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

[Andrea B.]

I'm not sure what the conclusion is here -- don't use Sony's own app to white balance unusual files from a Sony full-spec camera???

[Andy Perrin]

Andrea, that is exactly the conclusion I would draw. Sony’s apps have never impressed me with their quality, although in this case it may be unfair to disparage them for not working with a camera that has been modified. But regardless of the reason, PhotoNinja has always had the best color correcting abilities I know of for full spectrum cameras. (Although their RAW conversion has sometimes failed in other areas.) 

[Andrea B.]

Andy, thx. Yeah, that was a definitely weird result. But I didn't think the file was so far off that it couldn't be corrected in almost any converter! 

*****

 

 

Back to the point.....

Lars has shown that

A7 full-spec + external ICFx2.0 + Daylight in-camera setting does not produce correct Visible colors.

 

There's a reddish-cast in such photos because a bit of high red and low IR passed, both of which are recorded in the red channel. UV is also passed, and most likely in recorded in either the red channel or both red/blue channels.

 

However, if we take a photo made that way and correct the white balance during conversion, the result seems to be acceptable. Similarly, I would expect that pre-setting an in-camera WB prior to shooting with ICFx2.0 would also give an acceptable result. So I'd say that the ICFx2.0 seems to be reasonably useable to restore visible colors in a full-spec conversion as long as WB is applied before or after the photo session.

 

The question remains:  does color profiling add anything to the results?

Let's see.

 

Conversion in Photo Ninja with generic Daylight setting and

manual WB on 3rd patch from the bottom.

 

 

Conversion in Photo Ninja with color profile created from CC Passport.

WB again applied on 3rd patch from the bottom.

 

 

These two results are very very close!

See the next post for a more direct comparison.

[Andrea B.]

Comparison of the two methods.

 

NOTE:  The label should read ICF x 1.5 mm.

 

[Andrea B.]

Lars - thank you for letting me learn from your files. Almost all my tests for correcting visible colors have made use of Schott/Hoya filters and Nikon conversions.  I've always worried that might skew any advice I've given about UV/IR-blocking filters. But this first result is similar to my other tests. So that's good. 

 

I'm going to go ahead and work up a similar color comparison for the other filters you used. But I'll only present the final results.

[Andrea B.]

Here is the workup for the QB21 x 2.0mm.

 

 

Now then, is there a way to determine whether the ICF profiled colors are "better than" or "worse than" the profiled QB21 colors? I'll take any suggestions about how to attempt such a comparison.

[LarsHP]

48 minutes ago, Andrea B. said:

...

 

Back to the point.....

Lars has shown that

A7 full-spec + external ICFx2.0 + Daylight in-camera setting does not produce correct Visible colors.

 

There's a reddish-cast in such photos because a bit of high red and low IR passed, both of which are recorded in the red channel. UV is also passed, and most likely in recorded in either the red channel or both red/blue channels.

 

However, if we take a photo made that way and correct the white balance during conversion, the result seems to be acceptable. Similarly, I would expect that pre-setting an in-camera WB prior to shooting with ICFx2.0 would also give an acceptable result. So I'd say that the ICFx2.0 seems to be reasonably useable to restore visible colors in a full-spec conversion as long as WB is applied before or after the photo session.

 

Unfortunately, as I explained in the previous post, I don't think the ICF filter I used today is 2mm. It is most likely 1.5mm. I suggest you correct it by simply removing the thickness indication. 

 

I have been able to unmount the 52mm (2mm) and 58mm (1.5mm) filters, but not the 67mm ICF filter, so at the moment, I don't know the thickness of that. The smoking gun is the redder color balance than the 52mm "ICF"-named filter (that was supplied when ordering QB21 2mm, sigh). 

 

Regarding getting the color charts right, it is true that we get quite good results by creating a custom camera color profile with the correctly set white balance.

 

However, we don't photograph test charts! Subjects that have a high degree of IR reflection will still look bad even though the color profiled chart looks great. That's why I always include my black cloth in these test shots. If the chart looks great, but the cloth looks redish, then it's still bad, in my opinion at least. 

[Andrea B.]

Yes, I will amend the ICF x 2.0 labeling. Thanks for the reminder!

 

However, we don't photograph test charts! 

No, we don't. But we can apply the WB obtained from previously photographed white standards by saving that WB as a preset.

