Testing dynamic range of a camera ...

[Turgay]

As my learning process goes on about digital photography, I tested the dynamic range of my camera in an amateurish way.

Firstly, I took a shot of a plain surface (with spot metering, neutral picture style, RAW; without auto lighting optimizer) and checked the luminance histogram for the peak at the middle tone. After that, for the same setup, I measured where the clippings occur. They are starting at +2 1/3 for highlights and at -4 2/3 for shadows.

So, it seems that my camera has 7 stops dynamic range in RAW.

First of all, I would like to ask ... is this method for measuring the dynamic range is true?

And two more:

1. Why do we have more room at the dark side? Why the dynamic range is not symmetrical around middle?
2. If I encounter with a scene which has a dynamic range more than 7 stops and I don't want to loss any single pixel, I should choose HDR. Does this is the exact meaning of HDR?

Thanks a lot ...

[JonathanC]

It depends how you measured the clipping, for a start. Also you would need to take account of the colour of the plain surface.
It could be that the metering was off, either by design or by some error.
Could you elaborate on the test method so that the statements are a little more scientific please?
For example, did you open the RAW file in Photoshop and change the exposure in post processing or bracket the shot?
How did you know that clipping was occuring, warnings in Photoshop for example might be indications that you are close to the clipping point or way past it. I don't know how that works.
I'm interested to know what you did.

[Turgay]

Hi Jonathan,

I used gray card but it doesn't matter what the color of surface. If you choose a single color, plain surface, all the pixels are going to the middle point in the luminosity histogram.

My first shot was a standard exposure with a gray card for placing all the pixels at the middle of the luminosity histogram. Then, without changing my setup, I shot with the indicated exposure compensations above. At +2 1/3, the camera said there is a highlight clipping (highlight alert function, can be seen in the camera) and at -4 2/3 it gives shadow clipping. There is no shadow alert on my camera, but can be seen on luminosity histogram, all the pixels leaning towards left.

I don't have professional equipment for measuring the dynamic range of my camera and I'm also not a specialist on this topic. I'm just studying exposure now and trying to understand dynamic range. It was just a simple test for this respect

Is it too much to have 7 stops dynamic range for a camera?

[herbert]

Hi Turgay,

If you are only using the clipping highlight on your camera then you are just measuring the dynamic range of the in camera Jpeg tone curve. The jpeg format is non linear. This is why there is less room at the top end than at the bottom end. Effectively a curve is added to the linear data to boost the differences between dark tones where our eyes are more sensitive to changes. This curve flattens the highlights and also shifts the midpoint to be brighter. This is why you see less room in your exposure above the midpoint than below.

As to how to measure the dynamic range I do not know hiw i would do it in my own house. Take a look at the pages of DxO mark. They have a measure of dynamic range for more cameras.

http://www.dxomark.com/index.php/en/...tested-sensors

This is extracted by sensorgen for all the cameras:

http://www.sensorgen.info/

The create a top-trumps table to compare sensors. This should help you out.

Alex

[Glenn NK]

While the histogram is a very valuable tool to help in exposing to the right (ETTR), it is not completely accurate.

As Alex pointed out, the image on the LCD is a JPEG. The camera can't use the RAW file, because a RAW file has no colour information that an LCD could interpret - it can only project a JPEG.

There is more headroom in a RAW file than there is in a JPEG file (one good reason to shoot RAW, not JPEG), and this headroom can be recovered in a RAW converter. By this I mean that what appears to be clipped in a JPEG, may be recoverable and useful in the RAW file. I use the LCD extensively in adjusting exposure, and I let the RGB histogram clip a bit on the LCD, BUT in Lightroom, I can recover the clipped areas (to an extent).

I would suggest that using the camera's JPEG/LCD to measure DR will accomplish nothing - except perhaps frustration and misleading results.

There are two sources I know of where this information can be obtained:
1) Reviews of a camera by DPP,
2) Reviews by DxO.

Glenn

[Colin Southern]

Hi Turgay,

The 60D has a dynamic range of 11.5EV.

Hope this helps,

[PBelarge]

Colin
I am no expert, but from other reading I thought (I tell all my students if they say, "I thought", it means they do not know) that most cameras had somewhere in the range of 8 stops of DR.

