Panasonic Lumix S1R Conversion: An Exploration

[Andy Perrin]

Stefano, assuming it isn't an artifact of noise reduction, then my logic is that perhaps the green channel is much much more sensitive than the red and blue, perhaps because it's more transparent at shorter wavelengths. I was hopeful that this might lead to better UV-B sensitivity. If it's just some kind of noise reduction effect, though, then it's probably not any better.

[Stefano]

Isn't the green channel less sensitive in the Lumix S1R? It is made less sensitive by software?

[Andy Perrin]

Isn't the green channel less sensitive in the Lumix S1R? It is made less sensitive by software?

D'oh! I was reading Andrea's histograms backwards. Oh well! (But maybe the R and B are more sensitive then??)

 

Whether it is by software is unknown. David's relayed comments are not confirmed yet I don't think.

[Andrea B.]

Andrea what ISO are you using?

Do you see a difference in green response from ISO 50, 100, 200, 400.

 

ISO 100 only. I'll have to make more tests for ISO 50, 200, 400.

 

---

 

Here is the raw color set for D610 and S1R with the U-340 x 7.00 stack on the UV-Nikkor. The UV-flash SB-140 was used for lighting.

Again, the raw color of the U-340 stack is different between the two cameras.

 

I forgot to say above that the Raw Histograms are only for a central selected region of the Spectralon rectangle. Hand-held flash lighting is not always even.

Had I selected the entire Spectralon area for the histogram, I don't think anything would have changed in the relationships between the R, G and B channels. There would have been more histos in the gram though.

 

 

Nikon D610 + UV-Nikkor + U-340 x 7.0 mm + SB-140

Central rectangle over Spectralon 5" x 5"

This color is 29° on the color wheel, orange.

 

 

 

Lumix S1R + UV-Nikkor + U-340 x 7.0 mm + SB-140

Central rectangle over Spectralon 5" x 5"

This color is 10° on the color wheel, red.

In UV light, green has gone missing once again in the S1R capture.

[Andy Perrin]

Q. What do you call a gram without enough histos?

 

Ans. A has-bin.

[Andrea B.]

[Andrea B.]

NOW does everybody see why I'm always warning that false color can't be used to judge wavelength??

[Stefano]

Well, now I see that there are many more variables involved. Not even raw colors seem to be good!

[dabateman]

Ok so the green dye and blue dyes are clearly different. That is really cool.

Your UVB raw color would not be green like I see, but purple or magenta.

Very interesting.

That blue dye just keeps going. Imagine if we could request a sensor with that blue dye and the standard green and red dyes that Sony uses. White balance in deep UV would be so easy.

[Bill De Jager]

A spectral response to dye for?

[dabateman]

A spectral response to dye for?

 

Sounds like you have been watching too much James bond.

[Andrea B.]

O U guyz! my merry punsters.......This calls for a tankard of ale. Or a Monty Python skit.

---

Dabateman: Ok so the green dye and blue dyes are clearly different. That is really cool.

 

I don't know. I'm kinda upset. Because for years we have been using this (so-called) Standard White Balance for documentary work in order that results appear the same for the same subjects when photographed across different gear, i.e., Bayer-filtered cameras and UV-capable lenses.

---

Dabateman: Andrea what ISO are you using? Do you see a difference in green response from ISO 50, 100, 200, 400. If only the green channel is suppressed internally to reduce noise and increase readout time for the AF DFD system to work, that could partially explain what you're seeing.

 

I'm going out to make this test now. I'm thinking that de-noising the green data would not explain the different raw colors. De-noising the green data should not affect how the UV goes through the Bayer dye? But we'll see.

[Andy Perrin]

Dabateman: Ok so the green dye and blue dyes are clearly different. That is really cool.

 

I don't know. I'm kinda upset. Because for years we have been using this (so-called) Standard White Balance for documentary work in order that results appear the same for the same subjects when photographed across different gear, i.e., Bayer-filtered cameras and UV-capable lenses.

I think you are upset prematurely. The point of white balance is that it takes whatever the camera spits out to RAW and normalizes it. As long as the same wavelengths are being recorded somehow to the sensor, it shouldn't matter what the RAW values are, because the white balance should get rid of any cast and leave PTFE looking white. It's only if you see a cast AFTER white balance that there is a problem, and that could simply be that even (e.g.) photoninja has limits on what it can normalize.

[Andrea B.]

To be clear, the white balance step does make PTFE/Spectralon white in UV photos made by the S1R full-spectrum conversion. BUT, the false colors are different. Not a lot different. But different. You can see this in some of the Bouquet photos above made with D610 & S1R.

 

That U-340 x 7.0 mm photo, for example. In the S1R bouquet photo there is much less false color than would be the usual with the D610. Has nothing to do with my conversion tweaks. The same sat is applied to any of my conversions.

