Home · Register · Join Upload & Sell

Moderated by: Fred Miranda
Username  

  New fredmiranda.com Mobile Site
  New Feature: SMS Notification alert
  New Feature: Buy & Sell Watchlist
  

FM Forums | Leica & Alternative Gear | Join Upload & Sell

1      
2
       3       4       end
  

Archive 2013 · Dynamic Range vs. Tonal Range

  
 
carstenw
Offline
• • • • • •
Upload & Sell: Off
p.2 #1 · p.2 #1 · Dynamic Range vs. Tonal Range


Joakim, I am curious why you call the outside the "object side", and not the "subject side". Is that some specific vocabulary used in your field?


Apr 20, 2013 at 11:42 AM
AhamB
Offline
• • • • •
Upload & Sell: Off
p.2 #2 · p.2 #2 · Dynamic Range vs. Tonal Range


carstenw wrote:
Joakim, I am curious why you call the outside the "object side", and not the "subject side". Is that some specific vocabulary used in your field?


I believe that's the standard terminology in optics? "Subject" is related to intent, so more related to the arts, I think. The subject is the person making an image of the object with the objective lens.



Apr 20, 2013 at 11:58 AM
theSuede
Offline
• • • •
Upload & Sell: Off
p.2 #3 · p.2 #3 · Dynamic Range vs. Tonal Range


Yes, it's fairly textbook use of nomenclature in optics. "Object" is rather logical, it's a three dimensional world outside of the optical system. Subject is both in linguistics and in physics the active part, object is the passive part. I guess the people that defined the nomenclature thought the target you point the lens (the "working" part) at to be "passive" part.

The "image" is strictly the focal plane of the optics when you point and focus the lens system at a flat surface that's exactly perpendicular to the optical center axis. So the "image" in optical terms is like the "field" in photographic terms. As in "field curvature".

This definition is rather unclear in practical use though, even the most stringent textbooks often stray between two usages - either the one stated above, OR the image projected on to the flat plane at the point of capture (the sensor).

For practical purposes the second and less strict definition of image (the sensor plane) is used in normal photography. The world you shoot is 3-D, and the actual point of capture is simplified to being "flat". It really isn't though, the filter plates on top of the sensor actually curves the virtual sensor plane outwards, decreasing the effect of curvature of field (but increasing astigmatism, coma and SA).



Apr 20, 2013 at 04:50 PM
zhangyue
Offline
• • • • •
Upload & Sell: On
p.2 #4 · p.2 #4 · Dynamic Range vs. Tonal Range


I was trying to avoid this thread since it require brain power to read and comments. Now I am waiting for my daughter piano class. Here are some of my thoughts.

All based on my understanding so I am open to be corrected:

DR of camera is purely defined by the sensor technology. As others have point out nowadays all manufactures have chosen ADC have high bit to cover the range so that sensor itself usually is the limited factor. So here is not much I can talk about. We all know that Sony based sensor is the best as for now.

In terms of TR, I was thinking that we don't need 14 bit ADC or thought high bit ADC has no advantage given sensor is the limited factor. But now, consider TR, make me want take it back.

There is benefit to have more bits of ADC to start with.

First, let assume there is two camera with sensors capable capture 8bit DR (or 50dB SNR, so signal is 2^8 bigger than noise). one with 8 bit ADC, the other with 12bit ADC.

There is an image has low DR, say 4 bit. the first camera with 8 bit ADC can only capture 2^4 resolution/step of tonal level, The same image captured by 2nd camera will capture 2^8 resolution level for the same 4 bit dynamic range image. So you can do the math here, there is a lot more information capture by 2nd camera even the image has the same DR.

Once you convert captured image to 8bit jpeg, obviously you have more tonal range and freedom to PP with 2nd camera. To the extreme, you can even remap the image to full 8 bit DR.

