Jump to content

Difference between tint in camera and green/magenta light filters?


Recommended Posts

  • Premium Member

The signal becomes RGB after debayering so playing with green channel changes green and playing with the blue and red channel together changes magenta.

In theory, you'd have less noise if you filter in front of the lens rather than adjusting the levels in camera depending on how far you want to go, but you'd be better off using CC Green camera filters rather than lighting gel, which is not optically clear, it will soften the image somewhat.

Link to comment
Share on other sites

  • 2 weeks later...
  • Premium Member
On 11/19/2020 at 1:39 PM, David Mullen ASC said:

The signal becomes RGB after debayering so playing with green channel changes green and playing with the blue and red channel together changes magenta.

In theory, you'd have less noise if you filter in front of the lens rather than adjusting the levels in camera depending on how far you want to go, but you'd be better off using CC Green camera filters rather than lighting gel, which is not optically clear, it will soften the image somewhat.

Oh I never thought of it that way.

So theoretically if you were to shoot 5500k with tungsten in studio and used a blue glass to correct the incoming light, that would present less noise?

Link to comment
Share on other sites

7 hours ago, Stephen Sanchez said:

Oh I never thought of it that way.

So theoretically if you were to shoot 5500k with tungsten in studio and used a blue glass to correct the incoming light, that would present less noise?

Yep! A lot of Red shooters found that out on the earlier sensors that were noisy in Tungsten light.

Digital sensors are natively 5000K so getting the temperature as close as possible to that will maximize dynamic range and SNR. That is because the red channel will clip first in 3200K Tungsten light, but all channels will basically clip equally at 5000K.

So, you can gel or dichroic filter Tungsten to 5000K or use a filter on the lens to get to 5000K.

If you do that, though, you do need to increase the light hitting the sensor, otherwise you would just be increasing noise. (If you are right below clipping on the Red channel and filter the light, it's just going to increase noise across the board since a lot of light is now not making it to the sensor. However, even if you don't increase the amount of light, I strongly suspect you may still increase color fidelity at the expense of noise, depending on the neutrality of the blue filter and how well the blue filter and the sensor cuts infrared pollution.)

Going way further, blue light is actually physically smaller than red light. When we describe light as a certain nanometer, it is literally the length of the wavelength of light. Blue light starts around 450nm and Red around 625nm. By shifting more light to a higher Kelvin temperature, you are actually shrinking the overall size of the light. This actually increases the effective resolution you can capture. This won't be apparent on most sensors (unless way stopped down at or past diffraction) but it can play a role on smaller pixel sensors, such as on the Blackmagic 12K with its 2.2um pixels, since diffraction on smaller pixels starts at a larger f-stop. And this effect may be more apparent on future sensor designs with even smaller pixels, especially when cropping in.

 

Link to comment
Share on other sites

  • Premium Member
7 hours ago, Joshua Cadmium said:

Going way further, blue light is actually physically smaller than red light. When we describe light as a certain nanometer, it is literally the length of the wavelength of light. Blue light starts around 450nm and Red around 625nm. By shifting more light to a higher Kelvin temperature, you are actually shrinking the overall size of the light. This actually increases the effective resolution you can capture. This won't be apparent on most sensors (unless way stopped down at or past diffraction) but it can play a role on smaller pixel sensors, such as on the Blackmagic 12K with its 2.2um pixels, since diffraction on smaller pixels starts at a larger f-stop. And this effect may be more apparent on future sensor designs with even smaller pixels, especially when cropping in.

Do you have a resource on this subject? I'd like to understand more on this.

Link to comment
Share on other sites

15 hours ago, Mark Dunn said:

Have to say I'm sceptical too. Pixels are much larger than the wavelength of light.

When light passes through an aperture it forms an Airy Disk: https://en.wikipedia.org/wiki/Airy_disk. The Airy Disk is the point of light that we are trying to capture. As you stop down an aperture, the size of the Airy Disk gets bigger. At some point, the size of the Airy Disk can be larger than the size of a pixel. As well, the point where Airy Disks overlap is where diffraction starts.

How do you determine the size of an Airy Disk? An Airy Disk does not have a sharp cutoff, but there is a bright central region, similar to a Gaussian distribution, with additional rings around it. The bright central region is measured at about a 2.44 ratio (2.4393 to 5 significant figures.) In order to determine the actual width you also need the f-stop and the nanometer of light.

The math is actually really simple - just multiply everything together: 2.44 x f-stop x wavelength in nm. The interesting thing about this is that 2.44 and the f-stop are ratios, or dimensionless numbers, so the only dimension we are left with is the nanometer of light.

For instance, the Airy Disk size of far red 700nm light at f2.0 is about 3416nm in diameter or about 3.42um (micrometers). [The math is simply 2.44*2.0*700.] That is smaller than the 8.25um pixels on the Arri Alexas, but much bigger than the 2.20um pixels on the Blackmagic 12K.

On the opposite end, the Airy Disk size of violet 400nm light at f2.0 is about 1952nm or 1.95um. That now fits within one pixel of the 2.20um on the Blackmagic 12K.

The more light is shifted to the violet or blue end of the spectrum - which we can do by increasing the temperature of the light overall - the smaller the points of light get overall and thus the better they will fit within a pixel.

Another way to say this is that you reach diffraction quicker at warmer color temperatures. 