Similarly, we can apply the Color Profile created from previous CC Passport photos by saving that Profile or saving that Profile as part of a larger preset. Then when making Vis photos containing those pesky, potentially IR-reflective areas we have the non-IR-reflective parts correctly colored. And we can deal with the IR-reflective stuff separately.

 

I'll try that out in just a moment.

 

 

[Andrea B.]

Lars observed that the QB21 x 3.0mm filter was rather good when used only with an in-camera Daylight setting. See the third photo in this link:  LINK

 

I created a composite of color patches from a Photo Ninja generic daylight + WB conversion, from Lars original "as shot" version and from a Photo Ninja color profiled version. The two conversions are quite close to the "as shot" version. Quite interesting!!

I also encountered a couple of problems. See comments after this composite photo.

 

 

Observation:  I've been trying to figure out why my daylight, white-balanced version and my color profiled version both have so many grey patches which are not entirely neutral. I added some RGB labels so that you could see what I mean. I did not put the saturation values on those patches, but obviously some of them do not have 0% saturation. There were sat values like 1%, 3%, etc.

 

The only reason I can think of for this lack of neutrality is that in the original raw files there is some noise, both luminance noise and color noise and that perhaps that noise affects white balance. So, I am wondering whether the CC chart photos should be de-noised before being used to create WB presets or color profiles?? I am going to try to do that experiment. But for today, I'm out of time!

 

Here is an example of the noise in the grey patches. This is the 3rd patch from the top after a conversion with application of both white balance and color profile. No noise reduction was applied during conversion or before creation of the color profile. This patch is not neutral. The RGB values are not equal in almost all areas of the patch.

 

Here is the same gray patch with some color noise reduction (Noise Ninja in Photo Ninja). There's still some noise.  But now the RGB values do seem to be mostly the same throughout the patch. This should happen with all the blotchy colors being gone.

 

And finally the gray patch with both color noise reduction and luminance noise reduction. Now the RGB values are the same per sample and they have less variation between samples.

 

 

Seeing how smooth and how neutral that last patch is after both color and noise reduction, I'm wondering if we shouldn't somehow incorporate noise reduction -- esp color noise reduction -- into any process or preset involving white balance or color profiling?

A good experiment will perhaps tell the tale.

 

I learn new stuff every time I experiment with this stuff! Cool!!

[Andrea B.]

Ulf writes:  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.

 

Many newer cameras do not have an anti-aliasing filter. Example:  Nikon D810, D850, Z7, Z9.

[Andrea B.]

The IR-reflective black synthetic materials problem.

 

Here is the uncropped conversion of Lars' black cloth and CC chart photo made with the ICF x 1.5. Color profiling was applied. So everything but the IR-reflective black cloth should have reasonably accurate color. 

 

 

 

Now I'll put a TIF version of this photo into an app in which I can change the bad color of the black cloth. For this photo that will be easy. There will certainly be some photos for which this kind of color correction of supposed-to-be-black areas becomes a tedious problem! However, if the only UV/IR-cut filter you have also leaks some IR, it is still best to make a good white-balanced, color-profiled conversion before tackling the restoration of black areas. What do you think? YMMV, of course. 

 

I used a neutralizing brush over the black cloth and then darkened it a bit.

 

 

 

[Andrea B.]

I learned some interesting things from Lars' work and from his files which he has graciously supplied to us above.

• Noise on color patches might play a role in the outcome when CC cards are used for white balance or for color profiling. We should test whether denoising prior to creation of WB or profiles is useful.

 

• There *are* some UV/IR-blockers available which can provide a reasonable restoration of Visible colors in full-spectrum cameras without a lot of effort in color profiling. See above the QB21 x 3.0mm when used with Daylight setting. Thanks, Lars for this finding!

 

• Some UV/IR blockers are useable even if they leak a bit of high red or low IR. But then you must make good WB and color profiling presets and apply them to correct any non-IR-reflective regions of the photo. And then you must be prepared to isolate any IR-reflective areas for further correction.

 

We also have learned, unfortunately, that not all thicknesses are correctly stated on Chinese filter sites. INSIST on refunds or exchanges! This should not be happening.

 

*******

 

I'm looking forward to seeing more experiments from Lars, from Tony and from others who are interested in getting the best out of our full-spectrum cameras. 

 

 (I am a terrible editor of myself. Please let me know if I have erred or typo-ed in any of my comments. Thanks!!!)

