Shooting underexposed raws indoors?

[DaDeuce]

How many of you guys shoot your raw's underexposed when indoors to be able to boost the shutter speed? I know that photoshop can adjust the raw file for exposure, I was just curious how this compared to shooting at the correct exposure?

P.S. If this is the wrong forum once again, feel free to move it.

[cfpackerfan]

I intentionally underexpose on rare occasions. It has come in handy.... BUT....

Even in RAW, when you underexpose, you are losing detail in the shadows. So when you adjust in processing, some information is lost, and you are unable to recover it. (unable to recover it well. )

If you have to radically adjust, the quality of the pic suffers.

[liza]

I don't underexpose as it introduces noise. Even if you shoot in RAW format, it's best to try to nail the exposure, as it will minimize digital noise in the final product.

[Dellboy]

The only times I'd recomend underexposing is:

1, If you're at ISO 1600 and need more speed to prevent camera shake. Though you may choose ISO 3200 if your camera allows
2, If you deliberately want to underexpose the photo.
3, If the lighting is extreme and you want to prevent the highlights from burn out.

[carsanbballs]

Underexpose with film and overexpose with digital is what I think I read somewhere? I'm with liza and try to "nail the exposure" when shooting
RAW and btw I only shoot RAW.

[GyRob]

when you try to bring the detail back in the shadow's the noise starts to show so not recomended.
Rob.

[PacAce]

The only times I'd recomend underexposing is:

1, If you're at ISO 1600 and need more speed to prevent camera shake. Though you may choose ISO 3200 if your camera allows
2, If you deliberately want to underexpose the photo.
3, If the lighting is extreme and you want to prevent the highlights from burn out.

ditto all of the above except that I would reword #2 as "if you deliberately want an underexposed photo, i.e. you are not going to compensate for it later in PP".

[Curtis N]

I know that photoshop can adjust the raw file for exposure, I haven't done direct comparisons myself, but others have compared boosting exposure via RAW converter vs. boosting exposure with a higher ISO setting. Concensus seems to be that the higher ISO setting creates less digital noise than boosting the RAW image and results in better image qualtiy.

[Croasdail]

.... and of course I am the one who runs contrarian to that. I will underexpose by up to a stop before dropping to ISO 3200. I don't like the softness of 3200 - period, forget the noise. I would much rather have sharp with some noise then soft and then de-noised. Just my preference.

And you can adjust JPEGs just about as much as you can RAW.... the only exception is you have a little more lattitude if you are on the boarder of having crushed shadows or blown highlights with RAW - but not as much as many would have you beleive. Fine JPEG has a lot of detail. The Gamut is just a tad smaller. But on output your going to be going to a smaller gamut anyway so on final output it really doesnt make a difference. 1 stop really shouldn't be a big deal with either. After a stop.... you start needing to be really carefull to avoid too much color shift. It's workable, but makes workflow much more complex.

[PacAce]

.... and of course I am the one who runs contrarian to that. I will underexpose by up to a stop before dropping to ISO 3200. I don't like the softness of 3200 - period, forget the noise. I would much rather have sharp with some noise then soft and then de-noised. Just my preference.

And you can adjust JPEGs just about as much as you can RAW.... the only exception is you have a little more lattitude if you are on the boarder of having crushed shadows or blown highlights with RAW - but not as much as many would have you beleive. Fine JPEG has a lot of detail. The Gamut is just a tad smaller. But on output your going to be going to a smaller gamut anyway so on final output it really doesnt make a difference. 1 stop really shouldn't be a big deal with either. After a stop.... you start needing to be really carefull to avoid too much color shift. It's workable, but makes workflow much more complex.

Can you explain how softness comes into play here? I didn't know that setting ISO had anything to do with softness.

[Croasdail]

Upping the ISO makes the sensors more sensitive causing them get light bleed from one to another. The colors are less accurate and your gamut decreases. These all lead to a softer looking images. It's simular to the what you got with film as you pushed it, the grain expanded and gave less definition. At least that is what I have seen has been explained to me. Perhaps brighter minds will prevail here.

[fivefish]

From my experience, a high 1600 ISO with flash that's correctly exposed gives pictures that are great.

An ISO 100 , 200 or 400 that's underexposed will show MORE noise than the 1600 ISO that's correctly exposed.

Again, that's just my experience.

