Sunday, February 2, 2014

Mireds and Correcting the Color Temperature of a Light

As I've recently shown in blog posts, sometimes a video light doesn't exactly have the color temperature that the manufacturer suggests it does.  Case in point, the K55 bulbs from Kino-Flo that are actually closer to 7000K than 5500K.

There's a somewhat less well known method for using filters to compensate for the color temperature of a light fixture using 'Mireds' (pronounced "my-reds").  'Mired' is an abbreviation for "micro reciprocal degree".  Cinematography.com has a nicely written posting about it.  There's also a Wikipedia page, of course.

The concept is quite simple.  Let's say you have a tungsten light, which we measured at 3092K in our last blog post, and you want to compensate that light to 5600K daylight.  The Mired shift is calculated as the difference between the reciprocal of each color temperature multiplied by 1,000,000.

So...

Tungsten = 1,000,000/3092 = 323.4

and...

5600K daylight = 1,000,000/5600 = 178.6

thus...

Mired value = 178.6 - 323.4 = -144.8

Not surprisingly, companies that make filter products, such as Lee Filters, provide Mired calculators.  When I entered in an input light of 3100K and an output of 5600K I found  the following results.  So combining a 3/4 CTB with their 1/4 CTB would actually give me close to what I want.  Using a full CTB doesn't get me quite as close.  They also smartly suggest a plus green filter in case I need to use the tungsten lamp in combination with fluorescent lights.


I would have verified this with real world measurements, but SOMEONE (you know who you are!) in the shop returned the UPRTek spectrometer this week!  I'm curious how well this calculation works with discontinuous spectrums light LED and fluorescent.  Obviously it works well for tungsten, HMI, and plasma; all of which have nice continuous spectrums and color rendering.

Let's do another example out of curiosity.  How about we turn those Kino-Flo Diva K55 bulbs to actual 5500K.  The calculator is telling me that the nearest match is 444 - Eighth C.T. Straw or 212 - L.C.T Yellow.  I would likely try the 212 filter first since that biases the light toward 5600K.

 

One thing that Mired calculation does not do for you is compensate for green/magenta shifts on LED and fluorescent lights.  That's the type of thing that can be quickly analyzed and figured out if you have a device like the UPRTek spectrometer and know how to read (x,y) coordinates.  It's not as technical as it sounds.  In fact it's pretty dead-brain simple with that spectrometer...if I still had it.