Luminous Landscape Forum

Equipment & Techniques => Cameras, Lenses and Shooting gear => Topic started by: ErikKaffehr on April 02, 2013, 04:38:32 PM

Title: How much DR do we need?
Post by: ErikKaffehr on April 02, 2013, 04:38:32 PM
Hi,

I analyzed a lot of images with RawDigger recently and found very few images fully utlizing the DR my cameras. Here are some examples:

This is a typical landscape image with about 8EV DR. The image is not exposed to the right, probably due to some specular reflections being taken into account when checking blinking highlights.
(http://echophoto.dnsalias.net/ekr/Articles/HowMuchDR/Lighthouse_small.jpg)
The raw histogram:
(http://echophoto.dnsalias.net/ekr/Articles/HowMuchDR/Lighthouse_RD.png)
The full mage is here: http://echophoto.dnsalias.net/ekr/Articles/HowMuchDR/Lighthouse.jpg
Here is the DNG files: http://echophoto.dnsalias.net/ekr/Articles/HowMuchDR/Lighthouse.dng

This image taken at dawn has a wider dynamic range, at least 11 stops.
(http://echophoto.dnsalias.net/ekr/Articles/HowMuchDR/Dawn_small.jpg)
(http://echophoto.dnsalias.net/ekr/Articles/HowMuchDR/Dawn_RD.jpg)
The full image is here: http://echophoto.dnsalias.net/ekr/Articles/HowMuchDR/Dawn.jpg
The DNG file is here: http://echophoto.dnsalias.net/ekr/Articles/HowMuchDR/Dawn_RD.png
There is still some noisy detail in the foreground tree:
(http://echophoto.dnsalias.net/ekr/Articles/HowMuchDR/Dawn_crop.jpg)

Finally, if we include the sun we can have infinite dynamic range, except setting sun, it will always be clipped.
(http://echophoto.dnsalias.net/ekr/Articles/HowMuchDR/WinterSun_small.jpg)
(http://echophoto.dnsalias.net/ekr/Articles/HowMuchDR/WinterSun_RD.png)
The full image is here: ]http://echophoto.dnsalias.net/ekr/Articles/HowMuchDR/WinterSun.jpg
The DNG file is here: http://echophoto.dnsalias.net/ekr/Articles/HowMuchDR/WinterSun.dng

Here is a shadow detail crop:
(http://echophoto.dnsalias.net/ekr/Articles/HowMuchDR/WinterSun_crop.jpg)

So here were some Sony Exmor sensors (Sony Alpha 99 and Sony Alpha 77). Would be great if Nikon, Canon and MF shooters also posted some high DR image samples.

Hints for evaluation:

I use log scale on Y-axis and EV scale on X-axis. I also use a cut of around 5k pixels/bin, so I count DR from maximum exposure to the position where histogram goes below 5K. This is almost arbitrary choice.

MFD will produce a different histogram, as they have 16 bit readout but around 3-4 bits of noise. So the histogram will not roll off but continue to the left edge. But channels below 5K will probably be just noise.

Best regards
Erik




Title: Re: How much DR do we need?
Post by: IanBrowne on April 08, 2013, 12:46:49 AM
interesting; 200+++ hits and no reply  ;)

Personally Erik, I don't go into too much pixel peeping stuff. We can get bogged down by worrying about too many of the finer details of digital photography. I don't think I have ever thought about DR of any my digital cameras and in film days I didn't worry too much about this colour channel of that film and that colour channel of this film. I'm not that good of a photographer to notice the difference  :'(

So I guess I cannot help you. I just take photos using RAW format so I can fix my bad photography  :-[ on the computer. Apart from cropping I cant fix composition so that's what I worry about the most.  :)
Title: Re: How much DR do we need?
Post by: bill t. on April 08, 2013, 01:12:28 AM
Nice pictures Erik!

I think we've conditioned ourselves to avert our eyes from pictures that need lots of dynamic range.

Here are a couple of daring Coolpix A shots that really do make use of a heck of a lot of dynamic range, and to good effect.  Even these lowly jpg's can reveal a lot of useful detail in the inky shadows in PS or LR.

Look at the "original" sized DSC_0045 and DSC_0049 here...

http://www.dpreview.com/galleries/reviewsamples/albums/nikon-coolpix-a-preview-samples

You can push 'em pretty hard.  DSC_0128 has a certain tonal something, as well.

I'm happy to have all the dynamic range I can get.  I just love lots of glowering shadow detail combined with sparkly highlights and it's it comforting to know I can re-think an entire image long after I shoot it.  And I'm sick to death of HDR.  D800, yeah baby!

Title: Re: How much DR do we need?
Post by: Rob C on April 08, 2013, 03:37:59 AM
interesting; 200+++ hits and no reply  ;)




It's simply too removed from the art of photography. Were one interested in the why/how of digital camera workings, then yes, I suppose it means something, but it is anything but essential to the art of snapping a fine snap: that you do by instinct and, with digi, the easy peep at the rear screen. I had little idea of how Kodachrome was made and even less of its processing - it made no difference at all to my ability to use the film. It's just a subject about something quite else than pictures. Nothing wrong with that at all, just probably why not many people are interested enough to write.

IMO.

Rob C
Title: Re: How much DR do we need?
Post by: HarperPhotos on April 08, 2013, 03:49:54 AM
Hello,

This reminds me back in the mid 70ís my brother and I where getting into hifi equipment. We even tried different stylist and would listen to the same song on the record to listen to the differences the stylist would reproduce the sound.

After some months I came to the resolution I wasnít actually listening to the song for what is was but listing to the sound it made. That was the end of my foray into the world of hifi.

Do ya get what Iím saying.

Cheers

Simon
Title: Re: How much DR do we need?
Post by: Petrus on April 08, 2013, 03:50:27 AM
We need infinite DR, then cameras would not need exposure meters and all exposing could be done just for the required DOF and shutter speed.

In the audio side there already are analog to digital converters with 26 bit (stop in photo language) resolution, which require no level setting with mic preamplifiers.
Title: Re: How much DR do we need?
Post by: hjulenissen on April 08, 2013, 04:02:47 AM
In the audio side there already are analog to digital converters with 26 bit (stop in photo language) resolution, which require no level setting with mic preamplifiers.
I have never heard of more than 20-21 effective bits. Do you have any references?

-h
Title: Re: How much DR do we need?
Post by: hjulenissen on April 08, 2013, 04:06:34 AM
High DR sensors does have merits for shooting low-DR scenes without having to set exposure, and for "ISO-less" shooting where you set time/aperture from artistic vision and to avoid clipping but do not touch ISO.

For for capturing a single high-DR scene, you probably don't need much more sensor DR than the lense DR. So what is the lense DR? Can it be measured/analyzed in any sensible, repeatable fashion? Is the lense DR any different for a dark scene with a single, bright spot (moon/sun/...) than if the "brightness" is distributed over a larger area?

-h
Title: Re: How much DR do we need?
Post by: Petrus on April 08, 2013, 07:59:17 AM
I have never heard of more than 20-21 effective bits. Do you have any references?

Stagetec Truematch standalone converter, 28 bit conversion with 26.5 bit resolution.

http://www.stagetec.com/en/audio-technology-products/standalone-converter.html

If the end result is any better than using conventional 24 bit converters and adjustable gain mic-pres is another matter, though. I am happy with my Prism Orpheus and Sound Devices 722, they both give about 21 bit resolution. Recording venues limit the achievable clean resolution to around 12-14 bits anyway (ambient noise), unless the aim is to accurately record the air conditioning and traffic, also.
Title: Re: How much DR do we need?
Post by: ErikKaffehr on April 08, 2013, 08:27:23 AM
Hi,

We have only around 60 thousands of electrons per pixel, and we can not read out with less then 1 electron precision so we would do fine with 16 bits.

BR
Erik

Stagetec Truematch standalone converter, 28 bit conversion with 26.5 bit resolution.

http://www.stagetec.com/en/audio-technology-products/standalone-converter.html

If the end result is any better than using conventional 24 bit converters and adjustable gain mic-pres is another matter, though. I am happy with my Prism Orpheus and Sound Devices 722, they both give about 21 bit resolution. Recording venues limit the achievable clean resolution to around 12-14 bits anyway (ambient noise), unless the aim is to accurately record the air conditioning and traffic, also.
Title: Re: How much DR do we need?
Post by: PeterAit on April 08, 2013, 09:01:31 AM

It's simply too removed from the art of photography. Were one interested in the why/how of digital camera workings, then yes, I suppose it means something, but it is anything but essential to the art of snapping a fine snap: that you do by instinct and, with digi, the easy peep at the rear screen. I had little idea of how Kodachrome was made and even less of its processing - it made no difference at all to my ability to use the film. It's just a subject about something quite else than pictures. Nothing wrong with that at all, just probably why not many people are interested enough to write.

IMO.

