CCDs have plenty of sensitivity in NIR and IR, used for surveillance cameras and astronomy.

There is a big difference in quality between different filter brands.

For example, a few years ago I needed a cirpol filter for my XL-1, so I was going to buy the usual $50 Tiffen.

Fortunately the store clerk (at a pro store) suggested I look at B+W's ditto for $250.

The difference was so great I could see it with my bare eyes, and this was confirmed in shooting.

(And of course I was able to shave the $250 sticker...)

Bob, the note I've quoted was given in the context of setting up parameters for blacks in HD cameras so, although I too wondered about just doing corrections in post, I thought it might be of use to people here. And Coursedesign's input is a good reminder to think about what we're buying. Agree I should have known this, but then I wouldn't have posted it!

Serena,

I really appreciated the information you provided. This is not something so obvious that it would be easily found, and this may save me some aggravation.

I wonder how much the Z1's built-in ND filters suffer from this...

I think the camera would just white balance this away.

Film cameras don't have electronic white balance (and neither do some digital cinema type cameras).

That's assuming you're in auto WB mode, right? If you are doing manual white balance, would you suppose that flipping on the ND filter might require remeasuring for WB? Or maybe it's a relatively minor thing and not particularly noticeable on the Z1?

This is something of significance for ND filters 0.9 and denser. I think WB might not be a very useful remedy because the effect peaks down in the deep red (and is perhaps only significant at that end). Affects still digital cameras also. I'm told that the correction in post is quite "heavy". However this is in the context of using very dense ND filters.
This makes me curious about the longevity of organic coatings; one might expect these to change characteristics after long exposure to light, but I've never seen in the manuals anything about limited filter life. Guess I just assumed that a filter costing over $400 dollars is good until you drop it. Any knowledge?

Or maybe it's a relatively minor thing and not particularly noticeable on the Z1?
My guess is that. Plus autoWB or push-WB would hide any problems with that.

The white point of daylight changes throughout the day anyways.

The white point changes throughout the day, but does that create a strong peak or dip in one region of the spectrum? (Sunrise and sunset are different of course because the atmosphere effectively becomes a filter through refraction.)

It seems to me that if you put say a deep red filter on your camera, a lot of blue light is just going to be blocked.

I was just saying you'd get minor error in WB since it changes throughout the day, unless you set it very often.

The white point changes throughout the day, but does that create a strong peak or dip in one region of the spectrum?
I don't think so.

Glenn, what is the methodolgy employed for setting WB? I don't mean holding up a white card, I mean what goes on in the electronics. I presume that WB circuitry looks at portions of the spectrum, at least at red and blue and determines colour temperature and hence WB from that. Now if you have a ND that distorts the spectrum, which is the matter under consideration, and that affects WB, then the set WB will be in error. The ND filter hasn't changed the colour of the scene, the colour temperature of the light hasn't changed, so you don't want flesh tones to suddenly shift blue because your ND filter has given shadows reddish cast. So I would say the last thing you would want to do is use WB to correct for the fault discussed.

Sorry I'm not Glenn but,
I think all WB does is set the gain of the RGB channels so the outputs match. That's probably a pretty crude simplification but that's all that's available to the camera to work on. WB affects all parts of the curve the same so a cast in the shadows will not be affected.

That's why I made a comment about the relevance of casts in gels to video Vs film. I'd add lights to that as well. The Osram Studioline fluro tubes are considered good enough for video because the WB can correct for their spectral output. They're generally not recommended for film.
However even with film I'd imagine using a grey card the errors in lighting or filtering could be graded out in post.
Let's not forget that both film and video are very crude at recording color, try recording an accurate version of IKB or the emmision of sodium lamps to film or video.

Bob.

Thanks Bob. About the only way WB could work, really. No argument with your other points. I posted orginally as "information to store away", but probably it came out sounding like it was a significant problem. In the rare event that one is working with very dense NDs, that they are not colour-neutral is a good thing to have in mind. But even with film it can be corrected in post.

Sodium lamps emit a narrow spectrum, so no matter what you use to record the color, there isn't much information to extract.

Has anyone been able to compensate this with some "magic filter"?

Well NO!
Looking at the spectrograph, it's only two extremely narrow lines, there's nothing in any other parts of the spectrum, try to filter out the yellow and you're left with no light.
The result of using sodium vapor lights is that almost all colors are rendered as grey or else some other shade of yellow.

This illustrates the big problem of relying on color temperature as a measure of light. CT assumes a black body radiator but many light sources are anything but black body radiators. The figure that's a much needed additional metric is Rendition Index, this gives a measure of how close to white the light source is.

For example there's some lights out there that use RGB LEDs. They can look white but compare the spectrograph from these light sources to real white light and all you see are a number of narrow spikes. This means not all colors will render correctly to the eye. Video though seem to make it look OK.

Bob.

Well, you can take B&W pics under sodium lights. The low pressure lamps produce most of their light at light at 589 nm wavelength (two lines, 589 and much less at 598.6nM), so getting any colour but yellow out of that would be a miracle. The high pressure lamps (also containing mercury and neon) give a full visual width, but discontinuous, spectrum. Not good for photography and no WB or filters will do much good there (for normal balanced images).

