[Ian Stuart Forsyth]

Thanks for expanding on what I added as an aside. I particularly appreciate the note regard color temperature. People tend to use use something white for in-camera custom white balance when a gray card would be much more appropriate.

One of the problems is that when WB on 18% grey you are balancing it to that grey and not to a neutral white target . This seem like it would be doing the something but they are not, if you are using 18% you are WB to a target that is only reflecting only a fraction of the total light the sensor is able to store.

This seems like nothing much but a large component of that data you are setting your WB to is noise, more importantly the red channel so any WB will contain a lot of data based on the noise within the red channel.

One only has to photograph a color checker and apply a white balance to the varying neutral color patches and see the resulting changes to the color values of the remaining patches, simply using a 18% card can introduce color casting if you are trying to produce a accurate WB in colors that you are trying to capture.

[stevebrot]

One of the problems is that when WB on 18% grey you are balancing it to that grey and not to a neutral white target .

Ummmm...that grey is equal amounts CMY and any color cast is due to the incident light. That characteristic is part of what we pay for. True white targets, on the other hand, are difficult to find. White paper usually has brighteners and even those that don't are seldom "white". As far as total light to the sensor from the card for a custom white balance, the process goes like this:

• Put the gray card in the scene. It need not fill the scene and is probably best if it does not, though it is important that it be exposed to the same light as the subject.
• Spot meter off the gray card. Doing so ensures adequate light.
• Custom in-camera white balance is done by placing gray to an element in the frame. Place gray to the gray card and be happy. The noise component will be no worse than other elements of the scene.
• If one is shooting RAW and one's software allows placing gray in PP (most do*), it works just as well to place gray to the card then.

As for matching tiles on a color checker, that is a completely different beast and quite different than white balance Ideally, one would use the colorchecker to generate a custom color profile for the session.


Steve

* I don't have broad experience with the various tools, but I do know that Lightroom's custom WB eyedropper disallows attempts to use it on anything even close to white (white, after all, being max for all three channels and technically "clipped" and undefined).

[stevebrot]

if you are trying to produce a accurate WB

That short phrase says a lot. White balance is an incredibly blunt tool and unless the incident spectra is perceptually equivalent to that of the range of correlated color temperatures that may be placed on the Planckian locus, white balance will always be a best fit and will never be "accurate" to the incident light.


Steve

[Ian Stuart Forsyth]

As for matching tiles on a color checker, that is a completely different beast and quite different than white balance Ideally, one would use the colorchecker to generate a custom color profile for the session.

Has nothing to do with the issue for WB I was only referencing the color checker card so that one can see what the differences are when selecting the wrong target to WB on.

As I said above it has to do with how we capture the light data to preform a WB, for this we want to use the largest signal just shy of clipping. Not doing this will result in using data (mostly in the red channel ) with noise to create a WB for you image. When you apply this WB to your image you are affecting the colors and introducing colorcasts into that image.

Every

https://www.adobe.com/digitalimag/pdfs/ps_workflow_sec3.pdf
I would go even further and say to sample not the sample that is closest to clipping for the best results

While you can use it much of the time you are introducing color cast to your image WB using the noise from a lower signal


So here is white balance done to sample 1


Here I have done white balance using sample 3 not even 18% and you can see a colorcast that has been introduced into the color found elsewhere in the image

---------- Post added 03-28-2020 at 09:49 PM ----------

That short phrase says a lot. White balance is an incredibly blunt tool and unless the incident spectra is perceptually equivalent to that of the range of correlated color temperatures that may be placed on the Planckian locus, white balance will always be a best fit and will never be "accurate" to the incident light.

I don't know what to say



Here is a image shot in the shadows requiring WB 8700 +30

Here is another in mid day light



Only thing done is WB 4750 +11



Here is the difference in between the 2 with only WB applied and using the same camera profiles

If we look here setting your WB when we are not using the best signal we see a greater variance in color cast



Like I said I don't know what to say? the reason why you are maybe seeing WB setting as a blunt tool is that maybe using 18% grey for setting WB is not a very accurate way of setting WB if you need accuracy.

[stevebrot]

Like I said I don't know what to say? the reason why you are maybe seeing WB setting as a blunt tool is that maybe using 18% grey for setting WB is not a very accurate way of setting WB if you need accuracy.

I see it as a blunt tool because I have some knowledge of how it works. I gave a concise explanation, but what can I say?

Addendum: It should probably be noted that the "tint" component in your examples is added by the software and is not related to white balance (color temperature) compensation. Also, as noted before, WB adjustment is not the same as color calibration and is not a substitute for such.