-----

Going way further, resolution is often calculated using green light, since it falls between red and violet. For instance, green 550nm at f2.0 is about 2684nm in diameter or 2.68um.

You'll notice that that is still larger than the 2.2um pixels of the Blackmagic 12K. That means that if you want to get an average of all Airy Disks (an average of the full spectrum of light) to fit within a 2.2um pixel, you'll need to decrease the f-stop.

What f-stop would I need to get a 2.2um (2200nm) Airy Disks size at 550nm? To figure that out, you can do the math in reverse: f-stop = pixel size / Airy Disk / wavelength in nm. In this case, f-stop = 2200nm / 2.44 / 550nm = f1.64

That seems like way too low of an f-stop, until you try the math with a bigger pixel. With an Alexa pixel: f-stop = 8250 / 2.44 / 550nm = f6.15

If that still seems too low, that is because the Airy Disks can overlap a bit and still not cause too much of a hit to resolution. This is where MTF plays a role.

MTF is actually very simple - it is really just looking at the overlap of Airy Disks. It is measured by imaging black and white lines and looking at the amount of grey that is produced as the lines overlap (which they do from being funneled through glass, an aperture, and the pixel structure.) MTF(100) would be a solid black line next to a solid white line. MTF(0) is where everything is a complete wash of grey. 

Making the Airy Disk the same size as the pixel puts the MTF at just about MTF(74), but resolution is typically calculated at MTF(50), which is where there is a 50% difference between white and black lines (dingy white next to greyish black).

Since MTF deals with line pairs, we need to figure out the line pairs per millimeter of a sensor. To do that, we just take 1mm (which is 1000um) / the pixel pitch / 2 (in order to measure in pairs). The Blackmagic 12K would be 227.27 lp/mm and the Alexa would be 60.61 lp/mm.

The actual math for MTF is a little complicated. I'm using the equation found in the middle of the page here (you can relatively easily run it in a spreadsheet): https://www.edmundoptics.com/knowledge-center/application-notes/optics/introduction-to-modulation-transfer-function/

There is a shortcut for the MTF(50) math, though. It is the lp/mm of the sensor divided by 2.475 OR the pixel pitch times 2.475. (If you do the math, it just ends up that way [and at eight significant figures it would be 2.4751447.]) For the Blackmagic 12K sensor, that puts MTF(50) at f4.1 and for the Alexa sensors that puts it at f15.2.

Those figures are using 550nm green light, though.

If we instead used 700nm light, the numbers would be different. The Alexa would hit MTF(50) at f12 and the Blackmagic 12K would hit that at f3.2

At 400nm, the Alexa would hit MTF(50) at f20.9 and the Blackmagic 12K would hit that at f5.6.

So, regardless of the MTF values, it does make sense to shift color temperatures cooler in order to maximize resolution, especially when the pixel size is smaller, unless you want to deal with an f-stop that limits the depth of field you can operate within. 

 

Edited by Joshua Cadmium
Link to comment
Share on other sites

Oops! Had an error that I just realized when I went back to check the math. That 2.475 ratio is actually the ratio of lp/mm of MTF(50) over lp/mm of MTF(0) .

The actual MTF(50) ratio is 1.971. So you could think of it as multiplying the pixel pitch by that or multiplying the Airy Disk ratio by that, but the 1.971 is just another ratio that would be multiplied by everything.

So, f-stop of MTF(50) = 1.971 * pixel size / Airy Disk / wavelength in nm

For the Blackmagic 12K sensor at green 550nm, that puts MTF(50) at f3.2 and for the Alexa sensors that puts it at f12.1.

At red 700nm light, the Blackmagic 12K would hit MTF(50) at f2.5 Alexa would at f9.5

At violet 400nm light, the Blackmagic 12K would hit MTF(50) at f4.4 and the Alexa would at f16.7

These numbers are theoretical, though, which just means that you would need a perfect lens to hit MTF(50) at these numbers. You are almost certainly going to maximize resolution somewhere between having the f-stop create an Airy Disk the size of a single pixel and having the f-stop at MTF(50), which creates an Airy Disk the size of about two pixels.

Again, though, higher temperature light (colder, more bluer) is physically smaller - the Airy Disks have a smaller diameter - so you can capture more of it before the Airy Disks start to overlap too much and lower the resolution by blurring together. 

Link to comment
Share on other sites

If you want some more info this page is very helpful: https://luminous-landscape.com/do-sensors-out-resolve-lenses/ . Table 1 and onward goes into things a little more in depth with charts. (Also, I found the lp/mm numbers in that Table 1 chart to be off by a few percentage points from my numbers. I think it might just be rounding errors. I kept everything unrounded on the spreadsheet I have.)

This is a short article about the diffraction limit and has MTF(0) numbers for 520nm light at various f-stops at the end of it: https://www.edmundoptics.com/knowledge-center/application-notes/imaging/diffraction-limit . (The video is a waste of time, though).

If you want to understand the math for MTF, this helped me: https://spie.org/publications/tt52_151_diffraction_mtf . It has the math for both a square and a round aperture. 

 

Link to comment
Share on other sites

  • 3 weeks later...

Here's another article I forgot to include that is extremely helpful in visualizing the difference in sizes in different wavelengths of light: https://www.edmundoptics.com/knowledge-center/application-notes/imaging/from-lens-to-sensor-limitations-on-collecting-information/ .

Here are some figures from the article:

 

fig-1-fls.gif

 

 

fig-2-fls.gif

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...