Interior Filter / IR Pollution test

[Richard Bradbury]

Hello Folks

I've just finished a set of RED tests with a studio test of some filters (thanks to Carey at South London Filters). The tests were principally looking for IR pollution, both with and without NDs in front of the lens. We looked at standard Tiffen NDs, Tiffen ND / IRs, Tiffen 80 series blues (to moderate the native daylight chip colour temperature), and a Schneider 486 UV/IR cut (round photo filter). Quicktimes are below.

We tested under tungsten lamps (2x 2Kw), first naked measured at 3000K, then with CTB to get us to 5600K, then at 3200K for 80 series filters. The camera was shooting 4K 2:1, Redcode 28, 25FPS, 1/50th, 320ASA. White balance was adjusted to account for lighting / filtering. We exposed to a zebra set at 45-55 IRE, reading off the grey card in shot. Lens was a Century short zoom (T3) at about 40mm. All the details are written on the board in each shot.

3000K
No Filter: http://www.filmspace.net/RED/RedMagA002_C071_080328_.mov
ND 0.3: http://www.filmspace.net/RED/RedMagA002_C072_080328_.mov
ND 0.6: http://www.filmspace.net/RED/RedMagA002_C073_080328_.mov
ND 0.9: http://www.filmspace.net/RED/RedMagA002_C074_080328_.mov
ND 1.2: http://www.filmspace.net/RED/RedMagA002_C075_080328_.mov
NDIR 0.3: http://www.filmspace.net/RED/RedMagA002_C076_080328_.mov
NDIR 0.6: http://www.filmspace.net/RED/RedMagA002_C077_080328_.mov
NDIR 0.9: http://www.filmspace.net/RED/RedMagA002_C078_080328_.mov
NDIR 1.2: http://www.filmspace.net/RED/RedMagA002_C079_080328_.mov
NDIR 1.2 with 486: http://www.filmspace.net/RED/RedMagA002_C080_080328_.mov
ND 1.2 with 486: http://www.filmspace.net/RED/RedMagA002_C081_080328_.mov
No Filter with 486: http://www.filmspace.net/RED/RedMagA002_C082_080328_.mov

5600K
No Filter: http://www.filmspace.net/RED/RedMagA002_C083_080328_.mov
ND 0.3: http://www.filmspace.net/RED/RedMagA002_C084_080328_.mov
ND 0.6: http://www.filmspace.net/RED/RedMagA002_C085_080328_.mov
ND 0.9: http://www.filmspace.net/RED/RedMagA002_C086_080328_.mov
ND 1.2: http://www.filmspace.net/RED/RedMagA002_C087_080328_.mov
NDIR 0.3: http://www.filmspace.net/RED/RedMagA002_C088_080328_.mov
NDIR 0.6: http://www.filmspace.net/RED/RedMagA002_C090_080328_.mov
NDIR 0.9: http://www.filmspace.net/RED/RedMagA002_C091_080328_.mov
NDIR 1.2: http://www.filmspace.net/RED/RedMagA002_C092_080328_.mov
NDIR 1.2 with 486: http://www.filmspace.net/RED/RedMagA002_C093_080328_.mov
ND 1.2 with 486: http://www.filmspace.net/RED/RedMagA002_C094_080328_.mov
No Filter with 486: http://www.filmspace.net/RED/RedMagA002_C095_080328_.mov

3200K
No Filter with 486: http://www.filmspace.net/RED/RedMagA002_C096_080328_.mov
80D with 486: http://www.filmspace.net/RED/RedMagA002_C097_080328_.mov
80C with 486: http://www.filmspace.net/RED/RedMagA002_C098_080328_.mov
80B with 486: http://www.filmspace.net/RED/RedMagA002_C099_080328_.mov
80A with 486: http://www.filmspace.net/RED/RedMagA002_C100_080328_.mov

NB We measured the 80s colour temperature shift as follows (quite suprising that the 80B was stronger than the 80A):
80A - 1600K
80B - 1800K
80C - 900K
80D - 600K

The RedCine workflow on these clips was as follows:
Project settings: 1280x720, fit height, Camera RGB, Gamma Rec.709
Shot settings: Fit height
Color settings: No adjustment
Output settings: Format Quicktime, 1280x720, H.264, Filter sinc, Process standard, Quality normal, Bitrate 1000kb/sec.

