Manarola

Manarola, The Cinque Terre, Italy

Yashica Mat 124G, Fuji Velvia 50

Scanning Film with the Epson V500: Resolution

Over the last few months I've been shooting a lot of film, and I needed a way to scan the negatives so that I could fiddle with them using Photoshop. I still have an old Canon 4400F flatbed scanner that can digitise 35mm film, but of late I've been exploring the world of medium format. Long, long ago my very first camera shot medium format film - it was a Holga, a kind of modern-day Box Brownie made in China. Holgas are infamous for their characterful image quality, which is objectively terrible but has a voluptuous appeal:

 

Holga 120N / Shanghai GP3 / red filter

They're made out of plastic; even the lenses are made out of plastic. Once upon a time rollfilm was the most popular film around, but from the 1960s onwards it was pushed aside by 35mm and cartridge-based Instamatic films. Nonetheless it remained popular with professionals, because the negative was huge. Amateurs like it too, because it's romantic, and there's something fetishistic about handling the rolls, spooling the film, sticking down the little tab, opening the film packet, putting a plastic bag over your head, the sweet touch of rubber bands. The big negative also flatters scanners, because you can get away with using a cheap flatbed and still have a usable file size. Which is handy, because it brings film to the masses.

As of 2012 there are essentially four options if you want to digitise medium format negatives. You can in theory send the film off to a professional bureau, who will run it through a drum scanner and post back a CD or DVD containing a high-quality scan, but prices are out of reach of all but the most dedicated amateurs. Metro Imaging here in the UK quote £55 per frame, and that's quite reasonable; most pro bureaus are of the "if you have to ask, you've come to the wrong place" variety. I remember dealing with that kind of attitude a decade ago, when I had my Holga; I'm sure they were fed up with students asking if their scratchy, proto-Instagram rubbish could be digitised, but if they'd been less British they might still be in business. Instead of going bust, which is what happened to lot of these places as digital took over. Still, I have fond memories of Joe's Basement, who were nice. And also went bust. February 2003, my word. Seems like only ten years ago or so. Facebook didn't exist then, you know? Hell, Myspace didn't exist then. Ye Gods (waxes nostalgic)

Odd thing about the Holga. The lens is actually quite sharp in the middle. Goes to pot around the edges, but it's not a complete dead loss in the centre. If you cropped down to a 35mm frame the results would be hard to tell from a good 35mm camera. That's one of the advantages of shooting larger negatives - you can crop a lot. Still, I digress. Option two is an expensive desktop film scanner, such as the Hasselblad Flextight X1, which weighs twenty kilogrammes and costs sixteen thousand dollars, so you'd need a hefty desk, and sixteen thousand dollars. Alternatively, for a tenth the price, there's also the Nikon Coolscan 5000 and 9000, which are apparently fab, but were discontinued a couple of years ago, and there are questions as to whether they can be made to work with 64-bit Windows 7. This is a problem which seems to affect old film scanners in general, although in theory you could budget for a second-hand Macintosh G3 Powerbook as a dedicated scanner interface (I have an old Toshiba laptop hanging around just so that I can use its PCMCIA card slot).

Nikon doesn't make film scanners any more; I assume that the professionals who needed to digitise their archives have done so, and there isn't enough of a market to support new models. Some professional bureaus use high-end Coolscans instead of drum scanners, with much lower prices, and indeed a lot of film development places use Coolscans as well. But I prefer to do my scanning at home, because I want total control. And the freedom to scan in the nude, or at 03:00, upside-down, etc. The coffee cup must have a spoon in it, I must have a spoon. Stir clockwise, then counterclockwise, in order to create an eddy. Alternate directions. Wait, then stir again. Press down on the base of the mug, in order to gauge the quantity of undissolved sugar. Stir some more.

Options three and four are to buy a flatbed scanner. Either the Epson V750, or the Canon 9000F. They're about equally popular, as are their immediate predecessors, the V700 and CanoScan 8800F. Both will scan up to 5x4, which might be handy if one day someone gives me a Speed Graphic press camera, with movements an' everythin'. But ultimately though I chose option four, an Epson V500. By a fair margin it's the cheapest of the lot, and in my experience is perfectly adequate for the internet and modest prints.

Desktop scanners typically have grandiose but misleading specifications. The V500 will apparently scan at a resolution of up to 6400 dpi which, if true, would render a 6x6cm medium format negative as a 15,360x15,360 pixel image - that's 235 megapixels, over 1.3 gb for a 16-bit TIFF. The scanner will produce a file that large (assuming it doesn't overwhelm your computer), but it's not really recording 6400 dpi worth of information. It simply scales up the output digitally.

