Normal colors test: Tangsinuo QB21 & ICF ("hot mirror") filters on Sony a7 full spectrum camera

[photoni]

there are differences in exposure because the sun was hiding behind the light clouds
a fairer test would be with the flash... which has constant light and power

.

.

[ulf]

It is impossible to fully recreate a stock camera's response by a single BG-filer and profiling.

 

The filtering in a stock camera is done with a combination of a BG-filter of some kind and the dichroic filter on the dust shaker or antialias component in the filter stack. Those filters in combination normally block IR very efficiently.

For some cameras the filters are combined in one single stack. Others have them as separate glass sheets.

 

If the dichroic component is missing there will always be some response different if there is IR in the captured motif.

If the goal is to get the best VIS response a stock camera is optimal.

   

[LarsHP]

15 hours ago, photoni said:

there are differences in exposure because the sun was hiding behind the light clouds
a fairer test would be with the flash... which has constant light and power

.

.

 

Thanks. The white balance with the QB39 1.5mm is quite close to the stock camera, much, much closer than 2mm QB21 or ICF.

[photoni]

15 hours ago, ulf said:

 

The filtering in a stock camera is done with a combination of a BG-filter of some kind and the dichroic filter on the dust shaker or antialias component in the filter stack. Those filters in combination normally block IR very efficiently.

For some cameras the filters are combined in one single stack. Others have them as separate glass sheets.

 

I'm sending you this test done on a beautiful sunny day (after 15 days of rain)

EL-Nikkor 80 mm @f16

As you can see, the visible differences are few the dust-shaking dichroic filter makes little difference, it probably cuts the lower part of the UV below 400 nm

.

iPhone - X

....... look at the shadow of the dichroic filter

 

.

A7 standard

.

A7fs + QB39

.

A7fs + QB39+nikon L39

.

A7fs + internal BG Sony

.

A7fs + internal BG Sony + dust-shaking dichroic

.

 

[LarsHP]

1 minute ago, photoni said:

I'm sending you this test done on a beautiful sunny day (after 15 days of rain)

As you can see, the visible differences are few the dust-shaking dichroic filter makes little difference, it probably cuts the lower part of the UV below 400 nm

...

 

 

You would need a subject that has more IR reflection than the color chart.

Take your a7FS (without filter), do a white balance on white paper or plastic, and find some visually neutral black cloth that you clearly see has lots of IR in the a7FS.

Then do the test with the cloth included in the image.

[photoni]

8 minutes ago, LarsHP said:

 

You would need a subject that has more IR reflection than the color chart.

Take your a7FS (without filter), do a white balance on white paper or plastic, and find some visually neutral black cloth that you clearly see has lots of IR in the a7FS.

Then do the test with the cloth included in the image.

 

Some time ago I did a different test with a dandelion flower
I used QB29+QB39 and TSN575+QB29 the result is similar
the percentage of IR is negligible in the visible with QB39

[ulf]

That is a good advice.

If your only motifs were x-rite test targets you would get a good result by creating a profile.

Those test targets are designed for VIS only.

However as soon as you include components in the image with a high IR reflectance like some black black cloth or really dark chlorophyl rich foliage there will be some difference.

 

When designing the cameras there was a good reason to include the sharp-cutting UV and IR rejecting dichroic filter. 

[Andy Perrin]

I mean you could always stack a Baader IR/UV cut filter on there with it, Ulf? Or similar dichroic ir cutting filter. 

[LarsHP]

5 hours ago, photoni said:

Some time ago I did a different test with a dandelion flower
I used QB29+QB39 and TSN575+QB29 the result is similar
the percentage of IR is negligible in the visible with QB39

The ICF or QB21 2mm filters also look IR free when just looking at the color chart, as seen below.

However, when you know the black cloth should look neutral, it becomes obvious that there is IR leakage.

That is gone when using 3mm QB21.

 

[ulf]

1 hour ago, Andy Perrin said:

I mean you could always stack a Baader IR/UV cut filter on there with it, Ulf? Or similar dichroic ir cutting filter. 

Yes you could, but you have to find a dichroic filter that cuts in a similar way as the original filter in the stock camera.

If I remember correctly the  Baader IR/UV cut filter is a bit aggressive cutting too much of the last red wavelengths.

[Andy Perrin]

I think you are exaggerating the difficulty, Ulf. Human color perception is just not that good, especially at the far edges of the spectrum. The Baader in particular might cut a bit steeply, but there are many cheap filters on the market and I think a little experimentation would find one that clips the IR reasonably well without messing up red. Our eyes are just not very good (most of us).