The bitrate control setting didn't work (does anyone have any experience of this?), and as a result the files were huge. So I recompressed them with Compressor using the same settings, but limiting the data rate to 2500kb/sec.

The upshot was that, as expected, there was a magenta shift that became stronger the more ND was in front of the lens. This was very pronounced in certain black fabrics (as documented elsewhere on the forum) and red/orange colours. The Tiffen ND/IR filters didn't do very much to correct this effect, but a Schneider 486 did a brilliant job (we only had a photo size filter). In my opinion the magenta shift is less noticeable under 5600K lighting than 3000K, and a 80 filter helps in the same way (although gives you all sorts of other issues).

I'd be happy to hear peoples' comments & ideas.
Richard

 

[Gunleik Groven]

Uploading?

gunleik

 

[Richard Bradbury]

Now uploaded. You lot are too quick.

 

[ericyoung]

How much visible light did the IR filters lose? Not very much I would expect. Is there a case for permanent mounting of IR cut filter on all lenses intended to be used with Red?

 

[Miltos Pilalitos]

Very informative test. Thank you.

 

[Jean Déraps]

Interesting test. I'm just wondering what would be the benefits of correcting for this shift in camera as opposed to in post? Each additional piece of glass has the potential of degrading image quality and it may be difficult to accurately correct for the exact amount of shift depending on the lighting conditions. On the other hand, all footage will be colour-corrected in post, where it would be relatively easy to remove the colour-shift. Or is there something I'm missing?

Thanks for the informative test!

Jean

 

[Richard Bradbury]

How much visible light did the IR filters lose?

If you mean the 486 (IR/UV cut) then I don't think it absorbed any visible light. There's a well discussed thread about their use already here:

http://www.reduser.net/forum/showthread.php?t=9412

And yes, in many situations this is going to help a lot, although beware of super-wide lenses since light hitting the filter from a very oblique angle can begin to become cyan.

what would be the benefits of correcting for this shift in camera as opposed to in post?

If you mean the magenta shift I don't think there's any doubt that it's better to record the right colour to begin with! Removing stray IR isn't going to introduce noise/gain of any kind, since it has no effect on visible light. If you mean the white balance issue I'm not the best person to answer this, but it looks to me from the test that the camera is performing much better under 5000K - 5600K light. Have a look here for much more:

http://www.reduser.net/forum/showthread.php?t=7454

 

[Herring]

Thank you for this test!

A very importent argument for filtering on set is that you are not only correcting the meganta shift but also preventing the infrared to get to your sensor. The parts of the image that reflect a lot of infrared light appear to be blurry, out of focus and washed out. This is a feature you cannot correct in post.

 

[jbeale]

also, if you have (for example) black cotton and black nylon fabrics in the shot, and you have some IR present, one material will look black and the other purple, I don't think that can be simply corrected by an overall color shift. Best to eliminate the IR issue up front.

 

[Andrew Madu]

Thank you for this test!

A very importent argument for filtering on set is that you are not only correcting the meganta shift but also preventing the infrared to get to your sensor. The parts of the image that reflect a lot of infrared light appear to be blurry, out of focus and washed out. This is a feature you cannot correct in post.

It's interesting that DoP's like Oliver Wood have resorted to using no filtering whatsoever.

 

[Herring]

The infrared-problems we are discussing only occur with digital sensors, not with film. Is Oliver Wood already working with a Red?

It's not that you cannot film without an infrared filter (I did so myself on two commercials without problems), but you can run into problems. And it is good to know how to avoid them.

 

[Andrew Madu]

It's not that you cannot film without an infrared filter (I did so myself on two commercials without problems), but you can run into problems. And it is good to know how to avoid them.

Indeed.