So in reality. It is not just DR make the image, but also TR, color response, and many things else. I agree with Sony sensor is capable to capture huge DR compare to Canon, but, but do we really need information the 2X, 2^2 or 2^3 smaller black? More DR, obvious better, give people more freedom to PP their images. But it is hugely overrated I think once consider how much it impacts image quality.

Since all those 14bit DR has to map to 8 bit JPEG, I fail to understand what is so fussy about Canon's files. I might miss something here, welcome to explain to me.

For lower DR images, some camera just has more TR with carefully design ADC/RAW algorithm. Have more bit ADC does offer more freedom to remap tonal range digitally, give photographer more freedom to achieve final results they wanted. (I think)

Since my field is electrical engineering, I have good understanding for signal processing, ADC/DAC, but not much about how raw, jpeg, print, gamma etc...works. I am open to any correction.

Just offer some food for thinking.





Edit: edit my original iphone content to make it readable. A little detail to explain tonal range and benefit of more bit ADC in sensor.




Edited on Apr 21, 2013 at 01:46 AM · View previous versions



Apr 20, 2013 at 05:00 PM
theSuede
Offline
• • • •
Upload & Sell: Off
p.2 #5 · p.2 #5 · Dynamic Range vs. Tonal Range


Just thought to add a little something for Rusty.
"Tonal range" isn't a very clear definition - it's rather convoluted. But maybe I can try to write down my understanding of it in stead of what my knowledge of it...

Blank mind.
Imagine a full range of gray tones, from pure black up to something so bright that the raw file clips. Capture it with the camera.
In each and every point of that ramp or "gradient" you can find an exact noise error or "standard deviation".
The lowest possible noise error is the electronic noise added by the camera, set this point as the blackpoint
So, starting from that BP, you move up in brightness to [BP plus the noise error]. Add "one" to the step counter
Now, at that point you have a different noise error. Add that new noise error to the current level, and add "one" to the step counter again.

Keep on doing this until you hit raw clipping.
Do this for all three color channels.
The number of steps you managed to get on the step counter is the "Tonal range"!

Higher sensor efficiency (better light-to-electric energy conversion) lowers the noise in each step, making the steps shorter. Shorter steps > you fit more steps in the range, and you get better tonal resolution.

Better DR means that you start further down towards absolute black (D800 starts the first step at ~-14Ev, D3 start the first step at ~-12Ev), and this does also increase the Tonal range - but the effect is smaller than that of having a higher efficiency.



Apr 20, 2013 at 05:11 PM
rico
Offline
• • • • •
Upload & Sell: Off
p.2 #6 · p.2 #6 · Dynamic Range vs. Tonal Range


Particularly lucid explanation, Joakim! I have started, but not yet finished, an experiment to determine the d/r when light reflections in the mirror box are factored. The design includes direct measurements from the RAW capture, and a light source with calibrated pattern. I have reason to believe that 10 bits is an achievable worst case.


Apr 21, 2013 at 01:21 AM
zhangyue
Offline
• • • • •
Upload & Sell: On
p.2 #7 · p.2 #7 · Dynamic Range vs. Tonal Range


I did some edit for my above post about Tonal range. to make it readable. Sorry for all the typo and not so clear defined content.



Apr 21, 2013 at 01:44 AM
rico
Offline
• • • • •
Upload & Sell: Off
p.2 #8 · p.2 #8 · Dynamic Range vs. Tonal Range


Re 8-bit versus 10-bit tonality, I forgot to mention I have a suitable video path to a 30-bit LCD display (10 bits per channel). 8-bit ramps are indeed discernibly quantized in certain areas of the color cube, particularly gray and green. 9-bits is good enough, and 10 is plenty. Display is aRGB gamut.


Apr 21, 2013 at 02:01 AM
RustyBug
Online
• • • • • • •
Upload & Sell: On
p.2 #9 · p.2 #9 · Dynamic Range vs. Tonal Range


So ... let me see if I can overstate something I probably should have understood much sooner than now, but just trying to confirm.

