Red-Blue ratio?

[Timber]

I was testing the Nikon EL-Nikkor 50mm f4 vs f2.8, both old version. My finding is the f4 gives more contrasty but darker picture. Checking the channels (with same white balance settings) the f2.8 gives much more blue. Everything was exactly the same, light settings were very identical (there was roughly 15-20 seconds between the two shots... can't swap lenses faster than that B)), both 0.4 sec, ISO 800, f8.0. I've read somewhere that there is this red-blue ratio for lenses in UV. This made me think that the visible Violet color consists of Red and Blue (can say if they are 1:1 that's a perfect shade of violet?) so since our cameras are RGB sensors both the R and B pixels are recording them but probably not at full "power". That could be one of the reasons why they are more sensitive to IR as Red is much closer to Infrared than Blue and especially Red from Ultraviolet. Even if UV is only recorded by the Blue pixels the wavelength "distance" from UV to Blue is much further than IR to Red. I hope you can still follow my thinking. I know I could be absolutely wrong as I have not studied in depth the way sensors are recording the light, I am just trying to use common sense here which tells me that each color sensor records it's own color...

 

Anyway... what I would like to ask is if with the same white balance a picture has more blue than an other one made with different lens, then does it means it records more UV or less UV? Since if you compare the photo below there is almost no difference in Red channel but huge difference in blue and I've read that the f2.8 is a better performer than the f4.0 one. Am I correct on this one?

 

http://clancode.hu/!wrx/redblueratio.jpg

 

Both shots were made:

Sony NEX-6, lens as stated above, Hoya U-360 + BG40

0.4 sec, ISO 800, f8.0

[nfoto]

it's quite a bit more complicated than so.

 

Firstly, forget anything about a direct mapping of spectral band to a given RGB channel. This is due to the fact that the Bayer colour dyes have transmission side lobes outside their designated spectral area. Thus, a 'red' pixel also responds to UV and IR in addition to red. Our cameras tend to 'leak' most (or respond to, if you prefer that description) outside the visible range in the red channel. Some Canons I have used (early models) responded more in blue to UV, so this is partially model specific. Few if any have much direct response in the G channel, but due to the cross-overs etc. you can still observe green hues in the false-colour palette.

 

Secondly, because the various dye transmission spectral curves have cross-overs or overlap in parts of their range, a quite wide variety of false colours can result that has no unique bearing to the actual spectral range in terms of nm values. Thus it is not feasible with any degree of certainty to relate a given false colour to a defined spectral range. The inverse mapping is not unique.

 

Thirdly, a camera set to either Tungsten/Incandescent or a low Kelvin number, will give colder appearing files straight off the camera when the UV response of the lens is restricted compared to warmer reddish even orange colours for a wide-band UV lens such as a Coastal 60 or UV-Nikkor 105. However, this cannot be interpreted as it having less UV transmission per se, just that the lens probably won't go very deep into the UV range (which obviously answers two entirely different questions).

 

What we have established so far, however, is that by calibrating the camera against a know UV neutral (UV white, UV grey, or UV dark) standard reference, for example, a piece of virgin Teflon, most scenes take on a very similar false-colour rendition. A lens restricted to say the 380-400 nm range still will appear colder and the richer and warmer hues of the false-colour palette will be muted or even absent.

[Andrea B.]

I'd like to add here that white balance for an ultraviolet photograph is defined as a white balance achieved either in-camera or in an editor/converter by use of a white (or light grey) standard target which is invariantly reflective in both visible and UV light.

 

For a scientific experiment, you would use a white diffuse reflective standard such as those from Labsphere.

For practical photographic use, a piece of PTFE works well enough even though it is not diffuse

and can occasionally throw the white balance off just a little bit because of specular reflections.

 

Once you have standardized the way you set white balance, then you can attempt to sort out other UV false colour factors.

 

We have found, so far, remarkably similar false colours in our UV fotos. But 'similar' does not mean 'same'.

Read on.......

 

What can determine the UV false colour which you see after white balance has been performed?

Way too many variables!!

• lens construction (is the lens colour corrected? what are the elements made of? coatings?)
  
• sensor construction (CMOS/CCD/microlenses/backlighting/and more)
  
• Bayer filter pattern and filter dyes (are they uniform across camera brands?)
  
