Easier Aerochrome: The Triple Bandpass TB550/660/850

[Christoph]

Will try that, Andy! Thank you! ... Just tried to do an IR test, so find out if I have IR contamination in the channels. But it didn't even get to the actual test. I used a 830nm filter from B+W and the lamp from my IPhone. But it seems like it still lets visible light through. I can even see it faintly with the naked eye. Is this a known problem? That even IR filters of "Renommée" still let visible light pass?

[Andy Perrin]

It’s actually two things. One is that those filters can have a gradual cut on, so it’s letting through some stuff below 830nm normally. The other is that human beings can apparently see further into IR than books and the web would have you expect. Stefano claims he can see LEDs on tv remotes, but even I am pretty sure I can see into the mid-700nm range if the light is bright enough, as verified with some very good filters from Thorlabs with sharp cut-ons.

 

The OTHER possibility is that you have a cheap Chinese filter and it’s out of spec. That is also a known issue.

[Stefano]

Yes, I don't know if I can see the 940 nm peak, but in the darkness I do see those LEDs flashing red. I know for sure I can see 808 nm as I have a laser emitting there (I modified a green laser pointer).

 

I can easily see through a Hoya R72 filter, and I can see through a ZWB2 (2 mm) + Chinese BG39 (2 mm) stack. I don't have deeper IR filters to test this further though.

[Andy Perrin]

Yeah R72 is dark but definitely if you hold it to your eye most people can see through it.

[ultrainfra]

This post reminds me... I have one of these! A little 37mm one. I haven't had the chance to use it. Got it dirt cheap for about 20$ (USD) a while back, but haven't done much photography lately.

[Christoph]

Filming is definitely easier with this filter than with the orange-green filter method discussed in another topic. The blue vegetation, which turns reddish after the channel swap, is already present here, which greatly facilitates processing. Here are some impressions:

 

https://vimeo.com/520877120

[colinbm]

You are set for some fun in the summer.

[Christoph]

I always try around a bit with the triple bandpass filter. Maybe someone has a quick answer for the question I'm trying to answer: What would the transmission curve have to look like for the IR contamination of red and green to be smaller? Would this matter at all? What if the transmission of 850nm was only 50% instead of 90%? How would that affect the result? Midopt TB

 

My thought would be this: what if I used a strong cyan filter in addition to the TB, affecting primarily red and NIR? For example this one from Lee

[Andy Perrin]

Everything past 850 goes equally into all three channels. Between 720 and 850, there is no blue, some green, and twice as much red as green. If you block this region, you get the visible blue red and green plus equal amounts of IR in all three. So lowering the stuff past 850 equally should be a good move. 

[OlDoinyo]

Some sample camera response curves are published here. You may think of these similarly to filter curves. To get an idea of what a filter would do, perform a point-by-point multiplication of the filter curve and the sensor curve.

[Andy Perrin]

OlDoinyo, that doesn't look right? That data doesn't show the IR response becoming equal for above 850, which we all know to be the case. For example Thorlabs shows this behavior:

 

I think those old MaxMax plots may not match what current cameras do.

 

[dabateman]

I also think Dan might have over corrected the green channels.  They should all be lifted to match the red and blue above 850nm. Then those plots would look more correct. 

I can't remember if those are divided by 2, as we have two greens for every red and blue.

 

[Christoph]

Thanks for your answers! :-)

[OlDoinyo]

The response curves are congruent above 850. But they can still differ by a simple constant. That is why IR images come out of the camera tinted.

[Andy Perrin]

OlDoinyo, that tint is caused by the white balance actually. They don't really differ by a constant, you are just seeing the effect of the white balance multiplying the channels. The original RAW does not have that until after WB is applied. The unWB'ed curves look like that one Thorlabs posted above.

[Christoph]

discount-optics has a TB filter on sale on ebay ... TB/550/660/850 MIDWEST OPTICAL MIDOPT TRIPLE BANDPASS FILTER GREEN+RED+850NIR | eBay

[Christoph]

I think I have the perfect formula together slowly but surely. This is the Triple Bandpass - integrated into the 28-80 Nikon lens - filmed with the Nikon D7200... The processing is extremely simple: Set the Kelvin value in-Camera to 2500, then enter the following values in the channel mixer in Premiere:

R: 0, 0, 100
G: 100, 0, -20
B: 0, 100, -20

https://vimeo.com/manage/videos/673249410

That's all. Color changes fit. It's important to me to get as close as possible to the Aerochrome look of Richard Mosse - that means pink meadows, red trees, blue sky. 

PS: I would grab the remaining TB filter (benefit nothing from this recommendation). But it's quite possible that there won't be any left to buy "cheaply" for a while. And 27mm can be put behind the lens.

[Christoph]

I removed the ring on my filter and attached it to the back of the lens. Like this: Disassemble a filter | Hidden Realms

[Andy Perrin]

The colors look nice, but that video looks very underexposed, at least on my monitor.

[Nate]

Colors look great. The video seemed to be exposed fine on my end

 

[Christoph]

 

[Stefano]

Looks nice, it also has a natural feel to it (the blue sky probably helps in that).

[Christoph]

Yes, it's pretty close to my ideal. Finally there was good weather. I chose 2500 Kelvin as white balance. But it is not the perfect value. Perfect would be if you could WB on vegetation in-camera. But that does not work for me so far. So: take white balance either in-Camera from vegetation or in Post-Processing via eyedropper, as I have done.

Load the video into Adobe Premiere. Then enter the following WB settings in Lumetri Color:

Temperature about +75, Tint -45.

Then the channel mixer. Very simple:

R: 0, 0, 100
G: 100, 0, -40
B: 0, 100, -40

The values with 0 and 100 are mandatory. The -40 can vary depending on white balance and color preferences. Remember: The lower the value, the more intense the reds will be. So I recommend to set the value between -30 and -60.

That's it. I did not change the contrast or anything else.

[colinbm]

That is excellent Christoph

[Christoph]

Unfortunately I can't offer you more water...