The advances in new camera technology that is claiming greater DR along with the trend toward smaller pixels isn't really a byproduct of the smaller pixels by itself ... but rather an improvement in the read noise @ ADC ... independent of pixel size. Shot noise on the other hand is more directly related to pixel size. In varying situations/scenarios ... read noise is readily evident and shot noise isn't a critical factor. In other situations/scenarios, read noise gets masked/overridden by shot noise in other scenarios or "hidden" by EV levels involved, such that neither read, nor shot noise are readily evident.


I probably screwed that up on technical terminology/concept ... so please correct as needed, but is that essentially why an old camera with large pixels can have (slightly) better SNR, TR, CS, yet less DR (as a function of more read noise)?



Apr 21, 2013 at 01:51 PM
theSuede
Offline
• • • •
Upload & Sell: Off
p.2 #10 · p.2 #10 · Dynamic Range vs. Tonal Range


Define if the statement is "per pixel" or "per image".

When people answer in one definition while the person asking the question meant the "other", things get ugly rather quickly.



Apr 21, 2013 at 03:04 PM
pbraymond
Online
• • • •
Upload & Sell: On
p.2 #11 · p.2 #11 · Dynamic Range vs. Tonal Range


Tag for later reading / comprehension.


Apr 21, 2013 at 09:19 PM
philip_pj
Offline
• • • • •
Upload & Sell: On
p.2 #12 · p.2 #12 · Dynamic Range vs. Tonal Range


'Higher sensor efficiency (better light-to-electric energy conversion) lowers the noise in each step, making the steps shorter. Shorter steps > you fit more steps in the range, and you get better tonal resolution.

Better DR means that you start further down towards absolute black (D800 starts the first step at ~-14Ev, D3 start the first step at ~-12Ev), and this does also increase the Tonal range - but the effect is smaller than that of having a higher efficiency.'

The three measures - DR, Tonal Range and what I think might be called SNR ('higher sensor efficiency') would tend to be highly correlated?




Apr 22, 2013 at 11:44 PM
zhangyue
Offline
• • • • •
Upload & Sell: On
p.2 #13 · p.2 #13 · Dynamic Range vs. Tonal Range


philip_pj wrote:
The three measures - DR, Tonal Range and what I think might be called SNR ('higher sensor efficiency') would tend to be highly correlated?



SNR is refer 'signal to Noise' ratio, its definition is total Signal power vs total Niose power within certain bandwidth. (for example, for audio, usually is 20Hz to 20Khz what we human can hear) It is close related to DR. and many times, people treat them the same. (Though they are quite different pure technically) DR mean whitest to blackest range usually measured in binary. In audio is loudest sound and faintest sound ratio in system define DR. DR usually has to be measure with not only noise but also THD (total harmonic distortion).

I don't see direct relationship between DR, SNR with Tonal range. At least with no first order relationship I can see. But from my understanding: once Huge DR image, let say 14 stop (2^14) DR be captured, you can remap the signal with different black/white level into print or 8bits with more freedom, so indirectly improve your tonal range/response of image.

The same apply with more bits ADC architecture in sensor, essentially, you have more tonal step(not talking traditional 8bit JPEG here) in captured image with more bits. Given you capturing more information digitally, how you want PP is up to you, but there is no side effect from what I see other than camera’s CPU power and power consumption. With raw files, you can do more in the future if printer or display technology marches to give you more tonal response.

That is my understanding.



Apr 23, 2013 at 12:08 PM
Kolor-Pikker
Offline
• •
Upload & Sell: Off
p.2 #14 · p.2 #14 · Dynamic Range vs. Tonal Range


The sensor's performance has a large impact on color and tonality in part from the bayer filter, which itself is made of colored dye, the actual spectral qualities of this dye can affect everything from color accuracy to tonal range, since luminance is derived from color with bayer sensors. Ideally, the perfect CFA would would allow each pixel to only see the intended color and have no light loss, but in practice this doesn't really happen.