Rob C

I agree. It makes me think of a painter spending lots of time doing chemical analyses on his paints and brushes. Nothing whatsoever to do with the art.
Title: Re: How much DR do we need?
Post by: Slobodan Blagojevic on April 08, 2013, 10:14:12 AM
... It makes me think of a painter spending lots of time doing chemical analyses on his paints and brushes...

Oh, but they did.
Title: Re: How much DR do we need?
Post by: fredjeang2 on April 08, 2013, 12:25:13 PM
Oh, but they did.

Present me those painters !  ;) Most of the big bosses I have been working in still with hardly know their camera menus and rely
on assistants for that.

In fact, those graphics and science or pseudo-science were irritating me in the past. Now I have a different view on them.
It is simply due to the fact that there isn't one photographer's profile as ideal but many.

There are people who come from an engineering or technician background, that don't necesarly have to leave with photography
and there are more interested on the engineering side.
For them, graphics talk to them, 1 point DR talk to them, the life of the photons talk to them more than the life of the model.
It's chinese lenguaje for me but for others it's a of great interest.

Photography, as a form of art that is created from complex engineered machines,
will always attract people with clear tech orientation because it has this component inside. And their views and interests are as valid as others.

Title: Re: How much DR do we need? ( A general comment by the OP)
Post by: ErikKaffehr on April 08, 2013, 12:56:41 PM
Hi,

The reason I posted my observations was mostly that I feel DR is overemphasized. I have very seldom encountered subjects with very wide DR. So what I wanted was to check the DR of different subjects and also point a bit towards Rawdigger to check out raw files. In principle, I think it is good to know your tools.

Raw converters can do a lot of tricks. For instance, Rawdigger can present an accurate histogram from raw, neither Lightroom or in camera histograms are really accurate. In camera histograms are said to be based on a JPEG interpretation of the image and Lightroom does highlight recovery behind the scenes.

This posting was posted on the "Cameras, Lenses and Shooting gear" forum, so I think it's quite OK to start a discussion like this.

Best regards
Erik
Title: Re: How much DR do we need?
Post by: Slobodan Blagojevic on April 08, 2013, 12:57:45 PM
Present me those painters !...

Almost all classic painters, from Leonardo onward (and probably even before him), were very particular about the quality and longevity of their paints. They were known to experiment, mix and match, trying to find a perfect formula. Looks like most of them succeeded, although some artwork faded faster than others.
Title: Re: How much DR do we need?
Post by: Slobodan Blagojevic on April 08, 2013, 01:10:01 PM
The reason I posted my observations...

Hi Erik,

To tell you the truth, I am still struggling to understand the reason behind this thread. Are you suggesting that, in real life, scenes do not have more than 11 stops? In which case the 3-stop advantage of D800 over 5DIII would not matter? You showed one such example, but then you said "There is still some noisy detail" in the shadows. Well, if it is noisy, than additional DR would have help, no?
Title: Re: How much DR do we need?
Post by: fredjeang2 on April 08, 2013, 01:19:35 PM
Almost all classic painters, from Leonardo onward (and probably even before him), were very particular about the quality and longevity of their paints. They were known to experiment, mix and match, trying to find a perfect formula. Looks like most of them succeeded, although some artwork faded faster than others.
>:(
No fun.
I thought you would have sent me the phone number of some lady painters in high-heels in Chicago district in search of the technical remedy to longevity. Booo.
Title: Re: How much DR do we need?
Post by: ErikKaffehr on April 08, 2013, 01:42:36 PM
Hi Slobodan,

It started that I bought a new camera and I was interested in seeing the purported advantage in DR. I knew the advantage was there but I could not find any subject showing it. Also, have seen very few images where I felt camera DR was a problem. So I started to check out many images and found that very few of my images had a dynamic range exceeding nine stops.

Also, the cameras I have used since 2006 had quiet decent DR, but also images from other cameras I have seen seldom had issues with DR. Another issue is that a very wide dynamic range needs a lot of tone mapping to be pleasant on screen.

Take this example from Gotland, Sweden, in this case I used HDR tone mapping and some other things on a single image;
(http://echophoto.dnsalias.net/ekr/Articles/HowMuchDR/Gotland1.png)

Now, a straight import of the same image in Lightroom looks like this:

(http://echophoto.dnsalias.net/ekr/Articles/HowMuchDR/Gotland2.png)

I don't think the second image is very attractive, but I'll admit that the first one is "over the top".

If I check out the DR in Rawdigger we can see that the dynamic range is around 8-9EV.
This is possibly even more obvious if I use a linear scale on the vertical axis of the histogram, see below.