I can't find a formal test of the V500's resolution, but this chap had a look at the Epson V600, which shares the same imaging engine. He comes up with a figure of about 1,500 dpi. This means that the V500 should, at the most, render a 6x6 negative as a 3,600x3,600 file, which is roughly twelve megapixels. Less than most modern digital SLRs*, but more than adequate for my needs. This isn't to say that those pixels will necessarily be sharp or colour-accurate however; that's why people pay a premium for expensive film scanners.

* Having said that, a square crop from the middle of a 21mp file from my Canon 5D MkII - old, obsolete junk nowadays, but hot stuff once - is only 14mp.

Let's do something practical. I hate blogs that waffle on. They tire me so. Using Epson's scanning software you can pick a range of resolutions, but I've found that 2400dpi captures the most information; any value above that simply produces a physically larger file, with no more detail. Contemplate this scene:

Manarola, part of Italy's Cinque Terre, shot with a Yashica Mat 124G at f/11, using Fuji Velvia 50. The Mat's lens is nice and sharp at this aperture, and Fuji Velvia 50 is famous for its resolving power, and Manarola is pretty detailed, so I'm willing to bet that the V500 is the limiting factor in this case. I scanned the same negative at 4800dpi and 2400dpi, and then bicubically enlarged the latter to match the size of the former. Here are the results, without sharpening or other processing, viewed at 100%. Original 4800dpi scan at the top, upscaled 2400dpi scan at the bottom:

The colours are slightly different - that was me, sorry about that - but there's no extra detail in the high-resolution scan. It's just larger and takes up more space. An uncompressed 2400dpi 16-bit scan comes to 173mb, roughly 5500x5500, which I save away as an archive master; I then size the file down to 4000x4000 for editing, because that's a nice round number.

Now, bear in mind that I used the standard negative holder that comes with the scanner. The scanner's optics are designed to focus at a certain point just above the glass plate, and the negative holder is designed to hold the negatives roughly at this point, within a certain margin of error, 'cause it's just a plastic mould. The other issue is that some negatives are a bit curly, and don't stay flat within the holder. Shanghai GP3 film in particular is notorious for its flexibility. The film only bends by a millimetre or so, but that's enough to bring it out of the optimal plane of focus. A company called BetterScanning makes a custom negative holder for the V500 that pinches the film between two pieces of glass, and can be adjusted to raise and lower the negative from the scanner's optics. Does it make any difference? Apparently so, and for $79.95 (not including $14.55 p&p to Western Europe) I should hope so.

Of course, resolution isn't everything. I'm not so much interested in medium format for the file size - beyond a certain point you get diminishing returns - but for the narrow depth of field, the tonality of different films, the novelty of the equipment (which is shallow, but I need something to de-funk me, otherwise I'll solidify) and... the extra control over depth of field would be more accurate.

Medium and large format film has a certain look, a combination of sharp subjects isolated from a blurry background with a shallow depth of field, moreso with large format. There's an unconscious description of the look in this New York Times article from 2005.

With one transition on the screen, that changed. In an instant, the chatter stopped, replaced by gasps and a collective groan of appreciation.

[Top photographer David Burnett] was explaining why in this age of ever more plentiful megapixels, at this moment when the concept of "film" seems as old-fashioned as a rotary telephone, he has spent most of the last two years lugging around a 55-year-old 4-by-5-inch Graflex Speed Graphic camera, complete with tripod.


On the screen was a wide overhead picture of a John Kerry rally last fall in Madison, Wis., which Mr. Burnett shot with a Canon 20D digital camera, the same camera used by thousands of other professionals around the world. Not surprisingly, the picture looks like thousands of others that were shipped around the globe during the campaign.


The colors are bright. Every part of the image is crisp, so crisp that just picking the minuscule figure of Mr. Kerry out of the huge crowd takes a "Where's Waldo?" moment.


And then Mr. Burnett flipped to a photograph taken seconds later with the ancient Speed Graphic. Suddenly, the image took on a luminescent depth. The center of the image, with Mr. Kerry, was clear. Yet soon the crowd along the edges began to float into softer focus on translucent planes of color.


The effect is to direct the viewer's eye to Mr. Kerry while also conveying the scale and intensity of the crowd. In accomplishing both at the same time, the old-fashioned photograph communicates a rich sense of meaning that the digital file does not.

That said, assuming the photograph described in the fourth paragraph is this one, the article's also describing the tilt-shift effect, which makes John Kerry look like a teeny-tiny speck of a man, a minuscule insignificance, a pygmy politician, an utter nothing, a plastic doll of a person, an inconsequential waft of a man with as much substance and lasting impact as a fart in a dirigible hangar.

