[stevebrot]

I know...I know...this is a subject of endless debate except that yesterday's article (and linked video) at PetaPixel actually worked with comparison images of a real world subject. While questions might still be raised regarding methodology, the video's producer's conclusions mirror my own experience fairly closely...

Edit: As is pointed out in comments below, what is being demonstrated is exposure latitude, not dynamic range. The two are related, but clearly not the same. /Edit

“I think that digital has a usable dynamic range of minus 6 or 7 to plus 2,” Lawson says. “I thought that at minus 8 stops, the grain made the digital image unusable. Of course, this is subjective and your opinion may vary. And different digital cameras may yield different results.”...

...“I think film has a usable dynamic range from minus 2 stops to plus 10 stops and beyond,” Lawson says. “I thought that at plus 8 stops and beyond, the film images had a distinctive, aesthetic look that I consider desirable. Of course, this is subjective and your opinion may vary.”

Of course, much depends on the definition of dynamic range. I personally feel that using signal-to-noise ratio for digital is lame and much prefer actual performance in use. I have no problem reproducing a 9-stop step wedge (11-stops actually) with most negative films* and standard development and 12 stops or more with compensating development. While my K-3 provide good avoidance of highlight clipping when shooting in RAW and impressive shadow recovery (not 6 or 7 stops, but at least 4-5 with credit given to the sensor maker, but more so to Adobe's clever data synthesis).

PetaPixel | Film vs. Digital: This is How Dynamic Range Compares

...and the video...

You Tube

You Tube

Steve


* Positive (slide) films are a completely different story. Not only are they at dire risk for highlight clipping, the range into the shadows is typically limited as well, hence the forever statement that shooting slides requires accurate metering.

---

Last edited by stevebrot; 05-03-2019 at 01:11 PM.

[Alex645]

I should walk the talk (one day) but here is my take on the video's test:

a) Isn't he really testing exposure latitude and not dynamic range? Dynamic range would be shooting a normal exposure in high contrast lighting and seeing at what point do the highlights and shadows lose detail in the film vs. digital.

b) Why a black and white film comparison? It's not like he's using a Leica Monochrom black and white digital camera for the comparison. If 100 ISO was the goal, then I wish he'd use Ektar 100 which not only has a better dynamic range, but also scans better.

c) Why TMax? TMax is known to have a higher contrast characteristic compared to Tri-X or FP4+ or HP5+.

d) Why 100 ISO? Higher ISO films, such as 400 ISO, have a much greater dynamic range (and exposure latitude).

e) Why develop in TMax in D76? Other generics like Xtol would improve the dynamic range of TMax, but the best would have been Kodak's TMax developer.

f) Film negs are always going to handle over exposure better whereas a digital sensor with RAW processing are going to be able to boost or add gain to under exposures.

g) And then there is the whole question of post-processing analog vs. digital: "Straight out of the camera" a film negative will usually capture a greater dynamic range, but with RAW processing, digital highlights can be recovered and shadows can be raised. Adding darkroom dodging and burning vs. lightroom highlight and shadow adjustments is a different test.

h) Many film photographers (pros) had little choice whether their client/agency/publication demanded negs or slides. Negs always were more forgiving with a greater dynamic range and slides demanded exposure bracketing. Digital sensor characteristics are much more like slides, but with a greater dynamic range. Processed jpegs vs. unprocessed RAW vs. processed RAW vs. unaltered film vs. scanned film with post-processing.......

Just some footnotes from the peanut gallery.

[abruzzi]

I would add that his definition of “usable” and my definition are pretty far apart.

[RGlasel]

It's nice to see a video about photography that doesn't indulge in self-cheerleading and I found it very interesting, but I have a hard time understanding the benefit of 10 stops of latitude in over-exposing film. Shouldn't dynamic range be the ratio of the darkest details to the brightest details in a properly exposed photo?

[stevebrot]

a) Isn't he really testing exposure latitude and not dynamic range? Dynamic range would be shooting a normal exposure in high contrast lighting and seeing at what point do the highlights and shadows lose detail in the film vs. digital.

Agreed, that is the traditional way these things would be done. When I establish a working EI, I use a standard setup in direct sun to challenge the dynamic range based on detail retention. If I can't get a full eleven stops with reasonably smooth tonal gradation and detail retention in all but the extremes, I either decide against the film or look to a different developer.

g) And then there is the whole question of post-processing analog vs. digital:

That is the elephant in the room. A more reasonable assessment might be made comparing prints with prints, but that is not Internet friendly. Most film images suffer some from digitization and even more so if an attempt at raising values is made in post. A RAW capture on the other hand, gives the processing software (Yay, ACR!) a great amount of leeway in constructing a reasonable shadow representation from very sparse data. (There are only 8 integer values to represent all tones for the bottom 4 stops of capture data, zero excluded)*. I have done a few simple studies with synthetic images showing that our tools do a considerable amount of work creating expected tonality in the shadows, i.e. creating something out of almost nothing. Adding +2 stops "exposure" in Lightroom to a discrete step image (9 1-stop steps) results in an unexpected value distribution with much of the lowest values being retained, rather than abandoned, and invention throughout.


Steve

(...really needing to do more work with synthetic data, but need a few better tools and some skills with MATLAB...)

* Zero indicating values below the sensor detection threshold (i.e. no data) for a particular capture.

[TonyW]

It just goes to show because you can does not mean you should - why did he make this video? Good points all Alex645

I would add nonsense about exposure - may be I missed something as I could not take more after a certain point

What exactly was his exposure reading based on i.e. what was his meter actually reading and calibrated to?

