Canon EOS WB shift

[Bart_van_der_Wolf]

Here is my thinking on that, and I'd appreciate any commentary or advice.
First, I bought some large lots of Wratten Filters on eBay from someone who obviously has no more use with them. There is a large collection of some used, and some unopened, in varying strengths and colors.

I'm thinking of gelling my strobes with varying levels of Green and Magenta and taking shots of a large high quality grey card.

Hi Nick,

Just 2 things for consideration.

The Wratten filters are intended to be used in the optical path. I'm not sure if they will survive the strobe's light intensity, and they may significantly change characteristics with such use. You may need to consider some dedicated light modifier gels by Lee (some of which are available as High Temperature versions) if you want to modify your flash. Of course modifying the flash will also reduce it's output power.

The other thing is that many Grey cards are not necessarily spectrally neutral. Afterall, they are usually intended to determine exposure, not necessarily white-balance. So use a good quality grey card that has a known spectrally neutral reflection. I once measured the WhiBal cards, and they seemed to be reasonably uniform, good enough for white balance tests on digital cameras (which have limited UV sensitivity and usually a built-in IR rejection/absorption filter).

Cheers,
Bart

 

[Doug Kerr]

Hi, Nick,

Doug,
I read the articles you linked to, as well as many more in your photography collection. Really great stuff!! I have a much better understanding of the content. Thank you.

Thank you. This is all fun stuff!

I'm going to most likely rent a Sekonic C-500 to do my own tests as described in an earlier post of mine in this thread. As I mentioned, the important first step is to try to figure out what the transfer function is from the Sekonic CC Index (Wratten Filter designation) and the Canon WB Shift +/- 9 steps in the Green/Magenta axis.

Yes, that would be valuable.

Here is my thinking on that, and I'd appreciate any commentary or advice.
First, I bought some large lots of Wratten Filters on eBay from someone who obviously has no more use with them. There is a large collection of some used, and some unopened, in varying strengths and colors.

I'm thinking of gelling my strobes with varying levels of Green and Magenta and taking shots of a large high quality grey card. I will then remove the gels and take photos of the grey card at all 19 levels of the G/M axis of WB Shift.

It would be even better to move along both axes (even though I know that will be a lot more work).

How will you quantify the results of those test shots? Will you perhaps observe which pair of shift coordinates gives a "neutral" (R=G=B) chrmaticity for the image of the target?

What WB will you have set in the camera (there is no "none") or in the RAW processing (there is no "none")?

For all of these shots described above, I will also take a reading from the C-500 to see what the CC Index shift off of the CCT is being measured. I will then figure out how to plot all of this data to see where the equivalency levels are. Hopefully it is some sort of linear response so that future translations are easy to calculate. Once I have this mapping figured out, then I can do my other tests to see how the quality of using the Sekonic is.

It looks as if that is a good approach. It seems as if you will be able to follow the trail! Be sure to not miss any of the stages (including those I raised questions about above).

Let us know how it unfolds.

Best regards,

Doug

 

[Jerome Marot]

I had the curiosity to check the manual of my camera (Sony). It also has a compensation scale of ±9. The manual reads:

Based on using the set color temperature as the standard, the color can be compensated to G (Green) or M (Magenta). The color filter value can be selected between G9 and M9. The increment is approximately equivalent to CC filter number 5.

The manual continues with:

Note
Since most of the color meters are designed for film cameras, using the measurements obtained from them may not give correct results under fluorescent/ sodium lamp/mercury lamps. We recommend that you perform the compensation according to a test shooting or use the custom white balance.

On a Nikon manual:

Use the multi selector to fine-tune white balance. White balance can be fine-tuned on the amber (A)–blue (B) axis and the green (G)–magenta (M) axis. The horizontal (amber-blue) axis corresponds to color temperature, with each increment equivalent to about 5 mired. The vertical (green- magenta) axis has the similar effects to the corresponding color compensation (CC) filters.

I don't have a Canon manual to hand.

 

[Nick Pudar]

Bart,

The Wratten filters are intended to be used in the optical path.

