SL2 Dynamic Range - How good is good enough?

[Manicouagan1]

Advertisement (gone after registration)

From the time the SL2 results were posted on Photonstophotos I have tried to understand the test results and my picture taking with the SL2.   For reference I attach a link to a plot of data for the series of Leica color digital cameras I have used (M8, M9, M240, M10, M10-R, SL2) and what I consider the end of the line of Digital Single Lens Reflex cameras the Nikon D850.   

https://www.photonstophotos.net/Charts/PDR.htm#Leica M (Typ 240),Leica M10,Leica M10-R,Leica M8,Leica M9,Leica SL2,Nikon D850

I my experience taking the sorts of landscapes and travel pictures with my Leicas was constrained by the dynamic range and image quality of the M8.  The M9 was dramatically better for my image taking and the M240 was "good enough".   The M10 and M10-R made "better" prints, but not that much better.   When I looked at my SL2 images (mostly on the computer screen) the reduced dynamic range (relative to the M10,M10-R or D850) at low iso did not seem a problem.  I found that a -1/3 or at most -2/3 stop exposure compensation was needed in some situations, but the resulting shadow if any was a non issue.   I do think I see more detail in my SL2 images than with M series cameras, but I suspect that the improved lens quality and image Stabilization account for those differences.  

Edited December 23, 2021 by Manicouagan1
left of sl2

[MrFriendly]

You could perform this experiment to find out:

1. Pick a scene with very bright and very dark areas.

2.  Photograph the scene (single exposure), like you normally would.  Let's call this A.

3.  Photograph the same scene using exposure bracketing (three should be enough, e.g., -3, 0, +3), and then perform HDR merge on the RAW files.  Let's call the merged result B.

compare the two A and B raw files to see how your sensor would have performed had it had more dynamic range. 

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[FlashGordonPhotography]

While I have no doubt P2P's dynamic range measurements are empericly accurate I don't think they translate to real world usage very well. They describe a blanket measurement but don't really indicate how robust a files DR is and how it is distributed. DR measurements miss things like detail retention vs noise reduction or PDAF banding or other sensor driven artifacts. The M9 was measured as quite low but in reality could be pushed quite hard in the shadows. It was somewhat noisy (lower DR) compared to some other cameras, which turned out to have less detail due to noise reduction in the raw files. So they were less noisy but also less usable.

The Nikon Z7 looks really good against the SL2 and appears to match the X1D Hasselblad on P2P. In the real world this isn't the case. It's possible to get PDAF artifacts from thei sensor (and from most of the Sonys) when the shadows are pushed hard. So they have DR on paper but aren't more flexible in real life.

If you compare most M sensors to the SL2 sensor they respond quite differently to file manipulation. The SL2 is quite balanced. You have a reasonable amount of flexibiliy at either end of the file for recovery. M sensors seem to be much more biased to the shadows. Even the newer CMOS sensors in the 240 and M10, M10R and M10M. The Monochrom is the most extreem. You can't clip the highlights at all but the shadows have a vast amount of recoverable information. The M10 and M10R have a small highlight shoulder but it's pretty steep compared to the SL2. Over exposure isn't recommended with any of the M bodies.

In terms of usable DR the M10M has by far the most. The M10R has about a stop more then the M10 and SL2. The SL2 sensor is actually very competetive with the A7R3/4. While the measurable DR is lower it doesn't have any PDAF banding and it's noise structure is less offensive. The Sony sensors semm to be a bit better in the highlights but fall apart in the shadows earlier because of the PDAF pixels they have.

With the SL2 I find the best results come from playing with shadows and highlights and leaving the whites alone. It may have less measured DR but it also have excellent detail retention and reasonable colour retention at mid ISO's. With the M10M/R it's shadows and blacks. Definitely don't stress the whites. With the regular M10 its highlights and shadpws with a bit of room in the blacks. Mostly I have custom curves set up for each sensor type.

If you really wanted the best DR you'd do away with these small sensors and use the 7 year old sensor in the X1D and GFX, which still outperforms pretty much anything on the market, even after nearly a decade of production. The difference from it, to the best of the smaller format sensors when pushing files in post is stark. I've done recoveries of over five stops of shadows on the X1D and been able to print them at A1.

