Expose to the right (ETTR) or underexpose?

[Jaybob]

Advertisement (gone after registration)

...It's safer to just leave the ISO where it is and boost in the converter, assuming, like Jaybob mentioned, that you don't have to give the actual raw files to a client, because they might think you don't know what you're doing.

 

 

I don't ever give my clients RAW files, they only receive prints, or edited, color balanced, and exposure corrected files.

[douglasf13]

I don't ever give my clients RAW files, they only receive prints, or edited, color balanced, and exposure corrected files.

 

Sorry about that, I misunderstood your earlier post. I thought you were saying that you can't let clients see the raws with this method, because the exposure needed to be boosted, first. I guess it could also be a problem if you shoot tethered and/or let clients see the shots as you go, before processing them. There are certainly several practical reasons to boost the M9's ISO in camera. It just isn't really necessary from an IQ standpoint.

Edited October 19, 2012 by douglasf13

[k-hawinkler]

Hi. Yep, these were from DNG's converted identically out of Lightroom 4 at Adobe's standard baseline settings, except that the ISO 160 file had the exposure slider at +2.90 EV. Pretty close, huh? I'm pretty sure I wouldn't be able to tell which is which. Plus, apparently Lightroom isn't the best converter to do this with, either, because of the way that it handles shadows, according to those that understand the technology more than I do.

 

With cameras that have linear DR like the M9, lots of MFDBs, Sony EXMOR cameras, etc., the amount of light getting to the sensor, which I would call exposure, is the important part. Changing ISO, while useful in calculating what your exposure should be, isn't technically necessary. With other cameras, like Canon DSLRs and the pro Nikon sports DSLRs that don't use Sony sensors, it is more beneficial to boost in-camera ISO, at least up to a point, because of the design of their ADCs.

 

Actually, to take this to a more confusing level, most cameras fall somewhere in between, where the in-camera ISO performs better up to a point, and then the raw converter push is better past that point. It just depends on the camera. In fact, with the Fuji X100, if you set ISO 3200, it actually just shoots at ISO 1600, and then the raw metadata tells the raw converter to push the exposure +1.0 EV under the hood without any indication of it doing so. Tricky! Some MFDB cameras do that with all ISOs above base ISO.

 

While this info may or may not be practical in all cases, here is one case where not upping the ISO may be helpful. Imagine you're in low light, and your camera is set to ISO 800, f2 and 1/30. Your meter/histogram tells you that you're "underexposed," so you have to decide your next step:

 

1) You could open up the aperture, but you're already at max aperture, so it isn't an option.

2) You could slow down your shutter speed, but, you don't want to, because you'll risk too much subject movement.

3) You can raise ISO, which will move your histogram to the right.

 

With the M9, step 3 is not necessary, because your raw converter is just as good at pushing the exposure as your camera, and, by pushing ISO in-camera, you risk over compensating and accidentally blowing out highlights. That's the beauty of sticking with the lower ISO. You don't risk blowing highlights by allowing the camera to do the gain boost. When I shoot the M9 in lowlight, I usually know what aperture and shutter are my bare minimum, so, once I have those set, the ISO kind of doesn't matter, unless you boost the ISO too much and blow the highlights. It's safer to just leave the ISO where it is and boost in the converter, assuming, like Jaybob mentioned, that you don't have to give the actual raw files to a client, because they might think you don't know what you're doing.

 

Does that make sense at all?

 

p.s. picture 1 was the ISO 160 file boosted by +2.9 EV, so your friend preferred that over the native ISO 1250 file. it's a super close call.

 

 

 

Many thanks. Very interesting.

Last year I was trying to explore how to get the "best" exposure.

I did some tests with my M9 + TE 1:4/135, varying exposure time while keeping ISO=160 and f=5.6.

Exposure time was changed from 1/3000 s to 1/350s. That's slightly over 3 EV. However, no highlights were blown.

Here goes:

 

1/3000 s

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

 

1/2000 s

 

1/1500 s

 

1/1000 s

 

1/750 s

 

1/500 s

 

1/350 s

 

In the next post I will show what these images look like when exposure is corrected appropriately.

[k-hawinkler]

Now with exposure corrections in PS CS6.

 

 

1/3000 s, +1.55 EV

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

 

 

1/2000 s, +1.20 EV

 

 

1/1500 s, +0.55 EV

 

 

1/1000 s, +0.10 EV

 

 

1/750 s, -0.45 EV

 

 

1/500 s, -0.90 EV

 

 

1/350 s, -1.55 EV

 

Exposure at around 1/1000 s is probably most convenient in this case.