In regards to the LCD on the camera, can not one use the histogram as a guide, seeing the clipping but looking at the LCD to try and gauge if the clipping is actual loss of information or just close to loosing the information.

[herbert]

Hi Pierre,

The dynamic range of a camera is the difference between the maximum and minimum light readings it can measure. The maximum is a function of the maximum number of photons it can capture per pixel. The minimum is a function of the smallest amount of photons it can capture and differentiate from random noise on the sensor. These two things vary by sensor and there is a quite a bit of variety between cameras.

However if you have a standard 8-bit monitor then the best dynamic range you can display is 8 stops. One stop per bit. This is if your monitor is awesome and your eyes are awesome enough to see the difference between 0 and 1 and each step all the way up to 255. Some older monitors had 6-bit displays. Before that they may have been even worse at 4-bit.

What this fails to point out is that you can take an image with 15 stops of dynamic range and squash it into 8 stops. This is High Dynamic Range (HDR) Imaging. Often the image looks a bit flat but you can see right into the shadows. It is like shooting on expensive cinema film where you can get a strange washed out (and arguably in the case of film beautiful) image unlike anything on normal TV. To overcome the flatness you can use local contrast enhancement to improve the separation of objects in close proximity. Or if you overdo it you can make a very horrible looking image.

Cameras with more than 8 stops of dynamic range usually apply a simple tone curve to the image before creating a JPEG, squashing the highlights and shadows a bit to fit the data more closely into 8 stops. This is why if you have the raw data you can redevelop the image and save a bit more highlight or shadow detail that the default tone curve lost.

Alex

[Turgay]

Thank you all.

I have read a little today about the topics from which Helbert and Glenn mentioned and decided that I need to have a sound and comprehensive background on these technology related subjects.

I found this link, 123 of digital imaging. Other than this, I would be very happied, if there are some other book advises about this subject.

[PRSearls]

Here is a process using the Sekonic L-758DR meter and a Sekonic calibration target (I rented mine) to calibrate your camera's sensor. It works; I calibrated all my camera bodies. It was easy, accurate and the software stores and displays the dynamic range and (settable) clipping points in the meter so I can easily determine an optimum exposure (or exposures in HDR).

http://www.youtube.com/watch?v=_7SZ58CugpY

[Turgay]

I'm trying to understand the RAW creations at the sensor and I have a difficulty with the file sizes. I have a Canon EOS 60D, it has a 5184 x 3456 = 17.915.904 pixels on its sensor and it has 14 bit AD converter, so at the raw stage, a file should be a 29,9 mb file according to my calculations, but I get ~27 mb RAW files from the camera. I would like to ask this difference firstly.

And, secondly, RAW file format is a single channel, just holding brightness levels for each pixel (in the case of 60D, we have 14 bit depth). So, I think that demosaicing is not a part of a RAW file and I would like to ask when we get the 3-channels image, just after demosaicing?

I'm sorry for these simple questions ... Indeed, I was just reading Strobist for flash exposure ... and now where I am

[Turgay]

Thanks Paul. Very helpful video.

I have never used a handheld lightmeter before, but after watching the video, I think it may be the correct time for me to invest in such a functional device.

[Colin Southern]

I'm trying to understand the RAW creations at the sensor and I have a difficulty with the file sizes. I have a Canon EOS 60D, it has a 5184 x 3456 = 17.915.904 pixels on its sensor and it has 14 bit AD converter, so at the raw stage, a file should be a 29,9 mb file according to my calculations, but I get ~27 mb RAW files from the camera. I would like to ask this difference firstly.

5184 x 3456 is 17,915,904 pixels - and with 2 bytes being needed per pixel then you'd need 35,831,808 butes to store the data from the sensor (plus a bit more for the metadata) IN THEORY. In practice, the data is losslessly compressed. On a side note, the higher the ISO the larger the file because the extra noise can't be compressed as well.

And, secondly, RAW file format is a single channel, just holding brightness levels for each pixel (in the case of 60D, we have 14 bit depth). So, I think that demosaicing is not a part of a RAW file and I would like to ask when we get the 3-channels image, just after demosaicing?

Yeah - sorta. Each photosite has a red, green, or blue filter over it - so after a capture you're left with 3 monochrome images (notice I say "monochrome" and not black and white). One of the first thing a RAW converter does is combine these 3 monochromatic images into a single full colour image.