Steve

[Ian Stuart Forsyth]

Addendum: It should probably be noted that the "tint" component in your examples is added by the software and is not related to white balance (color temperature)

Sure it does as white balance is a software related item, most importantly for you is that it calculate a WB for a selected area, the software will measure the signals for that area and apply a multiplication to those signals to achieve a white balance ( making those channels equal in lightness). If I take readings for grey and for white and they multiply the channels differently ( meaning that they will measure a different WB setting for those areas) you will see a tint to the processing in your image and all resulting color within that image. And as I stated before the difference in this is related to a noise component because of the lower signal from that 18% patch while the patch with a larger signal has a lower noise component most notably the red channel.

Also, as noted before, WB adjustment is not the same as color calibration and is not a substitute for such.

Never said it was, but however it plays a relationship to how the colors are mapped into a working color space via the converter. Different WB different color tint to the mapped image

The examples that was shown was not about color calibration, it was to show that

Custom in-camera white balance is done by placing gray to an element in the frame. Place gray to the gray card and be happy. The noise component will be no worse than other elements of the scene.

and as you can see from my examples that we have calculated a different WB setting for both the white are and the 18% grey area resulting in a color tint even thou we have WB the same image in 2 different areas.

hite balance will always be a best fit and will never be "accurate" to the incident light.

The second part was to show that when you use a signal from a adequate source, (this not being a grey card as this introduces noise based inaccuracy's in the WB) that the accuracy of WB can be carried over from daylight image to an image taken in the shadows with very little to no issues color reproduction using the same color profile for both images.

Again this has nothing to do with color profile, the target was only used to show that just using the 18% patch and the white patch as a target of calculating the color channel multiplier to achieve a correct WB setting is different for both targets.

You may say this really doesn't prove anything but if we remove the noise component from the equations by taking a image of that same color checker grey target but the only difference is that we increase the signal (with a longer shutter speed ) coming from the 18% grey target and match it to the signal from the white patch from previous image like so.

calculated WB setting 8700 +29
Increase the signal and remove the noise component of the WB calculation

We have a very similar multiplier to signal with a similar WB setting, 8700 +28
All I have done is increase the signal, removed the noise component to the calculation of the WB setting


Here is the comparison of the calculated WB with the weaker signal
This calculated WB setting is 9700 and this is what has introduced the color tint to the resulting image in the raw conversion.

[stevebrot]

You may say this really doesn't prove anything

Yes, that is what I would say, but I will give you the last word here.


Steve

(...you measured off the wrong patch for the second image...)

---

Last edited by stevebrot; 03-29-2020 at 10:44 PM.

[Ian Stuart Forsyth]

(...you measured off the wrong patch for the second image...)

Nothing was measured incorrectly I metered off of the grey patch increased the exposure so that the signal from patch #22 was the same size of signal coming from patch #19 from the previous image, This was done by decreasing the shutterspeed.
This was to show that when the noise component was removed from the WB calculation by increasing the signal from the surface you are trying to set your WB you will calculate a more accurate WB.

Using 18% grey because the lower signal you will be applying WB multiplier with a larger component of noise giving you a less accurate WB, and if you are trying to reproduce or carryover any color accuracy between different light conditions it is best to use a WB multiplier calculated using the least amount of noise.

My sample images above was to show when the same signal was used to calculate a WB for that scene that it was near constant at 8700 +29, once we used a target for WB with a lower signal the WB changed it went from 8700 +29 to 9700 when measured using the %18 grey patch under the same light and this is what caused the color tint we see in that WB setting and I will state it again it was the noise that was being used to set the WB that caused this tint.

---------- Post added 03-31-2020 at 08:47 PM ----------

Addendum: It should probably be noted that the "tint" component in your examples is added by the software and is not related to white balance (color temperature) compensation. Also, as noted before, WB adjustment is not the same as color calibration and is not a substitute for such.

If we remove WB from the raw file converter there will be no recognisable image from a standpoint as we see it in photograph and color, we will be left with this.

zero WB

The software is WB while using color profiles to create our images you cannot have one without the other.

Image 1


image 2


Image one was using the same setting on the same photograph with the same color profile as was image 2 the only difference was that image 2 was set using 18% grey and like I said the software is both WB and the color profile.

If nothing has changed other than the target used for the WB calculation and we see color tint it should be clear that it was the target used for WB that changed the color tint of the image. And when we increase the signal for that target we lose that tint it then should be clear that there was something else that was contributing to that color tint in the raw conversion. Noise from the lower signal using 18% grey target

[stevebrot]

This was done by decreasing the shutterspeed.

...by two stops, not three.


Steve

[Ian Stuart Forsyth]

...by two stops, not three.