But after an exposure correction is applied, all the images look indeed very similar.

 

Full size images can be found here:

Edited October 21, 2012 by k-hawinkler

[dwbell]

Again, surprisingly similar! I would expect much more shadow detail in pulled 350 than in pushed 3000, but that seems to be not the case.

 

This is perhaps useful to understand a 3 stop latitude for street or reportage. Say sunny sixteen, f/4 1/4000 ISO160, open shadow is often 5 stops down from there but if you want to protect blown highlights in the sunny shots just leave it at 4000 and don't flirt with 2000, 1000, or 500 because you can pull those ranges in post with apparently no negative effect.

 

In open shade one could leave the shutter at 125 knowing you have 60, 30 and 15 covered. Combined that's a considerable simplification of shooting practice IF there's no appreciable IQ loss in the file.

 

I might need to do my own testing to confirm. Thanks k-hawinkler.

 

(Before the purists get offended, I'm not advocating sloppy metering technique, far from it, protecting highlights by shooting a pushed 4000 instead of a straight 1000 with no IQ loss seems like good technique to me? Especially relevant to MM users one would imagine)

[dwbell]

Just a thought. Perhaps we're seeing the exceptional quality of the sensor at base ISO. It might be interesting to see the results of such tests at subsequent higher ISO ranges to ascertain if the expected quality drop then starts to show?

[wlaidlaw]

Advertisement (gone after registration)

I would put forward a different acronym: ENTTR = Expose Nearly to the Right. This I would maintain is a more appropriate methodology for day to day photography. Unless you are on a tripod and carefully examining your histogram for the shot, adjusting it to run right to the blow out limit, with multiple shots, you will risk more blown out highlight shots than if you plan to leave a margin. The small histogram on the LCD of an M8/9 is not particularly easy to interpret, nor do I feel it is terribly accurate. If you constantly run up against the far right hand end, you will end up with a lot of blown out highlight shots. If you want to run right up to the limit, in order to maximise the information at the dark end (and I do understand that), then surely exposure bracketing and HDR is a more appropriate practical solution than ETTR.

 

Wilson

[Jaybob]

my 500th post!

 

My own test...

 

Sony NEX-5 Summicron 50

2 exposures in dimmer than usual conditions, custom white balance taken off the mat in the white area on the right.

 

one ETTR'ed at 3200 13th of a second set to f2.8

one at base ISO 200 13th of a second set to f2.8

 

the 200 ISO has been LR4 BOOSTED Four FULL stops, to 3200.

 

Imported into LR4 with no develop settings applied, and no import presets.

The ONLY correction is Sharpening is turned to 0 on both photos, and a 4 stop boost (3.95, actually) using the EXPOSURE slider for the 200 ISO photo.

 

On screen, yes they look startlingly similar, the Blacks and shadows of the histogram however, do not.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

Edited October 21, 2012 by Jaybob

[Jaybob]

Again, On screen, they look startlingly similar, however the adjusted image's Blacks and Shadows of the histogram however, do not. There appears to be a color tweak to me as well in those ranges. The leather strap and the handle of the jug look warmer to me, perhaps that's from the green and blue information starting to compromise at the bottom of that histogram. Granted this is a four stop push on a NEX...

If all the pixels are sacred, I contend that ETTR files are dealing with those shadow and the black pixels much better. I'm going to stick with that method for now.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

Edited October 21, 2012 by Jaybob

[Peter H]

Very interesting.

 

Do you think these differences that show up in the histograms would be more evident in print than on the screen?

[dwbell]

Thanks Jaybob, interesting indeed.

 

In your example one can clearly see the combing of the histogram in the lower tones. This is what we would expect to see and I have seen doing the same with my 5D2 a few years back.

[Jaybob]

Both the NEX files LOOK very good on the screen, even at 100%, but the mathmatical representation (the combed over histogram) tells what could be a different story. Splittling hairs? Maybe.

 

Stay with me, cause here's where it gets weird...

 

My second test

 

I've only spent ONE WEEK with an M9 (in Vegas, of course, where else?), and looking back through my files I only have ONE example of an exposure that was shot under these conditions (really underexposing at the 160 ISO setting) to compare with.

