DR has improved by 0.6 stops in ten years

[Karim Ghantous]

I am using DxO measurements for this comparison. Keep in mind that DR does not equal image fidelity. The Sony A9 (2017) has poor fidelity, yet a pretty good DR of 13.3 stops. That's over a stop more than the Leica M9 (2009), which has superior fidelity.

A Hasselblad X1-D 50c, with a 50Mpx CMOS 44x33 sensor has a measured DR of 14.8 stops. Now, that's a big sensor with nice, big photosites. It's not the newest sensor in the world, but it's very, very good. The Sony A7RV, with a smaller, more modern sensor, with slightly more resolution, has exactly the same DR.

The Sony A1, with the same resolution as the Hasselblad, has almost the same DR, at 14.5 stops. Not bad given that it was built for speed.

The A7, released in 2013, with 24Mpx, has 14.2 stops. This camera was definitely not built for speed. In 2013, this DR figure was very good indeed. It still is, in absolute terms, of course. That's more than a lot of medium format CCDs! But not much seems to have changed since then. These days, we have backlit sensors, better technology, and all the rest of it. And we gained only 0.6 stops if we compare across Sony cameras.

Just for kicks and giggles, here is a quick comparison between three DSLRs:

Nikon D610, 2013, 14.4 stops
Pentax K5 (APS-C), 2010, 14.1 stops
Canon EOS 6D, 2012, 12.1 stops

I expected more progress in the DR department, especially from the medium format sensors. This is not to say that things won't improve. Oh, they absolutely will. High end cinema cameras by ARRI and Red have more DR than 14.8 stops, but they are quite expensive.

Kodak Vision3 film has up to 15 stops in 35mm format. So it's not like film has made much progress either, although AFAIK you gain a stop with every doubling of surface area.

What are your thoughts?

[hjoseph7]

The Sony A7RV, with a smaller, more modern sensor, with slightly more resolution, has exactly the same DR.

I'm glad that camera manufacturer's are finally focusing on the DR issues rather than megapixel, Video, FPS issues. I have a Canon 6D that is notoriously weak when it comes to the Dynamic Range. Even some of my older Nikon and Pentax cameras record a better dynamic range than my Canon 6D. Good dynamic range is especially important in high-contrast scenes when the sky is much brighter than the foreground, or when taking indoor shots where you don't want to blow out the windows. Of course you can compensate with flash, but that is not always possible, or convenient. 

I'm really surprised that the Sony  A7RV has a dynamic range of 14.8 that is incredible for a small sensor camera !  A few years ago 12.3 DR was the max on some of the most expensive cameras. 

[rodeo_joe1]

The 'dynamic range' - AKA brightness ratio - that can be captured in real life has almost nothing to do with the sensor or electronics of a digital camera, but rather its mechanical construction. 

It's all about internal reflection of light from the dark chamber of the camera, including the rear of the lens and any additional 'extension tube' that the lens might incorporate. 

Capturing a 14 stop 'DR' requires a brightness ratio of just over 16,000:1 to actually hit the sensor. And in turn that depends on no more than 0.00625% of the highlight brightness reflecting back into the shadow areas of the picture. 

Now the best matte black paints or flocking materials in common use have a reflectivity of about 1%, integrated over a 180 degree solid angle. Worse than that at small angles of incidence. While the reflectivity of a digital sensor is probably in the region of 20%.

So, in order to get only 0.00625% (or less) of the highlights of a scene illuminating the shadow areas, you're effectively asking two surfaces with an aggregate reflectivity of around 0.2% to magically attenuate the highlights by a further factor of 30!? That's a big ask.

OK; that's a worst case scenario where the shadow area is a tiny percentage of an otherwise bright scene. Which begs the question - "What is an 'average' ratio of highlight to deepest shadow area?" 10%? 20%?

Even if it's only 5% that still leaves about 50% of the reflected light bouncing around between camera body, lens and sensor to somehow get entirely lost. 

In conclusion; achieving a 'dynamic range' of more than 14 stops looks like pretty much a wild pipe-dream for the majority of real world scenes. Unless you want to stick to subjects that are nearly all pitch black with only a tiny proportion of highlight area. Maybe the moon on a good dark night and shot with a 24mm lens? 

 

 

Edited December 5, 2023 by rodeo_joe1

[ilkka_nissila]

A commonly photographed theme in landscape photography involves a wide angle image of a scene which includes foreground features such as rocks and the sun just tingling to rise above the horizon. In this case the sun will create some ghosting, depending on the quality of the lens, but large areas of the frame are mostly unaffected by this, and the photographer can adjust the position of those ghosts in the frame by moving the camera relative to the subject matter and the sun. In this case some photographers often expand the dynamic range by bracketing the exposure by many stops and then paint masks on a stack of layers to reveal from each exposure the right part of the frame, to get to a natural-looking image with foreground details and the sun not blowing out outside of a small central core. Having a modern lens does help, but often the flare or ghosting can be used as an effect to create a feeling. It doesn't mean there is no detail revealed in the shadows; there can be. There are photographers who have made this process an art and their livelihood. But anyway, yes, the brightest part of the scene could only be less than 0.01% of the image area. People tend to love this kind of images.

[macwest]

On 12/5/2023 at 5:20 AM, rodeo_joe1 said:

The 'dynamic range' - AKA brightness ratio - that can be captured in real life has almost nothing to do with the sensor or electronics of a digital camera, but rather its mechanical construction. 

It's all about internal reflection of light from the dark chamber of the camera, including the rear of the lens and any additional 'extension tube' that the lens might incorporate. 

Capturing a 14 stop 'DR' requires a brightness ratio of just over 16,000:1 to actually hit the sensor. And in turn that depends on no more than 0.00625% of the highlight brightness reflecting back into the shadow areas of the picture. 

Now the best matte black paints or flocking materials in common use have a reflectivity of about 1%, integrated over a 180 degree solid angle. Worse than that at small angles of incidence. While the reflectivity of a digital sensor is probably in the region of 20%.

So, in order to get only 0.00625% (or less) of the highlights of a scene illuminating the shadow areas, you're effectively asking two surfaces with an aggregate reflectivity of around 0.2% to magically attenuate the highlights by a further factor of 30!? That's a big ask.

OK; that's a worst case scenario where the shadow area is a tiny percentage of an otherwise bright scene. Which begs the question - "What is an 'average' ratio of highlight to deepest shadow area?" 10%? 20%?

Even if it's only 5% that still leaves about 50% of the reflected light bouncing around between camera body, lens and sensor to somehow get entirely lost. 

In conclusion; achieving a 'dynamic range' of more than 14 stops looks like pretty much a wild pipe-dream for the majority of real world scenes. Unless you want to stick to subjects that are nearly all pitch black with only a tiny proportion of highlight area. Maybe the moon on a good dark night and shot with a 24mm lens? 

 

 

mind blown !

[rodeo_joe1]

On 12/5/2023 at 11:17 AM, ilkka_nissila said:

But anyway, yes, the brightest part of the scene could only be less than 0.01% of the image area. People tend to love this kind of images.

But most landscape pictures aren't just one small highlight surrounded by total blackness.

The 'contamination' of light into shadows consists of the partial and diffuse reflection of all the light hitting the sensor, bouncing off the walls of the 'dark' chamber. Not even considering lens flare.

Metering theory would have us believe that an average scene reflects about 13% of any highlight brightness. So if we take our sensor + dark chamber as reflecting 0.2% (20% x 1%) of the total image-forming light back into the deepest shadow, that's a minimum shadow brightness of 0.026% (13% x 0.2%) of the highlight brightness, given an 'average' scene. Which translates to a brightness ratio of just under 4,000:1. Or just over 2 full stops short of a 14 stop 'dynamic range'. 

So my contention is that a 14 stop DR is just not practically possible until the dark chamber of a camera - including the rear and internals of the lens - is made capable of reflecting absolutely zero of the image-forming light. 

BTW. Multiple images at different exposures will still have the same limit to their individual brightness ranges. And the light reflected back into shadow areas actually helps in revealing shadow detail, in the same way as the old trick of pre-flashing film did. 

 

Edited December 15, 2023 by rodeo_joe1

[ilkka_nissila]

59 minutes ago, rodeo_joe1 said:

But most landscape pictures aren't just one small highlight surrounded by total blackness.

The 'contamination' of light into shadows consists of the partial and diffuse reflection of all the light hitting the sensor, bouncing off the walls of the 'dark' chamber. Not even considering lens flare.

Metering theory would have us believe that an average scene reflects about 13% of any highlight brightness. So if we take our sensor + dark chamber as reflecting 0.2% (20% x 1%) of the total image-forming light back into the deepest shadow, that's a minimum shadow brightness of 0.026% (13% x 0.2%) of the highlight brightness, given an 'average' scene. Which translates to a brightness ratio of just under 4,000:1. Or just over 2 full stops short of a 14 stop 'dynamic range'. 

