Dynamic Range - What's the trade off?

[dje]

George

You are not the only one confused by this business!

The general thrust of the concept, as discussed on most web sites, is that with say DR200% the ISO is set to 400 and the camera metering uses this figure to set ss/f stop for "correct" exposure. However the actual ISO used by the camera is 200 and thus the shot is under-exposed by 1 stop. The discussion then goes on to say that this under-exposure allows more detail to be preserved in the highlights. The shadows are then lifted with some sort of tone curve.

I might be missing something but this explanation just doesn't sit right with me. Firstly, why does the image need to be under-exposed to preserve highlight detail? It should all be there provided you don't clip anything. Secondly, why not just bring the shadows up with a tone curve on a normally exposed image?

As a result of these doubts, I'm more inclined to think that the concepts discussed in the TechRadar article I referred to in post 32 (ie a "dual ISO" capture) have more credence. But I just don't know!

Robin would it be possible to do a simple test please to see just what exposure settings the camera chooses?
Shot 1 : ISO set to 200 , f stop set to a certain value, note ss chosen by camera
Shot 2 : ISO set to 400, DR200% selected, f stop as above, note ss chosen by camera

This would tell us if the metering is based on ISO400 or ISO200 for the second shot.

John(2), sorry to prolong this thread, you've probably got more than you bargained on!

Dave

[dje]

In DPR's method of measurement, the exposure for the test is set to put the middle part of the test target in the middle of the test image tonal range.

Would remind everybody that there are several definitions of dynamic range, so saying "DR" without any qualification often leads to misunderstanding. Any disagreement with this particular post should be made with reference to the link posted above and with reference to DPR's test method - not some alternative definition of DR.

Yes Ted I'm not sure that i like DPReview's way of defining DR but maybe I just don't understand where they are coming from!

Dave

[Inkanyezi]

I don't know whether I should open the can of worms regarding "under-exposure" that has been mentioned here.

However, there is not a "correct" exposure, to which "under" and "over" relate, provided by a behind the lens meter in a camera, only an evaluation based on what is in front of it. As I had the Active D-lighting explained long ago, it was a feature allowing a greater DR capture for the chosen jpeg processing in the camera, which held back highlights, by taking a highlight reading and setting exposure for the brightest highlights near clipping. That is a rather simple procedure, which can be made manually with most cameras by spot metering with positive compensation.

So whether exposure changes from normal Auto mode (any of the PAS settings) is a matter of available highlights. The camera does not set a higher ISO, but it adapts to the contrast of the scene. Bright highlights will be preserved by taking a highlight reading and make sure they don't clip. When there are no very bright highlights, the system will allow more of the available light to hit the sensor. The precision of highlight assessment depends on the number and size of fields that are evaluated by the meter.

In effect, this might mean that normal Auto could take the exposure in bright sunlight at ISO 200 at 1/500 f/11, while Active D-lighting might take it at 1/2000 f/11, in order not to clip very bright highlights, which would have been clipped with a normal Auto exposure. It is evident that as it is a change of exposure, the RAW file is less exposed when highlights are present and Active D-lighting is applied. The ISO however is not changed.

Some will say that the Active D-lighting image is "under-exposed", but the first part of that assessment, under, is a subjective evaluation, not objective truth.

It becomes a bit clearer, as we see that if there are no large very bright highlights, the exposure will not be different from normal with Active D-Lighting. If it were so simple as just setting the meter to a higher ISO than the actual one, it would have done exactly the same as applying minus compensation. Instead of just compensating the image darker, the "Active" setting is an instruction to the light meter system to read highlights and adapt exposure not to clip them.

In the other end of the brightness range of the image, D-Lighting can be set to different values. D-Lighting does not alter the RAW data; it lifts the darker parts of the tone curve when processing to jpeg, effectively raising ISO digitally.

[george013]

George

You are not the only one confused by this business!

The general thrust of the concept, as discussed on most web sites, is that with say DR200% the ISO is set to 400 and the camera metering uses this figure to set ss/f stop for "correct" exposure. However the actual ISO used by the camera is 200 and thus the shot is under-exposed by 1 stop. The discussion then goes on to say that this under-exposure allows more detail to be preserved in the highlights. The shadows are then lifted with some sort of tone curve.