 

I was pretty close to being AS COMPLETELY underexposed as the camera would let me at 160, in this situation exposing 1/3000 of a second (4000 is the shortest speed) at F16. This was the first frame I took after switching lenses. (a 1966 Elmarit 90mm is attached and not the 28mm lens it says is on, the concept of switch lenses in the menu was new to me, AND I was in Las Vegas...) The 90 is a pretty low contrast lens to start with, and the histograms are usually pretty forgiving. There are no other corrections made other than a zero sharpening and I gave it 3 and a third stops of Exposure Boost in LR to get it back to where it should have been to start with. Further adjustments could surely improve its exposure and color, as is the case with the NEX files as well.

 

The histogram surely held together better at the bottom, but it wasn't as much of a push, it was with a different lens, different camera, different white balance, blah blah blah.

 

So I suppose we all have to test, interpret and repeat...best of luck gentlemen.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

Edited October 21, 2012 by Jaybob

[Jaybob]

Completely unrelated...there's a girl up there on that deck, next to the folded blue umbrella.

 

I love that lens...

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[k-hawinkler]

Many thanks. Very interesting.

Last year I was trying to explore how to get the "best" exposure.

I did some tests with my M9 + TE 1:4/135, varying exposure time while keeping ISO=160 and f=5.6.

Exposure time was changed from 1/3000 s to 1/350s. That's slightly over 3 EV. However, no highlights were blown.

Here goes:

 

1/3000 s

[ATTACH]342274[/ATTACH]

 

1/2000 s

[ATTACH]342275[/ATTACH]

 

1/1500 s

[ATTACH]342276[/ATTACH]

 

1/1000 s

[ATTACH]342277[/ATTACH]

 

1/750 s

[ATTACH]342278[/ATTACH]

 

1/500 s

[ATTACH]342279[/ATTACH]

 

1/350 s

[ATTACH]342280[/ATTACH]

 

In the next post I will show what these images look like when exposure is corrected appropriately.

 

 

 

Here are the corresponding histograms:

 

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

 

Larger ones and images with histograms are posted here:

 

2011-03-30 Exposure Variation - winklers' Photos

 

and here:

 

2011-03-30 Exposure Variation - winklers' Photos

Edited October 21, 2012 by k-hawinkler

[k-hawinkler]

Now with exposure corrections in PS CS6.

 

 

1/3000 s, +1.55 EV

[ATTACH]342281[/ATTACH]

 

 

1/2000 s, +1.20 EV

[ATTACH]342282[/ATTACH]

 

 

1/1500 s, +0.55 EV

[ATTACH]342283[/ATTACH]

 

 

1/1000 s, +0.10 EV

[ATTACH]342284[/ATTACH]

 

 

1/750 s, -0.45 EV

[ATTACH]342285[/ATTACH]

 

 

1/500 s, -0.90 EV

[ATTACH]342286[/ATTACH]

 

 

1/350 s, -1.55 EV

[ATTACH]342287[/ATTACH]

 

Exposure at around 1/1000 s is probably most convenient in this case.

But after an exposure correction is applied, all the images look indeed very similar.

 

Full size images can be found here: 2011-03-30 Exposure Variation - winklers' Photos

 

 

 

Here are the corresponding histograms:

 

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

 

Larger ones and images with histograms are posted here:

 

2011-03-30 Exposure Variation - winklers' Photos

 

and here:

 

2011-03-30 Exposure Variation - winklers' Photos

 

 

Also, the full size JPGs still have Exif data.

[Lindolfi]

Thanks K-H and others.

 

Also ran some tests and found that a properly exposed image at 1250 ISO had slightly more noise (stdev of 5.43 versus 4.67 of the same dark noisy part) than one exposed in the same way, but now at a setting of 160 ISO (plus 3 stops correction in LightRoom). So that would mean an advantage of the method of underexposure when there are limits in aperture and exposure time in a certain situation. Exposure time of the example: 1/15 second, aperture f/2.0, camera M9. LightRoom 4.1, process 2012

 

http://www.photoplaza.nl/lindolfi/compareisopush.jpg

 

http://www.photoplaza.nl/lindolfi/compareisopush.tif

Edited October 21, 2012 by Lindolfi

[douglasf13]

Thanks for those tests, Jaybob. Something just hit me about the NEX cameras. Unlike some of their DSLRs, like the A700 and A900, Sony doesn't have an uncompressed raw option in any of the NEX cameras, including even the NEX-7. All of the raws are (semi)lossy compression in NEX cameras. My guess is that the lower bit, compressed raws of the NEX-5 aren't playing as well with this technique as they could be if they were uncompressed. Despite the 5N having the same sensor as the K-5, the K-5 has 14bit, uncompressed raws, and I'm betting that makes a difference.

 

I hadn't attempted to make a large exposure boost with my NEX cameras, but it looks to me that this technique isn't great for NEX, at least with a +4EV boost. I'd imagine that smaller adjustments would be ok, though. Either way, since the NEX cameras rely on live view, this technique probably isn't all that applicable in practical use, nor is it with other live view based cameras, since you wouldn't be able to see the screen to focus and compose.

 

It would be interesting to see the NEX test with a smaller boost, like +1EV, just to see if it would cover the need for ETTR.

 

It would also be interesting for someone to try this with an M8, since it also has compressed raws, although they still may be less lossy than the NEX raws. The M8 may not work as well as the M9 with this technique. Heck, while were at it, it would be interesting to see this technique with the M9 set to compressed raw, as well.

[Lindolfi]

Heck, while were at it, it would be interesting to see this technique with the M9 set to compressed raw, as well.

 

You got it!

 

Here a three way comparison with the M9. The standard deviation of the same dark area:

 

ISO 1250 uncompressed stdev 7.51

ISO 160 + 3stops uncompressed stdev 6.02

ISO 160 + 3stops compressed stdev 6.74

 

Exposure time of all three: 0.5 sec. at f/4 with the 75/1.4

 

Finally we find a difference in compressed and uncompressed! But only if you push 160 ISO by 3 stops. Still the 160 ISO is less noisy than the equally exposed 1250 ISO and uncompressed by quite a margin.

 

Tiff file at 100% of original, standard setting. LightRoom 4.1, process 2012

 

Edited October 22, 2012 by Lindolfi

[k-hawinkler]

You got it!

 

Here a three way comparison with the M9. The standard deviation of the same dark area:

 

ISO 1250 uncompressed stdev 7.51

ISO 160 + 3stops uncompressed stdev 6.02

ISO 160 + 3stops compressed stdev 6.74

 

Exposure time of all three: 0.5 sec. at f/4 with the 75/1.4

 

Finally we find a difference in compressed and uncompressed! But only if you push 160 ISO by 3 stops. Still the 160 ISO is less noisy than the equally exposed 1250 ISO and uncompressed by quite a margin.

 

Tiff file at 100% of original, standard setting. LightRoom 4.1, process 2012

 

 

 

:D

 

 

Thanks.

Very interesting. Definitely a first!

I think the .DNG files I used earlier in this thread were all compressed, each 17.5 MB in size.

[Advertisement]

Hallo,

Look here: Leica M

[Jaybob]

...All of the raws are (semi)lossy compression in NEX cameras. My guess is that the lower bit, compressed raws of the NEX-5 aren't playing as well with this technique as they could be if they were uncompressed. Despite the 5N having the same sensor as the K-5, the K-5 has 14bit, uncompressed raws, and I'm betting that makes a difference.

 

...Either way, since the NEX cameras rely on live view, this technique probably isn't all that applicable in practical use, nor is it with other live view based cameras, since you wouldn't be able to see the screen to focus and compose.

 

It would be interesting to see the NEX test with a smaller boost, like +1EV, just to see if it would cover the need for ETTR.

 

 

My secondary tests using this technique on the NEX5 I use (the 1st CMOS Exmor sensor, 14.2 MP) show exactly the same types of low end histogram combing when trying to Lightroom 4 boost my NEX files just one or EV from 200, combing I'm not getting by just changing the ISO in camera. Compressed RAWs cannot be helping, but the lower end seems to be where all of the damage is being done, at least in theory. On screen they look totally fine, with just a minimal hue shift.

 

The M9 file that I uploaded earlier (with the umbrella) was from a compressed DNG, and isn't exhibiting that type of combing with a 3 and a 1/3 stop boost, however, upon further review I'm pretty sure it wouldn't ever be mistaken for a 160 ISO file, as it needs some NR, especially in those lighter colored shadows.

 

CCD, CMOS, different lens signatures, lighting conditions, original WHITE BALANCE (underated, in my opinion) blah, blah, blah. There's mondo variables...

 

A live histogram combined with focus peaking and the ability to hold the camera up to my eye, when combined with this Hoodman cinema kit, make my NEX extremely practical.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

Edited October 22, 2012 by Jaybob