Hope this helps.

[Donald]

5184 x 3456 is 17,915,904 pixels - and with 2 bytes being needed per pixel then you'd need 35,831,808 butes to store the data from the sensor (plus a bit more for the metadata) IN THEORY. In practice, the data is losslessly compressed. On a side note, the higher the ISO the larger the file because the extra noise can't be compressed as well.

That's FAR too much information for someone who's 75% the way through a bottle of Guigal's 2003 Hermitage (Northern Rhone red wine). I really need to get more informed about and interested in, the technical side of photography.

[Colin Southern]

Colin
I am no expert, but from other reading I thought (I tell all my students if they say, "I thought", it means they do not know) that most cameras had somewhere in the range of 8 stops of DR.

I don't know of any modern DSLR that has a DR as low as 8. If I take a handful of modern cameras - say Canon 1100D - 60D - and 1Ds3 - we get DRs of 10.9, 11.5, and 12.0 (http://www.dxomark.com/index.php/Cam...(brand3)/Canon)

In regards to the LCD on the camera, can not one use the histogram as a guide, seeing the clipping but looking at the LCD to try and gauge if the clipping is actual loss of information or just close to loosing the information.

Yes and no. In theory it's not accurate because the histogram is based on an in-camera generated JPEG and not the RAW file, but in practice, if you push a sensor too close to the saturation point (keeping in mind that one channel will saturate before the others) then you can get into a non-linear response region that can in turn induce a weird colour cast thats impossible to null out because it's not consistant. So personally, I find that using the highlight alert feature when trying to optimise landscape actually works pretty well (even though it's based on a JPEG) because it gives me just enough safety margin) (keeping in mind also that highlight alerts generally aren't triggered if only 1 channel is blown - and the best landscape - shot at the beginning or end of the day - generally has a strong red colourcast due to the warmth of the light - and thus the tendency is to blow the red channel first).

Going back to DR for a sec -- I think a lot of folks get themselves confused about this because they forget that a photo can only display a DR of about 4 - and a monitor can only display a DR of about 6 - so they confuse this with the DR that's been captured. As a good example I'm forever seeing HDR images created by bracketing shots when all they needed to do was use a healthy does of fill light control to compress some of the detail captured by the camera into the range that's visible on their monitors. It's a bit like infrared radiation - just because we can't see it with our eyes, doesn't mean that its not there.

Hope this helps

PS: What's your first name?

[PBelarge]

Colin
This is exactly why I try to find forums/reading that will help me to understand the complex issues involved with this industry. I can say that what I have read on this thread has been somewhat easy to understand (well explained), and very interesting. Like Donald my technical skills can use some help, but over time I am sure they will improve.

[Colin Southern]

Colin
This is exactly why I try to find forums/reading that will help me to understand the complex issues involved with this industry. I can say that what I have read on this thread has been somewhat easy to understand (well explained), and very interesting. Like Donald my technical skills can use some help, but over time I am sure they will improve.

Glad you find us useful <insert first name here>

To be honest, there are a few topics in photography where there is a lot of mis-information floating around (with DR being one of them) (and sharpening, and HDR, and high-ISO "noise", to name just a few) - so we like to try to dispell the myths and set the record straight whenever we can.

[Donald]

Glad you find us useful <insert first name here>

To be honest, there are a few topics in photography where there is a lot of mis-information floating around (with DR being one of them) (and sharpening, and HDR, and high-ISO "noise", to name just a few) - so we like to try to dispell the myths and set the record straight whenever we can.

Pierre

Colin has signatures switched off (this is an option available on the forum), so he doesn't see your real name in your signature.

If you wish you can go to Edit Profile and enter your proper name under 'Real Name'. Then it will appear underneath your Username in all your posts.

[Colin Southern]

Pierre

Colin has signature's switched off (this is an option available on the forum), so he doesn't see your real name in your signature.

If you wish you can go to Edit Profile and enter your proper name under 'Real Name'. Then it will appear underneath your Username in all your posts.

Well spotted Donald

Pierre - don't worry about adding your name to the real name field - it's magically appeared all by itself all of a sudden

[PBelarge]

Pierre - don't worry about adding your name to the real name field - it's magically appeared all by itself all of a sudden

The magic of the internet...