing ISO, and hence noise, I'm not too sure. I suspect it will depend on the sensor.

Jeremy
Logged
Slobodan Blagojevic
Sr. Member
****
Offline Offline

Posts: 5002



WWW
« Reply #2 on: November 16, 2012, 11:31:13 AM »
ReplyReply

The ETTR is becoming less relevant with the latest generation of sensors (though not Canon's).
Logged

Slobodan

Flickr
500px
Isaac
Sr. Member
****
Offline Offline

Posts: 2332


« Reply #3 on: November 16, 2012, 11:48:07 AM »
ReplyReply

I took some pictures last summer of some mountain goats in fading light. I had to turn the ISO up to 800 on my D80 to establish a fast shutter.

As Jeremy implied with "exposure to push the histogram to the right" there's an implicit all other things being equal.

In this example, you were already letting-go some IQ to maintain a fast shutter.

Without that overriding constraint you would have had a choice between the camera's default exposure and "exposure to push the histogram to the right" -- and you would have been able to increase IQ by "exposure to push the histogram to the right".

Logged
Vladimirovich
Sr. Member
****
Offline Offline

Posts: 1149


« Reply #4 on: November 16, 2012, 12:20:35 PM »
ReplyReply

His logic for saying this is because the right fifth of the histogram contains 2048 of the available 4096 tonal levels, assuming a 12-bit raw image.

if you will search this forum for ETTR you shall find THE topic where it was shown that is was an error on his side, ETTR was/is about SNR (for sensor), not about tonal levels (in raw file written by firmware using data read off sensor).

as Mr. Reichmann says to do in raw-processing software to maximize the signal-to-noise ratio?

you maximize the SNR during the exposure, not during raw conversion (unless you are binning = smaller size).
Logged
digitaldog
Sr. Member
****
Offline Offline

Posts: 8064



WWW
« Reply #5 on: November 16, 2012, 12:51:16 PM »
ReplyReply

The ETTR is becoming less relevant with the latest generation of sensors (though not Canon's).

So if one uses ETTR and then what the meter recommends (let's just say non ETTR), both captures on these cameras will have equal quality (S/N ratio)?
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
Slobodan Blagojevic
Sr. Member
****
Offline Offline

Posts: 5002



WWW
« Reply #6 on: November 16, 2012, 01:36:28 PM »
ReplyReply

So if one uses ETTR and then what the meter recommends (let's just say non ETTR), both captures on these cameras will have equal quality (S/N ratio)?

A technical answer to that is "above my pay grade," but from a practical photographer's point of view, I would assume the answer would be yes (or at least the difference is negligible for all practical purposes). Especially taking into account that ETTR

1. requires fiddling
2. involves guessing
3. risks overblown highlights
4. requires extra step in post-processing
5. creates lousy previews for chimping

So, what I do in practice is I go for an automated exposure if I need to get that shot first. If I have the time, then I will go through all those motions mentioned above. Then again, I am a Canon user, so I am cursed to do so Wink

I am sure more knowledgeable folks will correct me, if wrong.

Logged

Slobodan

Flickr
500px
Vladimirovich
Sr. Member
****
Offline Offline

Posts: 1149


« Reply #7 on: November 16, 2012, 01:47:30 PM »
ReplyReply

I am sure more knowledgeable folks will correct me, if wrong.

you spot meter the most bright area in your image where you still want to have some details and then dial in the calculated exposure compensation... which for canon cameras might be 3 stops (as they probably meter below 12.x%).
Logged
digitaldog
Sr. Member
****
Offline Offline

Posts: 8064



WWW
« Reply #8 on: November 16, 2012, 01:48:35 PM »
ReplyReply

Especially taking into account that ETTR

1. requires fiddling
2. involves guessing
3. risks overblown highlights


That's exposure and photography 101. We did it in the old days with transparency film to a precise degree (snip tests anyone?).

Quote
4. requires extra step in post-processing

Once yes. Then you build a preset to normalize and the image is back to the same basic appearance as not using ETTR.

Quote
5. creates lousy previews for chimping

It does indeed. But then even bypassing exposure issues, the JPEG and what I want from the raw are quite different.

Quote
So, what I do in practice is I go for an automated exposure if I need to get that shot first. If I have the time, then I will go through all those motions mentioned above.

Exactly what I do and would recommend. But when considering the later part of the process, it seems there is an ideal exposure and then all others. I would submit that proper and ideal exposure for raw is largely based on the ETTR premise. Maybe it should not even be called ETTR. Maybe it should be called best exposure for raw data.

