Exposure in low light with D30

[gthorn]

I hope this is not a dumb question:

I have a D30 and have discovered that when I take pictures in low light (indoors) without a flash and on a tripod, the pictures are underexposed. I can get the exposure right by cranking up the dial on the back of the camara to uncrease the exposure (i.e. reduce shutter speed). The problem though, is that I first have to take a bad picture or three to know how much I must compensate for. (I use the histogram display to verify this)

In a sense, it is logical that a picture taken as described would be underexposed, but on the other hand, isn't this what having the camera measure it's own exposure levels and setting the shutter speed by itself is for ?

It's as if the camera doesn't know how to correct the exposure for different average light levels.

The lens is a Sigma 28-300. The first Sigma 28-300 lens I had, developed a problem where it would not change it's aperture and this resulted in exposures which were all over the place. I had it replaced, and maybe this problem I am now experiencing has to do with the replacement lens, but I can't see that.


Graham Horn

[Pekka]

A demo photo with EXIF data could help in replying, but anyway:

Maybe the lens you use is not sensitive enough? Do you see anything blinking in viewfinder?

For night/dark shots use manual mode and set exposure like this

- set the camera to Manual Mode.
- set desired aperture first (usually f/8-11)
- then set shutter speed to what you desire.
- if meter is not on center, use bigger aperture (if you can, based on what you shoot): lower aperture number until you have meter on center. if that does not help raise ISO and see if you can get meter to center. If raising ISO does not do it (or you don't want noise in) lower shutter speed so that the meter is dead on center.
-shoot

Thats the exposure "workflow" I use. That way you should get a decent exposure to start with.

D30 meters to 18% gray. This means if you shoot pure white the camera will make it a bit darker and if you shoot pure black it tries to make it lighter. The basis of this is that this way the exposure should use all the dynamic range of the media.

You can use the exposure compensation setting to tell the camera that "I need black" here and "I need white there". Or use plain manual settings and see what the meter tells you. With some experience you will know the settings as soon as you see the subject.

Try also different metering modes to see what suits you best. The camera will try to avoid any overexposure, and e.g. evaluative metering meters the whole scene - this has the drawback if there is even one bright or reflective spot in meterin area there the camera will make scene darker.

[gthorn]

Thanks for the prompt response.

I tried your advice and I see that when I set the meter on dead center, I get the same results as in program mode. i.e. the picture is still very dark.

The test scene I am using only uses about 50% of the range on the histogram, and I am pretty sure there are no bright spots anywhere. The histogram "mountain" is not centered, but almost all the way to the left. (hope you know what I mean).

I can't imagine what sort of a problem there could be with the lens, unless it is closing it's aperture more than it should when the picture is actually taken. As I said in my first posting, the problem I had with the first lens was that it didn't close it's aperture at all when the picture was taken so that when I took a picture with an intentionally small aperture, I would end up with an overexposed picture instead.

Graham

[Pekka]

Would it be possible that you post one demo shot of your test scene?

[gthorn]

Thanks, I will. Just need to figure out how to do that .

[Pekka]

gthorn wrote:
Thanks, I will. Just need to figure out how to do that .

If you don't have a server where to post, mail the photo(s) to me and I'll attach them here.

[gthorn]

I have placed the test picture at http://www.muzicman.com/code/IMG_3009.jpg

It was taken in program mode with default setings. As I said earlier, if I use manual mode and center the meter, the results are exactly the same. If you adjust the brightness of the picture up, you can get the picture I would have expected to see.

Thanks for your help and interest.

[Roger_Cavanagh]

Graham,

If the histogram is not obviously clipped at the left, you will still get good results with postprocessing:

http://www.rogercavanagh.com/images/temp/gthorn.jpg

This is just an autocolor adjustment in Photoshop 7.

Regards,

[gthorn]

Yes, the histogram is not clipped at the left as you say and I can get results with post processing, but my question is whether the result is correct ? I don't really want to have to post process my pictures if I can help it for a number of reasons. For example :

1. The picture used less than 128 levels of color (per RGB) which means that I'm not getting the full color level resolution range in midtones. (Sorry, I don't know how to explain this point properly).

2. I don't like to save my pictures as jpg once I have edited them so that I don't incurr the jpg loss a second time. This means my files are bigger and that I lose the EXIF information when I save them as TIFF etc.

3. My Nikon 995 does not exhibit this problem at all. Picture for picture, it always produces a better result than the D30 (much sharper pictures). The only problem with the 995 is that it's red is over saturated and it doesn't distinguish well between red and orange.

[Pekka]

gthorn wrote:
Yes, the histogram is not clipped at the left as you say and I can get results with post processing, but my question is whether the result is correct ?


Looks your D30 meters stop or two under - unless the wall to the left is really reflective and it fools the meter. How is it compared to what you see in real life?

I don't really want to have to post process my pictures if I can help it for a number of reasons. For example :

1. The picture used less than 128 levels of color (per RGB) which means that I'm not getting the full color level resolution range in midtones. (Sorry, I don't know how to explain this point properly).

Use 16-bit originals and you'll have 65525 steps per channel.

Raw truth: you will have to post process most shots with digital SLR's. See answer to question 3 below.

2. I don't like to save my pictures as jpg once I have edited them so that I don't incurr the jpg loss a second time. This means my files are bigger and that I lose the EXIF information when I save them as TIFF etc.

RAW originals are not very big. They can produce any output from 8-bit JPEG to 16-bit linear TIFF.

3. My Nikon 995 does not exhibit this problem at all. Picture for picture, it always produces a better result than the D30 (much sharper pictures). The only problem with the 995 is that it's red is over saturated and it doesn't distinguish well between red and orange.

Consumer cameras are designed to give users the "end product" for easy and quick albums. D30 is designed to give the "unedited raw material" so that the photographer has complete control over individual image processing.

[gthorn]

The wall behind the flowers is not any brighter than the flowers in my estimation. Are you agreeing that there could be something wrong with the camera's calibration ?

I thought RAW pictures had 12 bit color depth, not 16 bit? This would mean a resolution per color of 4096 steps which is of course still a lot better than 256. I don't wan't to use raw mode most of the time anyway because for me it's impractical. I do a lot of wild life pictures and I don't need anything that slows up the camera.

I bow to your superior knowledge and experience, but I don't think I agree that digital photography should have to be post-processed anymore than film. The CMOS sensor is the substitute for film and should even be more consistent than the consistency between rolls of film. Everything else about the camera is exactly the same as film (ok, film has no white balance).

Thanks for the information.

[Pekka]

gthorn wrote:
The wall behind the flowers is not any brighter than the flowers in my estimation. Are you agreeing that there could be something wrong with the camera's calibration ?

If the undexexposure happens on almost every shot, indoors and outdoors then you have a camera problem.

One thing: try the shot without AF assist light - perhaps it does not go off during shot and that messes exposure.

Also, there is a Q-16 test which you can do to determine the level of underexposure.

I quote Chuck Westfall here (he is Canon Manager/Technical Information Dept.):

------------------

1. Obtain a Kodak Q-14 Color Separation Guide and Gray Scale (CAT 152 7662) from your photo dealer.

2. Set up the Gray Scale indoors in typical overhead room lighting and mount your D30 on a tripod. Position the camera roughly 5 feet from the target and parallel to it. Zoom the lens to fill the frame from left to right. Aim the central focusing point at zone 8 on the Gray Scale. Make sure the subject is focused, then shut off the AF.

3. Set the D30 to Manual mode with Partial metering. (DO NOT use an automatic exposure mode.) Other camera settings will be manually selected center focusing point, auto white balance, single frame advance or self-timer, IS0 100, and any JPEG setting you like.

4. Set an aperture of f/8 and, while looking through the viewfinder, adjust the shutter speed until the exposure level indicator lines up in the middle of the exposure level scale.

5. Take a series of non-flash exposures at, over and under the D30's recommended exposure level, using shutter speed as your adjustment. (Note: the camera defaults to 1/2 stop settings, but you can change this to 1/3 stops if you like by means of C.Fn. 4.

6. Examine the resulting images in Photoshop, using the Info palette set for Grayscale (K) values. The K reading for Zone 8 should be at about 65%. Check your images and see which image matches this figure most closely.

-----------------------

If the test gives you consistent results that your camera is underexposing, have Canon fix it, under warranty of course.



I thought RAW pictures had 12 bit color depth, not 16 bit? This would mean a resolution per color of 4096 steps which is of course still a lot better than 256.

The file is 16-bits per channel and it has used only 12 bits. As soon as you touch it with levels, saturation or an other tool the rest of the 16 bits are "used".


I don't wan't to use raw mode most of the time anyway because for me it's impractical. I do a lot of wild life pictures and I don't need anything that slows up the camera.

That I understand. D60 might help there, it has faster and bigger buffer.

I bow to your superior knowledge and experience, but I don't think I agree that digital photography should have to be post-processed anymore than film. The CMOS sensor is the substitute for film and should even be more consistent than the consistency between rolls of film. Everything else about the camera is exactly the same as film (ok, film has no white balance).


Well, I work only with RAW, which is unsharpened so I always post process. But CMOS, it does not do it by itself is just a chip in a big computer system (D30).

CMOS builds colors from information that is not really there, does noise reduction per pixel, is a chip and like any other chip relies very much on what camera operating system tells it to do. It may act to heat or cold and may suffer from dust and power feed problems. It activates with shutter so it may suffer from shutter timing problems (the shutter is also controlled by computer).

Film is "just there" waiting from shutter to open. Much simpler stuff.

[Roger_Cavanagh]

Graham,

I have a number of comments on the recent posts, but they are in no particular order.

Flowers vs Background Brightness: checking the luminosity in Photoshop clearly shows a difference (on average) of around 50 points between the wall and the flowers. This is sufficient to change an exposure reading. As Pekka has pointed out already, exposure meters treat everything as 18% gray.

Saving EXIF data: as you say saving JPG's more than once is a bad idea. However, if you use BreezeBrowser you can apply the EXIF data from the original files to TIFF's.

Bits per channel: you are right the D30 is 12 bits per channel. You must shoot raw to get this and convert to 16-bit TIFF or PNG for use in PS. Here is a Bruce Fraser article that discusses the benefit of using high bit files http://www.creativepro.com/story/feature/7627.html.

JPG vs raw: the bottom line is that you are not getting the best out your D30 by using JPG. You say that you cannot afford to lose any speed, but are you using all the tricks to get the most fps. This link http://www.dlcphotography.net/ChuckTips.htm#anchor215580 describes techniques for minimising shutter lag. If you know you are in a situation that calls for more than six frames on the bounce before slowing to 1 frame every 2-3 secs (this is what I get with raw), then use JPG. Otherwise use raw. You can set CF12 so that the Set button can be used to change quality quickly.

Film vs digital post-processing: it is a misconception that film is requires any less processing. It's just that we don't see it. We bung the exposed film in a little envelope and mail it. The fancy lab equipment does levels and curves adjustments analogous to what we do with Photoshop. Just as different film has different characteristics that require different processing and give different results, so do the chips in digital cameras vary. It is quite certain that the CMOS chip in the D30 does not give the best results straight out of the camera (presumably by Canon's deliberate design decision). You _must_ shoot raw and you _must_ be prepared to apply substantial post-processing. This does not have to be done by hand. Pekka's LinearSharpen can be used in batch mode to give excellent results on most images without further adjustment.

Dynamic range: we cannot expect to have a nicely balanced, full histogram on every shot. The lighting conditions do not always provide this. So we "cheat" with levels and curves, colour and contrast adjustments to give an acceptable result. If you shoot raw, you have the maximum possible data to make these adjustments.

Sharpness D30 vs 995: this is another facet of the "post-processing required" issue. I prefer to shoot raw and convert to linear TIFF. Images will completely unsharpened and so I can use the power of Photoshop and my PC to give a much better result that the camera's much less powerful internal software would do.

The bottom line is that the D30 is not a point-and-shoot camera. If you try and use it as such then it should be no surprise that a purpose designed P&S performs better. Of course, the technology is still maturing and evolving, so may be when we get the the D120 or whatever we'll be able to use it both ways, but right now I never touch the auto settings on the D30 because I know I will get better results "doing it by hand".

You don't say whether you have had the D30 very long. You certainly wouldn't be the first D30 user to be disappointed with the results (I was one :) ). I'm not trying to say it is the perfect camera, it has definite flaws and it does take some getting used to. Stick with it and I am sure you will come to love the quality of the images it can produce.

Regards,

[lensecap]

I have a 30 and do not have this problem. I can shoot in any light and it works fine. always a little tweeking to be done but nothing like what your photo showed

[gthorn]

Thanks for all the information. I am taking the camera in to have it checked, although I am not hopeful that I will get a proper answer. I had two test pictures printed at the dealer, one underexposed and one where I cranked up the exposure. Although they appeared quite different when they came up on the printer's screen , the looked about the same when printed because the printer software appears to adjust the exposure. This is what the camera is supposed to do !

I have also been taking pictures in RAW mode to see how it goes, but I can already see this is inconvenient for general picture taking. For example: converting the CRW files is laborious, and working with 16bit TIFFS is slow because of their size. When I then finally save the file as JPG then all the EXIF info is lost.

I would rather have the camera just set it's exposure right in the first place.

Thanks for the useful input folks !

[Pekka]

Let us know if the camera got any better after checking.

[gthorn]

I spent some time analysing this problem further and I have concluded that it comes down to the level the camera chooses for a plain white subject. (The example I had with the flowers covers a small (ish) range of luminoscity intensities and therefore the results correspond with the results I found this morning.

I took pictures of white walls in poor light all the way up to bright sunlight. The shutter times varied from 1 sec down to 1000th of a sec and each picture has the wall looking exactly the same shade of grey. I have posted histograms of the pictures at http://muzicman.com/code/1.png (1 sec exposure), http://muzicman.com/code/45.png (1/45 sec exposure), http://muzicman.com/code/1000.png (1/1000 sec exposure), http://muzicman.com/code/1000.jpg (clip of the actual picture - you can see how dark it is).

I would appreciate it if a kind D30 owner out there could do the same test i.e take pictures of a white wall (or white paper I guess) at various light levels and check the histograms in photoshop and let me know. This should confirm whether or not I have a problem with my camera.

Thanks for a great and useful forum.

[Roger_Cavanagh]

Graham,

This relates to the metering issue that Pekka and I mentioned before. The camera assesses required exposure on the assumption that the luminance is equivalent to an 18% gray card. Since the white wall is giving off more light than this, the camera is fooled into underexposing the scene.

If you adjust exposure by a couple of stops, you will get a better exposed picture. I have taken three shots of a white wall (newly painted by yours truly in brilliant white vinyl matt emulsion :) ). I used Tv mode and used exposure compensation of 0, 1 and 2 stops:

http://www.rogercavanagh.com/images/ev_examples/ev.htm

The page shows a crop from each shot with the histogram from the full image. It's a cloudy day here, so the wall isn't exactly brilliant white. However, it is clear that the exposure compensation moves the image towards a "truer" representation of the white wall. In every case, the image can be adjusted in Photoshop by using the whitepoint eyedropper in the Levels command.

Your example image is not greatly different from my EV 0 example. The histograms show median luminosity of 110 vs 118. So by comparison your D30 is slightly underexposing compared to mine, but I would not have thought enough to matter. In any case, with digicams it is better to have underexposure rather than overexposure as it is generally possible to bring out shadow details, but a blown highlight is a blown highlight.

To test your camera properly you should apply the test that Pekka described or get hold a a gray card from your local, friendly camera shop. Take a few shots using meter readings taken from the gray card. Or take some shots including the card and use the card to set white balance during conversion.

Regards,

[gthorn]

Thanks Roger

After more experimentation, I find that the camera is very consistent with the histogram if I use a sheet of white paper in low light and bright sunlight, so the fact that you don't have a lot of sunshine today should have nothing to do with your results. The only important thing is that the lens must only see the paper and nothing else that could influence the metering.

So, compared with your camera, mine is out by about 3/4 of a stop.

One thing about post-processing an under-exposed shot is that noise will be amplified, so it should be preferable to have the shot exposed correctly by the camera.

I have experimented with the Nikon 995 as well, and it looks slightly brighter than your EV+1. What I also found with it is that it's white balance goes pink the lower the light, whereas the D30 stays true in in very low light. (Learning lots of things today !!)

Thanks for the information, I really appreciate it.

Regards

Graham

[Roger_Cavanagh]

gthorn wrote:
Thanks Roger

After more experimentation, I find that the camera is very consistent with the histogram if I use a sheet of white paper in low light and bright sunlight, so the fact that you don't have a lot of sunshine today should have nothing to do with your results. The only important thing is that the lens must only see the paper and nothing else that could influence the metering.

So, compared with your camera, mine is out by about 3/4 of a stop.

Are you sure it's that much? According to the histograms on my examples, 1 stop equates to around 40 points of the 0-255 scale. On our EV0 pix, we had a difference of 8.

One thing about post-processing an under-exposed shot is that noise will be amplified, so it should be preferable to have the shot exposed correctly by the camera.

True, but it's a question of how much under-exposed. As I said before, you can get a lot of shadow detail out of the D30. Although the bottom line is digicams don't have the dynamic range of colour negative film, so it's hard to get good pictures sometimes. My next door neighbour's black and white cat is really hard to photograph. Either the white fur is blown or there's no detail in the balck fur. ( :rolleyes ).

Cheers,

[gthorn]

Ok, I estimate mine is about 25 points lower than yours based on my worst case example (which I didn't post), so that would make it about 5/8 of a stop, let's say 1/2. I would be interested to see what anyone else gets with this test.

Graham

[Roger_Cavanagh]

Graham,

OK, it's Sunday afternoon, the World Cup Final's over. (Well done, Brazil! Couldn't have stood it, if the Germans had won. :D )

Anyway, I remembered I bought myself a new light meter (first one ever :) ) earlier this week. So I got it out, dug out my gray card and did this:

- Blue-tacked the gray card to wall

- Set the camera on the tripod (couple of meters away)

- Adjusted the zoom so that the gray card filled the viewfinder

- Used the camera in manual to get a reading of 1/4sec @ f/8

- Check the reflected lilght reading with the meter, got the same reading (yippee! :) )

- Took a picture of the gray card

- Removed the gray card and took a picture of the wall with same settings

- Converted to JPG in BreezeBrowser loaded the images to Photoshop and ran the histogram. Got median luminosity of 120 from the gray card and 239 from the white wall.

So my camera does seem to be working more or less OK. Using the gray card did give more consistent results than the wall, so go buy one and see what happens.

Regards,