When he says over or under exposed what does he mean?
Is it what the meter suggests?

What metering mode and precisely what areas being evaluated within the meters ROI?

Many DSLR’s metering systems are calibrated to around 12.5% matrix and centre weighted will interpret differently A 12.5% point will mean that we have a maximum of +3EV on a metered area before we reach saturation and loose all texture detail - very different from B&W.

Digital is rather easy to calculate. A 14bit system will at best have a usable DR of 2-3 stops less than bit depth (tolerances to acceptable noise may differ). So let’s assume that we have 11 stops usable and we measure and base exposure on a highlight area This gives us a DR of +3EV to -8EV. Figures already published

B&W analogue very different and probably 10 stop DR still relevant? What we do not know is if the guy knows how to expose process film correctly i.e. minimum exposure to place the low values just on the toe area of the curve and development times suitable for intended reproduction method.

Film does not have the rapid cut off in highlights if digital being a gentle rolling off and compression in the shoulder (dependant on proper pricessing)

Really the comparison should have been have been made including printing to photo paper.

[photoptimist]

It's also notable how film and digital fail in such different ways. Film seems to get frightfully muddy and soft on both ends. (Some might laud this as having a nostalgic old snapshot look).

Digital clips abruptly with over-exposure (some might laud this for a sharp high key look) and suffers from patterned-grain with under-exposure (some might laud this for it's techno-digital ambience).

Neither seems universally better (and I'd judge the latitude of digital to be better than film).

The real lesson is two fold: 1) know how your media responds to light; 2) different photographic intentions call for different tools.

[stevebrot]

I should walk the talk (one day)

BTW...I was thinking the same thing for me. This would be easy enough to do, but with a few more controls in place (same EV for both cameras, off-camera incident metering, in-frame gray reference, and so forth) as well as making sure both cameras were fully operational. The evidence of light leaks in the developed negatives should have fully negated the test, particularly since a -9 ND was being used (BTW...eyepiece blind?).

Oh...and ten stops each way is way overkill. The point could have been made with as few as +/- five stops.

Finally, label the test "Exposure Latitude"


Steve

[stevebrot]

Film seems to get frightfully muddy and soft on both ends.

Part of blame there may go to a serious light leak mentioned in the video and visible on the negatives. I wonder about back lighting through the viewfinder as well since a -9 filter was in front of the lens.


Steve

[Pentax Syntax]

Very interesting post. While all of the comments above are making valid points, I know of very few examples where exposure latitude has been compared. I still shoot both film and digital and this reinforces what I have observed but not examined systematically. I avoid overexposure at all costs on digital. Faster B&W film has even more overexposure latitude than the 100 film used here. Of course a better film scanner and push processing methods would have given the under exposed a better outcome than shown here. Thank you for posting. Maybe it’s my old analog eyes but I agree with the video’s concerning the different aesthetics of the film vs digital images.

[abruzzi]

While it's not as extreme, and it's not a analog vs. digital test, this guy, Kyle McDougall, did an interesting exposure test with three of the big color negative films: Portra 400, Fuji 400H, and Ektar 100:

You Tube

You Tube

You Tube

You Tube

You Tube

You Tube

Its not fully scientific and he doesn't cover a 21 stop range (10 stops on a Pentax 67) but to me its still quite interesting and useful.

[tuco]

I just posted a film picture with awesome dynamic range. More DR than any of those people testing came up with.

[Alex645]

Part of blame there may go to a serious light leak mentioned in the video and visible on the negatives. I wonder about back lighting through the viewfinder as well since a -9 filter was in front of the lens.
Steve

Agreed. That would either be caused by not shuttering the viewfinder on long exposures or the film pressure plate was insufficient which could have contributed to his comment about focus issues. Considering his subject was backlit, it may have been the latter.

The N90 he used was an entry level semi-pro camera and using something more like an F3 with a viewfinder blind makes more sense if the digital camera compared is a D750. Of course, we have to assume the apertures and shutter speeds were pretested and calibrated. I have shutter speed calibration tester for film SLRs and it is not uncommon to see more than +/- 1/3 EV error in the shutter speeds.

[Ranchu]

It's nice to see a video about photography that doesn't indulge in self-cheerleading and I found it very interesting, but I have a hard time understanding the benefit of 10 stops of latitude in over-exposing film. Shouldn't dynamic range be the ratio of the darkest details to the brightest details in a properly exposed photo?

The benefit is that you can shoot difficult subjects at will, without the knowledge while you do it that when you get home and look at them, all your files will have holes in them. Except a couple that you were super careful about.

It frees your mind.

---------- Post added 05-04-19 at 05:40 AM ----------

And it looks better.

---------- Post added 05-04-19 at 05:49 AM ----------

I mean all we have to do here is look at which one looked better.

[AntonioS]

I totally agree with Steve and Alex.

That guy choosed probably the worst film for this test.
For me what's he doing is to compare a very specific situation (specific film, developer, scanning and post processing).

Repeat this same test using an Adox CMS 20 developed in Rodinal, then scanned with a Heidelberg Tango and post processed by someone that knows how to do and it will be another totally different story.

Repeat again using an Ilford Delta 3200 developed in Caffenol, scanned with a flatbed scanner and again the story will be other.

Topics like this one from Petapixels are IMHO just click baits because they don't tell the whole story, just what the author wants to prove.