At some point all sensors will run out of DR. You'll have to amke a decision to either clip or bracket. Mostly, as an old Velvia shooter, I consider anything over 5 stops to be a bonus.

Gordon

[Manicouagan1]

My interest in dynamic range comes from photographing white birds against dark backgrounds.  The American White Pelican and the various white egrets and herons around Galveston Bay provide me with a lot of interesting subject matter.  One of my goals is to show feather detail.   Empirically I found no major problem with achieving this goal with the SL2.    Having looked at the Photonstophotos data I expected more of a problem than I encountered.  The SL2 image behavior in the bright range is such that moving the Lightroom Highlight slider to the left brings out all the detail I could possibly desire.   I suspect that all the steps in the analogue to final dng data digital conversion in the top of the sensor’s range of the three-color channels is tuned to facilitate recovery of these subtle variations in whites.  The collective effort of the sensor, analogue to digital converter software and firmware development people did some very good work and should be complimented.  The SL2 is a lot more than just an adequate platform for outstanding lenses, better than some tests would suggest.
 

 

 

[chrismuc]

any idea why the DR of the SL2 is about 1.2 stops lower than the S1R at ISO 100 (same sensor)

'Triangle up indicates scaling'

what is the meaning?

https://photonstophotos.net/Charts/PDR.htm#Leica%20SL2,Panasonic%20Lumix%20DC-S1R

[SrMi]

3 hours ago, chrismuc said:

any idea why the DR of the SL2 is about 1.2 stops lower than the S1R at ISO 100 (same sensor)

'Triangle up indicates scaling'

what is the meaning?

https://photonstophotos.net/Charts/PDR.htm#Leica SL2,Panasonic Lumix DC-S1R

The graph for S1R contains triangle down.Triangle down indicates noise reduction applied to the raw files. Noise reduction typically increases PDR.

The difference can be explained by differences in electronics and firmware. You see that S1R's curve is shifted to the right. This could be explained by differences in calibration.

Bill Claff wrote about scaling:

"If the Digital Number (DN) comes out of the Analog to Digital Converter (ADC) is then multiplied by some number greater than 1 then that is "scaling".
Usually this is pretty obvious because if you look at a histogram of the values in an image you'll see gaps because the data has been stretched out.

For example, ISO 51200 might really be ISO 25600 multiplied by 2."

 

[nicci78]

Advertisement (gone after registration)

Fun fact. 
Every cameras in normal mode is now lacking DR for display. 
 

For prints, any camera even 1 inch or m4/3 have more DR than the paper can handle. 

For display it is another story. OLED and miniLED panels are becoming more mainstream. Your mere 15 stops is nothing compare to the minimum of 22 in real HDR modes. 

HDR photos have now the means to be viewable. 
Sun will look like sun and shadows will be totally dark. Infinite contrast is now available and mainstream

 

[Photoworks]

I just want to point out that he sensor is only one factor that influences your dynamic range of the image.

many lens coating affects contrast and for example Zeus T* retain more highlights.

and then there is the processing program. In my experience Capture one has a better curve response than Lightroom. In the past C1P has a much better highlight recovery than Lightroom.

I find that the SL2 sensor compatible to the latest Sony and canon. I use all of them for different applications. on a bling test you would not be able to tell the difference when processed to match when shot true the same lens.

Medium Format is still king, with the higher 14-16 Bit raws there is so much more color information than any 35mm cameras.

 

[convexferret]

@Photoworkscan you point me to the study on how Zeiss coatings hold highlights better?

[SrMi]

9 hours ago, convexferret said:

@Photoworkscan you point me to the study on how Zeiss coatings hold highlights better?

This is the short version:

https://lenspire.zeiss.com/photo/en/article/technical-article-on-t-coating-and-reduction-of-reflections-in-lenses/

This is the 74 page version:

https://pixinfo.com/wp-content/uploads/2016/04/en_About-the-reduction-of-reflections-of-camera-lenses.pdf

 

[FlashGordonPhotography]

12 hours ago, nicci78 said:

Fun fact. 
Every cameras in normal mode is now lacking DR for display. 
 

For prints, any camera even 1 inch or m4/3 have more DR than the paper can handle. 

For display it is another story. OLED and miniLED panels are becoming more mainstream. Your mere 15 stops is nothing compare to the minimum of 22 in real HDR modes. 