Apr 23, 2013 at 12:41 PM
AhamB
Offline
• • • • •
Upload & Sell: Off
p.2 #15 · p.2 #15 · Dynamic Range vs. Tonal Range


zhangyue wrote:
The same apply with more bits ADC architecture in sensor, essentially, you have more tonal step(not talking traditional 8bit JPEG here) in captured image with more bits. Given you capturing more information digitally, how you want PP is up to you, but there is no side effect from what I see other than camera’s CPU power and power consumption. With raw files, you can do more in the future if printer or display technology marches to give you more tonal response.


Of course higher bit depth also creates bigger RAW files, but more bits only help if you're not yet getting the maximum information from the signal, otherwise you're just subdividing noise and creating useless extra data.



Apr 23, 2013 at 12:56 PM
RustyBug
Online
• • • • • • •
Upload & Sell: On
p.2 #16 · p.2 #16 · Dynamic Range vs. Tonal Range


The audio analogy, while I'm not technically proficient in it makes some sense.

It would seem then that DR is CLEAN range. The more you can clean up the lower end, the greater the range can be extended for usefulness (say from a clean 40Hz down to clean 20Hz). The former at a clean 40Hz may have been able to produce a 20Hz, but it would have been accompanied with noise/distortion such that you really only wanted to use it for 40Hz and above as it would be unpleasant to listen to @ dirty 20Hz.

Our post processing then takes on the role of a graphic equalizer to push/pull the input signal to our desired tastes. The 20Hz signal can be tweaked from either, the clean 20Hz and will tolerate the tweak much more pleasingly than the dirty 20Hz will. The amount of THD/noise associated to the input, and subsequent THD introduced by the manipulation(s) will have an impact on how much we can push/pull the original signal and still have it rendered "cleanly" within the limits of our perception tolerances.

So, if I'm shooting @ ETTR, process to the left ... it is for the purpose of getting my shadows above the "noisy" threshold into "clean" range. Then I can more cleanly pull them back down in post (with good processing not to crush) as warranted. (As long as I don't blow my highlights in the process.). But, if I capture them below the noise threshold, I may want to crush them (or at least not amplify them) rather than amplify the noise if they are not critical information to the scene.

Is that an appropriate "crude" perspective correlation to the audio?



Apr 23, 2013 at 01:02 PM
zhangyue
Offline
• • • • •
Upload & Sell: On
p.2 #17 · p.2 #17 · Dynamic Range vs. Tonal Range


AhamB wrote:
Of course higher bit depth also creates bigger RAW files, but more bits only help if you're not yet getting the maximum information from the signal, otherwise you're just subdividing noise and creating useless extra data.


My opinion is: no, we are not getting maximum information from signal if we limit ourselves with ADC bit. especially for tonality of lower DR image and uncertainty of future media device.


I want use the example I listed above:

First, let assume there is two camera with sensors capable capture 8bit DR (so max signal is 2^8 whiter than black). one with 8 bit ADC, the other with 12bit ADC.

There is an image has DR of 4 bit. The first camera with 8 bit ADC can only capture 2^4 resolution/step of tonal level, the same image captured by 2nd camera will capture 2^8 resolution level for the same 4 bit dynamic range image.

I am not expert on how sensor itself works, for the time being, I assume it is an analog device, has fixed noise threshold due to physics limitation. So we can’t have 1/2^14 black level captured. But once the signal above the noise threshold, it is analog nature has infinity digital step or infinity tonal step at any DR level above its physics limitation.

Now use again 4bit DR image above, the 8bit ADC will have only 2^4 step information, anything between step is truncated. The truncated noise itself to your image highest signal ratio is 1/2^4. Not to mention the ratio between this quantized noise to all the signal level between 2^4 white and 2^0 white.

12 bit ADC obviously give you a lot less quantization noise here with 2^8 digitized step capture within 4 bit DR.

So, once the sensor captured analog signal, the benefit of have more bit ADC to process this signal is obvious to me. Not to mention we don’t know what future device or printer’s capability limitation. There is no reason to let less bit ADC to truncate the signal which can be used for your PP or potentially display in the future.