EDIT: Bob beat me to it! Normal LEDs have a very narrow spectrum. Bob, have you data on "white" LED emissions?

EDIT #2: Ok, depends on the method of generating "white". The quantum dot technique appears to produce the most continuous spectrum. white LED

As far as i know the current white LEDs are very similar to fluro tubes, they use a UV LED to drive a fluroescent screen.
Those quantum dots are very interesting.

Bob.

The ND filter hasn't changed the colour of the scene, the colour temperature of the light hasn't changed, so you don't want flesh tones to suddenly shift blue because your ND filter has given shadows reddish cast. So I would say the last thing you would want to do is use WB to correct for the fault discussed.
An imperfect ND does change the colors coming in... so it's appropriate to WB through the ND.

2- Color is a spectrum of light at different wavelengths... you can arbitrarily divide the useful spectrum in 31 bands (of certain wavelengths). The eye has 3 types of cones that sense color... and they each have they own response to certain wavelengths of light.

Among the light, you can have different combinations of the 31 bands that will produce the same stimulus/output out of the 3 types of cones. This is one form of metamerism.

On the flip side, an object will have slightly different color depending on the light. i.e. if you have yellow light hitting it, the reflected light will be yellow. This applies to all light sources, since they all have a different proportions of the 31 bands (SPD / spectral power distribution would be a more accurate term). This is one source of metamerism. This is a slight disadvantage to HMI and fluorescent lighting, as the spikes in their SPD graphs indicate that they will cause worse metamerism.

Similarly, adding an imperfect ND will alter the spectral power distribution of the lighting hitting the camera's sensor... this can be considered a form of instrument metamerism (i.e. changing the ND changes the light-measuring instrument).

3- Life isn't very color accurate anyways, so your brain doesn't worry itself with small color inaccuracies. So metamerism issues from HMI and fluorescent lighting may not be a big deal for this reason.

4- I think if you plot the SPD of LED lights, you will see a "hump". You can see such a SPD graph at
http://en.wikipedia.org/wiki/Image:Red-YellowGreen-Blue_LED_spectra.gif

5- One way to guess the SPD of any light is as follows:
Take a CD, since the surface exhibits a phenomenon called diffraction. If you look at the reflection of a light source on the CD, look at the even-ness of the 'rainbow' that appears. Fluorescents typically have a "spiky" SPD, so they will have an uneven rainbow (they are typically soft light sources, so the rainbow will be a bit muddied). With LEDs, you won't see the extreme colors. With tungsten light, you will see an even rainbow and all of the possible colors.

A prism would be another way of breaking light up into its individual wavelengths.

SPD graph of a fluorescent light:
http://en.wikipedia.org/wiki/Fluorescent_light

You can also get SPD graphs for D50 (daylight-series illuminant; daylight is sometimes like it), HMI, black body radiators, sodium vapour, etc. etc.

6- Um... does any of this make sense???

Yes, Glenn, all correct. However general discussion of the colour perception doesn't deal with the issue of how cameras set WB and perhaps you're avoiding the point that the ND filter characteristics under discussion needs to be thought of a narrow band transmission filter (for far red). This doesn't affect the majority of the visible spectrum. It puts a lump at the bottom end of our region of interest. Little effect on near red, none on higher frequencies. Trying to flatten the curve using WB will broadly affect colour output, adding other distortion while trying to flatten far red.
How this would work out in practice requires us to consider specific systems, and the data isn't available. You'll note that the original comment was that compensating corrections needed to be made in blacks only. However I think Bob's comments are probably sufficient.

So the gist of this (if I understand it correctly, insert big question mark here!) is that WB is not by any means selective enough to deal with a relatively narrow spike or dip in the color spectrum that is captured by the camera. Perhaps, if the issue is even noticeable at all, this is something that could theoretically be dealt with using Vegas's secondary color corrector?

While perhaps some of this may seem a bit esoteric, it does bring attention (for me, at least) to one thing. If you were to intentionally put a color gel filter to the camera lens, I gather that you'd want to do any manual WB before you put on the gel filter? Pardon me if that sounds like a dumb question, but I haven't had much experience with using gel filters and it seems like that'd be a highly useful thing to know.

No, that's a good question. Presumably if you were to use a coloured filter, you would be wanting to get that tint into the image. In which case you would do WB before adding the filter and lock it. Otherwise WB would remove the tint.

Well actually from my experiments if it's a deep tint the WB circuits will do nothing. Try pointing a camera at a blue or green CK screen and hit the WB button. In theory the thing should go nuts but thankfully it doesn't. With no signal in two of the channels in theory it'd be almost trying to add infinite gain to those channels. I'd say those warm/cool cards are about as far as you can push the WB off kilter.

But getting back to the original question maybe there's something in this and the problem being in the NIR part of the spectrum. CCDs are very sensitive down there, for that reason a IR filter sits in front of them, on most cameras the Night Vision switch flips that filter out of the way. However if the not so good ND filter is letting through way more NIR than visible light and the RGB parts of the imager are all sensitive to that part of the spectrum even with WB it might not correct properly.
In other words the G and B channels get their gain wound down due to the IR hitting them then you'd introduce a red cast .

Bob.