HDR photos have now the means to be viewable. 
Sun will look like sun and shadows will be totally dark. Infinite contrast is now available and mainstream

 

This isn't really how it works. Capture DR and display/print DR are treated differently.

Firstly it's not actually true that prints have less DR than the files. That used to be bandied around by those arguing over whether printing was viable vs screen viewing. You'd need to do specific print by print measurements comparing every ink type with every LUT and every paper made. Then you'd need a way to accurately compare a reflected range measurement to a back lit one (can't be done yet). Currently opinion with commercial photographic printers and equipment is that the improvements in dithering to reduce *visible* noise means prints *visually* improve *estimated* DR over the source file. Note that's a LOT of **. Also note that inks, paper, LUTs and RIP's have improved dramatically over the last decade and continue to do so. Comparing a file on a display to a print is more apples and oranges really.

As for the HDR thing, the limitiation is going to be the human eye. Even if a screen can display 22 stops there's very very few adults who can perceive the same range in a still image. That's about what a 3 year old can see. HDR displays are mostly made for video content. Stuff that streams by at 120 fps. In the right viewing conditions (ie: not on your desk looking at a laptop display) your eye can adapt (iris) to appreciate HDR video content. Not so much with stills where the eye can't actually keep up with the display. You'll note that the new MacBook Pro displays do HDR dynamically (turn it on and off as they think you need it) for this reason. It's not always on by default. Apples new displays are really interesting, although they currently have issues with user calibration and colour control as currently colioromiters and other equipment needed to calibrate them is different to what's commercially available. About half the current harware available can't even calibrate OLED let alone micro LED.

I should note I work on hardware calibrated OLED displays and commercial level fine art printers every day. Currently one of the areas I specialise in is high end art reproduction and printing for the local art community as well as fine art printing for selected photographers. I first started printing commercially as a fine art b&w and cibachrome printer 25 years ago and moved to injet earlier than most. I've been doing this a looong time. The developments in print and display tech have improved like cameras. It's incredible how much things have improved even in the last few years. Old truisms aren't any more. When I print I'm not using limiting previews on my monitors. I'm using LUT's to my printer that hold the content to wht I see on screen.

The simple reality is the vast majority of us can't see the difference between last years tech and this years. Yes mini LED is great. But for most viewers it's a placebo. You think you seeit because you saw the ad on TV. In a blind test most wouldn't know. Kind of like walking into a TV store. Even the cheap TV's look good with optimised content and curated lighting.

Mostly DR is a limitation of what the eye can see, not the camera. That's not going to change.

Gordon

[FlashGordonPhotography]

21 hours ago, chrismuc said:

any idea why the DR of the SL2 is about 1.2 stops lower than the S1R at ISO 100 (same sensor)

'Triangle up indicates scaling'

what is the meaning?

https://photonstophotos.net/Charts/PDR.htm#Leica SL2,Panasonic Lumix DC-S1R

Firstly, The sensors are similar but NOT identical. The wafer is the same but the colour array and microlenses are different. For practical purposes it's impossible to separate the wafer with the stuff glued to the top of it.

For my SL2 and S1R I have found that they're basically the same in terms of usable DR and file robustness. The Sl2 files have *slightly* (and I mean not visible most of the time) better accuity than the S1R when using APO Summicron Sl primes. This could be due to hardware reasons (thinner cover glass or slight changes to the CFA) or software (noise reduction before raw output or demosaicising) or a combination of both. The colours are slightly diffferent and I use a slightly different curve when processing but noise and DR wise they end up being all but equal.

Personally I prefer both to either the Z7 or A7R3 I own. The base noise is *slightly* higher and DR *slightly* lower but the very rare appearance of PDAF banding on the Nikon and Sony mean the the usable DR is at least as good on the Leica and Panasonic in real world use.

Gordon

[jrp]

38 minutes ago, FlashGordonPhotography said:

The colours are slightly different and I use a slightly different curve when processing but noise and DR wise they end up being all but equal.

What processing curves do you use?

[FlashGordonPhotography]

7 hours ago, jrp said:

What processing curves do you use?

 

I make a custom curve for each camera I use in Lightroom as a starting point I like. It's not scientific. Just asthetic. I also do curves for lenses sometimes.