 

 

*******

 

It is possible that I have just learned that using full-spectrum conversions for Visible work is not exactly easy and maybe we should just grab the stock camera for Vis.

But then we wouldn't have any fun trying to figure things out, would we? 

[ulf]

4 hours ago, Andrea B. said:

Ulf writes:  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.

 

Many newer cameras do not have an anti-aliasing filter. Example:  Nikon D810, D850, Z7, Z9.

That is correct, but the important thing here is not the antialiasing function but the added filtering of the dichroic surface.
I do not think there are any newer camera that is relying on only a thin ionic BG-filter.

 

It is easy to think of a filter as a reasonably thick component, but a dichroic filter function is happening at the surface of some substrate.  

[LarsHP]

9 hours ago, Andrea B. said:

I learned some interesting things from Lars' work and from his files which he has graciously supplied to us above.

• Noise on color patches might play a role in the outcome when CC cards are used for white balance or for color profiling. We should test whether denoising prior to creation of WB or profiles is useful.

 

• There *are* some UV/IR-blockers available which can provide a reasonable restoration of Visible colors in full-spectrum cameras without a lot of effort in color profiling. See above the QB21 x 3.0mm when used with Daylight setting. Thanks, Lars for this finding!

 

• Some UV/IR blockers are useable even if they leak a bit of high red or low IR. But then you must make good WB and color profiling presets and apply them to correct any non-IR-reflective regions of the photo. And then you must be prepared to isolate any IR-reflective areas for further correction.

 

We also have learned, unfortunately, that not all thicknesses are correctly stated on Chinese filter sites. INSIST on refunds or exchanges! This should not be happening.

 

*******

 

I'm looking forward to seeing more experiments from Lars, from Tony and from others who are interested in getting the best out of our full-spectrum cameras. 

 

 (I am a terrible editor of myself. Please let me know if I have erred or typo-ed in any of my comments. Thanks!!!)

 

 

*******

 

It is possible that I have just learned that using full-spectrum conversions for Visible work is not exactly easy and maybe we should just grab the stock camera for Vis.

But then we wouldn't have any fun trying to figure things out, would we? 

 Thank you for doing the additional side-by-side work.

 

It makes the color differences more clear. Regarding the 3mm stack, your tests also point out that it is quite good hot mirror for Sony cameras. I think that, if I didn't have or wanted to spend the money for a Color Checker chart, and the extra work it requires to make use of it, I would live happily with just using 3mm QB21 on my Sony a7FS. It is dense enough to cut away IR contamination, and it doesn't change the colors much compared to the original filtering.

 

Using 1.5mm QB21, QB39 or ICF may work in daylight with a custom camera profile and no IR reflecting objects, but try shooting in incandescent / halogen light! Halogen lamps emit large amounts of IR and the color temperature is so low that it will be impossible to get a proper white balance in raw converters from Adobe, DxO Photolab and Capture One and possibly others, that are limited to 2000 Kelvin in the low end.

 

Interesting that Photo Ninja has a separate color setting for light source. That is neat, and something I would have liked to see in other photo editing software programs. Neither Adobe Camera Raw, DxO Photolab nor Capture One (according to memory) has that.

 

Today, I received my lens spanner wrench. As I expected, the 67mm ICF filter is in fact 1.5mm (measured 1.48mm).

 

Earlier you posted this comment in response to me:

Quote

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?

 

Actually, at one point I saw that Tangsinuo said they are the manufacturer themselves, so they should know what they are selling. Your last suggestion seems to be most likely, based on the response I got from Jason at the Tangsinuo eBay store when complaing aobut the 1.5mm 58mm QB21 filter: "Thickness is not workable?".

 

[LarsHP]

I just did a quick and simple test with a plant in my living room in order to see how much color difference there is from near IR leakage in the green leaves. To me, it is obvious that they have a red tint in the image with only 1.5mm filter. Of course, this is without creating a custom camera color profile, which will remove at least some of the red discoloring in the darker tones. And with a 2mm QB21, it will be even less. Still, my stance is that 3mm QB21 is the best filter for a Sony camera, even though one may "get away" with 2mm and a custom camera color profile.

[Andrea B.]

