Ad
Ad
Ad
Pages: [1]   Bottom of Page
Print
Author Topic: A question about dynamic range and exposure from a newbie  (Read 3657 times)
guypyetan
Newbie
*
Offline Offline

Posts: 8


« on: May 14, 2013, 02:53:59 PM »
ReplyReply

Apologies in advance because I'm asking what I'm is probably a monumentally fat headed question. I note that DX0 claims that a camera such as the Nikon D800 has ~ 14 stops of dynamic range. I note my own camera (a Nikon D700) has around 12 stops of DR according to DX0. The questions I have are:

1) Given that most scenes do have exposure ranges of 14 stops why then do we have such a problem with exposure and capturing the light in a digital camera? Except for capturing specular highlights most scenes should be coped with nicely by practically any digital camera.
2) If pure white occurs (digitally at around + 2 and a bit stops and black at around -2 and a bit stops off middle grey then what exactly is the sensor capturing with the other 9 odd stops (for the D800)?
3) My SLR has an exposure meter which covers around 4 2/3rds stops. e.g. the pure black to pure white range in 2 above which makes sense in the context of 2 above but none whatsoever in the context of 1.

What's happening to the additional dynamic range in 1 above? Is it that:
1) Most of the light being captured is non visible to the human eye and thus isn't that important? or
2) Is the camera mapping that huge DR to a 4 2/3 stop range? or
3) Is much of the DR being lost to other elements in the cameras imaging chain (noise etc.).
4) None of the above. there's entirely another reason and the reason is .....

What is going on? Nobody seems to explain this at all. I've asked a number of learned people and none of them have given me anything approaching a good answer.
I'm beginning to lose confidence in what I thought I'd understood about exposure. I'm beginning to suspect that I've totally failed to grasp something fundamental and everybody else is making the relevant inference from reading the established literature and doesn't need to ask these questions (in other words I'm not half as clever as I'd like to think I am which is most likely true even if not something I'd like to be confronted with :->). Perhaps I need to go back to playing with brightly coloured shapes and attempt to stick them in appropriately shaped holes rather than play with this photography lark.

As a mere beginner attempting to tread (somewhat clumsily) in the footsteps of masters I'm seriously hoping that someone who knows can spare the time to answer these questions.

Many thanks in advance for any help anyone can give

Logged
Tony Jay
Sr. Member
****
Offline Offline

Posts: 2157


« Reply #1 on: May 14, 2013, 04:01:29 PM »
ReplyReply

2) If pure white occurs (digitally at around + 2 and a bit stops and black at around -2 and a bit stops off middle grey then what exactly is the sensor capturing with the other 9 odd stops (for the D800)?
#2 is simply not true.
It holds if the dynamic range of a sensor is 4-5 stops but not if the DR is 14 stops.

Tony Jay
Logged
Paul2660
Sr. Member
****
Offline Offline

Posts: 1973


WWW
« Reply #2 on: May 14, 2013, 05:49:35 PM »
ReplyReply

I look at the DxO marks as a gauge to range of exposure for a single shot and that is the total Dynamic range of a particular camera.  How far can I push a single exposure without losing shadow details or totally blown highlights.  The MP reach is not that important to me, as I can always stitch for a large resolution image.  But how far can I work up a single exposure getting the best balance between shadows and highlights is to me the best gauge of Dynamic range.

Examples:  all at base iso

1.  Canon 5d MKII, at 100 iso, I found that I had to shoot this camera in brackets.  Extreme lighting situations, examples sunset, sunrise, water in slower speeds required at least 3 brackets preferably 5.  If you simply exposed for the sky, and then attempted in post to pull up the shadows usually more than 1/4 of a stop, you would start to see extreme noise in the shadows.  Noise that was destructive and made that part of the image non workable.  Canon noise tends to show up in bands and large blue/red splotches.  Knowing this I was still able to work within the confines of the Dynamic range of the 5d MKII.  I would say that the 5D MKII was a bit more forgiving on highlights, but you could not go overboard. 

2.  The Phase One P45+, required very precise exposure control.  Even at the base iso of 50, if you let a highlight go, it was pure white and ruined.  This was easy to do in a sky full of clouds as if you were not careful, you would allow a portion of the cloud to go beyond the range of the sensor, it went pure white and was no good.  The P45+ also seemed to have trouble in underexposed shadows in that if you were about 3/4 of a stop under exposed, the sensor would have trouble making out the details that were in the shadows and they tended to be smeared.  Working in most outdoor light, I tended to alway shoot the P45+ in at least 3 brackets.   Sure the P45+ could handle up to 1 hour long exposures at the base iso of 50, but in normal outdoor lighting, I never really had much luck with non bracketed exposures and you constantly had to check your histogram to see if you had blown out anything.

3.  Nikon D800.  It was pretty easy to see in the February 2012 Fred Miranda review that at iso 100, this camera had an amazing dynamic range.  It still amazes me in the base iso range, just how far you can push shadows with this camera and not have noise.  If you get noise, it tends to be more of a even grainy look which looks less digital and more film like.  I can easily let my highlights go over as much as 1 stop and then recover them in post or expose for the highlights and pull up the shadows as much as 2 to 2.5 stops.  The details and resulting resolution are very impressive.  With the D800, I have stopped most of my bracketing work.  If I need a bracketed series, 2 shots is all that is required.  Areas of a image that are pulled up as much as 2 stops will still have good color, and details. 

I realize other photographers may totally disagree with my findings.  I did not want to make a switch to Nikon after being a Canon shooter for 15 years, but after seeing the results from the D800, I made the change and I still don't regret it.  I will say the new Canon 6D, has some very nice capabilities.    The 6D still pulls noise in shadows at base iso when working with a underexposed image, but the noise is much less destructive.  I found the noise issues of the 5D MKIII to be about the same as all previous Canon sensors.  The 6D to me is totally different and more enjoyable to use. 

The best way to try to get your hands around this issue is side by side testing.  I realize this may not be easy to do depending on your location.   I was able to work with my local camera dealer in Arkansas, Bedford PHoto, and was able to do several side by side shoots. 

Paul Caldwell
Logged

Paul Caldwell
Little Rock, Arkansas U.S.
Photography > http://photosofarkansas.com
Blog> http://paulcaldwellphotography.com
Jim Kasson
Sr. Member
****
Offline Offline

Posts: 1047



WWW
« Reply #3 on: May 14, 2013, 07:32:52 PM »
ReplyReply

1) Given that most scenes do have exposure ranges of 14 stops why then do we have such a problem with exposure and capturing the light in a digital camera? Except for capturing specular highlights most scenes should be coped with nicely by practically any digital camera.
2) If pure white occurs (digitally at around + 2 and a bit stops and black at around -2 and a bit stops off middle grey then what exactly is the sensor capturing with the other 9 odd stops (for the D800)?

I think there are two important issues here. One is that your assumptions are, well, wrong would be maybe too strong a term, so how about unusual? The second is the definition of dynamic range.

Let's take the second one first. The sensor in your Nikon is essentially linear. The dynamic range of a linear system is usually defined as the ratio between the highest signal in the linear region (in this case, the number of photons that just produce the highest value that can occur in the raw file) and the lowest signal that produces an acceptable signal-to-noise ratio (SNR). To convert the ratio to stops, which is the way that you, and most photographers talk about it, you take the log base 2 of the ratio.

I believe DxO sets the signal-to-noise ratio at the low end to one. Most photographers wouldn't do that. An industry rule of thumb is that 100 electrons in the sensing element is the fewest that gives a "photographic" SNR.
With 100 electrons in an otherwise=perfect camera, you get an SNR of 10. So there goes a little over three stops off the DxO dynamic range. Now we're down to, for your camera, a bit under 9.

You really don't want to blow your highlights, and the in-camera histogram is, for Byzantine historical reasons, biased in ways to keep you from doing that, so if you're using that for your exposure, you've probably lost another stop off the top end. Now you're under eight.

For the rest, let's look at your assumptions. The key, er, unusual, part of your assumptions is defining anything below 2 and a bit stops below middle gray as black. Most would say that it's not black until the SNR gets so bad that you don't like the way it looks and you pull the black point up to cover up the grunge.  That's probably more like five or six stops down from middle grey, with your camera.

Does that make sense?

Jim
Logged

guypyetan
Newbie
*
Offline Offline

Posts: 8


« Reply #4 on: May 15, 2013, 03:02:37 AM »
ReplyReply

Firstly, thank to all who have taken the time and trouble to answer so far. (To Jim your answer kind of answers my question though some clarification may be required)

1) Permit me to explain how I arrived at the -2 stops = black and + 2 stops = white. Hopefully you can see how my thinking developed
2) Allow me to explain what I did to establish what my camera could apparently see. Both elements may be fundamentally flawed but they made sense to me at the time.

Tackling 1 first.

a) My exposure meter apparently covers a range of -2 1/3 to + 2 1/3 stops. Why would nikon do this if the camera can see 10 or 14 stops?
Reading about the Zone system reveal provides the following:
b) For us digital photographers, we are only concerned with zones III through VII (zones 3 through 7). The darkest part of a scene would fall into zone III, while the brightest part of a scene would fall into zone VII. Anything darker than zone III would render as pure black with no detail (under-exposed), while anything brighter than zone VII would render as pure white with no detail (over-exposed). Source: http://photo.tutsplus.com/tutorials/shooting/understanding-using-ansel-adams-zone-system/

Both a and b above led me to the conclusion that there is a useful 4-5 stops between black and white hence why does a wider DR in a camera matter???

2) I tried a little experiment. I took a letter (B&W) and pinned it to a wall.I established a distance which would allow me to see the text nicely when the letter was photographed with a suitable exposure. My thinking (and this is where I may have gone wrong) was the ultimate test is if you can pull back black text back from a badly underexposed photograph and still recognise it as such then you can still resurrect shadow detail. Black on black ought to be the stiffest possible test. When there is no more information left in the mage then thats the low point of the DR.

Sure enough when I tried this I got a cliff like highlight rolloff beyond +2 2/3. The text very quickly become unrecoverable.
On the shadows the rolloff was rather more gentle but still, I got nowhere near -6 stops before detail simply wasn't there. At - 3 stops the text looked like braille. In fact if you hadn't seen the photographed letter you wouldn't know that the info brought back was text (therefore I see precious little point in trying to recover it). I never got to - 6 stops. I was seeing a white surface long before then.

This is why I asked the question and this is why I still wonder what the point is of cameras with larger dynamic ranges as apparently we are still ultimately constrained to the real world practical limits of -2 to +2 or thereabout of pure white and pure black.

3) Additionally (For Jim) I didn't get this bit of your explanation. Where did the SNR of 10 come from in your text below?

I believe DxO sets the signal-to-noise ratio at the low end to one. Most photographers wouldn't do that. An industry rule of thumb is that 100 electrons in the sensing element is the fewest that gives a "photographic" SNR.
With 100 electrons in an otherwise=perfect camera, you get an SNR of 10. So there goes a little over three stops off the DxO dynamic range. Now we're down to, for your camera, a bit under 9.

4) Finally I should say I'm not arguing with you chaps. I am simply trying to really nail this and understand. I want to be able to use my camera to its fullest extent and know how far I can push it. I also want to understand as much of the theory as I can.

If you choose to answer this then great but if not well thank you for your kindness so far.

Guy

Logged
guypyetan
Newbie
*
Offline Offline

Posts: 8


« Reply #5 on: May 15, 2013, 03:56:41 AM »
ReplyReply

My apologies to you Jim. Is this how you arrived at the 9 stops in your text?

I believe DxO sets the signal-to-noise ratio at the low end to one. Most photographers wouldn't do that. An industry rule of thumb is that 100 electrons in the sensing element is the fewest that gives a "photographic" SNR.
With 100 electrons in an otherwise=perfect camera, you get an SNR of 10. So there goes a little over three stops off the DxO dynamic range. Now we're down to, for your camera, a bit under 9.

low end SNR = 1 = 100 electrons
for a Nikon 800 High end SNR = 14 = 1400 electrons
log base 2 1400 = 10.45 approx
So 10.45 - 1 = approx 9

If this is the case then I hadn't engaged by brain. Before askign for verification. That said, I'd still be grateful for answers to the rest of what I've said in the questions immediately preceding this.

MTIA

Guy
Logged
bjanes
Sr. Member
****
Offline Offline

Posts: 2869



« Reply #6 on: May 15, 2013, 07:50:09 AM »
ReplyReply


1) Permit me to explain how I arrived at the -2 stops = black and + 2 stops = white. Hopefully you can see how my thinking developed

a) My exposure meter apparently covers a range of -2 1/3 to + 2 1/3 stops. Why would nikon do this if the camera can see 10 or 14 stops?

I think you are misinterpreting the range of your meter. The analog display on the camera LCD or in the viewfinder typically shows a slider with the nominal exposure in the middle and plus or minus 2 or 3 EV on each side. This is not the total range of the meter, but merely a window into the range. If you take a reading in bright sunlight and then go into a dark room and take another reading, the exposure window moves to a new range. For the D800 Nikon lists the range of the light meter at 0-20 EV for matrix or center weighted and 2-20 EV for spot metering. Moreover, the DR of the sensor has little to do with the range of the meter.

Reading about the Zone system reveal provides the following:
b) For us digital photographers, we are only concerned with zones III through VII (zones 3 through 7). The darkest part of a scene would fall into zone III, while the brightest part of a scene would fall into zone VII. Anything darker than zone III would render as pure black with no detail (under-exposed), while anything brighter than zone VII would render as pure white with no detail (over-exposed). Source: http://photo.tutsplus.com/tutorials/shooting/understanding-using-ansel-adams-zone-system/

The zones in Ansel's system are highly nonlinear. There are only 2.5 stops from mid-gray (18%, Zone V) to pure white (100%, Zone X) and a camera with 12 stops of DR would have 7.5 stops of DR below mid-gray. The response a digital sensor is linear, whereas the H&D curve used for film is logarithmic. The zone system has limited application for digital photography, although the concepts are invaluable.

Regards,

Bill
Logged
Jim Kasson
Sr. Member
****
Offline Offline

Posts: 1047



WWW
« Reply #7 on: May 15, 2013, 09:42:11 AM »
ReplyReply

2) I tried a little experiment. I took a letter (B&W) and pinned it to a wall.I established a distance which would allow me to see the text nicely when the letter was photographed with a suitable exposure. My thinking (and this is where I may have gone wrong) was the ultimate test is if you can pull back black text back from a badly underexposed photograph and still recognise it as such then you can still resurrect shadow detail. Black on black ought to be the stiffest possible test. When there is no more information left in the mage then thats the low point of the DR.

Sure enough when I tried this I got a cliff like highlight rolloff beyond +2 2/3. The text very quickly become unrecoverable.
On the shadows the rolloff was rather more gentle but still, I got nowhere near -6 stops before detail simply wasn't there. At - 3 stops the text looked like braille. In fact if you hadn't seen the photographed letter you wouldn't know that the info brought back was text (therefore I see precious little point in trying to recover it). I never got to - 6 stops. I was seeing a white surface long before then.

I think I can answer all your questions. It looks like Bill has already answered the exposure meter one.

However, we need to get something out of the way first. I've been assuming that you are working with raw files, not JPEGs that the camera is making.  Please let me know whether or not that's the case.  The results that you report above make me think that you might not be working with raw files.

Thanks,

Jim
Logged

Jim Kasson
Sr. Member
****
Offline Offline

Posts: 1047



WWW
« Reply #8 on: May 15, 2013, 09:54:51 AM »
ReplyReply

3) Additionally (For Jim) I didn't get this bit of your explanation. Where did the SNR of 10 come from in your text below?

I believe DxO sets the signal-to-noise ratio at the low end to one. Most photographers wouldn't do that. An industry rule of thumb is that 100 electrons in the sensing element is the fewest that gives a "photographic" SNR.
With 100 electrons in an otherwise=perfect camera, you get an SNR of 10. So there goes a little over three stops off the DxO dynamic range. Now we're down to, for your camera, a bit under 9.

Guy,

I was hoping to avoid this calculation, since it's a bit too much "inside baseball" for most beginners, but here goes. There is an effect called shot noise, or photon noise, that limits the signal to noise ratio of any device that captures light to form an image. Please let me elide the conversion of photons impinging on the photosite, or sensel, or individual light detector in your camera to electrons that the camera can convert to numbers. The standard deviation of the number of electrons in the sensel is proportional to the square root of the number of electrons in the sensel. Call the stndard deviation the noise. Thus, the signal to noise ratio is the number of electrons divided be the square root of the number of electrons, or just the square root of the number of electrons.

The square root of 100 is 10. So, in a camera with no other sources of noise, the signal to noise ratio of an exposure that puts the average value of electrons at 100 is 10. Real world cameras have other sources of noise, so it might take a bit more light than that to get the SNR to 10, but that's a good place to start.

I hope that's not too confusing.

Jim

Logged

Jim Kasson
Sr. Member
****
Offline Offline

Posts: 1047



WWW
« Reply #9 on: May 15, 2013, 10:50:22 AM »
ReplyReply

2) I tried a little experiment. I took a letter (B&W) and pinned it to a wall.I established a distance which would allow me to see the text nicely when the letter was photographed with a suitable exposure. My thinking (and this is where I may have gone wrong) was the ultimate test is if you can pull back black text back from a badly underexposed photograph and still recognise it as such then you can still resurrect shadow detail. Black on black ought to be the stiffest possible test. When there is no more information left in the mage then thats the low point of the DR.

Sure enough when I tried this I got a cliff like highlight rolloff beyond +2 2/3. The text very quickly become unrecoverable.
On the shadows the rolloff was rather more gentle but still, I got nowhere near -6 stops before detail simply wasn't there. At - 3 stops the text looked like braille. In fact if you hadn't seen the photographed letter you wouldn't know that the info brought back was text (therefore I see precious little point in trying to recover it). I never got to - 6 stops. I was seeing a white surface long before then.

If you look at reply 130 in this thread: http://www.luminous-landscape.com/forum/index.php?topic=76446.120  you can see image information from a Nikon D4 capture where the exposure is roughly 14 stops (!) below clipping, or about 11 stops below middle grey.

Jim
Logged

guypyetan
Newbie
*
Offline Offline

Posts: 8


« Reply #10 on: May 15, 2013, 11:48:12 AM »
ReplyReply

Hi Jim,

Nope I'm strictly raw and manual exposure.

Guy

Logged
guypyetan
Newbie
*
Offline Offline

Posts: 8


« Reply #11 on: May 15, 2013, 11:52:08 AM »
ReplyReply

Cool makes perfect sense. Thank you.
Logged
guypyetan
Newbie
*
Offline Offline

Posts: 8


« Reply #12 on: May 15, 2013, 11:54:58 AM »
ReplyReply

Guy,

I was hoping to avoid this calculation, since it's a bit too much "inside baseball" for most beginners, but here goes. There is an effect called shot noise, or photon noise, that limits the signal to noise ratio of any device that captures light to form an image. Please let me elide the conversion of photons impinging on the photosite, or sensel, or individual light detector in your camera to electrons that the camera can convert to numbers. The standard deviation of the number of electrons in the sensel is proportional to the square root of the number of electrons in the sensel. Call the stndard deviation the noise. Thus, the signal to noise ratio is the number of electrons divided be the square root of the number of electrons, or just the square root of the number of electrons.

The square root of 100 is 10. So, in a camera with no other sources of noise, the signal to noise ratio of an exposure that puts the average value of electrons at 100 is 10. Real world cameras have other sources of noise, so it might take a bit more light than that to get the SNR to 10, but that's a good place to start.

I hope that's not too confusing.

Sorry I forgot to quote you when I thanked you. Thank you that makes sense.

Jim


Logged
Jim Kasson
Sr. Member
****
Offline Offline

Posts: 1047



WWW
« Reply #13 on: May 15, 2013, 12:03:45 PM »
ReplyReply

low end SNR = 1 = 100 electrons
for a Nikon 800 High end SNR = 14 = 1400 electrons
log base 2 1400 = 10.45 approx
So 10.45 - 1 = approx 9

The numbers for a D800 are more like this:

low end SNR = 10 which implies 100 electrons.
Full well capacity = 48000 electrons
Dynamic range = 48000/100 = 480
log2(480) = 8.9
8.9 - 1 = 7.9

Jim
Logged

Jim Kasson
Sr. Member
****
Offline Offline

Posts: 1047



WWW
« Reply #14 on: May 15, 2013, 12:14:47 PM »
ReplyReply

4) Finally I should say I'm not arguing with you chaps. I am simply trying to really nail this and understand. I want to be able to use my camera to its fullest extent and know how far I can push it. I also want to understand as much of the theory as I can.

In that case, at the risk of TMI, you might want to look at the following:

http://www.rawdigger.com/howtouse/pixel-capacity-and-amplifier-gain

http://www.clarkvision.com/articles/digital.sensor.performance.summary/#unity_gain

http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/noise-p2.html#readandshot

and last, and maybe least,

http://blog.kasson.com/?page_id=2387

Jim
Logged

Jim Kasson
Sr. Member
****
Offline Offline

Posts: 1047



WWW
« Reply #15 on: May 15, 2013, 12:24:06 PM »
ReplyReply

Nope I'm strictly raw and manual exposure.

Then I'm having a hard time understanding how things get black so fast for you as the exposure goes down. Are you using the shadow-boosting tools in your raw developer? If you tell me which one you're using and it's one I know I can be more specific.

Jim
Logged

guypyetan
Newbie
*
Offline Offline

Posts: 8


« Reply #16 on: May 15, 2013, 12:46:59 PM »
ReplyReply

I'm not sure how things are getting so black for me so quickly either :-<. Perhaps I need to carry out the exercise I went through previously with a little more precision and seriously firm up my numbers.

FYI I am using Lightroom and PS CS6 Either ACR or Lightroom raw.

I have to say thank you hugely. You've helped me a lot and advanced my knowledge.
Logged
guypyetan
Newbie
*
Offline Offline

Posts: 8


« Reply #17 on: May 15, 2013, 12:58:54 PM »
ReplyReply

I forgot to say. I think I have a lot of reading to do! On the bright side, apparently when I make a picture, I manage to achieve what I require in terms of exposure consistently so I guess that's a good thing (I might be talking nonsense but somewhere something has stuck). Trouble is when you realise what you need to know you then realise all the connected subjects you need to swot up on so that you can then study the new stuff (my maths is really rusty).

Logged
ErikKaffehr
Sr. Member
****
Offline Offline

Posts: 7919


WWW
« Reply #18 on: June 07, 2013, 02:46:18 PM »
ReplyReply

Hi,

I have some 65000 images in my Lightroom database and struggle finding images having issues with dynamic range, so I don't really see this as a problem.

http://echophoto.dnsalias.net/ekr/index.php/photoarticles/61-hdr-tone-mapping-on-ordinary-image

http://echophoto.dnsalias.net/ekr/index.php/photoarticles/63-lot-of-info-in-a-digital-image

The clue is really to use minimum ISO and expose to the right, expose so that nonspecular highlights don't blow out but still having a histogram going fully to the right.

Best regards
Erik
Logged

Pages: [1]   Top of Page
Print
Jump to:  

Ad
Ad
Ad