Keeping over-exposure under control

Just as I was going to put a film through a few different cameras with an assortment of lenses to check how they (the lenses) compare with each other, Yr ‘Umble Ed – Mike Evans – asked if I’d write about the differences between film and digital cameras, and their response to over-exposure.

The idea is that film is thought to be more ‘forgiving’ with over-exposure, allowing details to be ‘recovered’ from very bright, over-exposed areas, whereas digital cameras just ‘white out’ or ‘blow-out the highlights’ irrecoverably when over-exposed. And what about ETTR (Expose To The Right)? We will explore that theory later in the article.

Here’s what I mean:

• 

• 

• 

The digital picture (the third above) – shot on Auto White Balance, but adjusted to ‘Daylight’ in Adobe Camera RAW to match the colours of the film photos – goes to pure white in the brightest highlights, just as the film (in photos 1 and 2) – rated as ‘Daylight’ – also goes to irretrievably pure white in the brightest highlights.

But there’s fractionally more detail remaining of the roof frames – here in the Great Court of the British Museum – in the first two film photos than in the digital photo. The detail is “eaten away” more severely in the digital photo. There’s (marginally) less detail of the roof frames to retrieve from the digital pic than from the other two. Digital photos, apart from those taken with Sigma’s ‘Foveon’ sensors, get part of their information from adjacent pixels on the sensor. So, if any adjacent pixels are overloaded and ‘whited-out’, there’s no way to gather any extra detail from them. Hence a “sudden” transition to completely white at the brightest highlight:

• 

• 

• 

All were taken with the same Voigtländer 10mm lens, on an M7 (with ISO 100 and ISO 800 negative films) and with an M10-P (digital, at ISO 100). Note that the two films pick up slightly more blue in the sky than the M10-P, particularly the ISO 100 film, unless that’s a result of the commercial scanning of the films to get them here onto a computer screen.

The less-sharp cutoff of fine detail in the film images goes along with the accepted idea that film retains more detail in highlights than digital does, although here there’s not very much difference.

The smooth transition to ‘white-out’ in the film pictures is generally described by the film having a ‘shoulder’, where its response to bright light (at the top of the next graph) gradually flattens out, so that more detail, or differentiation of tone, can be gathered by the film emulsion before it reaches its maximum gathering of bright light, and then goes absolutely white – or really, being a negative, goes absolutely black.

The films – both kept cool, and both in-date – are ISO 100 Kodak Ektar (professional) and ISO 800 ‘Lomography’ (re-packaged Kodak 800, I understand, and it certainly behaves like it). The overall results from the two films are certainly more ‘grainy’ than the far smoother digital photo when they’re all magnified to the same size:

• 

• 

• 

The digital M10 looks far better in this dimmer region of the photos than either of the film scans. But the M10-P is nominally rated up to ISO 50,000 by its makers, though I usually set a ceiling of 25,000. The main reason that the lettering is more legible in the digital picture is that the film scans are only six megapixels, whereas the digital pic was shot at 24 megapixels, and so needs less magnification to produce the same size of crop. But the coarse ‘grain’ of the film, compared with the fine-detail-capable sensor, also makes the teeny lettering less legible.

The response curve of film, with its ‘shoulder’, generally looks like this:

• 

• 

As a result, once the sensor reaches maximum stimulation by light, it just can’t hold further detail and simply “whites out” instead of storing some fine detail within the brightest highlights, as those two film shots have, just about, done.

Doesn’t that first curve, the one with the ‘shoulder’, look like the adjustment option on Olympus cameras – say an E-M1, or PEN-F? So can some cameras be adjusted to respond more like film, and so avoid having these ‘blown’ highlights?

Let’s see. But first, I’ll say what my conclusion was after a whole load of testing, and after gathering expert opinions (no, sorry, one expert opinion – from the people who design cameras, film, etc. After taking shot after shot with different parameters, various settings, multiple lenses and multiple camera bodies there’s one conclusion: The best results came from setting all the cameras to auto-exposure (if available) and letting the cameras choose optimum exposure. Sounds like heresy, huh?

Auto gave unimpeachable results with a huge range of brightness in each photo. I used a digital Leica M10-P, a Nikon Df (for Dave Seargeant of course) an Olympus PEN-F, a Sony A7s and a Sony A7r MkII. The film cameras, which I didn’t use on auto-exposure (except once), were a Leica M7 and a Contax IIIa (because of the reputation of its Zeiss ‘Sonnar’ 135mm f/4 lens, which I used for comparison with a Leica 135mm f/4 on a Sony A7RMkII, as well as on the original 1960s Contax film camera). However, because this article is long enough already, I haven’t included those pics here.

