Fuji X100: the Pitch Black project (infrared and beyond the infinite)

In the corner of my office, there is a small cabinet full of old Persol sunglasses, almost all of which have Havana Brown frames and bottle-brown tempered glass lenses. They are brilliant for photography because that shade of brown makes everything look so beautiful, and it takes a lot more than the metal eyepiece of a camera to scratch glass. The problem is that virtually all of them are now derelict, with cracked frames, missing lenses (dropping onto cement causes tempered glass to shatter into a million blunt cubes), and general acetate decay.

I worried for a little while that my Fuji X100 would eventually meet this same fate, slipping bit by bit out of use until it became a paperweight or a bludgeon-style weapon for a small child to use. So I wondered, “why not make it see something I can’t?” And so the project began to build a full-spectrum X100 (UV + visible light + infrared). It’s a little like eyeshine in Pitch Black, the rather interesting indie movie that spawned the awful Riddick sequels.

Here are some examples of what the X100 can do with its intrinsic IR capability, a Polaroid 720nm filter (like a Hoya RM72), and difficult-to-handhold exposures (as in things like 1/10 sec at f/2 at ISO 3200 – which would not be too bad if you were used to using infrared film back in the day.

Standard landscape, shadow side, unmodified camera.

Standard landscape, lit side, unmodified camera.

Backlit tortoise-shell Persol glasses, unmodified camera.

The X100 actually turned out to be the best camera for a conversion for a number of reasons:

• It has live view for focusing and a glass viewfinder for framing. This solves almost every focusing and framing problem that plagues the conversion of DSLRs. You can’t really do a full-spectrum DSLR because sticking an infrared filter on the lens blocks the viewing and focusing mechnism. And putting a filter on the sensor not only limits your options – it also means that your phase-detect focus can only be optimized for visible or IR – not both.
• It can be fitted with 49mm filters. Pretty much every bandpass filter ever made comes in this size.
• It has a color balance fine-tuning facility that works on the red-magenta axis. This is important because it helps cut down on the number and strength of filters used to shoot visible light.
• Its conventional Bayer pattern and files are supported by a considerable number of platforms, and some are coming down the pipeline (like Accuraw Monochrome) that will be able to do uninterpolated conversions.
• It is a very small camera with a very big sensor.

The filter pack. The X100 has a package of filters that sits right in front of its APS-C sensor. These consist of a UV/IR blocking filter (interference type), two layers of filters that act as the antialiasing filter, and a thin sheet of S8612 filter glass, a bluish rare-earth glass filter that knocks out the rest of the UV and IR. Most cameras have a package like this, usually adhered to the sensor.

To convert a camera to infrared only, you pick your wavelength, usually 590, 650, 720, 830nm, etc. and install the appropriate filter. The lower the number, the more visible light gets in. More visible light means that you can differentiate colors better, leading to something called “false color.” At around 830nm, something really magic happens, and everything goes monochrome (well, at least mostly; see below). It does this because the R-G-B (or Bayer) filter on the sensor does not work at all with infrared light – it all goes right through. On film cameras, there was no infrared conversion – you would put some infrared film in and slap an infrared filter over the lens. That led to a very low-performance arrangement: focusing had to be adjusted, the film picked up a very large range of light (an “infrared” filter usually started well into the visible light range, so not everything could be focused perfectly), and so things always looked a little soft. With a closed-loop focusing system on a digital camera, you don’t have to stop down – and the advent of cheap filters 830nm and up, it is easier to concentrate only on IR light.

To convert a camera to full spectrum involves replacing the filter pack with a piece of colorless glass. Shot without a filter, this leads to a reddish picture – because you have dropped a cyan filter from inside the camera and because there is significant infrared contamination in the red channel. So for visible light shots, you need to stick something on the front of the lens to block everything but the visible. Your basic choices are (and they are by no means mutually exclusive, since the X100 can use two full-size filters with no vignetting):

• Interference filter: the X100 has a 35mm FOV (@35mm), which means that you can use a B+W 486 filter with no color shift. The 486 knocks out pretty much all UV and all IR. It cannot be used on wider-angled lenses. The 486 is not actually a bad idea with any camera; it does not screw up color balance and it kills the remaining IR bleed (the X100 attenuates IR 10 stops, but that’s still an amount you might want to cut).
• BG38, BG39, or S8612: these can be in many instances the only solution you need – they knock out most IR, a tiny bit of UV, and re-compensate the color balance of the camera. The problem is that these filters are made of fluorite glass impregnated with rare earth metals, which gives you fragility (or scratchability) combined with vulnerability to moisture (Schott publishes warnings with all of them that they will decay over time). To do a really good job, you also need to block UV.

It is of some note that if you can get away without an aqua filter, then you pick up some sensitivity in the red and green channels – meaning that if you are shooting b/w, you might get a speed boost.

Filters. The big complaint about full-spectrum cameras is that you always have to use filters. This is not such a big deal if you are like me and use a B+W MRC UV to seal off the end of the X100 lens tube anyway. And certainly, you are not cabined to a camera that shoots one band of IR and has zero ability to shoot normal pictures. This is more of a consideration when you start talking about more expensive cameras.

Should you go with an IR-only camera? On an SLR camera, the answer is probably yes, because you can’t slap an IR filter on the lens and still be able to see through the viewfinder (the filter would be behind the mirror and not in your line of sight). Additionally, an SLR can’t be set up to autofocus both IR and visible light (the focus points are different). The downside on an SLR done like this is that the metering is usually in the viewfinder, so you are not metering through the IR filter.  The other thing to consider is that, contrary to some statements you have seen on the interwebs, slapping an 830nm filter over a 720nm conversion is not the same as just using the 830nm (due in part to the fact that you have to multiply the transmission curves).

The conversion. The camera went to LDP LLC (MaxMax), an outfit in New Jersey that has a pretty impressive array of optical conversion services, almost like the armorer in The Man with the Golden Gun (Bond: “Of course, yet you make guns for fingerless hoodlums, bullets for assassins…”). Why LDP? The simple reason is cleanliness. I saw the dust test from another service that converted X100s, an I was not impressed. The other thing is turnaround; LDP gets your camera back within a week.  And indeed, the camera arrived there on a Thursday and shipped out on the following Monday. The conversion was more expensive than the $450 shown on the website; it was actually even more than the $500 someone reported paying for an X-Pro.

The conversion did not change anything visibly on the camera except a tiny amount of stress on some of the leatherette. Here are some preliminary observations (and some sample shots will come with the next installment).

General exposure: with 720nm and 830nm (B+W 093) filters, there is little predicting where exposures will end up (unless you pay very close attention to whether CFL bulbs are the light source – they emit very little usable light). The 720nm filters generally expose similarly to the uncoverted camera. That points to absolutely huge sensitivity to IR in the CMOS sensor, since you are basically lighting an entire scens with wavelengths that humans can’t even see. At 830, you lose about two stops in most situations that involve sunlight or incandescent light. Fluorescent lights produce very little IR, and exposure times rise radically.

093 filter, hot lights

Color. You’ll obviously want to pick a pleasing color balance, but this is where you land with the various permutations as they show in auto white balance and with some tuning:

• No filter – big red cast
• No filter (cyan +9, blue +4 to +9): reasonably good auto white balance; blue “wood” effect.
• 720nm (no trim): pink or blue to white. Many woods and some plastics look blue.
• 720nm (cyan +9): neutral plus blue.
• 830nm – monochrome red-violet to white. Don’t worry about false color; just put the camera in black and white mode or convert to b/w in Lightroom.

The AWB on the camera will generally get you to a place where for 720nm, things will look like Frankie Avalon in heaven in Grease. Strike that in part. The clothes look like Gene Kelly’s suit in Xanadu. Faces still render fairly normally; some synthetics look pinkish or bluish. Wood looks blue. With 830nm, it’s monochrome purple (easy to turn to grey).