(http://echophoto.dnsalias.net/ekr/Articles/HowMuchDR/Gotland3.png)

So you need no DR champ to handle it, just learn to expose to the right and how to process your images and chances are you can be happy with the gear you have!

Best regards
Erik

Hi Erik,

To tell you the truth, I am still struggling to understand the reason behind this thread. Are you suggesting that, in real life, scenes do not have more than 11 stops? In which case the 3-stop advantage of D800 over 5DIII would not matter? You showed one such example, but then you said "There is still some noisy detail" in the shadows. Well, if it is noisy, than additional DR would have help, no?
Title: Re: How much DR do we need?
Post by: Slobodan Blagojevic on April 08, 2013, 01:57:17 PM
Erik,

In the OP, you presented one image with "at least 11-stop DR." That's approximately 5DIII range. Now, if someone shot the same scene with a D800, using the same parameters, the resulting initial image would be the same. However, the ability to pull cleaner details from that tree in the foreground should be different.

Thus, if the scene has 11-stops in itself, and we do not want to change the apparent tonal relationships within it (like in your example, where leaving things in the shade as near silhouettes works well visually) , than an 11-stop DR camera would suffice. If, however, we do need/want to pull out details in the shade, than a higher DR camera would be preferred, right?

I know that, as a good friend, you are just trying to dissuade me from the financial disaster that switching to Nikon would be for me, but your examples are not helping ;D
Title: Re: How much DR do we need?
Post by: ErikKaffehr on April 08, 2013, 02:09:20 PM
Slobodan,

You are actually right on all three issues ;-)

Just so you know: My best friend also owns Canon, had 20D and 40D and has just recently upgraded to 5DIII. I never felt that his tools were lacking.

Best regards
Erik




I know that, as a good friend, you are just trying to dissuade me from the financial disaster that switching to Nikon would be for me, but your examples are not helping ;D
Title: Re: How much DR do we need?
Post by: TMARK on April 08, 2013, 02:10:07 PM
Almost all classic painters, from Leonardo onward (and probably even before him), were very particular about the quality and longevity of their paints. They were known to experiment, mix and match, trying to find a perfect formula. Looks like most of them succeeded, although some artwork faded faster than others.

True, but this was before manufactured paints.  They had to make their own paint.  I know some painters who still make their own paint, but its more of a fetish than anything else.  Liquitex, Utrecht, etc.  You have to mix your color but you know it will last.
Title: Re: How much DR do we need?
Post by: NikoJorj on April 11, 2013, 04:28:13 AM
Finally, if we include the sun we can have infinite dynamic range, except setting sun, it will always be clipped.
As said elsewhere, it's just in that case that I need more DR than my canon can give me : sunset or sunrise.
You have to be careful not to blow anything in the sun or that lovely hue will be gone (F! jpeg histogram!) and the foreground is solely illuminated by the dawn/dusk part of the sky, yet our eyes can see some detail and hues there, often complimentary to those in the skies.
I got examples in the 12EV territory from what rawdigger says ; alas, banding starts at 9-10EV and makes not so pretty shadows (see example, as it's a pano it barely shows at this size but it's only 1500px wide, it couldn't print well).


As for lens flare, I would also think it plays a part in the equation, but I couldn't even qualify it.
Title: Re: How much DR do we need?
Post by: PhotoEcosse on April 11, 2013, 05:26:03 AM
To answer the original question

Quote
How much DR do we need?

In my experience of landscape photography, the most that I have had to add to the 14.3 EV provided by the Nikon D800 sensor is 8 EV in shots which do not include the direct sun in the image. This has been achieved by taking 9 exposures at 1-stop intervals and merging in HDR Efex Pro2.

So I guess my answer to the question is 22 EV DR.
Title: Re: How much DR do we need?
Post by: BartvanderWolf on April 11, 2013, 07:27:17 AM
As for lens flare, I would also think it plays a part in the equation, but I couldn't even qualify it.

Hi,

Correct. Lens glare will seriously limit the effective DR that can be recorded on the Sensor (which has a much larger engineering DR). Each air/lens surface transition will add some stray light that will add to the shadows more as a percentage that to the highlights. A glare of 1% will potentially reduce the dynamic range to 6.64 stops (locally if you are lucky, globally if not so lucky).

The use of a properly dimensioned (= uncomfortable in use, because deep) lens hood can help a bit, but some internal reflections will occur in any lens, due to the air/lens surface transitions and internal reflections. Quality coatings, and mechanical lens design will help, but we cannot influence that. We can try to use better shielding of non-image forming light to somewhat improve the situation.