I might be missing something but this explanation just doesn't sit right with me. Firstly, why does the image need to be under-exposed to preserve highlight detail? It should all be there provided you don't clip anything. Secondly, why not just bring the shadows up with a tone curve on a normally exposed image?

As a result of these doubts, I'm more inclined to think that the concepts discussed in the TechRadar article I referred to in post 32 (ie a "dual ISO" capture) have more credence. But I just don't know!

Robin would it be possible to do a simple test please to see just what exposure settings the camera chooses?
Shot 1 : ISO set to 200 , f stop set to a certain value, note ss chosen by camera
Shot 2 : ISO set to 400, DR200% selected, f stop as above, note ss chosen by camera

This would tell us if the metering is based on ISO400 or ISO200 for the second shot.

John(2), sorry to prolong this thread, you've probably got more than you bargained on!

Dave

I think it's the same as ADL with Nikon and other brands with there own function.
You use it when there is to much contrast in the image, highligts and special clipping. It's not much more as getting the right exposure for those highlights, meaning underexposure for your subject, and than correcting the darker parts.
The difference is what I think more how the correction is expressed or calculated. Whether you call it a percentage, higher ISO or whatever, you correct with A and/or S.

Try it with your camera. And watch the exposure changing with or without DRx%. I don't have that camera.

George

[xpatUSA]

Yes Ted I'm not sure that i like DPReview's way of defining DR but maybe I just don't understand where they are coming from!

Hi Dave, it looks to me that they sought an easily-understood and repeatable method. Their back-lit target represents a quantifiable scene with a scene dynamic range that exceeds the capability of most if not all cameras. Quantifiable because the target itself is calibrated to have 1/3 EV steps. Only one image per camera 'DR' setting is needed to produce their curves (after they've got the exposure set correctly). Here is an example of one of their shots:



On my monitor screen and at a quick glance, I see steps below 10 and above maybe 31. Outside of that range, the steps have no contrast, i.e. they are blown or are at black level. So I personally am saying the camera that took this shot captured a 'DR' of 21 steps * 1/3 = 7 EV. Your monitor might show different but hopefully the method is clear?

As we know, other methods e.g. 'photon transfer curve' involve taking multiple images from cap-on-lens up to deliberately blown - often with arguable results.

[Black Pearl]

George

You are not the only one confused by this business!

Robin would it be possible to do a simple test please to see just what exposure settings the camera chooses?
Shot 1 : ISO set to 200 , f stop set to a certain value, note ss chosen by camera
Shot 2 : ISO set to 400, DR200% selected, f stop as above, note ss chosen by camera

This would tell us if the metering is based on ISO400 or ISO200 for the second shot.

John(2), sorry to prolong this thread, you've probably got more than you bargained on!

Dave

My first reply at the beginning of the thread seems to have proved a bit....well....wrong if I'm honest :roll eyes:
Actually I'm really glad this cropped up and it turns out I've learned new things about my X-T1 which is never a bad thing.

You can shoot raw or jpeg or both when using the DR Expansion modes so the camera is manipulating the data right at source and without having to manually correct the raw files when you open them they still have an extended DR and are presented with the correct exposure. This means (and I've since read this online) that shooting at 800iso using DR400 with a raw file can give you a great basis for processing. This I'll need to look into further to see how the noise is being dealt with (pretty good with Fuji typically so I'm not too worried) and see how it compares with a basic base 200iso file....something for another day.

As to that exposure - it turns out from the test shots below that the camera is displaying and using the correct shutter speed/aperture for the selected ISO. I had initially presumed it indicated an under exposure and then fiddled with the resulting dark file to produce a final output so holding onto lost highlights and evening out darkened shadows. It may be doing that under the hood as it were but you wouldn't know that when shooting. I shot at native 200iso then 400iso with/without DR Expansion and then ran a test at 800iso using 100% 200% & 400%.


