Dynamic Range

Blog Entree: 8th June 2008
Written by: Siegfried Seierlein

In a German camera magazine I was reading a review of the new Olympus E-420 plus the Sony Alpha 300. Comparing these two cameras the reviewer wrote that the dynamic range of the Sony is better than that of the Olympus and then he continues writing that the reason for that is the larger size sensor of the Sony. The review of the two products were interesting and the magazine is very proud with their lab testing ability and reported that the Sony has a DR of 9,5 and the E-420 a DR of 8,7. The same magazine also tested the G7 and reported a DR of 9,7. The magazine also tested the E-410 and reported a dynamic range of 8,3. 

While back I wrote a article about dynamic range and in the article I tried to remove some of the mystique of dynamic range. As with all my articles I gave a brief background plus easy to understand parallels or examples to help the reader better understand the significance of dynamic range. In the article I made a important parallel that said that one could safely assume that 1 bit in the digital world is equal to 1 full stop in the photographic world. This parallel is important because it will help the reader understand so many other aspects related to dynamic range.

With this assumption it is possible for us to quickly see what a camera should be capable off. For example the new 40D from Canon use 14 bit A/D (analogue to digital conversion). That means the 40D can safely manage 14 full stops dynamic range. The sensor could be a limiting factor meaning that the sensor could potentially not deliver a full 14 stops of dynamic range. Unfortunately very little information on sensors are generally available.

Today I like to take a deeper look at dynamic range and I like to give a little more information to help you better understand more aspects related to dynamic range. I like to start with a simple question. How does one fit more than 8 bits (8 stops) of dynamic range into the 8 bit JPEG file? In the article I wrote on dynamic range I said that 1 bit = 1 stop. That means one can only fit 8 full stops into a 8 bit JPEG file. Have you ever ask yourself the question, how does the typical reviewer measure a camera’s dynamic range?

Do they look at the physical size of the sensor, like the guy did with the review of the E-420 and the Alpha 300 or do they measure the dynamic range of the camera sensor? The G7 should have a larger sensor than the E-420 based on the above mentioned review, did you notice that?

What is a typical dynamic range for an average picture. A large number of articles I read on the subject agrees that the upper limit of dynamic range will typically be 16 full stops. The harsh African sun could demand a higher dynamic range of 16 to 20 full stops. What will the typical average be? It depends on the area or the continent you live in and without making it to difficult my experience is that a good average will be between 6 and 9 full stops. Most digital cameras are ”measured” between 7,5 and 10,5 full stops. The typical problem areas are the darker areas in a picture or the bright areas. What happens with these problem areas is that detail will be lost. One of the reasons is that when the limits of a camera is reached then the bright areas will turn white and the dark areas black. These white and black areas then have no picture detail and they are typically referred to as blown-outs.

The above picture is a illustration by Nikon and you can read more about it on the Nikon official web site. Nikon wrote the following on their web site:-

“14-bit A/D conversion and 16-bit image-processing pipeline

Nikon's FX-format sensor produces an incredibly wide dynamic range, with rich initial data that maintains its integrity through 14-bit A/D conversion. As professional photographers demand the smoothest tonal gradation and faithful color reproduction, no shortcut will do. So, to maintain impeccable accuracy, the D3 uses 16-bit data transmission throughout the image-processing pipeline.

Improved tonal gradation

The D3's dynamic range has broadened to such an extent that you can expect finer tonal gradation in highlight areas - even with extremely bright subject matter, such as skies, snowfall or white clothing shot under harsh sunlight. And thanks to more accurate treatment of data via 14-bit A/D conversion and a 16-bit image-processing pipeline, tone jump is negligible.”

The engineers at Nikon did a great job of clearly describing the challenge photographers typically have and they have designed and build a camera (D3 & D300) that can cope with said demands. I have not had the privilege of working with one of these great cameras but reading comments from owners its clear that Nikon has done a great job of mastering dynamic range within the constraints set by technology.

The picture on the left is with compliments from Canon and illustrate the excellent bit depth the EOS 1D achieves.