AWB for hot lights (720nm). Note the ghosting that occurs – lens multicoating is ineffective in IR light.

You can, of course, do what you want with false color to suit your mood.

720nm, false color

Monochromaticism. There seems to be a little bit of misinformation about monochromatic operation in infrared. You hear that in the high infrared range, Bayer filters become transparent. With the 093 filter, which may be as far as you want to go to shoot available light, the apparent effect is monochromatic in files, but a dump of the raw file using the Unix command-line program dcraw (use the -d flag) reveals that the various channels are not exposing exactly evenly.* Accuraw Monochrome promises to fix this and prevent the false noise that occurs.

*Why are all off-the-shelf OS X builds of dcraw so old that they can’t do the X100 (let alone the X-Pro1)? To get dcraw to work, you need to install xCode (1.72Gb plus) and then do a recompile. Getting xCode for a machine with OS 10.7 or earlier is a lot like pulling teeth. I will see if I can’t find a way to make this build available. That said, dcraw does not do a great job with 093 to monochrome conversions, particularly at high ISOs.

Take for example Mr. Spats (the insets are at 200%, so you can see the effect more clearly).

Overall view, made mono in Lightroom

Detailed view, LR conversion (200%)

Straight dcraw conversion (200%)

dcraw RAW dump (200%)

Cautions for social photography. IR photography has two uses that are interesting. One is landscapes, where you can help cut some haze and get more dramatic skies and plant tones. The other is for available-light social photography, where people’s faces light up much more brightly in IR than they do visibly (I have no idea why this is; my guess is that ceiling mounted can lights, even when apparently dim, can emit a lot of IR radiation that we just can’t see).  Aside from that, there are many practical observations about taking pictures of your friends.

• Red-eye is fierce when you use IR and flash. Everyone is a Replicant, and Lightroom does not see the red-eye to correct.
• IR pictures without flash can sometimes give “doll eyes,” depending on the iris color.
• Be careful with some clothes that go from black to white; nothing becomes see-through, but you can see undershirt lines and the shadows of other bulges that you wouldn’t notice on a black background.
• At 720nm, it’s very easy to spot where people have hit the Just for Men or the Clairol hair coloring products because they come out in unusual colors (most hair and some skin looks blue, but there is a hyper blue from the dyes). For some reason, dyed hair has a distinctive signature. I need to do more testing to confirm this.
• Note that the X100’s focus assist lamp does not transmit any usable focusing assistance for 830nm and up filters.

The icy effect of rebalanced 720nm. Or is that the cold shoulder? This shows you how you can get quite a bit of usable light in environments where you can barely see.

And another. Here is some fierce IR lens flare.

Not for X-Ray. There seem to be some pervs who think that infrared is great for “see-through” effects. Let me offer some observations on this:

• If your goal is pornographic, there are many better ways to spend $550.
• Most people wear more than one layer of clothing.
• Forget about fabrics – most materials in general are not sufficiently IR porous to allow light to go in one way, bounce off something underneath, and then make it back to the camera. That “fake check” thing is very hard to reproduce.
• I’m sure you can teach to the test by going outside in massively strong sunlight and make an attack on the thinnest, chintziest synthetics, but most synthetic materials actually reflect IR brightly, to the point that black becomes bright white.

My suspicion is that wherever these effects do exist, it’s at wavelengths that are very difficult to shoot anyway (950nm and up).

Upshot. This is still under heavy testing, but on balance, the better low-light capabilities of a converted camera are fun – and open up some doors you might nor otherwise see. That said, where the camera has a lot of IR capability already, you may be better off living with the longer exposures (or getting a tripod). We’ll update over the next few months.  The issues currently under study are false noise (or high ISO noise, as the case might be) and ways to improve 830nm focusing operation.  But it’s been fun so far.