[DocFrankenstein]

From my experience, a high 1600 ISO with flash that's correctly exposed gives pictures that are great.

An ISO 100 , 200 or 400 that's underexposed will show MORE noise than the 1600 ISO that's correctly exposed.

Again, that's just my experience.
That's the right answer from the physics point of view.

[PacAce]

From my experience, a high 1600 ISO with flash that's correctly exposed gives pictures that are great.

An ISO 100 , 200 or 400 that's underexposed will show MORE noise than the 1600 ISO that's correctly exposed.

Again, that's just my experience.

That's the right answer from the physics point of view.

From tests I did a year or so ago, I didn't really see any difference between a properly exposed image shot at ISO 1600 and one that was underexposed at ISO 800. And that makes sense when you consider that in both cases, you are using the same image senosr and both are underexposing the image in the same way.

However, what I wasn't sure about was how the image would look at ISO 3200. After all, you have to go out of your way to use it so I thought maybe there was a reason for that. Maybe the image doesn't come out as well at that ISO compared to the other ISO settings? I don't know. So, I did a test to see if there really are a differences in images shot at ISO 3200, ISO 1600 and underexposed by 1 stop and ISO 800 and underexposed by 2 stops.

See for yourself. The test photos were shot in RAW and converted in DPP all using the same parameters except that the exposure of the image shot at ISO 1600 and underexposed by 1 stop was correct by +1 stop to compensate, and likewise, the image shot at ISO 1600 was corrected by +2 stops.

Also, other than changes to the ISO setting, all images were shot in Manula mode @ 1/25, f/5.6 and custom WB with camera on a tripod and cable release.

Resized full images:

1) Shot at ISO 3200

http://tanseikai.com/others/Pic-1.jpg

2) Shot at ISO 1600

http://tanseikai.com/others/Pic-2.jpg

3) Shot at ISO 800

http://tanseikai.com/others/Pic-3.jpg


100% crop of focused area:

1A) Shot at ISO 3200

http://tanseikai.com/others/Pic-1_A.jpg

2A) Shot at ISO 1600

http://tanseikai.com/others/Pic-2_A.jpg

3A) Shot at ISO 800

http://tanseikai.com/others/Pic-3_A.jpg

[PacAce]

And here are more crops from other parts of the photos.


1B) Shot at ISO 3200

http://tanseikai.com/others/Pic-1_B.jpg

2B) Shot at ISO 1600

http://tanseikai.com/others/Pic-2_B.jpg

3B) Shot at ISO 800

http://tanseikai.com/others/Pic-3_B.jpg




1C) Shot at ISO 3200

http://tanseikai.com/others/Pic-1_C.jpg

2C) Shot at ISO 1600

http://tanseikai.com/others/Pic-2_C.jpg

3C) Shot at ISO 800

http://tanseikai.com/others/Pic-3_C.jpg


Upping the ISO makes the sensors more sensitive causing them get light bleed from one to another. The colors are less accurate and your gamut decreases. These all lead to a softer looking images. It's simular to the what you got with film as you pushed it, the grain expanded and gave less definition. At least that is what I have seen has been explained to me. Perhaps brighter minds will prevail here.

I don't know about you, Croasdail, but looks to me that your concern about image softness are unfounded. :)


BTW, other than the initial raw conversion followed by resizing or cropping, the images are as they came out of DPP. No noise reduction was used as is probably obvious by looking at the images and no sharpening was applied in PP.

[DocFrankenstein]

From tests I did a year or so ago, I didn't really see any difference between a properly exposed image shot at ISO 1600 and one that was underexposed at ISO 800. And that makes sense when you consider that in both cases, you are using the same image senosr and both are underexposing the image in the same way.
Nice tests. I don't see a difference either.

I'm just saying that theoretically the singnal amplification should be noiseless and you should see a definite improvement when the image is properly exposed at higher ISO.

I know the rebel and 10D, judging from the noise curves don't use analog amplification going from 1600 to 3200, so in that case the you really won't be able to see any difference.

On the other hand if you go from ISO100 underexposed by 3 stops to ISO 800 exposed properly... the ISO 800 should definitely win because the analogue amplification of the signal greatly outweighs the added noise.

That's my take on it.

[PacAce]

Nice tests. I don't see a difference either.

I'm just saying that theoretically the singnal amplification should be noiseless and you should see a definite improvement when the image is properly exposed at higher ISO.