So my contention is that a 14 stop DR is just not practically possible until the dark chamber of a camera - including the rear and internals of the lens - is made capable of reflecting absolutely zero of the image-forming light.

I don't know where some of your numbers are coming from, but I would like to make a few points. First, the black parts of the camera don't by default get exposed with light as the lens projects its image (which often is flagged by a mask not to go outside of the sensor area) onto the sensor and then a small part of that light is reflected on the black surfaces of the interior of the camera and lens, but those surfaces are matte and so they produce mostly a small diffuse reflection that causes a small part of that small part to be scattered in various directions.

 

Chasing more DR isn't necessary for a scene that is close to mid toned to begin with. But many images which people find interesting or fascinating have a large range of brightnesses, there might be the sun coming up in the horizon or between the trees, the sun could be hitting the hair of the subject etc. Those are the highlights of the image. In order to not overexpose those highlights, one needs to underexpose most of the image and then lighten areas of the image in post-processing. This is a standard technique of dealing with high-contrast scenes, and because the dynamic range of sensors is limited, many photographers put the camera on tripod and expose a set of different frames with different shutter speeds, covering, e.g., 6 stops in the brackets to be able to make a (relatively) noise-free image showing a full range of tones in each major area of the frame. Here is an article on the topic, with examples, though unfortunately the text is in Finnish:

 

https://www.maisemanlumo.fi/artikkelit/dynamiikan-laajentaminen-maisemavalokuvauksessa-hdr-ja-exposure-blending/

 

I think you can get the idea of what kind of scenes benefit from high dynamic range in the capture process and appropriate editing. If the scene were dominated by reflections, bracketing exposures would not help recover more scene detail (as giving the sensor more light would simply increase the brightness of those reflections), but it does.

 

In cinema, Hollywood seems to love ARRI cameras which are famous for their dynamic range. Most Oscar-winning movies were shot with their cameras. Arri Alexa 35 has about 16 stops of dynamic range https://www.cined.com/arri-alexa-35-lab-test-rolling-shutter-dynamic-range-and-latitude-plus-video/

 

Now why do the cinematographers care so much? They want to edit the images for stylistic effect and mood and so the final color might be quite different from the actual filming situation as seen with the human eyes. So in order to have that room for editing where the colours could be adjusted for dramatic effect, and the original scene can also have high dynamic range, it's beneficial to have the highest-quality data from the camera. Arri makes some lenses where the flare isn't even minimized but that produce artistic flare and ghosting. Often the scene is lit in such a way that the drama is emphasized by the lighting, and there can be highlights that are very bright compared to the main subject's brightness. In indoor scenes you may have contrasty artificial lights, and light coming from the window that is very bright in comparison to the shadows within the interior. Yet to look natural there should not be any clipping of the artificial lights themselves, or the scene behind the window. These situations can be best handled by a camera which captures as much scene information as possible. if you do photography in typical artifical lights e.g. outdoors or even in some homes, they're not lit like an office where you'd have very even lighting to make working easier. They're often moody and extremely high in contrast. The trick then is to make a natural looking image out of such situations as easily as possible.

https://www.arri.com/en/camera-systems/cameras/alexa-35

 

BTW. Multiple images at different exposures will still have the same limit to their individual brightness ranges. And the light reflected back into shadow areas actually helps in revealing shadow detail, in the same way as the old trick of pre-flashing film did.

 

The in-camera and in-lens reflections of the longer exposures in a bracketed landscape sequence would simply produce brighter levels with the same information if noise wasn't an issue. But it is, and more information is revealed using multiple exposures, because it overcomes the noise (despite increasing what reflections there are in the image).

Edited December 15, 2023 by ilkka_nissila

[macwest]

 

[rodeo_joe1]

On 12/15/2023 at 11:56 AM, ilkka_nissila said:

First, the black parts of the camera don't by default get exposed with light

Yes they do. Light is reflected diffusely from the image plane - be that film or digital sensor. 

That light is not totally absorbed. It bounces around the 'dark chamber' and makes its way back onto the sensor, thereby contaminating shadow areas with extra light and limiting the 'dynamic range'. 

This has nothing to do with noise. In fact, as I previously said, the extra light performs an optical compression of the subject brightness range and helps the sensor cope with extreme subject contrast. 

Edited February 10 by rodeo_joe1