• UV-pass filter (bandwidth/type/thickness/etc)
  
• in-camera demosaicing software (method for reading Bayer filter and more)
  
• in-camera white balance software (each camera has its own native white point)
  
• amount-and-kind-of-UV-in-sunlight
  (which depends on time-of-day/time-of-year/altitude/latitude/longitude/etc)
  
• amount-and-kind-of-UV-in-flash
  (which depends on filter/bulb type/etc.)
  
• editor/converter (camera brand specific or generic?)
  
• editor/converter settings applied after foto import (saturation/contrast/hue)
  
• editor/converter white balance tool type (not all alike)
  
• monitor type (LCD/backlit?/plasma/analog/etc)
  
• monitor calibration (white point setting and more)
  

I've probably left out a couple of the variables. But you get the idea I'm sure.

 

In spite of the aforementioned similarity of false colours in our UV work, there are simply too many variables which can affect the nuances of UV false colour to be able to make any judgements about what a particular false colour might mean spectrally.

 

Personally, I think there is probably something to the argument that a false colour might indicate a spectral property.

But --- the experiments thus far are not rigorous enough to make a spectral claim with any authenticity because there is no control of the variables.

 

As an example of false colour variability:

I have opened a Nikon UV photograph in Photo Ninja and in Capture NX2, performed a white balance step against a Labsphere white standard in each converter and then sometimes seen dark green in the NX2 conversion but not in the PN conversion.

Why are these two conversions different? I don't know.

Does that dark green false colour mean anything? I don't know.

 

Another example of false colour variability:

Look at the floral work posted here on UVP and you will see quite a variation in the UV false colour "blues", "yellows" and "dark greens" even for the same flower. Some fotos seem to have a tinge of pinks or purples that others of the same flower do not have.

 

It is all very fascinating.

[Andrea B.]

Timber - and other posters - when you post fotos please remember to clearly provide all relevant data about camera, lens, filter and illumination used to make the fotos. And it is usually nice to provide the exposure data.

 

We see a lot of photos and don't always remember which camera you might be using. And any casual reader won't have any idea at all what gear you are using for the displayed fotos.

 

Thanks!!!!

 

Example:

Use some statement like the following formatted according to your personal preference.

(My formatting is not a UVP standard.)

 

Equipment [Nikon D600 + Nikon 105/4.5 UV-Nikkor + Baader UV-Pass Filter + Sunlight]

Exposure [f/11 for 1/200" @ ISO 200]

[Timber]

But in this case I used the same white balance setting on both shots. The shots were made in RAW then I used Lightroom to set the white balance to 2000 / -150 and exported it to a 16 bit TIFF. Then on the TIFF file -100 / -100 on the white balance and that gave me the above results. It seems that the exposure on red channel is about the same but then on the green and the blue channel the f2.8 EL-Nikkor is brighter. As the filter and the camera (and settings) were exactly the same I assume the f2.8 was letting more light trough, hence the brighter picture. Basically my main question is not about the colours itself as I am aware it's all false color (although it's very interesting where are those colors are coming from...), but if all the settings are same and I assume that both lenses are transmitting the same amount of visible light, then this extra light we can observe in blue channel (in this case) is coming from better UV transmission? What really interests me is why only one or two channel is receiving more light?

 

I think what you mentioned Bjorn, is if a lens is less UV capable then it will need longer shutter speed and by that it will "leak" more visible light, in this case blue. But in my experiment I used the same exposure and filters for both shots, which should mean that the light entering the front lens is equal but the light leaving the back lens is more on the 2.8. But what makes me loose some sleep (oh there goes my afternoon nap) is why only one channel? I am sorry for having so many questions... sometimes I feel like I have the curiosity of a scientist but the intelligence of a lobster B)

 

Andrea: I did mention the exposure settings but you're right forgot the gear part... I updated the original post with it and in the future I will make sure to add as much information as possible.

[nfoto]

I wish the RAW files were available for examination. You have a very unorthodox work flow.

[Timber]

http://clancode.hu/!uv/redblueraw.zip I converted to DNG as well in case your software does not opens the ARW

[nfoto]

Thanks.