The picture on the right has more info on the 40D and again with both these examples one can see that the limitation today is 14 bit A/D conversion. By limitation I mean ideal would have been 16 bit A/D in a world that can reach as high as 16 stops dynamic range.

Canon wrote in a white paper on the 40D and said:- “The analog output signal from the imaging element is converted into a digital signal with 14 bits (16,384 colors) per channel instead of the previous 12 bits (4,096 colors). The result is digital data with finer gradation. With RAW/sRAW images recorded at 14 bits, one can use Canon DPP (Digital Photo Professional) software to process and save the image as a 16-bit TIFF image. This yields the maximum range of colors afforded by 14-bit processing. Also, because images recorded in JPEG (8 bits per color) are also generated from 14-bit RAW data, there are fewer blown highlights than before and gradation is excellent. Even in the case of highly saturated subjects or subjects with high contrast and clear differences between light and dark features, high quality color reproduction and smooth tone rendition from highlights to shadows are possible.”

I think it is important for the reader to see that both the above manufacturers refer to finer gradation as a result of using higher resolution analogue to digital conversion plus each manufacturer then also claim that the 14 bit A/D process delivers less blown highlights or alternatively said higher dynamic range.

All said and done please keep in mind that the high performance 14 bit engine can only process what is fed into the system. That means if the sensor can only deliver a dynamic range of 8 stops, then only 8 stops of dynamic range will be processed.

The above picture is from Olympus and it describes the key functions the Truepic III processor performs. The main difference with Olympus and the above manufacturers are the analogue to digital conversion. Olympus uses 12 bit A/D and Canon and Nikon 14 bit A/D. From all our discussions to date it is clear that both the Canon and the Nikon has the ability to provide finer gradation in terms of picture detail plus the Canon and the Nikon can potentially have significant better dynamic range when compared to the Olympus SLR.

It could be true that the current range of sensors are for example not yet at 12 bit level in terms of dynamic range and therefor Olympus could have correctly decided to stay at 12 Bit D/A conversion rate. What sparks the question is the fact that on average digital SLR cameras only measure a dynamic range between 8 and 10.

Conclusion.

One can conclude by asking, so what? Why discuss the subject and is it important? Will this knowledge help the average photographer to take better pictures? Folks this is important and yes this information or a good understanding thereof will help you to better understand your camera plus techniques like HDR.

Before closing the discussion I like to spend a few minutes on how does the manufacturer fit more than 8 stops of dynamic range into the 8 bit JPEG file? In the above picture the engineers at Olympus describe how they convert the signal coming from the sensor into the final 8 bit JPEG file.

Lets take an example. Lets say I take a picture and the dynamic range of that picture is 10,4. My Olympus E-410 is OK with that because we assume that the sensor is capable of reading the full dynamic range of the scene. The picture is then fed into the Truepic III picture engine via the 12 Bit A/D conversion process. Olympus then apply various techniques like tone mapping or gamma control to convert the picture into the final JPEG file. This final picture then only have a dynamic range of 8,3.

What is the reason for the dynamic range “loss”? Its a long discussion on its own and something we will discover in the future. Very brief one could say that similar to sharpening or saturation, if the camera applies to much then its at cost of picture quality. If you look at the three manufacturers discussed then you will see that the dynamic range difference on JPEG file level is not that big, at most 1,5 full stops from the lowest to the highest. On the other hand if you look at what is possible via their A/D processing engines plus the typical picture process path then its clear that all three these manufacturers are very discreet at pumping up their JPEG files.

Interesting and not something we discussed much here is the RAW files between the different manufacturers. Each of them offer the ability to display significant results when working with the RAW files. Canon for example recommend the user to convert the RAW file into 16 bit TIFF files to get maximum benefits.

Folks its wrong to say, this camera has better dynamic range than the other because their are to many variables without enough information. I think important to take notice off is the fact that manufacturers are responsible in terms of what they include in the typical 8 bit JPEG file. Remember that most reviewers use these JPEG files when reviewing cameras.

Then the next step will be to discuss High Dynamic Range (HDR) photography. In next articles I will give more detail on how a high dynamic range is compressed into the typical 8 bit JPEG file. That knowledge will then help us to better understand the typical software used to build HDR pictures.

Best

Siegfried