 

In the BaaderU photo, for example, the S1R discovers a lot of teal green which the D610 does not.

 

I'll need to make some D610 and S1R pairs in the next few days to illustrate this better.

[Andy Perrin]

We are going to end up with some kind of UV Color Checker, aren't we? It's basically the same situation as with visible light, where a white balance gets you 90% of the way, and then you have to fine-tune it. I know how to make a program that will convert colors from one camera to another by remapping it, so it's possible in principle if I had photos of the same scene with both cameras, and if the scene had the full range of false color in it.

[Andrea B.]

We are going to end up with some kind of UV Color Checker, aren't we?

 

Maybe.....but so much depends on those Bayer dyes. If some new dye is used in some new camera and that dye does not change the visible color but changes the UV false color, then it isn't so easy to decide what to do.

 

I've got my results about ISO levels and green in the S1R, so I'll post that next.

[Andrea B.]

S1R ISO test here.

 

Gear: S1R Full-spectrum + UV-Nikkor + BaaderU UV-pass Filter + Sunlight

Subject: Spectralon 5" x 5" and Color Checker Passport

ISO Settings: 50, 100, 200, 400, 800

Aperture: f/8

Exposure Length: I set each exposure time so that the right-hand side of the JPG

histograms matched at the different ISOs.

 

Raw composites were made for each of the 5 files in Raw Digger.

The same rectangular selection over the Spectralon area of the photo

produced raw data Histograms from Raw Digger.

White balance was applied in Raw Digger using RGB Render.

 

Observation:

The raw composites for all 5 ISOs are the same.

The white balances for all 5 ISOs are the same.

The raw data histograms for the Spectralon area for all 5 ISOs are the same

with the understanding that higher ISO histograms show the ISO boost.

 

Conclusion:

Noise reduction in the S1R green channel at lower ISOs does not appear to affect

the white balance.

 

Next I'll make some histograms over the Color Checker Passport.

 

5 Raw Composites

 

 

5 Raw Data Histogams on the Spectralon

 

 

5 White Balances on the Spectralon

[dabateman]

Well that puts that idea to bed.

 

You know this was bound to happen. Especially to you with the tools to properly test it because you kept warning Stefano that the raw color couldn't be used for wavelength determinations. Now you have a clear cut Anti-Stefano camera.

 

Just like the Anti-Cadmium lens, igoriginal 35mm.

 

I still wouldn't be upset due to color, some flowers might look better with this pallet.

 

What concerns me is the sensitivity. A sensor is 2 green for every 1 red and 1 blue. Your green seems to drop off like my blue. So I have green and red sensitivity (3/4ths of my sensor) for a deeper UV filter stack. Since you don't have green you only get about 1/2 your sensor being sensitive.

 

But too be positive, you found a sensor with a blue dye that is still UV sensitive into deeper UV. I can hope this isn't a one off like the CYMG sensors and that we might see this dye used with the regular green and red dyes. Yes UV color would be different again, but white balance and overall sensor UV sensitivity should be improved.

[colinbm]

Can you sketch a graph of what you think the CFA response is below 400nm please ?

[Andrea B.]

EDITOR'S EMBARRASED NOTE: I mislabeled the 2nd photo.

 

The white balance I made in Raw Digger is not the same as the white balance I made in Silky Pix Darktable 3.6.0+3.

More baffled than ever by this S1R.

 

I'll only show one example: ISO-400

 

 

[Andy Perrin]

White balance algorithms differ dramatically in sophistication, Andrea, and I can see from the raw histograms that there are dramatic sensitivity differences between channels, hence the WB algorithm must struggle to bring things into balance. It’s not surprising that this highlights differences between algorithms.

[Andrea B.]

Yes.

For example, we have seen WB problems with Adobe Camera Raw.

 

I'll try a couple of other apps.

[Andrea B.]

Luminar4 does a good job of white balancing the S1R UV photo made with the BaaderU.

This is the previous ISO-400 photo made with BaaderU and then white balanced in Luminar4.

 

[Andrea B.]

OK I am a complete doofus who mislabeled an example above. GEEZ, Andrea. I should NOT.do.stuff.late.at.night.

I labeled a white balance example above with "Silky Pix" when I should have labeled it with Darktable 3.6.0+3.

Fixed now.

Time to retire......

[Andrea B.]

Silkypix Developer Studio 8 SE was not quite as good as the Luminar4 white balance above, but it was much better than the Darktable white balance. With Silkypix it seemed like the photo became rather desaturated, so I tried to add back some sat. But there is this cyanish tint on the right third of the Spectralon. Where the heck did that come from? The Silkypix white dropper covers only a tiny bit of the photo, so you can't drag it larger to cure this tint.

 

 

 

 

Same conversion only without the sat push to see if the cyanish tint goes away.

It does seem to go away.