35mm and full-frame digital can generally do sharp or narrow with normal lenses, but not both at the same time, although the medium format advantage has been whittled away as lens technology improves. It's interesting to have a look at this article on JuzaPhoto, which profiles the Canon 50mm f/1.0, the fastest autofocus 35mm lens ever made. It was an extreme design that pushed the envelope in the late 1980s and is still unmatched to this day. Sized down for the screen it can do the medium format look, although the pictures still seem a bit soft, and there's a tonne of vignetting. The 50mm f/1.0 is so fast that there isn't a direct medium format equivalent - the nearest is the Mamiya 80mm f/1.9, a 645 lens that's roughly equivalent to a 50mm f/1.1. Whereas the 50mm f/1.0 is soft and veily and vignettes like a mutha, the 80mm f/1.9's wide open output looks razor-sharp sized down to the screen, and isn't too bad at 100% either (another example).

In the quote up above, David Burnett was using a Speed Graphic, a 5x4 camera (five by four inches, not centimetres); the standard press lens for many years was a Kodak Ektar 127mm f/4.7, for which you would need a 35mm f/1.2 if you wanted to duplicate the look in 35mm. Burnett was using an exotic Kodak 178mm f/2.5 Aero Ektar, originally designed as an aerial reconnaissance lens. It doesn't have a 35mm equivalent, which would be something like 47mm f/0.6. Used Speed Graphics aren't all that expensive. Burnett was also a Holga fan, too.

Having said that, 80mm f/3.5 is still conservative; the next step up is 80mm f/2.8, which is the medium format equivalent of a 40mm f/1.4. And in the next post I might well have a look at such a lens, bolted on to the front of a Mamiya C33.

EDIT: But what if 2400dpi is overkill, too? Let's have a look. Here's a shot of the train station at Manarola:

And here are three scans, at 1200dpi, 2400dpi, and 4800dpi respectively, with everything else turned off. I sized all the scans to match the 4800dpi version:

Notice the bit I've circled in red. There's definitely more detail in the 2400dpi scan - you can see two clear strands instead of a blur. Looking at this, and other details in the scene, I see nothing at 4800dpi that I don't see at 2400dpi.

The Cinque Terre

Manarola

590nm

Up in the north-west armpit of Italy, along the coast, there's a line of five villages built into cliffs that descend into the sea. The Cinque Terre, or "Five Lands". A couple of them look like backdrops from a science fiction film, so I was interested to see what they were like in real life. From south to north they go Riomaggiore, Manarola - picture postcards both - and then Corniglia, which is perched on a hill and looks relatively normal, Vernazza, which is like a flatter version of Manarola, and finally Monterosso, which has an interesting name but lets the side down a bit, because it's just a beach with some hotels. "Every July, peas grow there."

A train line connects them, but in practice you can walk between the villages along a set of well-marked hiking trails. I decided to do that, but therein lies a tale, because sadly the five villages are not what they were. Well, one of them isn't. Back in October 2011, Vernazza was hit by torrential rains and flooding, as documented in this alarming slideshow. The town is currently in the process of recovery, and I decided to leave them to it. They don't need someone wandering around taking pictures of everything. Monterosso didn't appeal to me, so I didn't go there, either. And although I planned to walk from Manarola to Corniglia, the trail was blocked by a rockfall. Looking down at the trail from a great height it seemed that some slate had slid.

 
In theory there are other paths, but - mindful of the downpour that totalled Vernazza, and the gathering clouds - I decided that the local mountain rescue people didn't need to practice their skills on yet another wandering tourist, and so I, er, chickened out. And restricted my adventures to Riomaggiore and Manarola instead.

I took an infrared camera because I have orders to wait here in the spacecraft, and that's exactly what I'm going to do. A cheap old Canon 10D, converted to record near infrared in the 590nm range. I have another 10D, converted to record 720nm, and I used the same lens, a Samyang 14mm, which becomes a nicely sharp 21mm-equivalent on an APS-C camera.

 
590nm infrared is a bit less stark than 720nm, a lot more vivid. The filter is deep orange instead of black-looking red. There's a lot more natural colour, although it's all bunched up in the red end of the spectrum and still has to be extracted and massaged with digital processing. Just like how certain animals massage the bellies of their young in order to aid digestion. Here's an example, with the original file at the top and the massaged version at the bottom:

For that image I swapped the red and blue colour channels, custom white balanced on the rocks, and once I had something to my liking I did all the dodging and burning that I would normally do to a normal photograph in order to make it look like one of the photographs they have in magazines.

Of course, the photographic process began when I took the pictures in the first place. The field work. The path between Riomaggiore and Manarola is trivially easy to walk along; the later paths are apparently harder, and assuming that the villages don't end up being swept into the sea I might one day find out how true that is.