Cheers,
Bart
Title: Re: How much DR do we need?
Post by: jrsforums on April 11, 2013, 09:47:54 AM
To answer the original question

In my experience of landscape photography, the most that I have had to add to the 14.3 EV provided by the Nikon D800 sensor is 8 EV in shots which do not include the direct sun in the image. This has been achieved by taking 9 exposures at 1-stop intervals and merging in HDR Efex Pro2.

So I guess my answer to the question is 22 EV DR.

I think this would only be the case if your base shot had captured the highest significant highlight at just below clipping in RAW....and then each add'l shot was +1EV, until the shadows were all captured on the left.

John
Title: Re: How much DR do we need?
Post by: MarkL on April 12, 2013, 06:36:59 AM
Other than the resolution, this is the biggest difference I noticed moving from D700 to D800. This really has made a difference shooting in hard contrasty light I like and issues like having blown skies on overcast days. For landscape work it means no moe (or less) eposure blending or grad filters and with the sun in the frame you can never really have enough.  Iíd rather have more than less even if I donít Ďneedí it all that often because it means there is more margin for error.
Title: Re: How much DR do we need?
Post by: PhotoEcosse on April 13, 2013, 03:08:54 PM
I think this would only be the case if your base shot had captured the highest significant highlight at just below clipping in RAW....and then each add'l shot was +1EV, until the shadows were all captured on the left.

John

Not quite like that, John.

If we take the "base" shot as the one that the camera's meter suggests was correctly exposed, then the highlights were hopelessly blown and there was no detail at all in the shadows. +4 EV was necessary to kill all the highlight clipping while -4 EV was required to reveal the shadow detail.
Title: Re: How much DR do we need?
Post by: jrsforums on April 13, 2013, 03:41:22 PM
Not quite like that, John.

If we take the "base" shot as the one that the camera's meter suggests was correctly exposed, then the highlights were hopelessly blown and there was no detail at all in the shadows. +4 EV was necessary to kill all the highlight clipping while -4 EV was required to reveal the shadow detail.

When you have the camera meter, you have no idea what "zone" of the scene it is setting as middle grey.  You are therefore shooting blind.  You have no idea if -4EV will capture the significant highlights...or be much too much.  Same with +4EV on the low end.  So your response on the number of stops need, overall, hold no basis of fact. 

Your method could work if you analysed each RAW image in something like RAWdigger to see what EV it clipped at and where the lows where capyured.  Unfortunately, if you have specular highlights or the sun, which you had no expectation in capturing and were going to let blow out, RAWdigger cannot distinguish between significant and insignificant.

You need to calibrate your camera to know the clipping point in RAW.  I have found, on my 5D3, it is ~+3.5 stops, if I spot meter the brightest significant highlight.  Once you have that, you can increase the EV until you see on the histogram that you have captured the darkest shadow.  (I will actually overdo the shadows to ensure that I have add'l detail). 

I find this method provide two possible benefits.  First, it will maximize the DR captured, which often will allow one image to sufice....whereas the camera matrix/evaluative or centerweighted metering will often not.  Second, it can allow you to minimize the number of shots required for blending/HDR.

John
Title: Re: How much DR do we need? ( A general comment by the OP)
Post by: Greg D on April 14, 2013, 05:11:30 PM
Hi,

The reason I posted my observations was mostly that I feel DR is overemphasized. I have very seldom encountered subjects with very wide DR.


I encounter this problem all the time, and I don't think my photography is that unusual.  I often shoot scenes with whitewater in deep forest, and such scenes will often exceed the DR of pretty much any camera, and in such cases, you simply can't sacrifice the highlights.  Exposure stacking may not be an option because of wind, etc.  Sometimes you just have to work around it and accept silhouetted shadows, but it's nice to not be backed into that corner.  I shot with Canon crop cameras for most of my shooting years, then switched to Nikon (D7000), which had 3 or so stops more of DR according to most analyses.  This reduced the incidence of exceeded DR a little, but the most important difference was what could be done with the underexposed shadows - pretty much nothing on Canon files, without creating ugly noise, but 2 or 3 stops worth of recovery from the Nikon.  I know this is common knowledge, and as I said, I don't think my photography is unusual, so why wouldn't any outdoor photog want all the DR available?
Title: Re: How much DR do we need?
Post by: Ellis Vener on April 14, 2013, 08:41:33 PM
The only correct answer is: As much as possible.