As a photographer interested in the weirder side of optics, I have long kept my eyes open for unusual techniques. Years ago, I saw another account on Instagram post some trichrome photography they took using CMY filters and B/W film. I had never seen the process before but I was instantly intrigued; I loved how the changes in the landscape over the seconds it took the photographer to exchange filters showed up in the different color channels during post-processing. Over the years, I have done my fair share of camera modding. A while back, I had converted a Pentax Q7 to full-spectrum for some good-old infrared (IR) photography. This got me thinking:

"What if I did trichrome but with colors we can't normally see?"

This is the story of my ongoing infatuation with hyperspectral trichrome photography, and the lengths to which I have gone to make the process more fun.

Background

You may be asking, "what is 'trichrome photography' and what might make it 'hyperspectral'?"

Trichrome Photography

Before color photography (and color film emulsions) many photographers still desired ways to make their photographs colorful using the tools they had at hand. Limited to black-and-white (B/W) film emulsions, many photographers used a technique that James Clerk Maxwell first proposed, known as the 'three-color-process', 'trichrome photography', or sometimes just 'trichromy'.

In this process, the photographer would set up a tripod on location and take three B/W photos, one each with a cyan, magenta, and yellow (CMY) filter in place. In the darkroom, they would then use special printing processes to overlay all three photographs using the three different colors corresponding to each filter in order to reproduce an approximation of the original scene's colors. Back in the day, this was a very long and expensive process and, while it is still harder than simply snapping a color photo, can be done today using RGB color mixing and layers in modern digital photo editing suites.

Hyperspectral Photography

In the realm of light and photography, any images taken using light outside the human visible spectrum are generally considered to be hyperspectral. This would technically classify the X-ray photograph of your compound fracture as a hyperspectral image which I will not dispute. However, I use the term to refer to photos taken using both the ultra-violet (UV) and the infra-red (IR) light spectrums. Light in both the UV and the IR are outside of normal human perception, by definition, therefore the hyperspectral label.

Hyperspectral Trichromy

The combination of the above two concepts, in which multiple hyperspectral images are combined to form a single human-visible image. Generally, my technique is to use UV as the blue channel, visible light as the green channel, and IR as the red channel.

Apparently, my 'aha' moment above is not an original thought. People have been doing infrared trichrome photography for years, even back in the days of film photography. There are photographers that did, and some that still do, take IR trichrome photographs on black-and-white IR film (like this one from Rollei), often to fantastic effect. I wanted to take it a step further and mix in UV as the blue channel for a true 'hyperspectral' technique. When I first started, the closest thing to a UV filter I had was a cheap ZWB2 bandpass filter which was not ideal, but sorta worked. Using it with my full-spectrum Pentax Q7 and some IR long-pass filters from Kolari, I got these photos of my college campus:

The ZWB2 filter was not ideal for the technique I had in mind since, while it passes a lot of UV, it also passes a lot of IR. Because the digital sensor is more sensitive to IR than UV, the IR (red) channel gets muddied out by stray infrared light in the UV (blue) channel. As a starving college student, I could not afford a more expensive filter like the UV filter from Kolari, but I did eventually pick one up about a year later. My results were much more satisfying:

I have been wanting to do more trichromy. I found the process of setting up the tripod, screwing, unscrewing and juggling multiple filters and then editing the photos together in post to be quite a hassle, but I have always loved the results. I think trichromy renders the most vibrant false color of any other hyperspectral technique. While the post-processing is part of the game, and the tripod is pretty vital, I thought the filter-juggling process could be streamlined.

The Lens

After searching and brainstorming, I found the vintage Minolta MC FISH-EYE ROKKOR-OK 1:2.8 f=16mm lens which has a built-in filter changer! This lens was first designed back in 'the good old days' when B/W film was popular enough to warrant such a thing as filter-based design considerations.

I decided to acquire one of these lenses and replace some of the filters in the carousel with UV and IR filters for quick-and-easy filter changes on the fly. I specifically sought out one that was in rough shape because I knew my project would leave the lens forever changed and I didn't want to feel guilty desecrating a mint-condition specimen. Before I modified the lens, I thought it wouldn't be fair unless I first gave the built-in filters a chance to impress. I tested it out with the Pentax and these were the results:

Drag the slider to compare images taken with the O56 filter before (left) and after (right) a Red/Blue channel swap:

The only filter out of the bunch that I was really satisfied with was the O56 filter, which performs admirably as a 560nm long-pass filter for near-IR photography. The 80B filter was not strong enough to block any of the IR and so would be awful for trichromy. The 1A filter might as well not have been there, and a full-spectrum can be passed by leaving the filter selector ring in between selections anyways (no filter in place) so there was no point in keeping it. The Y48 filter was also weak and thus not particularly useful; I especially did not appreciate the green cast it rendered in the sky after a red/green channel-swap.

The modifications would proceed.

Filters

In photography, a filter is used to allow only a subset of the available light in the scene through to the sensor. For this project, I chose a variety of long-pass and band-pass filters. Long-pass filters will only allow light with wavelengths longer than a specified cutoff frequency. A band-pass filter will allow a certain bandwidth of frequencies through, usually centered around the specified frequency.

