I was out in the woods this morning and because the light was boring and I had a tripod with me I shot a series to see how pushing or oulling affects image quality.
My conclusion based on this quick test is that with a low contrast scene there is no reason to expose to the right, what do you guys think?
I guess you already know the theory of it, but just to summarize:
- If you don't want to change the exposure after shooting, just meter to get the look you want, no need for ETTR.
- If you are not happy that the top stop in a 12-bit file has 2048 possible values, whereas the bottom stop has 2, then you need to ETTR, as long as you are not blowing out anything you would like to keep.
Modern cameras are good enough that I stopped using ETTR almost completely. The D800 in particular stuns me with its shadow detail.
Well, like all photographic techniques, ETTR is not a hard rule. By using ETTR, you want to maximize the DR of your sensor. Like you said, for a low contrast scene, it would not be needed. I have been doing +0.7 EV consistently with my A99 (I think it can go up to almost 2EV). But for an extremely high DR scene, I found myself under-expose the picture to preserve the highlight portion, often going from -1EV to -2EV. Depending on the print size, the shadow detail and noise might not be that perceptible. Most of prints are 13x19, and I have been happy with some -2EV on my 100MP+ stitch. The new PRIME noise reduction from DxO certainly helps a lot as well.
Modern day religion, unsupported by reality. Highly recommended to photographers who (i) use digital cameras from the dawn of the digital era, and (ii) those who value formula and theory over colour integrity - aka 'the numbers cannot be wrong' only my lying eyes, lol.
I never liked the look of 'close to the edge' highlights and they never come back from highlight hell looking any good, this for nature work. I blame Canon for the ETTR craze. And Michael Reichmann, a serial offender on several fronts.
Better way to test might be hold ISO constant and vary SS 'til she blows. One less variable.
philip_pj wrote:
Better way to test might be hold ISO constant and vary SS 'til she blows. One less variable.
ISO was a constant 100 and the only varying factor was shutter speed.
When my A7 arrives I will repeat the test and try to borrow my friends 5dII as well.
How would you compare dynamic range? I could simply take the same picture of a high contrast scene and see what happens but I would like to know how much better the A7s sensor is for everyday use.
OK, sorry Phillip, I got the impression you shot varying the ISO. I would look for one of your quite typical scenes - I know you shoot a lot of high subject brightness range (SBR) compositions, with a good spread of tones as shown in a histo.
Push content up to and a little past highlight loss (usually not all three channels) for your ETTR candidate; with your 'control' shot(s) having the meat of the histo in the middle or near to it. With high SBR the two shots might be 2-3 stops apart. Process both to suit your tastes then compare.
You are looking at differences in shadow quality (the raison d'etre for ETTR) and 'scorched' content in the range 200-255 after reducing exposure in post. ETTR suggests better shadows and undamaged other tones. The other side of the argument is non-ETTR gives very good shadows and better other tones in the mid-higher tone range.
It might be subtle to see differences...software is so good these days, largely (I speculate) because of this very problem the dominant market player Canon experienced.
I would expect more attractive colour with better separation and still clean shadows in any modern Sony shot at normal exposure and more flat output from the ETTR shot, along with some 'compressed' highlights maybe, depending on the scene - best to avoid specular highlights (leaves etc) when looking at this. The essence of it is that pixels don't respond well to too much tone shifting. And that is exactly what ETTR does in post. Use base ISO as you did before.
Always happy to be wrong however! The later Sonys are 2.5 stops better in handling SBR than the 5d2/3, you should see this easily in RAW not yet processed images. It's huge because some of the highlights are inevitably lost for good with less DR, plus at the other end you get an extra stop or so of tones - it's a more natural look with smoother shadow fall-off.
ETTR was THE big craze when I got my first digital camera in late 2006.
I tried it from time to time, but could never see a distinct improvement.
The sensors/processing have definitely improved since then, although Canon users may have more reason to still use it than those operating the newest Sony sensors.
I have Nikon D800s and also a Fuji XE1, and the highlight recoverability of the Fuji seems significantly better than the D800.
I'm not clear on your methodology for this test. And what camera was it?
I did the ETTR test (controlled lighting, one shot 2 stops under and pushed up in post, one shot properly exposed, one shot 2 stops over and pulled down in post, compare noise and detail especially but not exclusively in the blacks) with the D700 (which isn't, like, ANCIENT) and there was a very significant and readily visible difference.
Those who aren't seeing a benefit from ETTR aren't going far enough "to the right" - or have cameras with much lower noise floors than the D700.
philip_pj wrote:
[recommended by photographers who] (i) use digital cameras from the dawn of the digital era.
Disagree with this. My D1x and even D2x did not have enough top-end dynamic range to risk ETTR - hell, they often didn't have enough for proper exposure. I first had success with it when I got my D700.
Lee Saxon wrote:
I'm not clear on your methodology for this test. And what camera was it?
I did the ETTR test (controlled lighting, one shot 2 stops under and pushed up in post, one shot properly exposed, one shot 2 stops over and pulled down in post, compare noise and detail especially but not exclusively in the blacks) with the D700 (which isn't, like, ANCIENT) and there was a very significant and readily visible difference.
Those who aren't seeing a benefit from ETTR aren't going far enough "to the right" - or have cameras with much lower noise floors than the D700.
He took this with the NEX-6. All images were taken at ISO 100 with various EC settings: +2, +1, 0, -1, -2, -3, -4, -5 (from top to bottom). As stated, this is a low contrast scene, so the benefit of ETTR is not as apparent. It's interesting to see that the solid black area showed no noise whatsoever regardless.
hiepphotog wrote:
All images were taken at ISO 100 with various EC settings: +2, +1, 0, -1, -2, -3, -4, -5 (from top to bottom).
I'm missing something here ...