[photoni]

@Andrea B. : there is a problem, the files now saved on the site LOSE their color profile

.
@LarsHP : there is a problem, you keep putting overexposed files, the black of the table must not be 61-57-53 but close to 25-25-25

.

.

now I close,
greetings and thanks

Toni

[ulf]

15 hours ago, Andy Perrin said:

I think you are exaggerating the difficulty, Ulf. Human color perception is just not that good, especially at the far edges of the spectrum. The Baader in particular might cut a bit steeply, but there are many cheap filters on the market and I think a little experimentation would find one that clips the IR reasonably well without messing up red. Our eyes are just not very good (most of us).

Yes, you are absolutely right, Andy. 
Some of those filters cut too much red, but otherwise it isn't that critical to get very close to the removed stock cameras's transmission, as long as long as it helps the ionic filter blocking enough IR.

[Yves W]

I have Fuji X-T2 and X-T20 Full Spectrum by Sven Lamprecht

BG39 and BG38=QB21 are bad for red and the white contains a bit of green

The Hoya cut UV-IR filter is good but allows some IR to pass through.

Maybe double it.

The best result is achieved with the double-layered Fotga cut UV-IR filter. The black cloth is pure black, the red is red, and the white is pure white.

A single layer of Fotga is not enough.

Thank you Toni for your beautiful tests

[Andrea B.]

OK,  I have written about this in the past. But we can't expect everyone to be familiar with past writings! 

And I note that camera color profiling has been mentioned above. 

 

To restore proper Visible color to a converted full-spectrum camera, you must

1) Use the UV/IR-blocking filter to make a photo of a color checker card which includes a Visible white standard.

2) Create a white balance preset from the white standard for that camera+lens+illumination.

3) AND, create a color profile from the color checker card for that camera+lens+illumination.

4) Apply both the white balance preset and the color profile to any photos made with that camera+lens+illumination.

 

White balance alone can sometimes be sufficient to get your Visible colors looking better.

But Vis colors will never be at their best without that extra color profiling step.

 

ADDED LATER:  As alluded to above, the goal is to get as close to the original Visible colors as possible. 

Perfection probably cannot be achieved. But we don't let Perfection get in the way of Good Enough, right?

 

 

---

 

 

References
Colour Calibration in a Converted Camera

<> Sticky :: White Balance in UV/IR Photography <>

Five IR-Blockers Tested for Visible Color Use on Full Spectrum Camera

This reference contains comparisons between WB-Only and WB+Color Profiling

with respect to restoring proper Visible colors in a converted full-spectrum camera.

Both methods are applied to a series of Visible photos.

 

 

---

 

 

Some generic comments here for Lars

(which kindly ignore if you already know them! I write for the general audience when responding to some topics.)

• You most likely need a UV/IR-cut filter, not just an IR-cut filter.
  Typically, UV-light is less "contaminating" in Visible photos. But it can upset things.
• Adobe has never been the best at white balancing photos from converted cameras whether Vis or UV.
  (Don't ask me why. Who knows? Its preset camera profiles just can't handle conversions I suppose.)
• The white balance step does not necessarily restore proper saturation. But saturation is easily dealt with.
  If a Vis photo made with a conversion has wonky saturation after WB, just fix it.
• Before shooting your subject, make an in-camera white balanceon the color checker WB page for the given camera+lens+illumination.

 

Lars, with your permission, I would like to work with at least one set of your raw files to determine whether either the ICF filter of the QB21 filter can be used successfully if both WB and Color Profiling have been applied. If you are interested, please upload the raw NEFs to this topic and I will pick them up.

 

[Andrea B.]

Toni writes:  there is a problem, the files now saved on the site LOSE their color profile.

 

Andrea:  ????

This image was created with an embedded sRGB profile.

When I drag if off the page and open it, that sRGB profile is still there.

 

[LarsHP]

22 hours ago, photoni said:

: there is a problem, you keep putting overexposed files, the black of the table must not be 61-57-53 but close to 25-25-25

.

.

now I close,
greetings and thanks

Toni

 

I don't know where you get the 25-25-25 values from, but when I download the ColorChecker pdf-guide and open the image of the chart in Photoshop, the darkest gray patch is actually 55-53-55, which is lower than the values you point out in my test, but more that twice as high as you indicate.

 

[LarsHP]

56 minutes ago, Andrea B. said:

...

 

Lars, with your permission, I would like to work with at least one set of your raw files to determine whether either the ICF filter of the QB21 filter can be used successfully if both WB and Color Profiling have been applied. If you are interested, please upload the raw NEFs to this topic and I will pick them up.

 

 

I will re-photograph the chart again tomorrow (sun all day is promised) and upload here. That said, I know that for best possible color reproduction, a custom camera color profile is needed.

 

However, as I have stated a couple of times in this thread, my point with this thread is that the ICF and QB21 2mm filters are too thin for a Sony camera, which is easily demonstrated when shooting in sunlight with the daylight (sun symbol) in-camera preset. The ICF (2mm) filter yields a way too red image before WB in post-process, while the 3mm QB21 stack is quite close and the ICF + 2mm QB21 is too cyan-blue. When setting WB in ACR, the following values make the point clear, since daylight is about 5400-5600 Kelvin:

 

2mm ICF:               4400K / -36 tint (green)

3mm QB21:            5500K /  -4 tint (green)

4mm (ICF+QB21): 6800K / +30 tint (magenta)

[Andrea B.]

Yes, I do understand what you are saying. 

Many IR-blockers leak a bit of red and/or IR. We use them anyway because it is difficult to find the the correct filter glass or the best thickness to use as a replacement for removed internal filters. And of course there is the expense factor.

And sometimes color profiling can help clean up the problem a little, or some, or maybe even a lot.

 

*******

I have to jump in here and remind everyone that, yes, we all also know that preparation of a file for posting as a website JPG here on UVP will probably screw up our Vis colors a little bit more. We have to live with that. 

 

[LarsHP]

New test scene with the color chart indoor for minimum reflections from the patches and sunlight at approximately 45 degrees to the chart. This should enhance contrast and color fidelity. In addition, I used a modern lens (Voigtländer Nokton 50mm f/1.2 Asph) which has top quality coatings, and added a Zeiss UV T* UV cut filter (which cuts below 410nm). As before, I set and used the in-camera daylight white balance in order to clearly see how close the used hot mirror filter is to what the camera had originally.

 

All raw files are shot with manual exposure at 1/500s and f/5.6 and ISO 100.

 

For reference, I include my stock Nikon Z7 using custom camera color profile (DCP-file) and set white balance in ACR. So, that's the reference image. In Adobe Camera Raw, I adjusted exposure slightly to get the third-darkest gray patch to about 150-150-150, and used the QB21 3mm shot to match the reference regarding exposure and saturation, letting the other filter test images get the same adjustments as the 3mm QB21.

 

 

The reference image: Nikon Z7 (stock) w. custom camera color profile and WB set in ACR:

 

 

Sony a7FS w. ICF filter:

 

 

Sony a7FS w. 2mm QB21 filter:

 

 

Sony a7FS w. 3mm QB21 filter:

 

 

Sony a7FS w. 1,5mm QB39 filter:

[LarsHP]

Here are the same images as above (except the Nikon Z7 above, which already has WB done in ACR), but with WB done in post process (Adobe Camera Raw).

 

White balance values in Adobe Camera Raw for these images:

ICF:                  3550K / -31                        (Note that this is my 67mm ICF filter, not the 52mm I used in the first test.)

QB21 2mm:     4200K / -5

QB21 3mm:     5200K / +22

QB39 1.5mm:  3550K / -27

 

Sony a7FS w. ICF filter:

 

 

Sony a7FS w. 2mm QB21 filter:

 

 

Sony a7FS w. 3mm QB21 filter:

 

 

Sony a7FS w. 1.5mm QB39 filter:

 

It should be obvious that all but the 3mm QB21 leaks infrared and has too bright red patch. The 2mm QB21 is the closest after the 3mm stack. The others are simply leaking too much red and IR.

 

[LarsHP]

19 hours ago, Andrea B. said:

...

Lars, with your permission, I would like to work with at least one set of your raw files to determine whether either the ICF filter of the QB21 filter can be used successfully if both WB and Color Profiling have been applied. If you are interested, please upload the raw NEFs to this topic and I will pick them up.

 

Since the four RAW images are 145MB and takes a while to upload, I chose to upload them to Google Disk instead. If you want to make sure this forum has them available for the future, you can upload them somewhere where you think it's best. I will keep the folder public at least for the next year, though.

 

https://drive.google.com/drive/folders/1hhr8CD8grxGK41J8IIQQ-A53FkRaW0cg?usp=sharing

[Andrea B.]

Lars - Thank you for the additional work & tests. This is much appreciated by me and others as we continue to try to learn how to get the best from our converted full-spectrum cameras. It is also important (very!) to learn how the less expensive filters compare to the "standard" Schott/Hoya filters.  I want to be able to offer the best advice to folks wanting to try UV/IR photography without their having to spend too much $$$ for the gear. 

 

I'm going to read carefully through what you have just posted. And then I will experiment with with your shared files. I will return later with a report/comments. Thank you, again, for the work and for sharing your files!!