For the record I used OpenStreetMap's data to tell me where to go. OpenStreetMap is one of the few open content experiments that might save your life one day, and is well worth whatever money you decide to donate. Does Garmin sell an official map of Italy? I'm not sure; they seem to have a detailed map of the Dolomites and a city guide to Italy and Greece. Here's a smaller map drawn on one of the tunnel walls in between Riomaggiore and Manarola:

It explains quite well exactly where the Cinque Terre are in relation to the rest of Italy. Further down Italy's left coast, much nearer the boot, is the Amalfi Coast, which is conceptually similar to the Cinque Terre but posher and closer to Naples, which is unfortunate.

On the hardware side I traveled with a Garmin eTrex Venture HCx. The next model up, the Vista, adds a barometer and a compass, which I don't need; both models have since been replaced by the eTrex 20 and 30 respectively, which don't add very much. As a consequence you can pick up the Venture cheaply. Is it better than the GPS in your mobile phone? I have no idea; my mobile phone makes calls, it doesn't have a GPS. "Go to the Astra. Go to the Astra."

Riomaggiore

Ultimately I disagree with Sartre's emphasis on progression. By taking the position that existence precedes essence he locks his philosophy into linear time, and imposes the notion that existence without essence is somehow incomplete. In an essentially meaningless universe it seems absurd to argue that the essential man is superior to the man who merely exists; that the cardinal in his fine robes is superior to the idiot pottering around the garden in his underwear. The cardinal has simply cloaked himself in a meaningless, tissue-thin set of arbitrary rules, masquerading as purpose. Compared to the immense complexity of the physical world - of the mathematics involved in understanding elementary particles, quantum physics, and so forth, and of the structure of the universe - even the most complex of human bureaucracies is essence-less. The difference between the idiot in his garden and the cardinal is, on the universal scale, too thin to observe.

Besides which, the idea that human beings begin as inessential is nonsense. The human animal was not obviously designed to build cars, or write novels; it is instead a machine for living, a machine designed by a process of evolution to live. Man is born with that essence, everything it adds after that is just a transparent, prophylactic sheath, designed to shield us from the cold reality of indifferent spacetime.

It's interesting to apply the principles of existentialism to the world of computers. A general-purpose computer has no essence until it is given some instructions. Until that point it sits and waits, perhaps running some low-level housekeeping tasks. Suppose it's asked to take a series of random numbers and sort them into an ordered list. The computer is now a list-sorter, rather than a nothing. But the ordered list only has meaning insofar as the observer perceives it to have order. The universe itself, if it could perceive, would observe the computer simply transforming one set of arbitrary symbols into another set of arbitrary symbols, for no purpose at all.

On a lower level the universe would perceive the computer as a mass of subatomic particles differentiated from their surroundings by their arrangement and the temporal periodicity of their motion, if observed over time. Of course, the universe doesn't perceive, and over no time at all the computer, the programmers, the list, are erased by the great churning force of entropy. The programmers die; the machine breaks down. The cardinal's cloak falls apart, the church loses its followers. The cathedrals fall down and become a heap of stones again. One set of atomic arrangements transform into another set, of equal value.

But the universe is interconnected. A computer running a program pushes electrons in a direction and with a force they would not have taken otherwise. These electrons cause ripples in the fabric of spacetime; the ripples weaken as they expand outwards, until they are indistinguishable from the action of wind upon the water, and of the tides. But suppose we could build a computer so large and powerful that, purely by running a program, the motion of electrons within its circuits has tangible effects on the physical world - and suppose these effects were so strong that they could disrupt spacetime itself? It would be The Crasher.

The obvious problem is that, by disrupting spacetime, it would disrupt its own functioning, but if it could be built strong enough - or its construction could account for feedback, and perhaps use it to boost the power of its thoughts - then the computer would become a universal machine. Not just capable of running any mathematical problem, but also of performing any physical action. Its thoughts would resonate with and amplify spacetime.

The question is, if such a computer could be built, has an alien civilisation already built it? I have to assume that it isn't possible for a civilisation to build a machine that can quickly destroy the universe, otherwise we would all be dead. In all the universe there must have been at least one civilisation that could do so and that would be insane enough to activate it.

I have to assume that the bubble of altered spacetime this machine produces would be subject to the laws of our universe, at least along its boundary, and thus would not be able to radiate outwards faster than the speed of light. Therefore the universe might well have numerous bubbles of altered spacetime expanding outwards like lethal cancers, altering spacetime as they go. Indeed, perhaps a distant civilisation *did* activate a universe-destroying machine, but its bubble of destruction has not reached us yet. Given the continual expansion of the universe it might never reach us.

Also, there was a damp-looking cat on the trail. It looked utterly bedraggled.

Nosegay, now there's a word you don't see very often these days.