How are we taking an exposures with 7 stops of EC and still getting relatively the same "proper" exposures, except by the offsetting shutter speed? And how is that putting us at ETTR? It sounds like we are just have offsetting shutter speed for the EC @ push/pull ... not really ETTR?
Sorry, if I'm a bit slow on this one ... sometimes I need things drawn with a REALLY BIG CRAYON for me to get it.
How are we taking an exposures with 7 stops of EC and still getting relatively the same "proper" exposures? And how is that putting us at ETTR? It sounds like we are just have offsetting shutter speed for the EC.
Sorry, if I'm a bit slow on this one ... sometimes I need things drawn with a REALLY BIG CRAYON for me to get it.
Well, top setting (ISO 25) was set at +2 EV EC at ISO 100. Then in post, he used -2 EV to get back to the "normal" exposure (see the tiny text at the bottom of the image for his explanation). So unless I seriously mis-interpreted it, that is how he did it. Effectively, only the 1st two images demonstrate the ETTR technique. I think it would be clearer to label EV instead of ISO value.
Okay, so if I'm following correctly ... all images were shot at same ISO, same aperture, different shutter
ISO 100 represents the "normal" exposure.
ISO 50, 25 represent overexposed by 1-2 stops respectively (corrected for in post).
ISO 200, 400, 800, 1600, 3200 represent underexposed by 1-5 stops respectively (corrected in post).
OR
The first three were shot @ ISO 100 (EC was used to emulate 50, 25), and appropriate shutter speeds were used to offset the ISO (EC) variance and still record proper exposure.
Each of the remaining images were shot at the image ISO settings indicated and the shutter speeds were offsetting to record proper exposure.
I'm still not clear on which it was.
I'd probably have followed better if EXIF had just stated ISO, aperture, shutter (and EC, if used).
RustyBug wrote:
Okay, so if I'm following correctly ...
ISO 100 represents the "normal" exposure.
ISO 50, 25 represent overexposed by 1-2 stops respectively (corrected for in post).
ISO 200, 400, 800, 1600, 3200 represent underexposed by 1-5 stops respectively (corrected in post).
I find that the current generation of sensors simply do not need this massaging (but the early D series Nikons definitely did); I do overexpose by up to a stop if shooting light-coloured things, but that's more about getting correct exposure than protecting the shadows, as (as we all know) perfect exposure in a camera is metering ~18% grey.
The GX7 bodies I am using at the moment have excellent shadow details as is, and I only occasional want to lift these in post with this sensor. Only three years ago, it was a very different story, and I was finding I was adding quite a bit of fill to get the shadows the way I wanted (which for me is all about how I remember the scene).
I"m having a hard time believing that an image that was actually 5 stops underexposed would be lifted in post by 5 stops without incurring any noticeable noise or other effects (but I can be made a believer). I gotta wonder if the metering didn't already equalize the EC/ISO via shutter and thus the images were taken at the same exposure, but at different gain rather than truly 2 stops overexposed, 5 stops underexposed.
I appreciate the test / effort, but ...
I'd probably repeat the test with all manual settings @ ISO 100, same aperture of choice and manually change the shutter by 1 stop throughout the +2/-5 range to make sure that there isn't any metering going on that is keeping the exposures equitable ... in order to get a true two stops over and 5 stops under. I mean, isn't -1 stop EC (or ISO) and +1 stop shutter going to still give the same exposure ... is this really an overexposure/underexposure/ETTR test scenario?
I'm wondering what the raws look like as well?
Okay ... yup, I'm still as as I first was at what we are actually looking at. Please forgive me if I'm missing something obvious here ... guess I need an even bigger crayon from Phillip.
Phillip Reeve wrote:
ISO was a constant 100 and the only varying factor was shutter speed.
With EC only applied to emulate ISO 50 and ISO 25?
RustyBug wrote:
I"m having a hard time believing that an image that was actually 5 stops underexposed would be lifted in post by 5 stops without incurring any noticeable noise or other effects (but I can be made a believer). I gotta wonder if the metering didn't already equalize the EC/ISO via shutter and thus the images were taken at the same exposure, but at different gain rather than truly 2 stops overexposed, 5 stops underexposed.
I appreciate the test / effort, but ...
I'd probably repeat the test with all manual settings @ ISO 100, same aperture of choice and manually change the shutter by 1 stop throughout the +2/-5 range to make sure that there isn't any metering going on that is keeping the exposures equitable ... in order to get a true two stops over and 5 stops under. I mean, isn't -1 stop EC (or ISO) and +1 stop shutter going to still give the same exposure ... is this really an overexposure/underexposure/ETTR test scenario?
I'm wondering what the raws look like as well?
Okay ... yup, I'm still as as I first was at what we are actually looking at. Please forgive me if I'm missing something obvious here ... guess I need an even bigger crayon from Phillip. ...Show more →
Now I don't understand this. Assuming that Phillip did use the exposure lock (AEL) and the lighting remained constant throughout the test, a -1 stop EC would mean that the shutter speed is now twice faster than the "right" exposure's shutter speed. That's how an EC would work on a camera. I mean you can check the shutter speed to make sure that the camera actually under-exposed by 1 stop as you want it to. There is no equalizing EC/ISO in such scenario.
This is how I shoot all of my HDR panoramas: get an average exposure of the scene, AEL, then do a bracket of 0, -2, +2. The under-exposed has recoverable noisy shadow while the over-exposed has unrecoverable blown-out highlight. I haven't seen such equalization before.
ETTR was necessary in the early digital days to minimize shadow noise that was really a big problem. However, most modern sensors will actually benefit from correct exposure or even slight underexposure. Many sensors are isoless these days. My M9 is certainly in this category and I find my slightly underexposed files that are pushed in ACR produce gorgeous color.