Apr 23, 2013 at 01:55 PM
zhangyue
Offline
• • • • •
Upload & Sell: On
p.2 #18 · p.2 #18 · Dynamic Range vs. Tonal Range


RustyBug wrote:
The audio analogy, while I'm not technically proficient in it makes some sense.

It would seem then that DR is CLEAN range. The more you can clean up the lower end, the greater the range can be extended for usefulness (say from a clean 40Hz down to clean 20Hz). The former at a clean 40Hz may have been able to produce a 20Hz, but it would have been accompanied with noise/distortion such that you really only wanted to use it for 40Hz and above as it would be unpleasant to listen to @ dirty 20Hz.

Our post processing then takes
...Show more

ETTR has two merit here:
1. like you said, you are push the signal away from noise. (analog push)
2. it essentially maximize the ADC used in sensor: so for 4 bit DR image example aboe, you remap it to 8bit tonal range, so you can PP it as desired later. but your truncated noise is down to only 1/2^8 level.

Though, I feel there is a lot more than that consider how your camera will response analog push (ETTR) or digital push, or using both. Depend on sensor design and ADC algorithm. You may not have benefit to do this kind of push, color can get weird looking. it is really a case by case.

Though, for audio, you have many concepts is not very accurate but there are off topic and long talk.

Equalizer is controlling the gain of different frequency band, for example: you amplify low end, or mid range, suppress the high frequency to suit individual's listening preference or amplify some frequency say 1892 Amati or Guarneri







Apr 23, 2013 at 02:08 PM
theSuede
Offline
• • • •
Upload & Sell: Off
p.2 #19 · p.2 #19 · Dynamic Range vs. Tonal Range


SNR does have a VERY direct relationship to tonal range. The SNR at any given intensity in the exposure gives the step length, the minimum required signal change to get a statistically secure brightness change... Large SNR means low noise strength, giving short steps = good tone resolution

SNR = log20(signal / noise)
DR = log2(saturation limit / electronic noise)

SNR at the DR lower limit = 0dB, the recorded signal is exactly as strong as the noise.
-So tonal range is a direct function of the SNR curve.
............................

The audio analogy is not quite coherent, as it is posed here by rusty it actually describes the signal conversion method rather than the signal quality.

It would be much more reasonable to model the frequency as color (since frequency = wavelength = color...)
Raw value is output signal strength to the speakers - on time-average we want to keep this constant to get a comfortable listening level. Just as an image has a comfortable or "natural" brightness level.

An equalizer would be the color profile for the camera. The closest audio signal processor to model the tone curve, the contrast and highlight/shadow curves in the raw conversion, would be a compressor that lifts the weak signals and holds back the transient peaks in strong signals.

ISO would be the preamplifier gain, if we say that 1v audio signal is 0dB or "ISO100", the 0.5v would demand ISO200, 0.25v input would be ISO400 and so on.



Apr 23, 2013 at 05:27 PM
zhangyue
Offline
• • • • •
Upload & Sell: On
p.2 #20 · p.2 #20 · Dynamic Range vs. Tonal Range


theSuede wrote:
SNR does have a VERY direct relationship to tonal range. The SNR at any given intensity in the exposure gives the step length, the minimum required signal change to get a statistically secure brightness change... Large SNR means low noise strength, giving short steps = good tone resolution



I forget about this noise level define tonal step. and I agree that minimal noise level define minimal tonal step at deep black.

but I am not quite sure that once we reach mid or high tonal range, the step still is defined by this noise level? We can amplify the signal, though, without change S to N ratio, but amplified signal can be divided into more bits/step to process in digital domain to achieve finer tonal response.

You have a very good explanation about audio and drawing a nice bridge between the two




















Apr 23, 2013 at 06:05 PM
1      
2
       3       4       end




FM Forums | Leica & Alternative Gear | Join Upload & Sell

1      
2
       3       4       end
    
 

You are not logged in. Login or Register

Username       Or Reset password



This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.