Clip-in filter for full spectrum conversion on Lumix MFT cameras

Background

I've always wanted a full spectrum camera. But I also want it to be convertible, that is, it can both be used as a standard camera, and by swapping out a filter, be used as a full spectrum camera. I chose GX850/GF9 for: a) it's cheap, b) it doesn't have IBIS, so less complexity and no risk of messing up the IBIS by altering the weight of the sensor stack. Which turned out kinda a mistake, I'll explain later.

For most cameras, there are 2 pieces of glass in front of the sensor: a blue-colored filter that corrects the color. And an IR-cut filter as the outermost layer. There's typically a piezoelectric ultrasonic dust removal element bonded to it.

Pic. 1 sensor stack break down. Bottom row, from left to right: sensor, blue filter (already cut), IR-cut filter

Full spectrum conversion

To do the conversion, I have to remove both the blue filter (it also absorbs infrared), and the IR-cut filter (dust removal function will be lost as a result). For the camera to function in 'stock' mode, I have to put the two filters back into the optical path. This is where the clip-in filter system comes in. I cemented the blue filter and the IR-cut filter together with Norland optical adhesive, ground it down to 15.55×20.15mm. And glued it into the 3D printed filter holder. It doesn't seem to introduce vignetting.

I'm reusing both the blue filter and the IR-cut filter. This is not just for cost. It's also because the total thickness of the filter is critical. Since this filter resides on the image side of the lens where the incoming light isn't parallel, a flat glass plate will introduce spherical aberration. This may not be a problem on slower telephoto lenses, but will definitely cause noticable drop in image quality on faster wide angle lenses. For reference, the blue filter is measured 2.75mm in thickness, and the IR-cut filter is 0.75mm. For full spectrum operation, an additional 3.5mm clear glass is necessary. 

Pic. 2 the sensor module with both filters removed

Pic. 3 a dial gauge is used to position the sensor correctly after reinstallation

I originally intended to use 400-1000nm AR coated glass. Which turned out surprisingly difficult to get. So I used uncoated glass instead. DON'T use common 400-650nm AR coated glass. Optical glass like BK7 has ~90% transmission throughout visible and NIR range. Those visible spectrum AR coating can knock transmission down to 75% or even less @1000nm.

As a side note, you may notice the sensor itself has an optical window on top of the actual silicon die. For my GX850, as for many other cameras, it's uncoated. However, Panasonic advertises G9 and GH5 II as having coated sensors. This may (or may not) impact infrared performance.

Pic. 4 filter glued to the 3D printed holder

Pic. 5 after the conversion in 'stock' mode, with the original filter installed.

Pitfall

I mentioned at the beginning that choosing GX850 turned out to be a mistake. On Kolari Vision's website, it's mentioned that GX850 has an IR monitoring LED that may cause IR leakage issue during long exposure at around 30s and 6400 iso. I strongly suspect this is a boilerplate statement. Here is a 15s test shot at 3200 iso, pretty bad. In fact, it's ruining my shots taken with 1/50s exposure.

Pic. 6 

I managed to get hold of a faulty shutter unit from another body. Here's the teardown. As you can see, there're actually two IR LEDs in the shutter mechanism.

Pic. 7 the IR LEDs responsible for the IR leakage problem

As a last ditch effort, I tried flocking the camera. I used stripes of black felt on copper tape. And it sort of worked. Here's another 15s shot at 3200 iso.

Pic. 8 

Pic. 9 tapes are used to block the internal IR monitoring LEDs

This level of IR leakage is not noticable for general photography, but it's still a problem for long exposures. In theory, you can subtract it out with a dark frame, but certainly not an ideal solution. Long exposure noise reduction helps, but cannot eliminate the IR glow.

Test shots

Pic. 10

Print instruction

I used ASA for rigidity. I recommend cranking up the number of bottom layers to print it solid. I included the FreeCAD project file, so you can customize it to your filter size.