Gordon

[convexferret]

15 hours ago, SrMi said:

This is the short version:

https://lenspire.zeiss.com/photo/en/article/technical-article-on-t-coating-and-reduction-of-reflections-in-lenses/

This is the 74 page version:

https://pixinfo.com/wp-content/uploads/2016/04/en_About-the-reduction-of-reflections-of-camera-lenses.pdf

 

I'm well aware of how coatings work. I'm asking where the assertion that Zeiss coatings hold highlights better (than everyone else's?) comes from. It wouldn't just be an anecdotal assertion, the likes of which we see so much around here would it?

[tom0511]

I dont find the sl2 limiting in regards of dynamic range in real world. I agree that mediumformat like leica S or x1diii have more flexibility in post processing and a little better midtones. The M10r is great in regards of highlight rolloff, I am not sure how these measurement numbers really reflext real world use.but if I choose to take sl2 or m10r with me it. is never influenced by the sensor, both are fully suffient for me.

its about manuel prime shooting vs AF and more lens flexibility and iq is great from both sensors.

the other thing is if we really allways want need to see details from deep shaddow to bright highlights. Peronally I am not a fan of the hdr-look.

 

[beewee]

On 12/23/2021 at 9:41 AM, Manicouagan1 said:

my experience taking the sorts of landscapes and travel pictures with my Leicas was constrained by the dynamic range and image quality of the M8.  The M9 was dramatically better for my image taking and the M240 was "good enough".   The M10 and M10-R made "better" prints, but not that much better.   When I looked at my SL2 images (mostly on the computer screen) the reduced dynamic range (relative to the M10,M10-R or D850) at low iso did not seem a problem.  I found that a -1/3 or at most -2/3 stop exposure compensation was needed in some situations, but the resulting shadow if any was a non issue.   I do think I see more detail in my SL2 images than with M series cameras, but I suspect that the improved lens quality and image Stabilization account for those differences.  

In my experience shooting landscapes with the M8.2, M10, and SL2-S: in high contrast scenes (mountains with snow and heavy shadows and/or people dark clothing), 10 stops is beneficial as I tend to underexpose by 1-2 stops to ensure I get good colour fidelity in bright areas and avoid blowing highlights. For low contrast scenes (astro and overcast days or flat light in a snow storm), 6 stops is sufficient. 

The M8.2 was fine during the day in good light, as long I as I shoot at base ISO and exposed to the right. The M10 benefitted from about 1-2 stops of under exposure to avoid blown highlights. The SL2-S benefits from about 1 stop under exposure to maximize colour fidelity in bright areas. I tend to adjust shadows and tone curves after the fact to get a balanced look in Lightroom.

With about 10-11 stops of DR on the M10 and SL2-S (at base-ISO), I tend to process my photos in a way similar to video grading with a log profile. I’ll push the highlight/shadow recovery and black point, to compress the tonality of the image and recover as much detail as I can on both ends, then play with tone curves to get a natural looking image while maximizing details at all grades.

Below is the final image.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

 

This is the image I started with before processing.

[Jon Warwick]

On 12/24/2021 at 8:03 PM, FlashGordonPhotography said:

If you really wanted the best DR you'd do away with these small sensors and use the 7 year old sensor in the X1D and GFX, which still outperforms pretty much anything on the market, even after nearly a decade of production. The difference from it, to the best of the smaller format sensors when pushing files in post is stark. I've done recoveries of over five stops of shadows on the X1D and been able to print them at A1.

 

Absolutely, I couldn't agree more. And even with a more challenging medium format camera that has smaller-sized pixels (ie, the 100mp GFX that I own), I have found the DR to be remarkable in that I don't tend to ever find a shadow / highlight "limit" for really any images I've taken with it .....and I don't recall that always being the case for smaller sensor cameras (such as the SL2) that I used to own.

[SrMi]

One can increase DR of your camera "on demand" by using frame averaging. Some camera support in-camera frame averaging.

[Advertisement]

Hallo,

Look here: Leica L-Mount

[FlashGordonPhotography]

4 hours ago, SrMi said:

One can increase DR of your camera "on demand" by using frame averaging. Some camera support in-camera frame averaging.

Definitely a useful technique but there are too many scenes where this simply won't work. Anything with waves or fast flowing water, for example. Leaves on a windy day. Unfortunately, too often, the world refuses to stay still.

Gordon