A huge amount of scientific and artistic effort has gone into devising the response to light of digital cameras (Nikon, Leica, Sony, Olympus, and all the rest) and nowadays relying on cameras’ own auto exposure gave me the best results. Every time. (Although that could just be because I’m a hopeless photographer or, maybe, because twiddling over- and under-exposure settings comes naturally, through practice, to me.

But now, back to the story..

These next three photos were all taken with the same Leica 50mm APO-Summicron lens, supposedly the best ‘standard’ lens that money can buy, on factory-set M7 and M10-P cameras, so the shutter speeds (as well as the constant aperture, and thus the exposures) should be accurate for each photo:

• 

• 

• 

Note that these images were all processed to gather maximum detail from that over-exposed highlight where the sun is bursting through.

It’s conventional wisdom that the emulsion – the light-sensitive coating – on a roll of colour negative film has some kind of god-given capacity to absorb light, even with over-exposure, without ‘blowing out’ those highlights (thus being able to keep more detail in clouds and other bright regions of a photo), and much more so than digital cameras. Reversal, or slide or positive film, though, such as Ektachrome or the late lamented Kodachrome, is not so forgiving. It does need (or needed) to be shot with absolute avoidance of over-exposure to keep details in the brightest areas.

Is that over-exposure tolerance an innate capacity of some light-sensitive silver compounds, or do/did the chemists, who formulated film emulsions, use some special trick to produce that characteristic ‘shoulder’ curve of film’s response, so that details in highlights are preserved better than the sensors, and the rest of the electronics, in digital cameras can achieve?

I went to the horse’s mouth to ask the best-qualified and most knowledgeable person I could think of regarding film: Bob Shanebrook, who wrote the textbook “Making KODAK Film” which describes the genesis of Kodak’s film products, and which includes “..detailed descriptions of photographic film image-forming mechanisms for black-and-white films, color negative films, Ektachrome Films, and Kodachrome Films”.

If anyone knows about the formulation of Kodak film – and that soft ‘shoulder’ response curve – it’s Bob.

He tells me that “The extended density range” (the capacity of film to absorb light without ‘overloading’ and thus failing to hold highlight detail) “is accomplished by combining multiple emulsions, typically named ‘fast’, ‘mid’, and ‘slow’. The need is to have the densities ‘blended’ so the curve shape is very smooth”. So it’s not just one emulsion, but a blend of several with different responses!

Incidentally, anyone my age may remember ‘Verichrome Pan’ black-&-white film, a great advance on previous films of the 1950s, because it was ‘double-coated’. Two separate coats of emulsion were layered onto the plastic base, so that very fine detail could be stored on the film, even in really bright light.

A diagram from Bob’s book – actually a scanning electron microscope image – shows the several layers of different emulsions applied to Kodak Ultra MAX 400 colour film (the so-called slow, medium and fast emulsion layers) so that the film won’t just ‘overflow when the light bucket’s full’, but will keep recording fine detail – in, for instance, a bride’s white dress in very bright sunlight.

There are eight separate light-sensitive emulsion layers here, not counting the various protective layers, and the acetate film base itself. With three magenta layers, three cyan emulsion layers and two yellow emulsion layers, all balanced to give the truest possible colour reproduction, you can see how that smooth bright-light-detail ‘shoulder’ response has been carefully formulated and physically engineered

Now for a bit of history..

Back in 2003, the late Michael Reichmann, who started the ‘Luminous Landscape’ website years ago, begat the ‘ETTR’ movement: ‘Expose To The Right’. This means that with any digital camera you’d get the best performance (both in highlights and in shadows) by using the ‘histogram’ graph which many cameras offer on their display screen (and/or in their electronic viewfinder) so as to adjust the exposure (before taking the shot) to make it increasingly brighter until the brightest highlights appear very close to the rightmost edge of the graph. But, it should be added, not so far that the highlights slide right off the edge

That, supposedly, gives the best capacity – or it did in 2003 – to retain maximum detail in bright areas of a photo, and also to gather maximum light in the darkest areas, so that shadow detail can later be retrieved (or made visible) using Photoshop or some similar editing program.

Michael R began his explanation by saying, “let’s assume for the purposes of illustration that a digital SLR has a dynamic range of 5 stops. It’s usually closer to 6 stops, but let’s not quibble“. However, current 2019 digital cameras generally have a dynamic range of about 10-12 stops according to DPR and DxO. So do Michael’s observations (well, actually, he was quoting Thomas Knoll, creator of the original Photoshop) still have relevance today?