Great point! And this is more of my lack of knowledge showing. That is why the filters are that small -- they are supposed to be put in front of the lens. That also slightly changes my testing procedure, but it hopefully will not degrade the results too badly.

many Grey cards are not necessarily spectrally neutral. Afterall, they are usually intended to determine exposure, not necessarily white-balance. So use a good quality grey card that has a known spectrally neutral reflection.

I think I'm ok here -- I will be using the full grey side of the Sekonic Exposure Profile Target II, which I'm pretty sure is spectrally neutral. (Although I could not find any documentation about that fact anywhere.)

Doug,

It would be even better to move along both axes (even though I know that will be a lot more work).

I will describe my testing process below, but I'm struggling in understanding how the Blue/Amber shift information would help me. The Sekonic C-500R only outputs the Color Temperature (actually it is CCT -- thanks for the tutorials!), and the Color Correction Index (in Wratten Filter numbers) for the Plankian Shift. I would be entering the CCT value which implies a zero shift along the Blue/Amber axis. And my whole testing process is designed to figure out how to convert the Sekonic CC Index to the Canon WB Shift steps.

How will you quantify the results of those test shots? Will you perhaps observe which pair of shift coordinates gives a "neutral" (R=G=B) chrmaticity for the image of the target?

I will be opening the images in Photoshop in Lab mode. Using the eyedropper tool, I will check to see the values of the a and b channels. The b channel is for Blue/Amber, and I should always see zeros for all positions on all of the the grey card photos. The a channel is for the Green/Magenta reading, and it will show the appropriate values based on the filters I was using. When the image should be properly White Balanced, the a channel should also read zeros.

What WB will you have set in the camera (there is no "none") or in the RAW processing (there is no "none")?

This is a great question, and it really got me thinking about whether the whole concept of the test was even possible. But I think it is. I will begin the whole testing process by using the camera's Custom White Balance function with my grey card to establish the camera's interpretation of the correct white balance. (I will measure the a and b channels of that image to ensure that they read all zeros.) Then when I add the green and magenta filters to the front of the lens durin the testing, the CCT won't shift, and I will be doing an offset only in the Green/Magenta axis. I'm really glad you asked that question!

Jerome,

I had the curiosity to check the manual of my camera

Thanks for looking that up. I really appreciate it. In my Canon manual it says that "One level of the blue/amber correction is equivalent to approximately 5 mireds of a color temperature conversion filter." But it does not say anything about the green/magenta axis.

So after much thinking and some great input from OPF members, here is my testing design. Any further input on the testing design would be appreciated.

Phase 1: Establish Base Reference Data

1. Use Sekonic L-758DR light meter to establish correct exposure settings for camera.
2. Use Sekonic Exposure Profile Target II grey card to use the camera's Custom White Balance function to set that as the starting reference point White Balance.
3. With the camera set up, take a reference photo of the grey card.
4. With the Sekonic C-500R color meter, take a reference reading of the light.
5. Open the reference images of the grey card in Photoshop in Lab mode, and using the eyedropper tool, confirm that the a and b channels read "0" at all points on the image.
6. Compare the Custom White Balance settings from the camera with the color meter reading.

It is my expectation (and hope) that this set of reference data and images will be very close to being identical. If it is not, I will have to think about what the offsets mean.

Phase 2: Record and Measure Modified Light

1. For a full range of Wratten gels (70M, 60M, 50M, ... 10M, 0, 10G, 20G, ... 70G), take a photo of the grey card at each level of light modification.
2. For each of the gels, also take a reading with the color meter and record the results. (Since the light strobes are not being gelled, cover the color meter sensor with the gels to modify its incoming light.)
3. Open each image in Photoshop (Lab mode) and record the a and b channel values.

Phase 3: Record and Measure Camera's White Balance Shift Modifications

1. For each step in Canon's White Balance Shift settings of the Green/Magenta axis (-9, -8, -7,... -1, 0, +1, +2,... +9), take a photo of the grey card (with no modifying gels).
2. Open each image in Photoshop and record the a and b channel values.

Phase 4: Determine Transfer Function

1. Plot the data from phases 2 and 3, and determine the transfer function that will convert the color meter's CC Index reading to the equivalent Camera's WB Shift value.

