Project: Creating a multispectral camera

_5015948 copy

Those of you who also follow the site’s Facebook page will have noticed some images posted of late by a mystery camera, one un-purchaseable, and un-available to the general public. (And no, it’s not the new Leica M 240 or a Hasselblad Lunar.) You’ve oohed and ahh’d at the tonality, and wondered why the output was solely monochromatic. Several people speculated that it might even be from a Phase One Achromatic medium format back! The camera is in fact a Sony NEX-5, with the kit lens. The images you saw were almost un-processed; just shot RAW and desaturated.

Sometimes, I get donated photography-related things by generous readers. One of the more generous things I’ve received was this cameras: mint, almost-new-in-the-box and hardly used. Originally, the donor suggested I run a competition to give it away; though a generous offer, I’m pretty sure the camera wouldn’t interest too many of my readers – being nearly three years old and all. So, I hatched a plan: why not make it into something a little more interesting? I’ve been paying a lot of attention to black and white tonality both in the past, and of late in conjunction with my serious re-exploration of film; there’s something about the way film responds that gives it wonderful quarter and three-quarter tones. The look is achievable in digital, but it requires a lot of post processing simply because sensors do not natively respond to light in that fashion.

_N5MT_DSC0112 copy

The why

In fact, almost all sensors are optimised for accuracy of colour reproduction in the visible spectrum; this is a quantitative, tangible thing that can be done through measurement and iteration. It’s fairly consistent: accurate colour is the way most of us perceive things, to within the limits of our output media. Black and white, on the other hand, is hugely subjective: some people like mega grain and huge contrast with no midtones, others won’t settle for anything less than an over-HDR’ed mess that’s all mid tone and not much else. Personally, I’m a big fan of Ansel; it’s just that it requires a lot of work and careful exposure to achieve. But what if you could have a camera that made wonderful B&W images without much work, and better yet, had a bit more sharpness and sensitivity to boot?

Returning once again to the rationale for me shooting film – more upfront thinking, less work, something in the tones – I decided to see if such tonality was really possible natively out of a digital camera. I recall the Leica M8 creating raw files which were excellent candidates for B&W conversion because of their luminous quarter tones – this was thanks to the camera’s weak native IR filtration. These files too required some work, but not as much as heavily-filtered cameras. I wondered if there was something of a nugget here. Not only would you get more luminous blacks, you’d probably get a bit of a sensitivity boost, too – given that the sensor would be seeing light in UV, IR and the visible spectra.

_N5MT_DSC0128 copy

The how

Infrared, and to a lesser extent, ultraviolet, photography have been done for some time. There are companies out there which offer (not cheap!) conversions to either or both; there are even companies which offer services removing the anti-aliasing filter – though oddly, not both. But to create what I envisioned as the ultimate black and white camera*, all of this would have to go: no UV or IR filters, no AA filter. Just bare naked sensor. After several days of monkeying around with dozens of tiny ribbon connectors, and breaking one (caveat: the camera of course still works, but that connector will never be able to be opened/ released again) and nerve-wracking moments with various sharp implements, I’m pleased to report that this particular Sony NEX-5 has no filtration at all in front of the sensor, except for the Bayer filter, which is part of the sensor itself and thus cannot be removed. It’s about as close as you’re going to get to bare silicon – in fact, what you see when you take the lens off is the bare silicon of the sensor surface. Note that silicon is a very hard material – it’s used in portions of mechanical watches that require extreme precision and zero lubrication, and move/ interact against other parts at 8Hz or higher (the escapement) – but I still wouldn’t recommend touching it. This sensor has been properly cleaned, but may have a small dust bunny or two on it from swapping lenses. Use a blower.

*I admit I was disappointed when I learned that the Leica M-Monochrom retained its UV and IR filtration, but it turns out that decision actually makes a lot of sense, as does the 50/2 APO-Summicron – I’ll explain why later.

_5015962 copy
What you don’t see in this shot are the hundreds of little screws, connectors and oddly-shaped parts that somehow go mesh together neatly to hold things in.

The NEX-5 is a very densely packed camera indeed; there isn’t a single cubic centimetre of free internal space for any additional components. For the most part, it’s quite well designed, and there are even some bits that are quite over-engineered. Some of the connectors are incredibly tiny indeed; I wonder how long it takes them to assemble one of these things. The first time I took it apart, it was a very cautious three hours – about an hour for reassembly. By the third time**, we were down to fifteen minutes in-and-out. The machining tolerances for this thing must be extremely tight indeed: not only are there no shims in the mount, but there are only (thankfully) three very small washers between the sensor board and the main frame of the camera itself, which is magnesium. The lens mount bolts directly to the other side of this – it doesn’t really get a lot more rigid.

_5015963 copy
UVIR and AA filter pack, removed. Also, damaged – they used some insane glue to stick it to the sensor frame.

**The reason why it took three tries at all was because the camera didn’t work after the first one; it turns out I forgot a very small ribbon cable that got trapped under the main board which controlled the shutter timing. The third time was to do with focusing: the rather thick (nearly 1.5mm) filter pack was glued to the sensor heat sink frame with some very tenacious adhesive. What I didn’t realise was that this component was part of the optical formula of the system, and the lenses do not have enough additional focus travel to deal with the missing bit of glass; as a result, it was impossible to achieve infinity focus. The solution was to move the sensor closer to the mount; and here I’m thankful that there were those three little washers between sensor and internal frame, because if there wasn’t, then we’d have a serious problem – the Sony engineers did not provision a way to reposition the sensor (they must have been pretty confident of alignment and machining tolerances). Curiously, those washers were precisely the optical thickness of the filter pack (it’s almost as though they intended for something like this to be done), and the camera now focuses to infinity with all lenses.