Never said I deceased the shutter by 3 stops but the fact remains that the signal for patch#22 18%grey was within -0.15 of the signal for patch #19 of the other photo, so they are within around 1/16 of a stop ( not the end of the world). We can see this as all I had to do was +0.15 to equal the same output lightness of 250,250,250. This drop in exposure did not even change the calculated WB setting for the 2 measured signals, as it shouldn't for where the images are taken.

I will further add that for the image it was taken using a modified EC for ETTR and placed 18% grey at 155,155,155 for this color space it should be closer to 100,100,100. and white 9.5 should be at 237,237,237

[stevebrot]

Never said I deceased the shutter by 3 stops but the fact remains that the signal for patch#22 18%grey was within -0.15 of the signal for patch #19 of the other photo, so they are within around 1/16 of a stop ( not the end of the world).

Sorry to assume that...I thought that was why you included the detailed photos showing your actual exposures and the values for each patch.

While we are at it, just what do you mean by "calculated WB setting"? I don't see any calculations.

I was typing, the conclusion came to me as a flash of light that I am obviously out of my depth and am guilty of serious under-thinking on this subject. Thank you for your efforts, but I fear they were in vain.


Steve

[Ian Stuart Forsyth]

Sorry to assume that...I thought that was why you included the detailed photos showing your actual exposures and the values for each patch.
While we are at it, just what do you mean by "calculated WB setting"? I don't see any calculations.
I was typing, the conclusion came to me as a flash of light that I am obviously out of my depth and am guilty of serious under-thinking on this subject. Thank you for your efforts, but I fear they were in vain.

Let start from the beginning, this is not to be sarcastic.

While many find no issues with using a 18% grey as a target to set a white balance from, the problem lies with why we are doing the WB in the first place and what we mean by applying a correct WB.

The calculation of setting the WB is done by the selection of your WB target with your WB tool, many times these targets are not even a neutral target. You can see the results of the WB calc by how it has placed the WB sliders, or if done by the camera what it has done to the RGB multipliers in the raw data.

Here you can see in the As shot neutral 0.475,1,0.697( RGB) this is telling the raw converter how to multiply the 3 channels to a the desired WB for Red 2.10 times green 1 and the blue 1.43 times. This is what WB is, setting a target to white and yes this target may not even be white

This is one of the colors I like to use for much of my work as it produces a WB I use very often
#e5f0f3 Color Hex
When selected as a target it produces the warm tones I like for much of my work, A correct WB calculation will allow me to carry this over independent of the light source. When this target is shot in a different light source and I use this as my WB target it will calc a new WB setting to produce this WB. This is one of the main reason why a person would go about trying to set their WB manually.



So here I have selected the patch labeled #7, the raw converter has calculated a what it takes to create that WB. If you take a look around #7 you will see 9 other WB targets all of which are not neutral and if you notice are not 18% grey.
One of the reason of this is for the very reason as I have stated before that noise from that weaker signal will cause the user issues with obtaining good WB calculations ( noise). Just look at a 18% target that has been photographed and view it with NR turned off and see the different colors within that grey or just take multiple samples over different areas of the target and see how it adjusts your WB sliders.

The image above was taken in direct sunlight while the image below is takin in the shade and selecting the very same WB target and the raw converter has calc'd a new settings for the same WB as you can see it has moved the sliders

This is the goal for setting WB to carry how the subject appears across different sources of light

Following Jeff Schewe for sometime I have learned to trust what he says as he has worked along with Thomas Knoll

page 5
https://www.adobe.com/digitalimag/pdfs/ps_workflow_sec3.pdf

Re: What they say is:: Retouching Forum: Digital Photography Review

[StiffLegged]

... Not all questions need a deep technical answer.

But this is the Pentax forum site - there’s gonna be a multitude of deep, technical answers. Really deep, tremendously technical, such as the above dozen or so posts. Probably a lot more to come...

[stevebrot]

Let start from the beginning, this is not to be sarcastic.

OK...I got it. My approach is much simpler, but like yours, it tends to work to my satisfaction and I override as needed. After all, if one is shooting in the blue hour, better that things look a little on cool side, right? One small note...the DNG color matrices are not the WB multipliers per se.* Those are found in the makernotes. An explanation of DNG color matrices requires a deep dive into the DNG specification and I ain't gonna go there!!


Steve

* They are the same for all DNGs from a particular camera model.

[Apet-Sure]

But this is the Pentax forum site - there’s gonna be a multitude of deep, technical answers.

I was just trying to be sensitive to the OP who is relatively new to the site. I do enjoy the deep technical info; it has helped me a lot. Just wasn't sure about the OP's desire/hunger to be thrown into the deep end, that's all. It's all good.