Until I plot the data, I do not know what the transfer function will look like. I am hopeful that it will be reasonably linear in nature, and something like "each step In Canon's WB Shift is equal to 10 CC Index values".

Phase 5: Test the Transfer Function

1. Go to an environment with different lighting conditions (gymnasium).
2. Take a light meter reading to correctly set the manual exposure.
3. Take a color meter reading.
4. Set the camera's Custom Temperature value to the Kelvin value from the color meter.
5. Set the camera's WB Shift value after converting the color meter's CC Index value using the transfer function from phase 4. This is for the Green/Magenta axis only.
6. Take a photo of the grey card with these settings.
7. Change the White Balance setting of the camera to Custom White Balance mode using the grey card. After establishing the Custom White Balance, take another reference photo of the grey card.
8. Open the images in Photoshop, and read the respective a and b channel values.

If the conversion went well, I would expect that the a and b channel values would all read "0" (in both images).

If I had time, I would also try all of the above steps at a couple of other ISO values, but I do not think the ISO should make any difference.

Anyway, that is my current thinking of a test design. If the test goes well, someone could probably do the same test with a Nikon camera for Nikon's WB Shift function.

Thanks again for everyone's input.
Best regards,
Nick

 

[Doug Kerr]

Hi, Nick,

I will describe my testing process below, but I'm struggling in understanding how the Blue/Amber shift information would help me.

I will be opening the images in Photoshop in Lab mode. Using the eyedropper tool, I will check to see the values of the a and b channels. The b channel is for Blue/Amber, and I should always see zeros for all positions on all of the the grey card photos. The a channel is for the Green/Magenta reading, and it will show the appropriate values based on the filters I was using. When the image should be properly White Balanced, the a channel should also read zeros.

Sounds good.

Best regards,

Doug

 

[Doug Kerr]

Hi, Nick,

Here is a link to (almost) the figure from my earlier post:

http://dougkerr.net/illustrations/20d_wb_shift_uv-pl6000.jpg

In fact, this is just a little different. It shows the Planckian locus based on the premise that the origin of the chart is planted at a blackbody chromaticity at 6000 K.

Note that this figure does not show or imply the L*a*b* a* and b* axes.

Best regards,

Doug

 

[Nick Pudar]

Doug,

I plotted the results on several different coordinate planes, along with vectors showing the "directions" of the chromaticity of the three color space primaries (R, G, and B) and the complementary hues C, Y, and M. The result for the "0,0" setting was used as the reference, arbitrarily plotted at the origin of the coordinate system.

The results on the CIE u-v plane were the most telling. There, as a matter of fact, the six "primary" vectors are almost equally-spaced angularly.

Thank you for posting that picture. It helps a lot, and is most interesting. I think I have to go back to your articles again, but how did you determine what the du and dv values were for each of the 8 readings around the origin?

If I could plot the results of my test on the CIE u-v plane, it might be the most revealing.

Thanks,
Nick

 

[Doug Kerr]

Hi, Nick,

Thank you for posting that picture. It helps a lot, and is most interesting. I think I have to go back to your articles again, but how did you determine what the du and dv values were for each of the 8 readings around the origin?

The basic test scheme is described in my post at the head of this thread.

Then, for each setting of the two color shift coordinates (with a uniformly-illuminated target), I read the L*a*b* a* and b* coordinates of the target in the adjusted image in (I think) Photoshop.

Then I converted those into other various other coordinate systems, including CIE u-v, and plotted them using my illustration program. I found the u-v plot the most "illuminating".

A good tool for doing such transforms is Bruce Lindbloom's CIE Color Calculator, available here:

http://www.brucelindbloom.com/index.html?ColorCalculator.html

If you go to Bruce's home page:

http://www.brucelindbloom.com/

and select the Math button, you will get to a page that has a matrix for all transforms from one color space to another, each cell giving links to pages where the equations used to do the transforms are given.

There is much other good info on this site as well.

It has been a while since I worked with any of this, but it will no doubt come back if needed! If you have any problem with the Lindbloom calculator, let me know. (Be sure to follow the "Learn more" link at the foot of the calculator page.)

I think you are on a good track.