Here’s a link to Reichmann’s original essay; but you may have to take out a (minimal) subscription to his website to read it. An updated version of his article appeared eight years later, in 2011, but that’s still eight years ago, and that’s online here. This updated 2011 version says, “most current DSLRs and digital backs have a dynamic range somewhere between 8 and 12 stops”.

Essentially, Reichmann said (to paraphrase), “expose RAW images to the right, even when taking pictures of dark scenes, and they’ll have the best possible signal-to-noise ratio (a cleaner image). Then, afterwards, move everything back towards the left of the histogram (make the image darker again) when processing the RAW file”.

This rear screen of an Olympus PEN-F shows that instead of a straight-line response to light, as in that Picture 8, up above, the camera can be adjusted. In this case, it is to give a gentler ‘shoulder’ response to the brightest light, similar to film, and also a gentler ‘toe’ response to capture more detail in shadows.

The broad ‘histogram’ graph shows that there’s plenty of bright white (that’s the tall columns jammed up against the rightmost edge of the graph, which correspond to the brightest white areas of the picture) and there’s a good range of mid-tones in the centre, too, but with no absolute black — which would have been shown by a tall section of bars way over to the extreme left of the graph.

The science

Now “Here comes the science bit..” (Jennifer Aniston, in the L’Oreal ads)

(This, as an aside, reminds me of the Dolby noise processors in cassette tape recorders. Remember them? When a tape is dragged past erase and recording heads, that dragging always recorded a hiss on the tape. So how to get rid of that? If you just turn down the treble when replaying the tape, you lose the highest frequencies which you actually wanted to hear (an orchestral triangle, say, or cymbals). The famous Ray Dolby’s solution was to ADD extra hiss (in that same particular frequency range of the unavoidably-recorded hiss) to every recording, and then, on replay, to REMOVE that added hiss (with a Dolby hiss filter) which also removed the unwanted hiss created by the tape being dragged past the heads during recording! Essentially, he added more, then took it all away afterwards).

‘ETTR’ adds extra light into the darkest areas of a photo – that’s light as distinct from the inescapable spurious hiss-akin electronic random ‘noise’ of the digital circuitry – and then by sliding the whole picture to the left (darkening it again) when processing the RAW file, that gives improvements in the dark parts of the image by having boosted its ‘signal-to-noise’ ratio through adding more light when the photo was taken.

ETTR is all well and good, if you’re an architectural photographer with a 2003-vintage 6-megapixel 5-stop camera on a tripod. But when you want a quick snap, you just can’t achieve ETTR – well, I couldn’t – while juggling focus and shutter speeds. However, I do visually assess each photo before I take it, and I use the +/- over/under-exposure option (which is there on most cameras) to apply less or more exposure based on my own experience. For instance, when shooting something black, under expose; when shooting something meant to look especially bright and white, over expose.

Here are a few domestic photos of what’s on our windowsill, including a crinkled sheet of shiny plastic. I shot these photos to just teeter on the edge of over-exposure. These are all jpeg photos (because the film scans are jpegs) all shot at ISO 800, with a 50mm (or equivalent) lens, at f/11 (but f/8 with the little Pentax Q-S1, as its lens doesn’t close down to f/11) and at 1/1000th (Pentax Q-S1 at 1/2000th).

The Nikon 50mm f/1.8G (designed to match the retro Df digital camera) and the teeny Pentax 8.5mm f/1.9 lens both have significant barrel distortion in RAW photos, but the Nikon Df and the Pentax automatically correct that in jpegs. The Leica 50mm APO-Summicron (on both the M7 and M10-P) and both the Leica/Panasonic 25mm f1.4 and the Olympus 12-100mm f4 (for micro-four-thirds) have no visible distortion in RAW photos.

• 

• 

• 

• 

• 

• 

See much difference between them? The colour’s slightly “warmer” on the “Daylight” film, and the darker areas of the window-frame surround are darker than on the other photos because film is less responsive to dimmer light than is digital.

The digital photos, all on ‘Auto’ white balance, give a less “warm”, more neutral, result.

There’s a teeny bit more detail on the top of the plastic sheet in the film photo than in the digital photos. A very teeny bit, but nothing that you couldn’t get by under-exposing the digital shots by less than minus one third of a stop. The film photo is a little darker, with a little less detail, in the shadows. Er, that’s it.

Two of the cameras have a sensor that’s one quarter the size of most of the others – is that in any way visible? – and one of the cameras has a sensor that’s about a twentieth the size of the “full-frame” cameras. Is that noticeable in any way? Apart from the smaller the sensor, the sharper the skylights look on that distant roof.

HDR merging

Of course, Michael Reichmann’s 2003 suggestion to ETTR was made two years before Photoshop introduced HDR (High Dynamic Range) merging, which allows separate ‘bracketed’ photos – shots at different exposures – to be fused into one, thus providing extra detail in the shadows (from added over-exposed shots), and extra detail in the highlights (from under-exposed shots) all blended together. Traditional film photographers, though, have ‘dodged’ and ‘burned’ since the first age of negative/positive printing, to increase the dynamic range of their photos when transferred onto positive paper. So HDR is nothing new.

Things persist on the internet; articles written years ago can appear to be brand new. For instance, look at Thorsten’s 5-page paean to the Leica Digilux 2, begun in 2004.

Although old ideas can persist, that doesn’t mean that they’re still relevant. ETTR was a good idea in 2003. But that’s now sixteen years ago.

Here’s the crunch – and we’ve been a long time getting to it – how do over-exposed film and over-exposed digital shots compare in real life?

First stop was (the very first photos, above) the Great Court of London’s British Museum. Norman Foster, the architect, created a webbed roof over the courtyard so that it is now an enclosed space where you can buy a cup of tea and a bun. (I’d shot for the stonework and the roof lattice, to see what happens in that bright and over-exposed spot where the sun drills in.)

But now let’s go outside to the front portico:

In summary..

I used to write for ‘Practical Photography’ magazine — not ‘Theoretical Photography’ nor ‘Imaginary Photography’. And in all practical ways I really don’t think that Expose To The Right makes much difference, or much sense, in 2019.

In 1968, my advice to myself was always “keep the starting handle under the seat” because the battery of my 1947 Ford Popular (or was it a Prefect?) wasn’t reliable. So I always used to park on a hill, if poss. Does that advice make sense in 2019? No. No current cars, nor even twenty-one-year-old cars, have a hole in the front for a starting handle. So a handle would be useless. It’s no longer appropriate.

Similarly, taking the trouble to ‘ETTR’ doesn’t make sense, to me, in the face of the extraordinarily good and reliable auto-exposure modes of every current digital camera. I usually choose my shutter speed and aperture, and then just ‘ride’ the ISO, letting the camera, any camera, adjust its own sensitivity to match the brightness of whatever I’m pointing at. I do, though, manually adjust the +/- over- and under-exposure dial to give me roughly what I want of any particular photo (for example, minus 1.7 for sunsets).

Current cameras are good enough to reliably choose the right exposure for themselves, no matter that negative film may have a bit more tolerance of overexposure. But few digital cameras generally overexpose anyway.

Current digital cameras produce photos with hugely recoverable detail, both in shadows, as has generally been the case with digital, but also reasonably well in highlights. And the ‘spread’ of detail, in both highlights and shadows is generally better than that of current negative film.

It’s no longer 2003.

What can we take away from all this?

People right at the top of the company which they’ve founded, or taken over (Dr Kaufmann at Leica, the late Akio Morita of Sony, the late Steve Jobs of Apple) generally respond enthusiastically to requests for info and technical details – as Bob Shanebrook, who was in charge of Kodak film products, has done here. Incidentally, Stefan Daniel of Leica couldn’t be bothered replying to my queries and Rob Macneice of Nikon couldn’t get anyone at head office in Japan to explain how the company configures the light response of their products.

It makes little difference which camera you use – unless you want one which is especially good in low light, like a Sony A7S for instance – because just about any current camera will respond well to both light and shade and will give well-exposed photos, with plenty of retrievable detail in dark areas. (See the results above, for instance, from the tiny interchangeable-lens Pentax Q-S1 which has only a minuscule 1/1.7” sensor.)

I choose cameras in order to use a particular lens, not vice versa – because, as we’ve seen, it often makes very little difference to the results which camera you use.

Negative (black-&-white or colour) film has very slightly better tolerance of bright light and doesn’t get as ‘overloaded’ by bright light quite so much as a digital camera can. But there’s actually very, very little difference between film and digital, and even low ISO film can look ‘grainy’ compared with digital.

Don’t obsess over very slight differences between film and digital. There are advantages to each. Digital is better with shadow (low-light) detail, and negative film is (only very slightly) better with highlights. But pretty much any difference between them is invisible on a limited-tonal-range computer screen anyway and may show up only on well-printed actual physical, tactile, stick-on-the-wall continuous-tone photographic paper prints. And that may depend anyway on who’s printed them, and on which paper, and whether they’re ink or photo-chemistry prints.

Finally, and essentially, instead of ETTR, I’d say DOBH — Don’t Obsess, Be Happy!