2. Explanation
Neutral density(ND) filters are used to cut the amount of visible light transmitted to the camera sensor in order to achieve an
acceptable exposure in brightly lit conditions. However, not all ND filters are created equally and no matter how expensive
the manufacturing process, not all of the visible spectrum can be affected, allowing small amounts of near infrared(IR) to
contaminate dark or black objects. This contamination becomes more prevalent at higher ND strengths, and is typically
observed above 0.9ND (3 f stops) when used outdoors under daylight.
The reason why we can start to see a red cast in dark tones is twofold:
1) Dark objects tend to reflect more IR and less visible light
2) NDs cut the amount of visible light and therefore change the ratio of visible light to IR light. Because it’s near impossible
for an ND to be manufactured to attenuate all of the visible spectrum, the small amount of IR light that can pass through an
ND filter is now in a greater proportion to the visible light reaching the sensor.
The visible spectrum is transmitted between 400nm(blues) and 750nm(reds). Most cameras have a built-in IR filter that
transmits most of the light between 400nm and 680nm. From 680nm to 700nm, the IR filter begins to reduce light
transmittance and therefore the camera loses some red sensitivity. From 700nm to 730nm the IR filter quickly cuts the
transmission of light to the sensor but not all of it. Generally the amount of IR picked up by the camera is minimal and is
washed out by the abundance of visible light.
3. Hypothesis
The original impetus and explanation occurs here at provideocoalition.com:
http://provideocoalition.com/index.php/aadams/story/red_ir_filter_test/
The author, Art Adams, found that the RED camera had large amounts of IR contamination when used out doors with large
amounts of ND above .9 strengths (3 f stops).
I contend that this red contamination can occur just as readily indoors under tungsten lighting but at lower ND strengths.
Recently while shooting Jewelry, I constructed a large tented surface lit with about 7000 watts of tungsten lighting. I did
this because we were using diopters and needed a large amount of light to achieve a stop of f16 to hold depth of field. This
was successful and I needed to add a .3ND (1 f stop) to achieve proper exposure. I noticed that when the fabric used for the
background (Velveteen) was at about a 45’ angle of incidence between the camera and the light, it would appear reddish in
color. The camera used was a Panasonic HPX-500.
4. Testing Methodology
Target setup:
I set up three common black cloth backdrops that we have in inventory – A studio curtain, duvetyne and velveteen. The three
samples were lit using two 650 watt Arri units .
Camera setup:
The camera was white balanced using an ExpoDisc Digital White Balance Filter (www.expoimaging.net) then black balanced. The
camera recorded the color temperature as being 3000’K. Exposure was judged using an 18% kodak gray card set to 50% IRE
using the camera’s internal spotmeter.
Test workflow:
I shot the backdrops using zero ND for control then .3, .6, and .9 strength ND Tiffen filters. The clips were imported into Avid,
frame grabs were exported as 8 bit TIFFs and imported into Photoshop where I analyzed the RGB values.
5. No ND control target
The white balance appears to be spot on. The gray card values are mostly equal and exposure of 50% IRE is very close to the
ideal R128 G128 B128 values for gray.
The studio curtain has good black level values and the duvetyne is reflecting the most light out of the three black targets.
However, the studio curtain and the velveteen already seem to contain more red compared to the duvetyne, although visually
the studio curtain appears to be the truer black of the three targets.
6. 3 ND
The gray card values are mostly equal and exposure of 50% IRE is very close to the ideal R128 G128 B128 values for gray,
perhaps just a little under exposed.
The duvetyne is still reflecting the most light back out of the three black targets, however, the red levels in the velveteen have
spiked considerably and it has taken on a red cast.
7. 6 ND
The gray card values are mostly equal and exposure of 50% IRE is very close to the ideal R128 G128 B128 values for gray,
although it is interesting that the blue levels have dropped.
Of note is that the duvetyne RGB levels haven't varied much from the initial 0ND readings. However, the red levels in the
velveteen and studio curtain have spiked considerably, with a slight red cast now visible in the studio curtain. Note that the
green and blue levels are consistent with their .3ND readings.
8. 9 ND
The gray card values now vary with an 11 point difference between the highest and lowest value. it is interesting that the blue
levels have dropped as red increases.
Of note is that the duvetyne RGB levels haven't varied much from the initial 0ND readings and the spread of values is minimal.
However, the red levels in the velveteen and studio curtain have spiked considerably, with a visible red cast in the studio
curtain. Note that the green and blue levels are consistent with their 3ND readings, and the greatest change has been in the red
level.
9. Conclusion
It is worth considering that the amount of red light reflected by black fabric is probably more a function of the dyes used than
the fabric itself. Although, It is probably safe to assume that certain fabrics are probably consistently dyed using the same
formula.
Of the three targets, the Duvetyne had the most stable RGB levels, but had the most reflectance. The studio curtain had the
lowest reflectance levels of the three but exhibited a red color cast at .6ND (2 f stops). The worst offender was the
velveteen. Red can be seen even in the control clip.
I think I was successful in demonstrating that near IR contamination is possible in a studio situation at lower strength NDs than
has been demonstrated in outdoor situations. This is probably due to the amount of IR given off by tungsten incandescent
lighting and the CCD sensor’s sensitivity to visible and infrared wavelengths. The point at which the ratio between visible
light and IR leads to contamination of black objects is reached much faster and at lower ND strengths in a studio situation
than it is outdoors.
Methods to avoid red contamination:
Master pedestal - During my Jewelry shoot, I didn’t notice the red contamination on the velveteen as much because I had
crushed the master pedestal to -10. This wont always be desirable, then again the chances of needing a filter beyond .3 ND
in a studio situation is rare.
Flag light off of black backgrounds – This is usually possible except when its in the immediate foreground close to key light.
Hot mirror filter - Another way to control IR contamination is by using a Hot Mirror http://en.wikipedia.org/wiki/Hot_mirror. All
of the major filter manufacturers (Tiffen, Harrison, Schneider, Formatt) have versions of hot mirrors, but they cost about
double the price of regular filters.
10. Conclusion
It is worth considering that the amount of red light reflected by black fabric is probably more a function of the dyes used than
the fabric itself. Although, It is probably safe to assume that certain fabrics are probably consistently dyed using the same
formula.
Of the three targets, the Duvetyne had the most stable RGB levels, but had the most reflectance. The studio curtain had the
lowest reflectance levels of the three but exhibited a red color cast at .6ND (2 f stops). The worst offender was the
velveteen. Red can be seen even in the control clip.
I think I was successful in demonstrating that near IR contamination is possible in a studio situation at lower strength NDs than
has been demonstrated in outdoor situations. This is probably due to the amount of IR given off by tungsten incandescent
lighting and the CCD sensor’s sensitivity to visible and infrared wavelengths. The point at which the ratio between visible
light and IR leads to contamination of black objects is reached much faster and at lower ND strengths in a studio situation
than it is outdoors.
Methods to avoid red contamination:
Master pedestal - During my Jewelry shoot, I didn’t notice the red contamination on the velveteen as much because I had
crushed the master pedestal to -10. This wont always be desirable, then again the chances of needing a filter beyond .3 ND
in a studio situation is rare.
Flag light off of black backgrounds – This is usually possible except when its in the immediate foreground close to key light.
Hot mirror filter - Another way to control IR contamination is by using a Hot Mirror http://en.wikipedia.org/wiki/Hot_mirror. All
of the major filter manufacturers (Tiffen, Harrison, Schneider, Formatt) have versions of hot mirrors, but they cost about
double the price of regular filters.