I ended up custom-ordering some high-quality optical filters from a supplier in China which was kind enough to provide a selection of several different types, cut to the exact dimensions I requested. For this particular lens project, dimensions of 1.1mm thick by 13.5mm diameter were perfect. Many IR long-pass filters are the absorptive type, meaning the visible light is absorbed and converted into heat. I spent extra to get the reflective types which use thin-film deposition of metallic and semi-metallic elements onto a glass substrate to create a mirror that selectively reflects visible light while allowing non-visible wavelengths through. In my experience, these types of filters produce better images.

Filter Selections

These are the filters I ended up selecting for the project. I have included the transmittance curves provided to me by the manufacturer at the links in the wavelengths column. Unfortunately, due to my own clumsiness, the SP700 filter broke before I could install it. So, I left the original O56 orange filter since it works as a mediocre near-IR long-pass filter.

Filter Removal

The process of changing the filters was challenging and fraught with several impediments. I had originally thought it would be easy enough to unscrew the front element to get access to the filter changer mechanism underneath. I knew this was possible after first seeing a lens repair video detailing the process, but they failed to mention that a normal spanner wrench was not thin enough to reach the tiny tabs recessed deep within the front of the lens, protected by the built-in tulip lens hood. I ended up needing to make my own custom spanner inserts and then 3D-printing a 'wrench' to do the job:

Once the front element group was safely removed and set aside for safekeeping, the mechanism behind the filter changer feature was laid bare for me to study. It was not exactly as I had expected, but upon closer examination, the operation of the mechanism was obvious: an indexed ring with a cam on the inside is positioned such that when, moved to any of the 4 indexes, it would displace one of 4 pivoting followers arranged in a ring about the perimeter of the lens body. This follower, when displaced, would pivot about a shaft that transmits the rotation to an armature inset with the corresponding optical filter. The displacement actuates a sweep of approximately 90° allowing only one of the 4 filters to eclipse the irising diaphragm aperture at a time while the other 3 are stowed. While a simple rotating filter platter would have been a more simplistic solution to this engineering problem, I cannot help but admire the lens designers for going above and beyond in their implementation of this mechanism.

In the end, I had to settle for a destructive approach to the filter swap; they were cemented into the filter trays too well to remove them while preserving their integrity. I used my flex-shaft rotary tool with a diamond-coated burr to painstakingly cut holes in the centers of each of the filters I wanted to remove. After the hole was made, I used a hooked pick to reach through the hole and pull the broken filters out the front of the lens. Before any of that, though, I masked off the aperture iris to avoid getting glass dust in between the blades or the lower lens mechanisms. The process was just as horrific as it sounds and the whole time, I thought I was about to make an expensive mistake. Somehow, though, I pulled it off.

Filter Replacement

With the old filters removed, it was as easy as simply applying adhesive to the filter trays and popping in the replacement filters. Even though the adhesive should theoretically only need to contact the perimeter of the filters I chose to use an optically-transparent UV-cured polymer cement that I have used to stack optics before. This stuff is great because it is viscous enough not to get everywhere during application and it cures up very quickly (~5mins) under a UV light. If there is a mistake when applying it, you can simply wipe it back up and start over again since it won't cure in the absence of UV light (don't work with it outdoors or near a window). Finally, considering the most important factor for optical applications, this adhesive does not off-gas during curing, since it is a cement and not something dissolved in a solvent that needs to evaporate off; Cyanoacrylate glue, for instance, will cause a white haze to appear on optics around where the glue is applied, irreparably ruining many types of optical coatings and the surfaces of refractive elements.

My personal advice is to NEVER use superglue or related types of adhesives in any optical application.

With the new filters in place, I only needed to do some clean-up before the modified lens was ready for reassembly! The video below shows the 3 filters I installed + the O56 that I left in.

Mounted on my full-spectrum Pentax Q7, I was ready to go test some trichromy.

Gallery

The next day, I was able to take these test shots:

The shooting experience was much improved with the new setup, making filter changes a lot quicker and easier. I think the #1 benefit is not having to risk breaking the fragile but expensive filters as I unscrew them and fumble about in my camera bag.

Final Thoughts

I enjoyed the challenge of this project, though the number of unknowns and the high stakes definitely made this one of my more stressful projects in recent memory. There were several points during the process that could have gone terribly wrong and effectively killed the project forever. Though I was out of my comfort zone on this one, I am still really glad I decided to do this.

One thing I still plan on doing is replacing the O56 filter with a good SP700 hot mirror filter (which I have re-ordered) since I broke that one while installing it improperly. Doing so will hopefully enable me to take some good 'normal' photos in the human-visible spectrum to use as the green channel; while using the full-spectrum photos for green works alright, it muddies out the IR/red channel quite a bit, leading to yellow foliage instead of the robust red foliage I love. I will probably either amend this article with an update once this happens or make a shorter supplementary post.

Thanks for reading,
~Joseph