 

*****

 

I do have that Zeiss T* UV-cut filter listed as the current "best", recommended UV-cut filter.

 

****

 

Lars, has your Color Checker card experienced any wear & tear or degradation? I currently am using my second CC Passport. The first CC became worn and would not stay open anymore. Now the 2nd CC has begun to show some marks and wear on the color patches because I do use it regularly outdoors in the sun and wind. No matter how much I try not to damage it, there is the inevitable fumble, or dust blown onto it, or other such events. I am trying to decide whether or how much minor marks or wear on the patches matters for the relatively simple color corrections we do for Vis photos made with filtered full-spectrum conversions (...as opposed to the color fidelity required by product manufacturers, etc.) 

[LarsHP]

1 hour ago, Andrea B. said:

Lars - Thank you for the additional work & tests. This is much appreciated by me and others as we continue to try to learn how to get the best from our converted full-spectrum cameras. It is also important (very!) to learn how the less expensive filters compare to the "standard" Schott/Hoya filters.  I want to be able to offer the best advice to folks wanting to try UV/IR photography without their having to spend too much $$$ for the gear. 

 

I'm going to read carefully through what you have just posted. And then I will experiment with with your shared files. I will return later with a report/comments. Thank you, again, for the work and for sharing your files!!

 

*****

 

I do have that Zeiss T* UV-cut filter listed as the current "best", recommended UV-cut filter.

 

You are most welcome!

 

My Color Checker Passport is very, very sparingly used and is only two or three years old, I think. I feel confident it hasn't lost its colors or anything like that.

 

After looking at my initial and more casual test shots (before posting here), it appears that the 67mm ICF filter is thinner (probably 1.5mm), than the 52mm "ICF" named filter I have, which is 2mm. Unfortunately, it shouldn't be a surprise to me, since I ordered 2mm QB21 filters in 52mm and 58mm size respectively, and after measuring them, I discovered that the 52mm size is in fact 2mm, but the 58mm is 1.5mm only (as reported in the first post). I should emphasize that the "ICF filter" I used in the first test (it has printed "ICF" on it), was supposed to be a 2mm QB21 in 52mm size according to my order. After seeing the ICF designation, and complaining about the only 1.5mm thin 58mm filter, Jason answers me: "Our ICF filter use the QB21".

 

However, Tangsinuo has posted different graphs for ICF and QB21 (both in 2mm thickness). In other words, we have to recognize that we can't trust the spec'ed filter thickness, even though it is written clearly in the eBay listing by Tangsinuo. We need to unmount each filter and measure the thickness.

 

Regarding ICF versus QB21, we may get one or the other, which may not be as grave a problem as the thickness variation. I am apparently not the only one to order two sizes of the same filter, while receiving two different filter types or thicknesses. Look at the below graph, which was posted here. It is supposed to be QB21 in 1.5 and 2mm thickness, but it is obvious that the upper graph is QB21, while the lower is ICF, when comparing to the official graphs (see my first post in this thread). 

[Andrea B.]

That is disappointing that Tangsinuo is not supplying the requested thicknesses for their filters.

 

It is nothing new to note the lack of quality control in manufacturing by some Chinese companies. But it is difficult to know whether the Tangsinuo seller is aware of the discrepancies in the filter thicknesses? Tangsinuo may be going by the manufacturer's specs without checking? Or perhaps Tangsinuo thinks that  .5 mm or so difference between thicknesses makes no difference?

 

I think I should add another cautionary warning to my recommendations about Tangsinuo filters. 

 

 

---

 

 

I am working on Lars' files. 

Right away I found a small problem which I cannot explain.

 

CASE:  Sony A7 full-spec with ICF filter (as described in earlier posts).

There are 2 successful conversions (with WB) shown here and one failure.

 

 

The  CC Passport as shot by Lars with A7/ICF in Daylight setting, resize, no edits.

 

 

 

The A7/ICF file converted in Photo Ninja with

• generic Daylight setting, and
• white balance applied to 3rd neutral patch from the bottom.

The WB was successful. The patch reads neutral at approximately 180/180/180.

• 

 

 

The A7/ICF file converted in Sony Image Edge with

• Sony Daylight setting, and
• white balance applied to 3rd neutral patch from the bottom.

The WB was not successful. The patch reads blue at approximately 144,190,224.

What went wrong here? This is Sony's own app. I can't find anything I did wrong.

 

 

As a sanity check, I ran this file through ACR using a Camera Standard profile

and again applied WB to the 3rd patch from the bottom.  

The patch reads neutral at approximately 193,193,193.