 

I'll try the DNGs but native format (ARW) would be better. Let me have a look at the files first. Andrea might give them a spin as well.

[Timber]

Oh sorry... you misunderstood me... I posted both DNG and ARW files.

[Andrea B.]

Here's the two photos as converted in Photo Ninja.

White balance was performed on the street.

Blown highlights were pulled back.

Color Enhancement = Plain.

Detail = 15.

Shadow = 100.

That's all.

 

One photo does have considerably more blue in it. I don't know why. I'm wondering if you had any light leak? I would reshoot to see if this is repeatable.

 

DSC01677

 

DSC01678

[Andrea B.]

An examination of the R, G and B channels shows that 1677 has more green and much more blue than 1678. The red channels of both photos are about the same.

[nfoto]

i got comparable results in PN. These two lenses clearly handle the UV band in a slightly different manner. Both would be considered adequate performers in UV though, as they do show some UV-related clues (headlights and wind screen in car are darkened for example).

 

On a tangential note, sharpness demonstrated is not the crispest I've seen in a UV capture and there is shadow noise as well. The frame 1678 has a little touch of motion blur too. 1677 appeared better although the colours came out slightly muddled.

[colinbm]

I am wondering about the custom WB & how they were taken.

Perhaps a single custom WB without any lens, or a custom WB for each lens on exactly the same target ? Just a suggestion.

Has the viewfinder been covered properly for both shots ?

Col

[Timber]

Hi Col, I have a Sony NEX-6 which has no OVF (just a fantastic EVF) and in my post the pictures are having exactly the same White Balance (copy/paste settings), but yeah, later on I wanna do a better test with a "static" environment as I have some black-light bulbs at home.

[Timber]

Just one thing I was thinking about... I took the image from Enrico's 35mm lens test page (http://www.savazzi.n...phy/35mmuv.html) and checked it as separate channels. It seems to me that the shorter bandwidths are the ones that getting this greenish/yellowish colour while the longer ones getting the purple/blue one. On the images the white balance seems to me "set" on the 370 one and compared to it the shorter bandwidths have no blue and the longer ones have much more blue than red. So could it mean that if with the same white balance settings with two image, if an image has more greenish tone then that lens transmits better on the lower bandwidths (so it's a better UV lens) or if the image is more blue/purple toned then it transmits worse? You can also see that the frame around the test circles (which is black I suppose) gets a heavy blue tone with lenses performing worse in UV (I suppose due to longer shutterspeeds and more VIS or IR contamination). I am only guessing but later on I will try to compare lenses with good and terrible UV transmission. I can't make such extensive tests like Enrico (I don't have the same equipment and definitely not the same knowledge) but will do my best with my limited equipment :D

 

This is his test-strip for the UV-Rodagon 65mm

[Andrea B.]

I've cautioned time and again not to overinterpret the false colours as being representative of spectral bands.

 

There is some truth to what is shown in the filter board photo when such tests are made with fluorite/quartz dedicated UV-lenses.

 

But let me show you what can happen with false colours by posting two photos I just made yesterday.

 

The Vivitar 35/3.5 lens, which is not the best transmitter of UV, for some crazy reason gave me green grass in this UV photo.

And in a not particularly uniform manner either.

 

The UV-Nikkor 105/4.5, which transmits at 70% between a very wideband 220-400nm, shows no green grass in a similar photo made in the same shoot. (No, there isn't as much grass to be seen on the right in this UV-Nikkor photo, but it certainly is not green.)

 

******************

 

I cannot for the life of me figure out where that green grass is coming from in the Vivitar photo? Vignetting?? Coatings??? Both photos had the same white balance treatment when being converted.

 

Added: Of course I'm going to look further into this. I need to try a different converter to see if for some reason that has anything to do with this.

[JCDowdy]

I cannot for the life of me figure out where that green grass is coming from in the Vivitar photo?

 

I have seen that same green grass with my Vivitar 35/3.5 (and other lenses), apparently it also grows here too!

 

I've cautioned time and again not to overinterpret the false colours as being representative of spectral bands.

 

So, I am formulating an explanation which may help understand this observation, or at least perhaps begin to form a testable hypothesis.

 

In my line of work there is the term "spectral mismatch" which describes the error of using a broadband radiometer which is calibrated for a light source different than the source being measured. If the spectral response curve of a radiometer is not essentially flat then sometimes significant measurement error results.

 

Without over-interpreting, as you caution, I know that the typical Baader-U filtered false color response of most of the modified Bayer array systems can be shifted to the green or to the blue by a spectral shift to shorter or longer wavelengths in the source being imaged (measured) relative to the source used to create a custom in camera WB (calibration). This is both predictable and reproducible.

 

Taking the filter transmittance curve and spectral CFA response curve as constants, the two other major factors influencing the false colors recorded must be 1) the spectrum of the UV source, and 2) spectral cut-off of lens transmittance. Use of a lens with essentially flat spectral transmittance across the transmittance band of the filter removes the lens as a source of mismatch. A lens with a spectral transmittance that cuts off around the same region of the spectrum as sunlight and/or the short wavelength side of the filter transmittance could be a significant source of variability.

 

This is I believe often the case, The solar UV cut of is ~50% @ ~320nm (we use Schott WG-320 filters with 50% cut-off @ 320nm on Xe-arc lamps to simulate this in the lab), The Baader-U has a HBW 320-380nm so again ~50% @ ~ 320nm, and many non-UV-specialist lenses have transmittance that will be cutting off around that ball park too.

 

​Could subtle changes in solar UV, such as increasing proportion of UV-B to total UV from sunrise to solar noon, or increasing total solar UV intensity lifting less intense short wavelengths above detection threshold, be sufficient source of WB spectral mismatch to observe as this false UV-green cast?

 

A cut off filter could be used to test this with a good UV lens. At solar noon set a proper Spectralon or PTFE custom in camera WB with a simple piece of window glass over the lens and then remove this glass and look for green grass. I have WG-320, WG-335, WG-345 and WG-360 filters with which to conduct this test. Probably a good idea to include the WB standard in any photos of what is observed as the color cast should show up there too.

 

Another possible test would be for someone with a UV-Nikkor to set a proper Spectralon or PTFE in camera WB with it and then switch to a simi UV lens of comparable focal length, perhaps something like that KURIBAYASHI PETRI ORIKKOR 1:3.5 F= 105MM that I recently lost an auction for (I hope someone here got it!), compared as well to a more modern less UV capable 105mm of some sort. WB with each of the other lenses and shoot all three against each WB.

Enough, I will stop here before I circle back on myself..........

[Timber]

Andrea don't get me wrong, I am far from jumping into any conclusions on this matter, I am just trying to gather all the information on this subject and trying to ask questions (first to myself :D). It's a kind of a brainstorming rather than scientific article :) I know that my Sony 18-55 transmits terrible UV, so I can set it as one reference and let's say the EL-Nikkor 50mm f2.8 seems to me my best lens so far, so that can be my second reference. But obviously whatever my tests will result could mean anything and even the opposite of it. But it would be nice if we were able to set some kind of standards, that regardless of location, time of the day or anything could give not an exact result but a guideline...

[Andrea B.]

But it would be nice if we were able to set some kind of standards, that regardless of location, time of the day or anything could give not an exact result but a guideline...

 

All of our informal determinations of UV-capability are not without merit. The evidence begins to add up from the informal tests. But without actual spectral measurements we can only have an indication of UV-capability and not an assurance.

 

Unfortunately it is almost impossible to design an experiment for testing UV transmission in sunlight because of the variability of UV across latitude, longitude, altitude, time of year and time of day. Not to mention the other variables (listed elsewhere) from the various pieces of gear.

 

It is also difficult to design such a set of informal standards for UV transmission in the studio because it would require the use of some kind of serious studio UV lighting. Our typical small UV flashes tend to be variable. Large studio flashes such as Bjørn's Broncolor are quite expensive. Even with studio lighting nailed down using Xenon arc lamps, we are still left with the aforementioned gear variables.

 

Oh well.

[Andrea B.]

Could subtle changes in solar UV, such as increasing proportion of UV-B to total UV from sunrise to solar noon, or increasing total solar UV intensity lifting less intense short wavelengths above detection threshold, be sufficient source of WB spectral mismatch to observe as this false UV-green cast?

I don't think so as the weird green was not observed from other lenses during the test shoot.

 

I think the Vivitar anomaly is probably a pile up of center/edge mismatch and inaccuracy in white balance conversion. As I wrote somewhere else (lots of writing these last few days!), the white balance is something that is never perfect even with use of standards.

[Timber]

I've made another experiment today. Took 10 lenses and shot the clear sky with them, f8.0 and ISO800, let the camera choose shutter-speed with metering set to spot/center. Then arranged the results by shutter-speed and load them in PhotoNinja. Then I took the first image, set the white balance in the middle and didn't changed anything else (all other options unticked). Then copied the the settings on all other photos. Exported them and load them in Photoshop as a strip. Then did the same with all the rest of the images, so I got 10 strips in total. Also loaded all the gray (white balanced) images in one strip and loaded the images in Lightroom and moved the sliders all the way left on the white balance sliders (2000/-150) and made another strip. In the attached image you can see the 12 strips.

 

The lenses and their shutterspeed:

(A) - E.Ludwig Meritar 50mm f2.9 - M42 - 1/25

(B )- Nikon EL-Nikkor 50mm f2.8 (old) - M39 - 1/25

(C )- Nikon Series E 50mm f1.8 - Nikon F - 1/20

(D) - Auto Chinon MC 50mm f1.7 - Pentax K - 1/13

(E) - Sigma DX EX 30mm f2.8 - Sony E - 1/13

(F) - Tamron BBAR 28 f2.8 - Adaptall - 1/10

(G) - Jupiter 9 85mm f2.0 - M42 - 1/6

(H) - Samyang 12mm f2.0 - Sony E - 1/5

(I) - Porst UMC 50mm f1.2 - Fujinon X - 1/5

(J) - Sony SEL 1855 f3.5-5.6 @55mm - Sony E - 0.6"

 

The letter on the top (for the columns) shows the result from the given lens, the letter on the side (for the rows) shows which lens' white balance was used for the strip.

Gr means that's the "correct" white balance strip and LR means that's the most I can get in Lightroom (which still gives some hints about UV performance)

 

 

Not conclusions just observations:

- Same white balance on 2 images caused the better performers to be more yellowish tone or bluish tone the worse

- There is hotspot in UV as well not just IR (i was not 100% sure on this until now)

- in LR setting the white-balance to 2000 / -150 made the better performer lenses a bit closer to red/orange while worse performers a bit closer to violet/pink

 

What I think is since my filter-stack lets through some visible violet I believe that causes the white balance difference as worse performers letting through more at the longer wavelengths. It also backs up what I noticed on Enrico's tests with different wavelengths, where the strips were white-balanced on the 370nm part and shorter ones were yellow while longer was purple.

 

If someone could do the same test (exact same white balance with different lenses) that could help us solve the yellow-purple colour mystery :D

[enricosavazzi]

We have already established that, at least with some lenses (including CoastalOpt 60 mm Apo) it is necessary to use a tight lens shade to avoid the central hotspot in the NUV and VIS ranges. I always use lens shades with front opening diameter as small as 28 mm with this lens. My UV-pass filter is always protected by the lens shade as well, never wide-open to all light. I wonder if this is why I never noticed the mystery dark-green shades that have been discussed recently, and other false-color shifts and hotspots.

 

Perhaps it would be good to repeat the outdoors test with and without a tight lens shade (= long enough to eliminate off-axis light and narrow enough to just barely avoid vignetting) protecting lens and filter from off-axis illumination, to see if these phenomena disappear.

[Timber]

unfortunately I don't have lens shades for all of the lenses, but it seems to me mostly the wide angle lenses (Sigma 30mm, Tamron 28mm and Samyang 12mm, which also shows heavy vignetting with the filterstack) and the Porst 50mm f1.2 (which has "huge" front and back elements). Luckily the Meritar does not shows any hotspots. Do you recon a lens shade can help IR hotspotting as well?

[enricosavazzi]

If the hotspotting is caused by off-axis illumination, then it should help also in the NIR. This is not the only cause of hotspots, however.

[Andrea B.]

Timber, I need a moment to figure out your chart. I haven't quite "gotten it" at first glance.

 

Enrico, apologies! I have been so busy that I have not dug out the photos with dark green yet. I've just hooked up the external hard drive to find some examples.