I know the rebel and 10D, judging from the noise curves don't use analog amplification going from 1600 to 3200, so in that case the you really won't be able to see any difference.

On the other hand if you go from ISO100 underexposed by 3 stops to ISO 800 exposed properly... the ISO 800 should definitely win because the analogue amplification of the signal greatly outweighs the added noise.

That's my take on it.
No matter how the signal is amplified, you are still going to be amplying the same noise that's inherent in the sensor so theoretically, IMHO, it really shouldn't make a difference. However, I think it does make a difference how the ISO 100 -3 stops is adjusted, i.e. whether it's done during raw conversion or during post processing. If it is shot in raw and exposure adjusted by 3 stop, then you shouldn't see a difference. If it's been shot in JPEG, however, then I think I would tend to agree with you only because of the JPEG factor being introduced into the picture.

Maybe an "ISO800 vs ISO100-3" tests, one with RAW and another with JPEG, is in order here, assuming ACR can adjust exposure up to +3 stops. I know DPP can not. I'll get to it when I get home unless somebody else wants to give it a try. :)

[Rubi Jane]

PacAce - I'd say from your tests it's pretty evident that a properly exposed image at ISO3200 fairs better than the underexposed ISO1600 & 800 images, not by a lot but enough to make the choice fairly simple...up the ISO in low lit scenarios and nail exposure unless you're looking for a darker mood.

Nice test - you've reassured me that I should continue to use higher ISOs when needed.
Oh BTW, CS2 allows adjustment for exposure by +/- 4 stops

[René Damkot]

Not a big difference IMO. The 'ISO 800 +2 stops' shot is worst, other two are fairly close. I think the 'ISO 1600 +1stop' shot is marginally better.

[MagicallyDelicious]

crikey i always used to do that!

after reading this thread never again!

[In2Photos]

And here are more crops from other parts of the photos.


I don't know about you, Croasdail, but looks to me that your concern about image softness are unfounded. :)


BTW, other than the initial raw conversion followed by resizing or cropping, the images are as they came out of DPP. No noise reduction was used as is probably obvious by looking at the images and no sharpening was applied in PP.

I see some softness in the pig and the vase. Don't know if it is from compression though. It isn't enough to matter really.

Not a big difference IMO. The 'ISO 800 +2 stops' shot is worst, other two are fairly close. I think the 'ISO 1600 +1stop' shot is marginally better.
I agree with Rene, here.

[Croasdail]

PacAce - very cool test. It's funny, there is only the very slightest difference in the 100% crops - the only area I see in a quick review is the black line around the top of the three flower cluster above her head near the edge of the plate where the ISO 1600 image has some extra black artifacts along the top of the line. Now with that said, if you need a 100% crop to see the difference between the three, well then that is nothing and could just be an anomily of how the file was written - or something. So ISO 800 + 2 stops adjust will yeild the same image quaility at ISO 3200 shot spot on. Very informative.

Question for you though... why the heavy artifacts in the images.... The digital compression is quit noticible. Was this introduced when converted to JPEP for display?

I learn something new everyday... I will have to so a little experimenting on my own. This is good stuff to know. Cheers and thanks for sharing.

[Dellboy]

I'm with Rene and Croasdail on this one too. If setting your ISO didn't really matter then surely we would just have ISO 100 and we could all just underexpose to suit. I would expect to see masive differences with the lower ISO settings.

I belive ISO 3200 too be an exception though because the camera just simulates this ISO by taking the shot at 1600 and underexposing by a stop. The exposure correction is then done in camera. Most Raw converters can do a much better job of this convertion and this is why Croasdail gets better results by keeping this conversion out of camera. I also find this to be true and it is not an uncommon pollisy amonst many togs.

This may not be true for JPegs though, I've never tried it.

[PacAce]

PacAce - very cool test. It's funny, there is only the very slightest difference in the 100% crops - the only area I see in a quick review is the black line around the top of the three flower cluster above her head near the edge of the plate where the ISO 1600 image has some extra black artifacts along the top of the line. Now with that said, if you need a 100% crop to see the difference between the three, well then that is nothing and could just be an anomily of how the file was written - or something. So ISO 800 + 2 stops adjust will yeild the same image quaility at ISO 3200 shot spot on. Very informative.

Question for you though... why the heavy artifacts in the images.... The digital compression is quit noticible. Was this introduced when converted to JPEP for display?

I learn something new everyday... I will have to so a little experimenting on my own. This is good stuff to know. Cheers and thanks for sharing.

Yes, it is. I set the JPEG quality to 30, which is what I always use whenever I post images here. I guess in this case, I should have set it at a higher quality since we are supposed to be comparing image quality here. Sorry, I wasn't thinking clearly when I posted this last night. It was passed my bed time. :confused:

I'll redo the resized full image and the 100% crop of the plate at a much higher resolution and put them up in a another post when I get home this afternoon.

[Croasdail]

hay... just redid your test and came up with very simular results. Used studio lights, then lowered ISO keeping everything else the same for ISO 3200, 1600, 800 - then reprocessed back up using Adobe RAW converter. The images were amazingly alike. I don't have them loaded to my site yet but the only difference was the Red channel was more noisy by a very small tad at ISO 3200 - but other then that, in a blind test - no one would be able to identify the differences. Send net-net on this is pick your poison, up the ISO or under expose at a lower ISO, the results will be near identical. Dang... who knew... obvioulsy you did... cheers.

[PacAce]

Yes, it is. I set the JPEG quality to 30, which is what I always use whenever I post images here. I guess in this case, I should have set it at a higher quality since we are supposed to be comparing image quality here. Sorry, I wasn't thinking clearly when I posted this last night. It was passed my bed time. :confused:

I'll redo the resized full image and the 100% crop of the plate at a much higher resolution and put them up in a another post when I get home this afternoon.

I took a look at the original 100% crops and they look just like what you are seeing posted here so what you are seeing is not JPEG artifacts but high ISO luminous and color noises. :(

[E-K]

I would underexpose the shot if:


I really want/need the shot
there is no other way to get the shot given the circumstances


For example, trying to get a shot of my daughter on a stage in a school gymnasium with my "not so fast" lenses at a distance where a flash would be ineffective :)

I wouldn't however underexpose an image at ISO 100 and then push it to get the exposure I would have had at ISO 200 or ISO 400.

If I'm at ISO 1600, which is the maximum for the XT, and needed a faster shutter speed I would push it to 3200 or even 6400. Granted the picture isn't something that I'm going to go and get enlarged to 16x20 but with some noise reduction software its passable.

e-k

[narlus]

i was forced to use the underexpose/high ISO trick this past monday, where i was shooting a band in fairly crappy light, yet wanted a shutter speed high enough to freeze the drummer. for shots @ -1 stop EC and 3200, w/ no noise reduction, they look alright to me. i did crank the contrast in PP (and may have adjusted exposure upwards too, i can't recall) so that the shadows were black; i didn't care if i lost detail.

http://narlus.zenfolio.com/img/p1025703046-4.jpg

[René Damkot]

Nice. Lot better then 'okay'. A lot cleaner then my effort:
ISO 3200, -2/3 (but obviously underexposed more BC of the white jacket). I think EC in post was about + 1 2/3 stop.
http://www.moonglade.net/~rene/061014%20Skatak/source/image/rhd_20061014skatak0228.jpg

[René Damkot]

Okay, update. Since narlus posted the drummer shot (and some more) in performing arts (http://photography-on-the.net/forum/showthread.php?t=243122), I noticed they looked cleaner then mine.
Did a quick check, and I think I need to look into the Manyk SRS I use for sharpening web ouput; it amplifies high ISO noise IMO...
I took some screenshots from a few 'yet to be processed' files in ACR:

Sharpness and NR is set the same in all. (Too high sharpening for these files, and luminance NR could be higher), but these are my defaults, since I only use ACR for recovering highlights...

I found a few files with about the same BG color, WB is set the same everywhere, only difference between the files is ISO and EC in ACR.

ISO 1600, no EC in ACR (http://www.moonglade.net/~rene/POTN/ISO1600_0EV.jpg)

ISO 3200; no EC in ACR (http://www.moonglade.net/~rene/POTN/ISO3200_0EV.jpg)

ISO 3200; + 1 EC in ACR (http://www.moonglade.net/~rene/POTN/ISO3200-1EV.jpg)

Last image was opened in PS using these settings, Manyk SRS applied, converted to sRGB, and saved as jpg, quality 10: Tadaa:

http://www.moonglade.net/~rene/POTN/ManykSRS.jpg

Not all images are crisp, but lighting was low (as you can see)

Thoughts?

I think I need to look into the shapening for web (again) ;)