The best exposure for ISO 100 (you name it) color neg was rarely based on ISO 100 printed on the box. If one did exposure and development testing, as I was forced to do, one finds the best (ideal) quality capture is based on something quite different.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
Slobodan Blagojevic
Sr. Member
****
Offline Offline

Posts: 5002



WWW
« Reply #9 on: November 16, 2012, 02:01:24 PM »
ReplyReply

you spot meter the most bright area in your image where you still want to have some details and then dial in the calculated exposure compensation...

That's what I called 1. fiddling

Quote
which for canon cameras might be 3 stops (as they probably meter below 12.x%).

Thats what I called 2. guessing

 Grin
Logged

Slobodan

Flickr
500px
bjanes
Sr. Member
****
Offline Offline

Posts: 2714



« Reply #10 on: November 16, 2012, 02:07:05 PM »
ReplyReply

Hi,

In the book "Photoshop CS5 for Nature Photographers: A Workshop in a Book", there's a section on exposure titled "Expose Right", written by Michael Reichmann. In it he says to "...bias your exposures so that the histogram is snuggled up to the right but not to the point that the highlights are blown." His logic for saying this is because the right fifth of the histogram contains 2048 of the available 4096 tonal levels, assuming a 12-bit raw image. This makes perfect sense, but should one always do this regardless of the degree of brightness in the image you're trying to capture? For example, I took some pictures last summer of some mountain goats in fading light. I had to turn the ISO up to 800 on my D80 to establish a fast shutter. Anyway, the resulting histograms had a range from all the way to the left (but not clipped) to just short of the middle; the images definitely were not pushed all the way to the right, as Mr. Reichmann suggests. Yes, the images were dim, but it was dim out when I took the photos. Should I have set my exposure to push the histogram to the right and then "...use the available sliders to change the brightness level and contrast so that the data is spreadout appropriately and the image looks right", as Mr. Reichmann says to do in raw-processing software to maximize the signal-to-noise ratio?

Thanks,

Mac

The rationale of ETTR is to maximize the signal to noise ratio by collecting the maximum number of electrons. The number of levels is a red herring (as explained by Emil Martinec here). ETTR works best at base ISO where a histogram with the highlights just short of clipping indicates that the sensor is just short of saturation (assuming the histogram represents the sensor saturation, which it usually does not, since the histogram and blinking highlights are usually conservative). For each doubling of the ISO over base, the maximum effective number of photons that can be collected is cut in half due to clipping in the analog to digital converter. With the same camera exposure, an ETTR histogram at ISO 400 with a camera whose base ISO is 100 would be 2 EV to the left of that with the camera set to ISO 100. The former situation does not necessarily give a better SNR, depending on the camera.

With an ISO less camera such as the Nikon D7000, with the same camera exposure, it doesn't matter with light starved situations if one increases the ISO on the camera or the exposure in the raw converter. This is because the read noise of the D7000 does not vary with ISO. With other cameras such as the D3 and most Canons, read noise diminishes as one increases the ISO up to a certain point, often ISO 800 or so, but doesn't change much thereafter. With such a camera, one should definitely increase the ISO on the camera in light starved situations. However, increasing the ISO above the point of diminishing returns gives no improvement in the SNR but does reduce highlight headroom. Emil explains all this in his post.

My own approach in this situation with the D3 would be to increase the ISO to 800 or so to minimize read noise and also improve the appearance of the LCD preview. Any further brightening of the image could be accomplished in the raw converter.

Regards,

Bill
« Last Edit: November 16, 2012, 02:15:54 PM by bjanes » Logged
ErikKaffehr
Sr. Member
****
Online Online

Posts: 6921


WWW
« Reply #11 on: November 16, 2012, 03:31:20 PM »
ReplyReply

Hi,

I agree with Bill, excellent advice.

I would just add that exposing to the right essentially minimizes noise. Increasing ISO tweaks readout noise on Canons and Nikons not using Sony Exmoor based sensors. The real reason that you want expose to the right is that you want to increase the number of photons captured by the sensor. More photons less noise!

The reason that increasing ISO does not help Sony Exmoor based sensors is that ADC (Analog Dialog Conversion) is on chip and it seems that ISO doesn't affect ADC read noise (or helps very little). The Sony Exmoors have also very little readout noise.

That is about the shortest explanation I can come up with.