Best regards,

Doug

 

[Nick Pudar]

Then, for each setting of the two color shift coordinates (with a uniformly-illuminated target), I read the L*a*b* a* and b* coordinates of the target in the adjusted image in (I think) Photoshop.

Then I converted those into other various other coordinate systems, including CIE u-v, and plotted them using my illustration program. I found the u-v plot the most "illuminating".

Doug,

Thanks for the great reference links to Lindbloom's work. Great stuff!

I took a look at his CIE Color Calculator, and I'm trying to figure out how I will calculate the delta u-v values from my Lab readings. When you converted your readings to CIE u-v, is that the same thing as Luv on his calculator? (There was no read-out called CIE uv.) Assuming that it is, when I generate all my data points, I will just calculate the delta from the reference grey card and make it the origin on the delta u-v scale as you did.

The other thing that has me perplexed a bit are the Reference White and Adaptation settings. Do I just leave them at his default settings of D50 and Bradford?

Thanks,
Nick

 

[Doug Kerr]

Hi, Nick,

Thanks for the great reference links to Lindbloom's work. Great stuff!

I took a look at his CIE Color Calculator, and I'm trying to figure out how I will calculate the delta u-v values from my Lab readings. When you converted your readings to CIE u-v, is that the same thing as Luv on his calculator? (There was no read-out called CIE uv.)

Yes, u and v are the u and v of the Luv coordinate system (L being the "quasi-luminance" coordinate). I should have mentioned that.

Assuming that it is, when I generate all my data points, I will just calculate the delta from the reference grey card and make it the origin on the delta u-v scale as you did.

The other thing that has me perplexed a bit are the Reference White and Adaptation settings. Do I just leave them at his default settings of D50 and Bradford?

This is very confusing to me, and I will have to ponder that a bit.

But you should get some practical insight from his page of explanation, which shows the data flow involved in the various transforms.

But when all else fails, go with the defaults!

I'll look at it too when I get a minute.

Thanks.

Best regards,

Doug

 

[Nick Pudar]

I did some more reading and learned that Photoshop uses then CIE D50 White Point setting, and that the CIE preferred Chromatic Adaptation transform function is the Bradford approach. So the defaults were the safe bet after all.

Incidentally, the fine folk at Sekonic are interested in my test results and have kindly allowed me to borrow a color meter for my tests. This is wonderful since I will not have to rush the work through a 3-day rental period.

Nick

 

[Doug Kerr]

Hi, Nick,

I did some more reading and learned that Photoshop uses then CIE D50 White Point setting, and that the CIE preferred Chromatic Adaptation transform function is the Bradford approach. So the defaults were the safe bet after all.

That's good.

But insofar as converting from L*a*b* to L*u*v*, the adaptation choice doesn't have any effect.

Incidentally, the fine folk at Sekonic are interested in my test results and have kindly allowed me to borrow a color meter for my tests. This is wonderful since I will not have to rush the work through a 3-day rental period.

That is terrific.

Let me know how the project unfolds.

Best regards,

Doug

 

[Nick Pudar]

Doug,

Between lots of business travel, I was able to make some progress on my project. I was able to execute Phases 1 and 3 of my experiment as described earlier in this thread.

Phase 1 is about getting some reference starting points.

Phase 1: Establish Base Reference Data
Use Sekonic L-758DR light meter to establish correct exposure settings for camera.
Use Sekonic Exposure Profile Target II grey card to use the camera's Custom White Balance function to set that as the starting reference point White Balance.
With the camera set up, take a reference photo of the grey card.
With the Sekonic C-500R color meter, take a reference reading of the light.
Open the reference images of the grey card in Photoshop in Lab mode, and using the eyedropper tool, confirm that the a and b channels read "0" at all points on the image.
Compare the Custom White Balance settings from the camera with the color meter reading.

I used the Color Meter to measure my light, and it read 5500K with 0 shift in the Green/Magenta direction. So, after setting the camera to the right exposure based on my Light Meter reading, I set the White Balance to 5500K. Here is a link to the photo of my reference card. When I opened the RAW image of the reference gray card in Lightroom, it reported that the White Balance was at 5300K and there was a Tint of +2 (towards Magenta). When I converted the image to L*a*b in Photoshop, the gray card read 44, -2, 0. The -2 of the a channel might correspond to the +2 in Lightroom to make the White Balance look right, but I'm not sure. I decided to go with this initial reference card setup, since I was going to be using offset calculations later anyway. (I am curious how to think about the 5500 vs 5300 White Balance readings. Any ideas?)

Since I do not have sufficient Wratten Gel Filters (still waiting on some to arrive from an order), I decided to move on to Phase 3.

Phase 3: Record and Measure Camera's White Balance Shift Modifications
For each step in Canon's White Balance Shift settings of the Green/Magenta axis (-9, -8, -7,... -1, 0, +1, +2,... +9), take a photo of the grey card (with no modifying gels).
Open each image in Photoshop and record the a and b channel values.

As you had recommended, I took photos for all combinations of White Balance Shift settings, even in the Blue/Amber dimension. I did all that (361 photos!) and recorded all of the L*a*b data from photoshop. I then used Bruce Lindbloom's CIE Color Calculator to convert the L*a*b data to L*u*v data as you had earlier. I then offset all the data points so that the Canon White Balance Shift setting of (0,0) actually fell on the du,dv values of 0,0.

After all that, this link shows the data plotted. (The red plot points are those where one of the White Balance Shift values are 0 -- to make it easier to read the chart.) It looks similar to the small data set you generated in your earlier test. Although my du and dv units are not exactly the same, and they are off by a factor of 1000. I may be doing something wrong, but I'm not sure what it is.

From here, I just need to wait for some of the filters to arrive. I was only able to get a very small set of pure Green and Magenta filters from eBay (I have lots of Blue and Yellow filters!). I'm buying a few more of the Green and Magenta and I will have a small sample set to run my other experiment phases. There will have to be some extrapolation going on later. Hopefully the filters come in the mail soon, as I only have the Color Meter on loan until the end of June.

While this is all interesting and I'm learning a lot, it certainly feels like the run will be way too long for what may turn out to be a very short slide! I'm keeping at it though.

Nick

 

[Doug Kerr]

Hi, Nick,

Between lots of business travel, I was able to make some progress on my project. I was able to execute Phases 1 and 3 of my experiment as described earlier in this thread.

Thanks for the update.

Your procedure sounds very good overall.

Phase 1 is about getting some reference starting points.

I used the Color Meter to measure my light, and it read 5500K with 0 shift in the Green/Magenta direction. So, after setting the camera to the right exposure based on my Light Meter reading, I set the White Balance to 5500K. Here is a link to the photo of my reference card. When I opened the RAW image of the reference gray card in Lightroom, it reported that the White Balance was at 5300K and there was a Tint of +2 (towards Magenta). When I converted the image to L*a*b in Photoshop, the gray card read 44, -2, 0. The -2 of the a channel might correspond to the +2 in Lightroom to make the White Balance look right, but I'm not sure. I decided to go with this initial reference card setup, since I was going to be using offset calculations later anyway. (I am curious how to think about the 5500 vs 5300 White Balance readings. Any ideas?)

Since I do not have sufficient Wratten Gel Filters (still waiting on some to arrive from an order), I decided to move on to Phase 3.

As you had recommended, I took photos for all combinations of White Balance Shift settings, even in the Blue/Amber dimension. I did all that (361 photos!) and recorded all of the L*a*b data from photoshop.

Yes, the "hundreds of test [photos, files, etc.]" phenomenon is a staple of life here!

I then used Bruce Lindbloom's CIE Color Calculator to convert the L*a*b data to L*u*v data as you had earlier. I then offset all the data points so that the Canon White Balance Shift setting of (0,0) actually fell on the du,dv values of 0,0.

After all that, this link shows the data plotted. (The red plot points are those where one of the White Balance Shift values are 0 -- to make it easier to read the chart.) It looks similar to the small data set you generated in your earlier test. Although my du and dv units are not exactly the same, and they are off by a factor of 1000. I may be doing something wrong, but I'm not sure what it is.

Yes, there is very good agreement with the general result I reported.

I'm not sure where the 1:1000 discrepancy comes from. If I get a chance I will poke around a little and see what that might come from.

While this is all interesting and I'm learning a lot, it certainly feels like the run will be way too long for what may turn out to be a very short slide!

Important work often has that feel to it. Do not let that feeling derail you in your quest for understanding.

Thanks again for keeping in touch.

Best regards,

Doug

 

[Doug Kerr]

Hi, Nick,

Just for reference, this shows the CIE u-v plane, showing the scale of the coordinates:

The figure was taken from this most useful paper:

http://www.efg2.com/Lab/Graphics/Colors/Chromaticity.htm

Best regards,

Doug

 

[Doug Kerr]

Hi, Nick,

There seems to be something funny in Bruce's calculator re Luv. I'm going to work my way through it and see why it seems funny.

Later.

Best regards,

Doug

 

[Doug Kerr]

Hi, Nick,

Here is where I led you astray.

The color space called "Luv" in Bruce's calculator is actually L*u*v*.

In some cases, the "*" in a color space designation means a nonlinear coordinate, and here it means that for L*, but here for u* and v* it means something else: that these coordinates are scaled with "L" (so that they essentially define pseudo-chrominance, not chromaticity). (True u and v define chromaticity.) *

The relationships are these (and for rigor I will use L*, u* and v*, the formal designations, for the values reported in Bruce's calculator in the three columns of the "Luv" color space):

u* = 13L*u
v* = 13L*v

where L*, u*, and v* are the three values reported in the Lindbloom calculator for the "Luv" color space, and u and v are the chromaticity coordinates as in the CIE u-v chromaticity plane.

Thus:

u = u*/13L*
v = v*/13L*

*If you are unsure as to the subtlety of the distinction between chromaticity and chrominance, you may find this helpful:

http://dougkerr.net/pumpkin/articles/Chromaticity_Chrominance.pdf

Best regards,

Doug

 

[Nick Pudar]

Well, I have run out of time with my experiment, and I had to return the loaner C-500R to Sekonic. My summer will be hectic, and I just do not have the solid blocks of time necessary to do this justice. Part of my dilemma is that I just can't get my hands on the right Wratten Filters. I ordered a lot of large bundles of used filters on eBay, but after a couple hundred dollars, there was not much quality stock of green and magenta filters. (LOTS of blue and amber.) I had also ordered the remaining necessary filters new from B&H Photo, but they seem to be on perpetual backorder, and I fear that I will never get them. Come to think of it, I should really just cancel the order.

I was able to complete most of the foundational elements of the experiment, but the transform function from the Color Meter to the Color Shift parameters in the camera eluded me.

I'm just going to have to stick with the Custom White Balance process with my camera.

Thanks to everyone, especially Doug Kerr for all the fabulous help and encouragement on this little experiment process.

Best regards,
Nick

 

[Doug Kerr]

Hi, Nick,

Well, I have run out of time with my experiment, and I had to return the loaner C-500R to Sekonic. My summer will be hectic, and I just do not have the solid blocks of time necessary to do this justice. Part of my dilemma is that I just can't get my hands on the right Wratten Filters. I ordered a lot of large bundles of used filters on eBay, but after a couple hundred dollars, there was not much quality stock of green and magenta filters. (LOTS of blue and amber.) I had also ordered the remaining necessary filters new from B&H Photo, but they seem to be on perpetual backorder, and I fear that I will never get them. Come to think of it, I should really just cancel the order.

I was able to complete most of the foundational elements of the experiment, but the transform function from the Color Meter to the Color Shift parameters in the camera eluded me.

I'm just going to have to stick with the Custom White Balance process with my camera.

Thanks to everyone, especially Doug Kerr for all the fabulous help and encouragement on this little experiment process.

I well understand your situation - I have been there many times. I have this very day in fact put several lines of investigation into the "banked fire" situation you describe.

I know you are disappointed not to be able to bring your investigation to a full conclusion. But take heart in the enormous amount of understanding you have gained so far.

Remember, we never understand anything completely! Our "wisdom" is a "grand triangulation" among many things we understand perhaps 20% or maybe even at best 80%!

Best regards,

Doug