All shots are with a Fuji X-T2 + 60mm f2.4 Macro mounted on a tripod taken within a few seconds of each other on an unchanging overcast day. I shot on aperture priority (lens set to f4) changing ISO and the DR Expansion from shot to shot. All the raw (Fuji RAF) files were imported into Lightroom with no presets and using the standard Adobe Profile. I made sure LR displayed lost highlights and just screen grabbed each file straight from the Develop screen.

This is shot at the base 200iso with standard DR100


This is shot at 800iso with standard DR100


This is shot at 800iso with DR200 selected


This is shot at 800iso with DR400 selected

[george013]

There is no exif in your photo's. I still don't know what is the exposure. I don't want to know good or bad, but how much, S and A and the ISO.
And try it with more dominant highlights. LR has a tresh hold.

George

[Black Pearl]

There is no exif in your photo's. I still don't know what is the exposure. I don't want to know good or bad, but how much, S and A and the ISO.
And try it with more dominant highlights. LR has a tresh hold.

George

All the exposure details are right there on the photographs.

This was a quick test to see what the exposures were when the DR was changed and to see if they followed the expected iso change - which they do. If you want further examples then try asking nicely

[george013]

All the exposure details are right there on the photographs.

This was a quick test to see what the exposures were when the DR was changed and to see if they followed the expected iso change - which they do. If you want further examples then try asking nicely

Please, could you make a same serie with real blown out higlights? Please.

But serious, I think with really more clipping parts in the picture you can get a better idea of what the camera is doing. If real clipping disapears with no exposure correction, then that's magic. I think.

Also the ADL of Nikon doesn't do much if nothing has to be done.

George

[dje]

Robin

Many thanks for that, most interesting. In the DR expansion shots, the shadows appear to have lifted a bit according to the histogram and the highlight clipping has been reduced a bit.

Could I ask one more favour please - is it possible to post the raw files on a host such as Dropbox so I could download them? I would like to have a look at them in RawDigger. Also if you happened to shoot RAW plus jpeg, it would be interesting to see what the jpegs were like.

Dave

[dje]

Hi Dave, it looks to me that they sought an easily-understood and repeatable method. Their back-lit target represents a quantifiable scene with a scene dynamic range that exceeds the capability of most if not all cameras. Quantifiable because the target itself is calibrated to have 1/3 EV steps. Only one image per camera 'DR' setting is needed to produce their curves (after they've got the exposure set correctly). Here is an example of one of their shots:



On my monitor screen and at a quick glance, I see steps below 10 and above maybe 31. Outside of that range, the steps have no contrast, i.e. they are blown or are at black level. So I personally am saying the camera that took this shot captured a 'DR' of 21 steps * 1/3 = 7 EV. Your monitor might show different but hopefully the method is clear?

As we know, other methods e.g. 'photon transfer curve' involve taking multiple images from cap-on-lens up to deliberately blown - often with arguable results.

Thanks Ted. I didn't have a problem with their test method using the Stouffer step wedge but more with the use of a "middle grey " reference point. But I'll give it some more thought, I'm coming around

Dave

[Black Pearl]

Robin

Many thanks for that, most interesting. In the DR expansion shots, the shadows appear to have lifted a bit according to the histogram and the highlight clipping has been reduced a bit.

Could I ask one more favour please - is it possible to post the raw files on a host such as Dropbox so I could download them? I would like to have a look at them in RawDigger. Also if you happened to shoot RAW plus jpeg, it would be interesting to see what the jpegs were like.

Dave

There you go:

https://www.dropbox.com/sh/byv9rpsoc...DST7YoBqa?dl=0

File sequence runs 200iso/DR100 - 800iso/DR400 - 800iso/DR200 - 800iso/DR100 for both RAF and JPG

[dje]

Thank you sir, I'll have a play.

[Black Pearl]

A quick comparison between the JPG output from the camera showing DR100 against the DR400. Both are shot with exactly the same ISO - Aperture - Shutter Speed so the difference (in the highlights) is down to how the camera is manipulating the data.

[george013]

A quick comparison between the JPG output from the camera showing DR100 against the DR400. Both are shot with exactly the same ISO - Aperture - Shutter Speed so the difference (in the highlights) is down to how the camera is manipulating the data.

http://www.fujirumors.com/how-to-expand-dynamic-range/

It does change exposure.

It analyzes the subject and determines if its range of contrast exceeds the sensor’s dynamic range.
If it does, the camera exposes the image one (DR200%) or two (DR400%) f-stops (exposure values) more conservatively than normal in order to retain detail in the bright areas of the image (highlights). This is why it is necessary to have an ISO setting of at least 400 for DR200% and an ISO setting of at least 800 for DR400%.
The camera processes the JPEG by remapping the tones in the RAW image (that was intentionally exposed too conservatively) to bring the shadows and middle tones back to “normal.” In turn, the amplification of the sensor signal is reduced by one (DR200%) or two (DR400%) exposure values.

It uses a lower ISO to calculate an underxposure of 1 or 2 stops.


George

[Black Pearl]

It changes the processing of the data but it doesn't change the aperture or shutter speed both of which are exactly the same for a given chosen iso regardless of the DR gain chosen.

Have a look at the DR examples I've posted - the ISO is 800 on each - the aperture is f4 on each - the shutter speed is 1/480's on each - what has changed is the way the camera then takes that exposure and manipulates it before recording it. It may be altering the iso (arguably changing the exposure) but without showing you and it has no affect on the way you make an exposure.

[dje]

I had a look at the raw files in RAWDigger. Here are the raw histograms. Click on the first to bring it up in Litebox and then scroll through the four screen shots to get a good comparison. The first two are ISO200/DR100% and ISO800/ DR100% respectively and the histograms are pretty much identical (as you would expect). The third and forth are ISO800/DR200 and ISO800/DR400 respectively. With these, the shape is pretty much the same but for ISO/DR200, the histogram is shifted to the left by 1EV and for ISO800/DR400, the histogram is shifted to the left by 2EV.

This confirms in my mind that for DR200 and DR400:

Exposure is set according to the ISO setting selected (800 in this case) but the actual ISO used for capture is 400 and 200 for DR200 and DR400 respectively, thus resulting in the underexposure shown in the raw histograms. Note also how the clipping has gone in the DR400 shot.

When these raw files are opened in ACR, they all appear to have the same exposure but I believe exposure compensation is being applied automatically by ACR, probably as a result of a DR expansion flag setting in the raw file metadata. With the DR400 file, there is less clipping and this can largely be removed (except for some blue) by reducing the exposure a bit.









The next four photos are 100% crops of the four images in the same order as above. These were produced from ACR (no noise reduction or sharpening) with the shadows lifted to 100% to emphasise the noise. The first shot (base ISO) has slightly lower noise whereas the other three have about the same noise each. This is because of the higher ISO and/or exposure compensation used. The fact that the last three are all about the same indicates ISO-Less behavior.









Dave

PS: I think I am starting to get Fuji's approach. Initially the concept of improving DR by under-exposing seemed counter-intuitive. However I think the key factors are the ISOLess nature of the sensor and the very good noise performance at ISO800. This means you can take a shot at ISO 200 , underexpose it by 2 stops and then increase exposure by 2 stops and end up with the same noise performance that you would achieve with a well exposed ISO800 shot. The advantage with this approach is that instead of increasing exposure in post by exactly 2 stops, you can increase exposure by just the right amount so that the brightest highlights are just on the verge of clipping.

[William W]

^ Nice job THANKS! - I bet that used up your lunch hour.

[dje]

^ Nice job THANKS! - I bet that used up your lunch hour.

Bill you are forgetting that we worry about the curtains fading up here and are not on daylight saving time. Lunch time beckons soon I hope.

Dave

[george013]

I don't have a Fuji and am not interested in one. I'm just surprised how arguments are set away.

The example of Robin might not be the best. There is hardly clipping in it. Highlights doesn't mean clipping by definition.
When there's clipping in the image, it's impossible that to correct in pp and get back details that where not there. The only way to get back details in the clipping area is exposure correction.
As I stated many time before, one can not see a RAW-file as an image. You see the result of the converter.

George