Lars wrote:  Actually, at one point I saw that Tangsinuo said they are the manufacturer themselves, so they should know what they are selling. Your last suggestion seems to be most likely, based on the response I got from Jason at the Tangsinuo eBay store when complaining about the 1.5mm 58mm QB21 filter: "Thickness is not workable?". 

 

So it must be that Tangsinuo does not understand that he should supply the *exact* thickness which the customer has ordered? That he does not understand that a .5 mm difference between the filter thickness requested and the filter thickness received can completely mess up the photographic results for some types of filters??

 

I don't know what to say! The only thing you can do is to insist that ordered specifications must be met. And if they are not met, then you will give a bad review to Tangsinuo, file a complaint with Ebay and request your money be returned. 

 

I put a warning in Best Basic Gear: Goggles, Filters, Torches about Tangsinuo not sending the requested filter thickness.

---

 

Lars writes: Still, my stance is that 3mm QB21 is the best filter for a Sony camera.

 

I'm assuming you mean "Under the assumption that the in-camera Daylight setting used"? But given that it is so easy to make an in-camera WB under the QB21 x 3.0mm, then why stop at just the Daylight setting?

---

 

The two foliage fotos look very much alike! It is not so easy to see any reddish cast in the 2nd foto as I sit here looking at my laptop screen in the ambient warm lamp light this evening.  I'll pull the two fotos off the page and measure a few spots.

 

I have not yet had time to add a QB21 x 3.0 comment to Best Basic Gear. Will try to get to that soon!

[photoni]

@LarsHP  is right when he says the QB39 has IR leaks.
I have seen that black cloth (as he says) is a very good indicator
...but the differences in the colors of the flowers and the landscape are minimal.

Only the purple flowers (which turn black under UV) become pinker.

last year I tried QB39 1.5 mm and 2 X QB39 and TSN575 the most natural colors cone with 1 QB39 1.5 mm
(P.S. I don't have 2mm S8612 or BG40 or BG39 or BG38)

[Andrea B.]

Toni:  I have seen that black cloth (as he says) is a very good indicator

 

Yes, synthetic, dark fabrics used for clothing, backpacks, or upholstery are very sensitive to IR light. I remember having some problems with that when using early non-modified DSLRs such as the D100. And the first digital Leica M (the M8) was notorious for bad blacks due to IR reflectivity. Leica had to send 2 free IR-blockers to everyone who bought the first M8s.

 

Also in earlier DSLRs at the other end of the spectrum, the visible violet colors were quite difficult to capture correctly with non-modified cameras. (I still get the occasional problem with dark purples or violets!)

 

 

---

 

 

I did not have time to do a thorough analysis of the colors in Lars' foliage photo, so I only measured 4 spots. But I did not find much difference between the two photos. IR-reflective greens don't seem to exhibit the kinds of problems that IR-reflective synthetic black/dark materials present. So, like Toni has noted, the QB39 x 1.5mm colors, are quite natural. And we can see that the QB21 x 3.0 colors are also.

 

 

 

 

 

This has been a very enjoyable topic. As mentioned, I learned a couple of things. I just hope Lars doesn't feel that we "hijacked" his topic!! 

[LarsHP]

14 hours ago, Andrea B. said:

...

Lars writes: Still, my stance is that 3mm QB21 is the best filter for a Sony camera.

 

I'm assuming you mean "Under the assumption that the in-camera Daylight setting used"? But given that it is so easy to make an in-camera WB under the QB21 x 3.0mm, then why stop at just the Daylight setting?

...

In my opinion, the best hot mirror filter from Tangsinuo for correct visual color reproduction with Sony cameras is 3mm (2×1.5mm) QB21 in any condition. 2mm QB21 is not bad, but neither sufficient, and 4mm is too much, while not having issues with IR leakage and incandescent lighting. 

 

Even when using auto white balance, setting WB in post processing or making an in-camera WB, the colors will be off when using 1.5mm (2mm too, but less so) ICF, QB21 or QB39 filters.

 

If a custom camera color profile isn't used, red colors will be too light even though greys are neutral, and other colors with red channel content will be affected too. A custom camera color profile will help this, but IR leakage will show itself when there are subjects that reflect a lot of IR. A 3mm thick QB21 solves all these issues and may even be used without a custom camera color profile, unless color fidelity is critical. 

 

As seen in the first post in this thread, hot mirror filters in Sony cameras cuts the most, so for a full spectrum converted Canon camera, 2mm QB21 may be enough.