Public Lab Research note


A Wratten 25A replacement from Rosco?

by mathew | April 29, 2014 00:59 29 Apr 00:59 | #10376 | #10376

What I want to do

Find a cheap polycarbonate or polyester film filter made by Rosco that can replace the more expensive and less durable Kodak Wratten 25A gelatin filter

My attempt and results

I downloaded Kodak's absorption graph, and wanted to compare it to Rosco's transmission graphs, so I flipped it upside down:

Screen_Shot_2014-04-28_at_3.53.03_PM.png

Its distinguishing features are that it begins to pass light above 580nm and somewhat below 380nm. with that in mind, I found these Rosco filters:

Roscolux #19 Fire

19.jpg Originally suggested by Chris Fastie as pretty darned close.

Roscolux #25 Orange Red

25.jpg

Roscolux #26 Light Red

26.jpg

Questions and next steps

Beyond the transmission curve, there is another factor, the % transmission, which determines how saturated with dye the filter is. The existing Infragram filter, Rosco 2007, has 10% transmission. I'm not sure the exact transmission of the Wratten, although I found this chart:

Screen_Shot_2014-04-28_at_5.52.38_PM.png

While the curve of #19 is very close to the Wratten, the total transmission is 20%, twice the #2007. That makes me lean towards the #26, which has 12% transmission. We'll have to prove it in the field.

I have to mount in my camera and go shooting with these. I printed Chris Fastie's filter mount:

chris fastie's 3d printed filter tester for the Mobius action cam

Why I'm interested

Good NDVI on the Mobius.


6 Comments

All three of those filters will probably be good replacements for Wratten 25A. Doing some tests with them is the only way to tell, and I suspect it will be very hard to differentiate between them without extremely careful tests. Their performance is likely to vary with sky color, light quality, plant type, exposure, etc., so it will take some effort to know which is best in most situations. The curves Rosco publishes are almost certainly hand drawn from the 20 nm data under the graph (and those data may not be from the most precise device). Those three filters are so similar that even the data may not characterize the differences well.

It's a bit of a mystery how the Rosco 19 can be 20% transmission and the Rosco 25 is 14%. The data don't look that different. But in general, the more transmission the better. The photo quality will be better with more light.

Rosco 19 and 25 look identical except for the blip in the UV in the 25. One nice thing about a red filter is that the blue channel can record only pure NIR light. So it might be nice to have as little transmittance around 400 nm as possible to keep UV and blue out. So Rosco 19 might have an edge there.

Rosco 26 cuts out some of the red near 600-620 nm compared to the 19 and 25. So the proportion of red to NIR in the red channel will be smaller than in the 19 or 25. We want that proportion to be as big as possible, so the red channel can represent red light better. So 26 might be the worse choice of the three.

So go with Fire.

Reply to this comment...


Dear Chris & Mathew, Does the Rosco 19 Fire allows the light spectrum from only 560 to 740 nm range? The Wratten 25 A transmission curves from 600 to 1000 nm, so i was wondering how these two filters are identical. Also, i saw some literature discussed on the leakage of NIR into red channel after the modification. I suppose this is the reason we apply custom white balance.Am i right ?

I look forward to your reply.

Best Regards,

Suman Ghimire

I look forward to your reply.

Thank you

Best Regards, Suman

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


The spectral response graphs supplied by Rosco don't go past 740 nm. So we are assuming that there is some transmission farther into the near infrared -- we don't have any solid data about that.

Contamination of the red channel with NIR is sometimes compensated for by subtracting an amount thought to represent the amount of NIR. The primary effect of the custom white balance is to exaggerate the brightness of the blue channel (which we use for NIR). That is required because camera sensors are not as sensitive to NIR as they are to red and without the exaggeration would report that there is not very much NIR. In fact, sunlight reflected from foliage has several times more NIR than red, but a camera will never record that result unless you fool it into making the blue channel artificially really bright. Making the blue channel really bright also compensates for the fact that the red channel is brighter than it should be because some NIR is recorded there with the red.

Chris

Reply to this comment...


Dear Chris, Thank you so much for your reply. So the accuracy relies on the custom white balance settings. I found one interesting paper which applied mobius modified infragram from public labs ( Wratten 25 A filter), if interested you can have a look.However not much focus on the validation part has been presented. Also, Is there any way you suggest that we could find the exact spectral response of the filter? Link to paper: http://ieeexplore.ieee.org/abstract/document/7857255/

Best Regards,

Suman

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


Dear Chris, After some thoughts , i believe Rosco 19 is not accurate enough, i mean it collects only few spectrum of infrared. If the specification is accurate , then is there any practical advantage of this filter for NDVI? I think shifting to some other filters like Wratten 25 A could be much convenient and its not that expensive ( ~ 8 USD ) I look forward to your answer.

Best Regards,

Suman

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


Although the full range of near infrared transmission of the Rosco 19 filter has not been measured (by anyone we know about ), we are not without the means to find evidence for or against NIR transmission.

  1. Lots of people have used the Rosco 19 filter to make NDVI images and they are very similar to NDVI images made with a Wratten 25 or other filter that passes red and NIR.
  2. A well exposed photo of healthy foliage in daylight taken through a Rosco 19 filter has a blue channel image with brilliantly bright foliage, just like a photo taken with a Wratten 25 or similar filter, and just like a pure NIR photo.
  3. An NIR LED is very bright in a photo taken by a camera with a Rosco 19 filter replacing the IR cut filter. NIR LEDs with emissions in different parts of the NIR range could be used to learn more about the NIR transmission of the Rosco filter.

So our assumption that the high transmission up to 740 nm depicted in Rosco's graph continues well into the NIR region is probably a good one. More important differences between the Rosco 19 filter and the Wratten 25 include:

  1. The Rosco 19 filter might leak some visible light so the blue channel will not be as pure an NIR image compared with the Wratten 25.
  2. The steepness of the cutoff between green and red (around 600 nm) is greater in the Wratten filter than in the Rosco. This could allow more green and/or less red into the red channel when using the Rosco filter.
  3. The Wratten filter might have higher transmission of red and NIR and provide a brighter image which makes photography easier (higher shutter speeds, smaller apertures).

These three issues are probably not as important as the inherent weakness of the single camera system which uses the red channel with a (mostly unknown) mixture of red and NIR as its visible light channel. Compensating for that problem will do more to improve the meaningfulness of NDVI results than switching to a Wratten 25 filter.

But as you say, the Wratten 25 filter is not that expensive and has a well known and widely published transmission curve. I have used it a lot and have never used a Rosco 19 filter.

Chris

Reply to this comment...


Login to comment.