Best regards
Erik


The rationale of ETTR is to maximize the signal to noise ratio by collecting the maximum number of electrons. The number of levels is a red herring (as explained by Emil Martinec here). ETTR works best at base ISO where a histogram with the highlights just short of clipping indicates that the sensor is just short of saturation (assuming the histogram represents the sensor saturation, which it usually does not, since the histogram and blinking highlights are usually conservative). For each doubling of the ISO over base, the maximum effective number of photons that can be collected is cut in half due to clipping in the analog to digital converter. With the same camera exposure, an ETTR histogram at ISO 400 with a camera whose base ISO is 100 would be 2 EV to the left of that with the camera set to ISO 100. The former situation does not necessarily give a better SNR, depending on the camera.

With an ISO less camera such as the Nikon D7000, with the same camera exposure, it doesn't matter with light starved situations if one increases the ISO on the camera or the exposure in the raw converter. This is because the read noise of the D7000 does not vary with ISO. With other cameras such as the D3 and most Canons, read noise diminishes as one increases the ISO up to a certain point, often ISO 800 or so, but doesn't change much thereafter. With such a camera, one should definitely increase the ISO on the camera in light starved situations. However, increasing the ISO above the point of diminishing returns gives no improvement in the SNR but does reduce highlight headroom. Emil explains all this in his post.

My own approach in this situation with the D3 would be to increase the ISO to 800 or so to minimize read noise and also improve the appearance of the LCD preview. Any further brightening of the image could be accomplished in the raw converter.

Regards,

Bill
Logged

Vladimirovich
Sr. Member
****
Offline Offline

Posts: 1149


« Reply #12 on: November 16, 2012, 03:33:37 PM »
ReplyReply

Thats what I called 2. guessing
you calculate that once, no guessing.
Logged
Tony Jay
Sr. Member
****
Offline Offline

Posts: 2029


« Reply #13 on: November 16, 2012, 05:13:24 PM »
ReplyReply

ETTR is still subject to much misunderstanding as originally conceived.
When this is combined with the current shifting of the goalposts created by the newer Sony sensors (the point Slobadan was making) greater confusion is possible.

The OP, in introducing the topic of ETTR, gives an example of a shooting situation where trying to apply ETTR principles does not make sense.
There was never any chance, due to the combination of light and subject (moving goats), of shooting at base ISO.
Shooting at base ISO (as already stated) is absolutely fundamental to the point of ETTR - which is to reduce noise generation.

Technically, Slobadans point about guessing what the real highlight exposure cutoff for one's sensor is incorrect since this information can be easily sought. In practice I have learn't from trial and error how far I can push exposure in the ETTR context with the cameras I use with excellent results.

The postprocessing involved in "normalizing" the exposure, with a little bit of practice is simplicity itself. I do not use a preset in this process. I always use a 'season to taste' approach.

It seems that many individuals are trying to apply principles of ETTR to situations that make no sense.
So, if you are shooting at high ISO for any reason trying ETTR is a nil sum game.
It is possible to use ETTR in poor light scenarios if the camera is on a tripod and the subject of the image is static - shoot at base ISO and use the appropriate (long) shutter speed.
Shooting live action in poor light (the OP's example) ETTR cannot work, simple as that, since any sort of image quality will require fast shutter speeds and hence appropriately high ISO's.
Also don't try ETTR if one has to shoot RAW and JPEG simultaneously (seems a no-brainer but this does confuse some).

Could the principle of ETTR become redundant going forward? Yes, possibly.
I don't own a D800, or any other camera that contains a recent Sony Exmoor sensor, however it is clear that the image quality obtainable from these cameras at high ISO's, not to mention future developments, may make the concept of ETTR redundant.
I am not sure if anyone who owns a D800 and is intimately familiar with ETTR has yet contributed to this post.

Slobadan may be pointing us to a time where ETTR will be of interesting but historical interest.

Tony Jay
Logged
FranciscoDisilvestro
Sr. Member
****
Offline Offline

Posts: 323


WWW
« Reply #14 on: November 16, 2012, 05:23:28 PM »
ReplyReply

Shooting at base ISO (as already stated) is absolutely fundamental to the point of ETTR - which is to reduce noise generation.

There are exeptions to this in cameras that have high read noise at low ISO. You have to check the characteristics of your camera/sensor (as in DxOMark) and see up to which ISO it makes sense to do ETTR. Rememer it is not the amount of noise but the maximun signal to noise ratio (SNR) that matters.
Logged

bjanes
Sr. Member
****
Offline Offline

Posts: 2714



« Reply #15 on: November 16, 2012, 05:48:49 PM »
ReplyReply

I would just add that exposing to the right essentially minimizes noise. Increasing ISO tweaks readout noise on Canons and Nikons not using Sony Exmoor based sensors. The real reason that you want expose to the right is that you want to increase the number of photons captured by the sensor. More photons less noise!

The reason that increasing ISO does not help Sony Exmoor based sensors is that ADC (Analog Dialog Conversion) is on chip and it seems that ISO doesn't affect ADC read noise (or helps very little). The Sony Exmoors have also very little readout noise.

Erik,

What you is quite true for such sensors as in the Nikon D7000, but the Exmoor sensor as implemented in the D800
does appear to have reduced read noise at higher ISOs. See this post for details.

Regards,

Bill
Logged
Slobodan Blagojevic
Sr. Member
****
Offline Offline

Posts: 5002



WWW
« Reply #16 on: November 16, 2012, 06:00:11 PM »
ReplyReply

So if one uses ETTR and then what the meter recommends (let's just say non ETTR), both captures on these cameras will have equal quality (S/N ratio)?

In all this back-and-forth, we (ok, you techies) kind of forget to answer the above? In one sentence first, please (for us, non-techies), and then you can explain it further for all the geeks out there Wink
Logged

Slobodan

Flickr
500px
ErikKaffehr
Sr. Member
****
Online Online

Posts: 6921


WWW
« Reply #17 on: November 16, 2012, 06:10:47 PM »
ReplyReply

Hi Bill,

I have seen Marianne's note but I don't think it agrees with what I can see in DxO measurements, see enclosed screen dump.

I also see very small change in readout noise in the sensorgen data: http://www.sensorgen.info/NikonD800.html

"Bclaff's" data indicates a small reduction of read noise with increasing ISO on the D800 but much less than non Exmoors.
http://home.comcast.net/~NikonD70/Charts/RN_e.htm

Best regards
Erik



Erik,

What you is quite true for such sensors as in the Nikon D7000, but the Exmoor sensor as implemented in the D800
does appear to have reduced read noise at higher ISOs. See this post for details.

Regards,

Bill
Logged

ErikKaffehr
Sr. Member
****
Online Online

Posts: 6921


WWW
« Reply #18 on: November 16, 2012, 06:18:05 PM »
ReplyReply

Hi,

It is not an easy question to answer!

Midtones will only be affected by exposure, so whatever your ISO is the midtones will depend on light, aperture and fstop, not much else. At least the way I see it.

If you increase ISO, so you expose more to the right shadows may improve due to reduction in shadow noise, but it depends on the camera you have. Also I'm pretty sure that changing ISO also affects the histogram, don't forget that he histogram is based on a "jpeg processed preview" and not actual sensor signal.

How things work out may depend much on the camera one has. So I can make an experiment with my Sony to find out but that observation would not be valid for your Canon. Sad but true.

Best regards
Erik


In all this back-and-forth, we (ok, you techies) kind of forget to answer the above? In one sentence first, please (for us, non-techies), and then you can explain it further for all the geeks out there Wink
« Last Edit: November 16, 2012, 06:20:33 PM by ErikKaffehr » Logged

bjanes
Sr. Member
****
Offline Offline

Posts: 2714



« Reply #19 on: November 16, 2012, 09:20:20 PM »
ReplyReply


I have seen Marianne's note but I don't think it agrees with what I can see in DxO measurements, see enclosed screen dump.

I also see very small change in readout noise in the sensorgen data: http://www.sensorgen.info/NikonD800.html

"Bclaff's" data indicates a small reduction of read noise with increasing ISO on the D800 but much less than non Exmoors.
http://home.comcast.net/~NikonD70/Charts/RN_e.htm

Best regards
Erik


The DXO DR data for the D800e do show a flattening towards the left as Ray describes. Going from nominal ISOs of 100 to 200 causes only a 0.48 stop decrease in DR rather than the expected full f/stop decrease. This means that increased read noise at base ISO has thrown away about a half f/stop of DR. Readers should see the explanation by Emil, who interestingly was responding to a question that you had posed. The take home point is that one can determine the ISO where read noise approaches a minimum by noting the ISO where DR decreases by one f/stop for each doubling of ISO as shown on the DXO DR plot. For the D800e this is about at ISO 200, and for the Nikon D3e, it is not reached until 3200 or thereabouts as shown by Emil's analysis.

Regards,

Bill

Logged
Pages: [1] 2 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad