Too much dynamic range?

[NormanBates]

We are saying the same, I guess I didn't make myself sufficiently clear.

What I mean is that if you add 1 bit to your ADC and instead of having

00 = 0-100 photons in a pixel
01 = 100-200 photons in a pixel
10 = 200-300 photons in a pixel
11 = 300-infinity photons in a pixel

You now have:

000 = 0-42 photons in a pixel
001 = 42-84 photons in a pixel
010 = 84-126 photons in a pixel
011 = 126-168 photons in a pixel
100 = 168-210 photons in a pixel
101 = 210-252 photons in a pixel
110 = 252-294 photons in a pixel
111 = 294-infinity photons in a pixel

...and you have the same read-out noise, you still have the same DR, because neither your full-well capacity nor your read-out noise have changed.

<----- is writing the VHDL code for the FPGA of a motion picture camera

[nightbreath]

There are some usage models for which DR is important, there are many usage models for which it doesn't matter at all. If the scene you have in front of you doesn't require more than 8 stops of DR, you're fine with a camera that can capture that, no point in going for one that is the same in every respect but will record 14 stops of DR.

If your portraits are in a studio, with a standard backdrop, your DR needs will probably be pretty modest. If your portraits happen in other less-controlled locations, you may have very high DR needs (e.g. if you want to take a portrait of someone in their bedroom, and there's a window in an interesting area). Wedding photographers take lots of portraits, and, not having a lot of control over their shooting scenarios, they usually need a lot of DR (for this reason, a friend of mine was still using his Fuji S3 pro as his backup body up until the D800 came out: a 12 mpix camera from 2005... with 13.5 stops of DR as measured by dxomark).

In any case, ADC precision will only be a problem if the manufacturer screws up the sensor-ADC matching. No current camera has that issue AFAIK.

I have only few times felt lack of DR in my camera. And that was before I really used to getting the pictures I won't delete later. There are several techniques to get the shot you need the way you want to see it.

Is there a web-site where I can look at your friend's photos to check why someone chooses high DR cameras?

[jukka]

You need not look at any photos, with 14 stops DR you have more  exposure options and with for example d800 you have no banding  and  pattern noise in lower levels. You can use one raw file and develop the raw file  after highlight and  one after shadows and mix them together, With a Canon you must take two or more exposure and have the Camera on a tripod and no moving objects, or you can also develope one raw file after the high lights and lift the  areas  in shadows  and a  contrasty motive with out pattern noise or banding.

[bycostello]

can you have too much...

[@!ex]

I hope that someday I could get a shot with this much dynamic range in a single exposure.  I bracketed 7 shots at 3 EV spacing per bracketed shot.  That is 18 EV spread, but each shot has it's total EV range (minus clipping) so the DR spans almost from pure black to pure white.  My eye saw these images like this, but with a single exposure (including using ND filters) I could never get these shots without increasing the DR of the camera via multiple exposures.  Sorry, not trying to go off topic, just thought it was relevant to the subject.


End of the Road by @!ex, on Flickr


Everything Peels... by @!ex, on Flickr

[helpful]

We are saying the same, I guess I didn't make myself sufficiently clear.

What I mean is that if you add 1 bit to your ADC and instead of having

00 = 0-100 photons in a pixel
01 = 100-200 photons in a pixel
10 = 200-300 photons in a pixel
11 = 300-infinity photons in a pixel

You now have:

000 = 0-42 photons in a pixel
001 = 42-84 photons in a pixel
010 = 84-126 photons in a pixel
011 = 126-168 photons in a pixel
100 = 168-210 photons in a pixel
101 = 210-252 photons in a pixel
110 = 252-294 photons in a pixel
111 = 294-infinity photons in a pixel

...and you have the same read-out noise, you still have the same DR, because neither your full-well capacity nor your read-out noise have changed.

<----- is writing the VHDL code for the FPGA of a motion picture camera

Right on! This is exactly what I'm trying to say, and you explained it much more clearly. DR is not the same as the number of gradations and also not the same as the bit depth (which actually just counts the number of "possible" gradations, whether or not the camera actually is capable of resolving all those gradations).

If the number of gradations accurately recorded within a 10 stop dynamic range is the same as the number of gradations accurately recorded within a 14 stop dynamic range, then the 10-stop camera has more precision and better image quality _within that 10-stop interval of light intensity_ versus the 14-stop camera. But outside that 10-stop range, the 10-stop camera has zero image quality, and so the 14-stop camera wins hands-down.

DR is not something to get angry about, just a trade-off between obtaining either greater differentiation between subtle shades of colors (like slide film with lower DR) or greater exposure latitude (like negative film with higher DR).

[nightbreath]

If the number of gradations accurately recorded within a 10 stop dynamic range is the same as the number of gradations accurately recorded within a 14 stop dynamic range, then the 10-stop camera has more precision and better image quality _within that 10-stop interval of light intensity_ versus the 14-stop camera. But outside that 10-stop range, the 10-stop camera has zero image quality, and so the 14-stop camera wins hands-down.

And that's why I've started the topic. To my understanding my near-12-stop DR camera is perfect for my work and I would think twice before getting next Canon released camera that might have bigger DR with the same number of gradations resolving power.

[NormanBates]

Is there a web-site where I can look at your friend's photos to check why someone chooses high DR cameras?

He's a wedding photographer, I don't think he publishes his pictures online, he gives them to his customers.
But the usual scenario he was referring to was: very sunny day, bride in shiny white, broom in matte black suit with subtle stripes, anything except his fuji (or, now, D800) will result in said suit looking like a black blotch, and there's nothing he can do about it.


Now, back to the technical discussion...


Let me add a twist: the ADC works linearly, but what you see is log

So, if you have a 14-bit ADC (you can count up to 16384)and can record 14 stops of DR, here is how those values will be distributed:

14th stop: 8192 to 16383
13th stop: 4096 to 8191
12th stop: 2048 to 4095
11th stop: 1024 to 2047
10th stop: 512 to 1023
9th stop: 256 to 511
8th stop: 128 to 255
7th stop: 64 to 127
6th stop: 32 to 63
5th stop: 16 to 31
4th stop: 8 to 15
3rd stop: 4 to 7
2nd stop: 2 to 3
1st stop: 0 to 1

So you may actually have very serious issues in the shadows... which I see in the Canons, but not in the D800!

* if you're going to have issues with "too much DR, not enough gradation", they'll be in the very deep shadows, which you wouldn't see anyway if you were shooting with a camera with the same ADC but less DR; your skin tones are unlikely to land anywhere below the 5th stop fro the top, so for them you have way more values than you need (anything above 50 gradations per stop is usually smooth even after heavy grading)

* how come I don't see this in samples from the D800?

[nightbreath]

Is there a web-site where I can look at your friend's photos to check why someone chooses high DR cameras?

He's a wedding photographer, I don't think he publishes his pictures online, he gives them to his customers.
But the usual scenario he was referring to was: very sunny day, bride in shiny white, broom in matte black suit with subtle stripes, anything except his fuji (or, now, D800) will result in said suit looking like a black blotch, and there's nothing he can do about it.

Something like the one I've attached? Shot in the middle of the day. So it's another reason why I've started the discussion. Because I don't understand why everyone is so tempted about DR possibilities when everything depends on technique.

P.S. It's not one of the best shots from this day, I've just used one with hasrh shadows.

* if you're going to have issues with "too much DR, not enough gradation", they'll be in the very deep shadows, which you wouldn't see anyway if you were shooting with a camera with the same ADC but less DR; your skin tones are unlikely to land anywhere below the 5th stop from the top, so for them you have way more values than you need (anything above 50 gradations per stop is usually smooth even after heavy grading)

As far as I understand each camera applies it's own tone curve to the image, or am I wrong?

Initially I wanted to be brand-agnostic and instead of discussing specific sensors, I want to identify what really matters for my needs (and maybe many others). I'm not able to tell what it is right now, so everyone's input is appreciated 

[nightbreath]

I hope that someday I could get a shot with this much dynamic range in a single exposure.  I bracketed 7 shots at 3 EV spacing per bracketed shot.  That is 18 EV spread, but each shot has it's total EV range (minus clipping) so the DR spans almost from pure black to pure white.  My eye saw these images like this, but with a single exposure (including using ND filters) I could never get these shots without increasing the DR of the camera via multiple exposures.  Sorry, not trying to go off topic, just thought it was relevant to the subject.

Shot #1
Shot #2

No offense, but these scenarios look uninspiring to me. And I believe it's not about how you or I see it, it's about everyone's way of thinking towards DR that makes HDR overused by lots of photographers around the world.

I believe that HDR imaging has its own niche, but it should be used when the result doesn't tell you whether it's HDR or not. So better scenes is what really matters for me (rather than increased DR):

[Danmix]

Touché nightbreath.... Great shot

[nightbreath]

Touché nightbreath.... Great shot

Sorry, I didn't sign the last photo. It's not mine, I've found it in a social network. It was said that "Baktiar Sontani" is the author of the photo 

[NormanBates]

As far as I understand each camera applies it's own tone curve to the image, or am I wrong?

Initially I wanted to be brand-agnostic and instead of discussing specific sensors, I want to identify what really matters for my needs (and maybe many others). I'm not able to tell what it is right now, so everyone's input is appreciated 

Not really: each camera applies its own tone curve, but that happens *after* the ADC has done its job, and, as far as I know, all cameras have linear ADC (ok, some have piecewise linear, but that's actually a change for exposure, not for the ADC, which is still linear; and I'm not sure any of the ones we're talking about actually does that).

The fact that light is linear but you see it as log makes this very inefficient, and it is the reason that, for example, some Nikon cameras (low to mid-end) apply a log curve even in their RAW files: with linear, you have way more gradation than you need in the highlights, and may still be struggling in the shadows. But even this curve happens after the ADC, so it's not what we are talking about, I think.



As for the need for DR, as I said, it depends on what you do, how much time you have to do it, and how much margin of error. My friend was carrying an extra 5-year-old 12 Mpix camera just in case, because he thought he needed to do so. If you don't think you need it, well, good for you, what can I say?

Now go ask any cinematographer if they think 11 or 12 stops of DR is enough.
(hint: I hardly ever shoot stills, I'm a vidiot)

[molnarcs]

I don't quite understand all the technical background, but I can attest to the advantages of high DR in my work. I do a lot of interiours, and most of them are high-contrast situations, often necessitating trips to PS for better noise control (I bought Noiseware Professional which used to be better than LR, maybe still is), occasionally blending in different exposures (I always do bracketed shots for interiour work), etc.

With the d7000 and now the d800, my trips to PS noticeably decreased. In fact, I don't recall any situation in the past year where I had to use blending. My only comparison is the 5D MK II and Rebel T2i, and my gut feeling is that you can push the d7k at least a stop more without image degradation (colour shift, noise, etc.) In Lightroom terms, this is about 30-40 points on the shadows slider, or being able to push both blacks and shadows on the Tone Curve significantly more. For me, it means staying in Lightroom for 99% of my workflow. You can just feel how much more malleable are NEF files then CR files.

I'm not sure any of this matters if you're shooting JPEG. Except maybe if you have ALO enabled. With Canon's, I never used ALO (and I didn't use ADL it with the d7000), but when I bought the d800, I just left ADL (the Nikon equivalent to ALO) on Auto setting. I'm a raw shooter, but lately I've been experimenting with JPEG+RAW because I started shooting events now (not much interiour work lately). As far as I can tell, ADL works well, I don't see any image degradation, and the change is quite subtle actually, but for the better as far as I can tell. Where it truly matters is when you shoot RAW.

I heard a good description of RAW somewhere - RAW is like a box of light. When you look at a RAW shot from either Canon or Nikon you basically see the same image and DR. The difference becomes prevalent when you're start messing with it, changing exposure or using curves. With a high DR camera, you can push shadows more without significant image degradation. It is as simple as that. You have a bigger box of light with a high DR camera

By the way, your photos are magnificent! I don't even know what most of my favourite photographers shoot. I know some shoot Canons, others shoot Nikons - and the end results are all magnificent. Once you have invested in either system, I don't see much reason to change. I didn't have a huge investment (Sigma 10-24 and a Rebel) when I bought the Nikon d7000 (and the new Sigma 8-16 which was to replace the 10-24 anyway). The 5D MK II wasn't mine. For the things I shot back then, the high DR of the d7k did make a difference, made my life easier. But for the things you shoot, it might make zero difference. As I said, what you get with high DR is a bigger box of light, but if you don't see the limits of your current box, then you don't need a high DR camera And I think for JPEG shooters it doesn't matter at all.

[noisejammer]

...Let me add a twist: the ADC works linearly, but what you see is log

So, if you have a 14-bit ADC (you can count up to 16384)and can record 14 stops of DR, here is how those values will be distributed:

14th stop: 8192 to 16383
13th stop: 4096 to 8191
12th stop: 2048 to 4095
11th stop: 1024 to 2047
10th stop: 512 to 1023
9th stop: 256 to 511
8th stop: 128 to 255
7th stop: 64 to 127
6th stop: 32 to 63
5th stop: 16 to 31
4th stop: 8 to 15
3rd stop: 4 to 7
2nd stop: 2 to 3
1st stop: 0 to 1
...

This is incorrect - if you replaced "bit" with "stop" then it is correct (and it's then obvious why ETTR works too.)

It's easier to consider a 3- or 4-bit digitiser. Let's say it offers 4 bit resolution then the possible counts are 0000 through 1111. This translates to 2^4 or 16 levels. Written the way that Norman stated it, there would only be four distinct levels - this is incorrect (but I understand he meant there would be 16 levels.)

Some things appear to have been glossed over in the discussion. First, the ADC operates on a per-pixel basis.

I've read the DxO tests on various sensors. It is important to remember that the notional dynamic range is referred back to an 8 mp standard. This means that the D800's quoted 14 bits dynamic range is significantly less than 14 bits at a per-pixel level. The 36 mp > 8 mp conversion gains the sensor sqrt(4.5) = 2.1x notional improvement in dynamic range. The 2.1x is slightly more than 1 stop in quoted dynamic range. This means that the true per-pixel dynamic range is about 12.9 stops.

In order to read those 12.9 stops, the ADC needs a bit more resolution than the 13 bits required by the pixel. I'm quite surprised because the 14 bits in the ADC suggests that the entire detection chain has ~1 bit of noise.... It sounds improbable.

Photon shot noise has been commented on briefly. If we assume 13 bits dynamic range and (say) 10% quantum efficiency (pidooma), then the number of photons required to fill a pixel is 10 x 2^13 or 80k. Since shot noise varies with the square root of the number of photons, the pixel could have shot noise of up to 280 photons (rms).

Since 1 bit translates to 80k/8192 = 10 photons and we must have about 6 bits of photon noise at the upper end of the sensor's dynamic range. At the bottom end, the quantum efficiency sets the performance and there must be ~3 bits of noise.

Shot noise alone suggests that the true dynamic range of an image cannot be more than about 8-10 bits. It seems that the only way to improve on this is by greatly enhancing the sensor's quantum efficiency.

To answer the OP's question - photon noise alone suggests that there's not a whole lot of benefit to high resolution ADC. It does allow for more sophisticated noise filtering - presumably at the expense of resolution.

<---- physicst / astronomer