_N5MT_DSC0121 copy

Results, notes and cautions

I’m going to be blunt here: the camera doesn’t hit full marks across the board. From a tonal viewpoint, the results are fantastic – just shoot raw and desaturate, and that’s all you have to do for almost every situation. All of the images in this post have had almost no work done on them at all – just desaturate. They came out of the camera 99% there, with this wonderfully filmic quality – even at high ISO. Is the more dynamic range? Not really. Skin tones are smooth yet delicately textured; deep shadows have that glow thanks to IR reflectance; and the detail is definitely better than a standard camera – I owned one for several months, and never saw this degree of sharpness. I wouldn’t use JPEG though, simply because it doesn’t retain as much information as the RAWs, and this will certainly affect tonal subtlety and resolution, to a lesser extent.

_N5MT_DSC0100 copy

Yet where the camera falls down is also resolution: it’s not because of the sensor; in the centre of the frame, there’s clearly a great degree of fine detail. The edges, however, look like crap. (Bear in mind the only E-mount glass I have now is the kit lens that’s bundled with the camera.) This is not because of the sensor: it’s because of the lens, which although it resolves quite well in the visible spectrum, is clearly nowhere near apochromatic enough in the corners; they look like a smeary mess in some cases. The smearing is caused by UV and IR spectrum image forming rays from the subject – cut out by the filter pack, normally – being registered on the sensor at a different physical location to visible light. There is still more visible light, of course, which means that focus is mostly where autofocus puts it, but not for all subjects – warm subjects in low ambient light – people indoors, for instance – tend to be a little back-focused because of this. Outdoors, things are fine (visible light > IR again).

_N5MT_DSC0049 copy

Results are better at the telephoto end of the zoom, stopped down, and of course with better glass. I’ve got a few adaptors lying around from my NEX days; unfortunately they’re discount Chinese items off ebay, and they perform as you would expect: crap. My M-adaptor doesn’t focus close up and has planarity issues; my F adaptor won’t focus past a couple of meters. (It would seem that their tolerances aren’t enough to deal with short flange systems.)

In my preliminary testing, the Zeiss ZF.2 2/50 Makro-Planar, ZF.2 2.8/21 Distagon and ZF.2 2/28 Distagon are all excellent performers – even into the corners. Surprisingly, the old Noct-Nikkor does very, very well from f2 onwards, and is quite passable with a little IR-induced glow even at f1.2. The Nikon 85/2.8 PCE and 45/2.8 P are also excellent, even if the former is rather impractical because of its tight focusing helicoid and inability to be stopped down without electrons. You’re probably wondering about the M-mount glass: forget the wides, they’re lousy on any mirrorless camera not optimised for them (i.e. anything that isn’t a digital Leica M). I don’t have anything telephoto. I suspect the 50/2 APO ASPH would be the lens to use on this camera, with its true optical potential seen outside the visible spectrum; and now it makes perfect sense why the Leica M-Monochrom retained its UV and IR filtration: without it, the other lenses – especially older wides – would appear very soft indeed, thanks to their property of focusing non-visible light at a different distance to visible light.

_N5MT_DSC0134bw copy

So how would I describe the tonal characteristics of this camera? In a nutshell, it produces B&Ws that are warm and rounded, if there’s such a thing. The sharpness is there but it’s not biting; the tones are rich and deep. If used with better glass, I think it would really sing – especially for portraiture. Skin looks baby-soft. Don’t use it in colour, it looks horrible due to pollution of the blue and red channels by UV and IR respectively. However, note that with a visible blocking filter over the lens, you could shoot either IR or UV without issue. The camera also gains some sensitivity – about 1-1.5 stops depending on the situation – because of the extra light it’s collecting. With the right filters, it might be an interesting tool for astrophotographers or voyeurs, for instance. Of course, you can always use one of the B&W or Leica UVIR filters and then have a regular NEX-5 again, but this time without the anti-aliasing filter.

An interesting experiment? Undoubtedly. Would I do it again, with a more interesting camera, better sensor, and something I have better glass for? Sure, why not? Though I think mirrorless makes an ideal candidate because a) you don’t have to mess around with a mechanically complex camera, and the attending realignment issues associated with disassembly (mirror, AF system etc.); b) the LCD/ EVF gives a great live B&W preview, so it makes it easier to visualise how the results will appear after conversion – the colours really are pretty funky – and c) these are just cheaper to experiment with.

Now, if somebody would like to donate a D800E to the cause, I think some very interesting results might ensue…in all seriousness though, if anybody would like to donate a camera to be experimented on (you will of course get it back afterwards, but no guarantees that it can be done) then please send me an email. MT

Note: We’re still giving this camera away. Tomorrow, I’m going to explain how – there will of course be a photographic competition involved! Update: Full details on how to enter here.

Coda: since I was asked by a couple of people over email and in the comments, here’s how the color images out of the camera look: heavy pink-magenta casts due to IR and UV pollution in the red and blue channels respectively. I suppose some (of the hipstagram persuasion) might like the look…

_N5MT_DSC0208 copy

_N5MT_DSC0128 copy

____________

Visit our Teaching Store to up your photographic game – including Photoshop Workflow DVDs and customized Email School of Photography; or go mobile with the Photography Compendium for iPad. You can also get your gear from B&H and Amazon. Prices are the same as normal, however a small portion of your purchase value is referred back to me. Thanks!

Don’t forget to like us on Facebook and join the reader Flickr group!

appstorebadge

Images and content copyright Ming Thein | mingthein.com 2012 onwards. All rights reserved

Follow

Get every new post delivered to your Inbox.

Join 23,844 other followers

%d bloggers like this: