Jump to content
UltravioletPhotography

Fluorescence-Mediated Sensor Response in Ultraviolet Photography: A Short Case Study


OlDoinyo

Recommended Posts

It might be easier for people to give you feedback, if you re-post this interesting information here. Some may do not even use Flickr at all and will not be able to reply to the original post either.
Link to comment
enricosavazzi

In the past, I had given some thought to the possibility of fluorescence by the Bayer filters. However, in the cameras I tested and with an excitation wavelength of 365 nm, I was not able to view or photograph any excited fluorescence in the VIS range emitted from the sensor.

 

This does not exclude that there is sensor fluorescence in some sensors and/or with shorter excitation wavelengths, or that there is some fluorescence emission in the NIR.

 

I think, however, that the research departments that investigate the pigments to use in Bayer sensors would be aware of the problem and try to avoid fluorescent pigments.

Link to comment

I don't know why it would be seen as a "problem."

 

The sensors for digital consumer cameras were designed for ordinary RGB photography in an environment where the overwhelming majority of the incident radiation falls between 400 and 700 nanometers. We tend to lose sight of the fact that, when we intentionally use these sensors in environments where this is not the case, we are using the equipment in ways for which it is not designed, an "off-label" use, if you will; and if they behave capriciously under such circumstances, that is not the engineers' fault nor problem. (I have also seen some truly wacky firmware antics when entire light channels are filtered out, but I digress.) Sensor fluorescence would be utterly insignificant in the context of everyday RGB use, because of the poor UV transmission of many lenses and because the hot mirrors of many cameras now incorporate UV blocking.

 

Red filters are usually quite opaque to UV, and presumably this would include the capping filters in Bayer masks. Unless these filters happen to have a fortuitous transmission window below 350 nm, I have difficulty explaining why the red channel of many cameras has the highest UV sensitivity (particularly at shorter wavelengths) without invoking fluorescence as an explanation. In fact, I have trouble explaining why the red channel would have any response at all. Although I no longer have the camera mentioned in the link above, my present Sony cameras seem to act in a very similar manner: without white-balancing the most noticeable characteristic of their UV images is red tint.

Link to comment

"In fact, I have trouble explaining why the red channel would have any response at all [in UV]".

 

The response curve for each dye has side lobes outside the dedicated passband. Most cameras (Nikons at least) will respond in red to a UV input, but some such as the Fuji also registers in green and a Canon used some years ago (forgot the model 1D something) produced mainly a blue image when my UV-Nikkor was macgyvered onto it.

Link to comment

I'm still waaaaay back there trying to sort out the experiment.

 

Camera: Canon 1DsMarkII [QUESTION: broadband or unconverted??]

Filter: Baader-U, 330-365nm halfmax with peak @ 350nm

Illumination: Sunlight

Lens 1: Pinhole: 30 seconds @ ISO-1600.

Lens 2: Zeiss 50/1.4: f/22 for 30 seconds @ISO-100. UV transmission characteristics unknown.

 

The Pinhole photo shows almost all R in the right-hand 4/5 (approx.) of the histogram.

The Zeiss photo shows more green in the left-hand 2/5 than the Pinhole photo.

The Zeiss photo shows more blue in the midtones than the Pinhole photo.

The Zeiss photo shows less red in the right-hand 3/5 of the histogram than the Pinhole photo.

 

Therefore we conclude something something about fluorescing Bayer dyes? I'm not sure I understand why? Why don't you conclude instead that the Zeiss might be truncating some portions of the UV passed by the Baader-U?

Thanks in advance for some help with understanding the conclusion.

 

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

 

it is possible that some of these dyes may fluoresce when exposed to ultraviolet light. The visible photons thus emitted.....

 

UV can also induce IR fluorescence in some subjects. Why are you assuming Visible fluorescence? And, if you are assuming Visible fluorescence, then wouldn't we be recording a Visible + UV photo instead of a UV photo?

 

Well, maybe not. If the UV photons are output as Red photons via the fluorescence, we are still somehow recording the UV??

 

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

 

The other common type of digital sensor, known as complementary metal-oxide semiconductor (CMOS), is said to have very limited response to light of wavelength shorter than 400 nanometers.

 

I'm not sure this is correct according to what I've read. Anybody got any references?

Link to comment

None of these references I was prowling around for tells anything about whether the dyes fluoresce. But here they are anyway because at least you can see some dye names and one spectral chart.

 

Bayer filter dye reference: Patent US4416961 https://www.legionpa...atents/4416961/

Patent holder: Karl Drexhage of Eastman Kodak

Drexhage used cationic, photo-bleachable dyes such as chromylium, thiachromylium, pyrylium, thiapyrylium cyanine, flavylium and thiaflavylium cyanine.

 

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

 

Good old Wikipedia mentions some other dyes used in colour filter arrays - but they aren't listed by their chemical names. http://en.wikipedia....es_used_in_CFAs

 

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

 

The link below is about CMOS sensors.

But scroll down to see a spectral chart of Bayer Filter transmission which gives us a tantalizing glimpse below 400nm. It goes a bit further into the UV than most such charts, but unfortunately doesn't go far enough for us. "-)

It does look like the red dye UV transmission is greater than the blue or green dye transmission starting around 380nm or so.

http://www.olympusmi...agesensors.html

Link to comment

Thanks Andrea

 

The Wiki article cited above has a VERY interesting two sentences under "Manufacture of the CFA (Color Filter Array)"

Diazonaphthoquinone (DNQ)-novolac photoresist is one material used as the carrier for making color filters from color dyes. There is some interference between the dyes and the ultraviolet light needed to properly expose the polymer, though solutions have been found for this problem.

 

As far as I can see (as an ancient chemist) this means that during the course of processing the CFA on the CCD or CMOS chip, UV light is used to cure the photoresist. SO all our sensors have been exposed during manufacture to UV... We are all using CMOS arrays now in our DSLRs I think and they seem to respond to UV with or without fluorescence of the red dyes or pigments.

 

Dave

Link to comment
enricosavazzi

I don't know why it would be seen as a "problem."

[...]

Because, with lenses and built-in barrier filters that leak some UV, pigment fluorescence excited by UV is recorded as false color superposed with (and not separable from, in post-processing) VIS radiation. In other words, in sunlight the camera will not faithfully record VIS color, which is what a consumer camera is intended to do.

 

Of course, pigment fluorescence is not necessarily a bad thing for us UV photographers, and may even be a good thing in certain cases, but consumer cameras are not designed for us.

 

On the other hand, pigment fluorescence excited by UV wavelengths so short that they are known to be completely blocked by all ordinary barrier filters is not a problem, unless the camera is modified. This type of pigment fluorescence can be regarded by sensor designers as not relevant in the choice of pigments for Bayer filters.

Link to comment

QUESTION: broadband or unconverted??

 

Unconverted; the hot mirror was, however, not a particularly strong UV blocker in this case. This is of course not true of all cameras.

 

Why don't you conclude that the Zeiss might be truncating some portions of the UV passed by the Baader-U?

 

I do not dispute this assertion; in fact, the whole line of reasoning would collapse without it. Although I don't have an actual curve for this particular lens, we all have a general idea of what such graphs tend to look like. Think of the lens as a crude long-pass filter with a cut-on somewhere near 350 nm; that is how I effectively used it, having no actual filter in my possession with those properties. Conveniently, the lens and the pinhole had similar fields of view.

 

The Pinhole photo shows almost all R in the right-hand 4/5 (approx.) of the histogram.

The Zeiss photo shows more green in the left-hand 2/5 than the Pinhole photo.

The Zeiss photo shows more blue in the midtones than the Pinhole photo.

The Zeiss photo shows less red in the right-hand 3/5 of the histogram than the Pinhole photo.

 

Indeed. In other words, the imbalance between the red channel on the one hand and the blue and green channels on the other hand is markedly reduced by blocking the shorter end of the wavelength band. The key question is, why? Clearly, the red channel is sensing not only more total signal, but shorter wavelengths than the other two channels. Either the red Bayer filter has an unsuspected transmission window extending below 360 nanometers, or we are dealing with some sort of secondary response mechanism. The typical red filters that one screws on to a lens are about as transparent as the proverbial brick wall to UV, so I was disinclined to believe the first possibility. The Olympus reference, however, does give one pause: perhaps some red Bayer pigments are not brick walls after all, since a true red channel needs to sense some blue as well, due to the idiosyncrasies of human vision. But we need a graph with more UV range to decide this.

 

Why are you assuming Visible fluorescence?

 

Well-spotted, a bit of mental laziness on my part. The word "visible" has now been deleted from that sentence.

 

If the UV photons are output as Red photons via the fluorescence, we are still somehow recording the UV??

 

I leave that point to the philosophers to debate. Is a medical X-ray recorded with the aid of an intensifier screen "really" an X-ray?

 

Anybody got any references?

 

It was something that Mr. Rørslett posted in the Nikongear forum. The link no longer works. I don't know if it is still posted anywhere.

 

Because, with lenses and built-in barrier filters that leak some UV.....

 

But we have been led to believe that they really don't much, nowadays; and many photographers from the film era had the habit of keeping a skylight filter on for physical protection of the lens. I suspect quite a few still do; and those that do not may not be the type to care if their snaps are lily-pure visible.

Link to comment

Substitute "fotozones" for "nikongear" in that link. The site was renamed.

 

*****

 

Either the red Bayer filter has an unsuspected transmission window extending below 360 nanometers...

 

Well, there are lots of UV photos "out there" made below 360nm with converted cameras and the appropriate filter.

I think Klaus has done some. Enrico was playing with it too IIRC. I've made a couple myself with a 340BP10 filter.

However, I don't recall just at the moment in which channels the UV landed.

 

*****

 

Thing is, if the Bayer dyes fluoresce then why don't we see that when we shine a UV-LED onto the sensor (with internal filters removed) ??

 

*****

 

But we have been led to believe that they really don't much, nowadays;

 

You can always force enough UV through a non-converted camera and a (so-called) non-UV lens to make a UV photo. It just takes forever and may lack some detail. I've made some UV photos with an unconverted D3 mounted with some newer Nikon lens (don't recall which one just now.)

 

*****

 

I think being suddenly at 7500' has temporarily altered my capacity to remember.

I need some oxygen !!! ;) :lol: :lol:

Link to comment
Even my 300/2.8 ED-IF can capture UV. But I do need to be patient and mount it on a sturdy tripod for the quite long exposures required.
Link to comment

First Statement: Either the red Bayer filter has an unsuspected transmission window extending below 360 nanometers...

Edited Statement: ...capping filters in Bayer masks. Unless these filters happen to have a fortuitous transmission window below 350 nm,

 

Here is a UV photo, before and after the white balance step, made with the 340BP10. UV is recorded in all three channels with most falling in the Red channel. This photo does not, of course, refute the possibility of red fluorescence in the Bayer dyes. It just shows that the Bayer filter red areas do transmit below 360nm.

 

Equipment: Nikon D300-broadband + 105/4.5 UV-Nikkor

Exposure: f/5.6 for 1/6" @ ISO-800

 

No white balance is applied in the first photo so you can see that the red channel predominates.

Next the green and then the blue.

The right side of the photo was cropped.

Not the best photo due to some breezes.

Brassice oleracea var. acephala = Kale

brassicaOleraceaVarAcephala_UV340AFSun_050611wf_18668origpn01.jpg

 

In the second photo, the white balance step has been applied.

No favors done to the photo with the wb step, however. Seems to bring out the noise.

There is vignetting in the lower and upper left corners because the UV-pass filter was so small.

Brassice oleracea var. acephala = Kale

brassicaOleraceaVarAcephala_UV340AFSun_050611wf_18668origpn2.jpg

Link to comment

Still thinking about this !!

 

If a user of a camera such as the Panasonic GH1 (or similar model) always uses a preset white balance, then that user will frequently see the histograms of their UV photographs with little or no recording in the Red channel.

So he/she might never suspect that there is any Red fluorescence???

(I edited that last sentence because I did not correctly phrase it the first time.)

Would these folks suspect some kind of boost in the blue channel perhaps?

 

Oh well. I'm giving up on this one now.

 

ADDED LATER: That was totally stupid what I wrote!! The white balance used is irrelevant to the recording of the raw data. We would (eventually) examine the raw data for any UV photo to determine where the UV was being recorded using an app like Raw Digger or Dcraw.

Link to comment

It just shows that the Bayer filter red areas do transmit below 360nm.

 

Sorry, it just shows that some mode of detection operated in that band. It says nothing about whether the detection was direct (which requires a transmission window at incident wavelength) or indirect (which does not.)

Link to comment

Yes - I agree that "it just shows that some mode of detection operated in that band".

However, I still don't think the case for Bayer filter fluorescence has been made. :D

 

So let's look at the following...a typical Bayer filter transmission chart:

 

cmoschipsfigure4.jpg

 

Reference:

Introduction to CMOS Image Sensors

Microscopy Resource Center

Olympus America Inc 2012

http://www.olympusmi...agesensors.html

 

Note that around 380nm on the chart - which is approximately where UV begins - transmission in the blue and green filters drops sharply under the toe of the red filter transmission. This is telling us that this particular Bayer filter records UV mostly in the red channel below about 370nm or so - and mostly in the blue channel between 370-400nm.

And indeed I have seen this with my converted cameras. When using lenses which do not transmit much below 370nm or when using a narrowband 385nm UV-LED for illumination, I do see more blue in my UV fotos. But when using a Baader-U filter on such a lens as the UAT or UV-Nikkor, quite a lot below 365nm is captured, and I see more red in the raw foto.

 

There are a few (but not many) other similar charts floating around the 'net. In another such chart that I looked at, it can be seen that the red dominance continues down below 350nm. And that somewhere below 350nm, the green comes back up a bit. [i'm trying to find that chart again.]

 

********

 

I have been enjoying thinking about this. I'm not trying to be argumentative, just trying to present some points for discussion as I think it all through. :)

Link to comment

The first link in that post is broken. But I think it references the product in this Pdf: http://www.ccd.com/pdf/ccd_29050.pdf

There are charts in that Pdf but they don't go below 400nm. Still interesting read though.

 

The chart in the 2nd link from Dave's post is very similar to the one above. It is labeled Figure 3 and clearly shows that the blue and green responses dip under the red response starting around 380nm.

 

In the below linked discussion of sensor terminology there is reference made to enhancing the quantum efficiency in the UV spectrum by coating a CCD sensor with Lumogen, which fluoresces. However, I wonder if our camera CMOS sensors are given this Lumogen treatment? I doubt it. Nevertheless, it is nice to find a reference to a process similar to what Clark has postulated for the Bayer dyes.

http://www.truesense...sor-terminology

 

Here is more about Lumogen - and also Metachrome - from Princeton Instruments:

http://www.princeton.../162-uv-coating

 

I wonder if we could get some of that Lumogen stuff and prime our camera sensors??? :D

Well, it is probably not feasible because there is some kind of microlens "binder" over the sensor.

Link to comment

I have two things to say on this subject:

 

1) The "mismatch in sensitivity between the red channel versus the green and blue channels" inferred from the pictures may well indicate the fact that the image processing software (in the camera or in the computer) amplifes signal from red "pixels" (sensils) more than the signal from green and blue "pixels". Here is a picture taken with de-bayered Sony NEX-5N (see here http://www.ultraviol...ersion-results/) and processed with Sony dedicated RAW converter with WB set to daylight.

 

post-29-0-21015200-1415210202.jpg

 

Naturally, signal recorded by all "pixels" is the same in de-bayered camera, but the image set to Daylight WB shows the Red channel to be more "exposed" than the other two.

 

2) Actual observations of UV-induced fluorescence response of the Sony NEX-5N sensor (with sensor cover glass removed)

 

post-29-0-57735000-1415210320.jpg

 

From left to right, same area of the sensor:

Picture of the sensor lit with visible light taken with unmodified camera (ICF intact), AWB.

Picture of the sensor lit with UV light (Nichia 365nm LED filtered through UG11 filter) taken with unmodified camera (ICF intact), AWB, close to how my eyes see it.

Picture of the sensor lit with UV light (Nichia 365nm LED filtered through UG11 filter) taken with full-spectrum modified camera throught Baader U2 filter, UV-specific WB.

Picture of the sensor lit with UV light (Nichia 365nm LED filtered through UG11 filter) taken with full-spectrum modified camera throught Hoya R72 infrared filter, UV-specific WB.

 

Exposures are marked in the picture, everything else is the same. UV-specific WB was set with Baader-U2 filter and UV capable lens (close to 330-340 nm transmission) on the PTFE standard on lightly cloudy day.

 

Areas marked with numbers on the sensor surface:

1 – undamaged sensor surface

2 – area with microlenses removed

3 – area with bayer filter array removed

4 – area with last transparent layer removed (layer between bayer filter array and photodetectors).

Link to comment

To be extra clear, in (1) you are referencing your De-Bayered Sony NEX-5N which is a monochrome camera

 

and in (2) you are referencing a Sony NEX-5N with internal filters removed but with Bayer filter intact. and the Bayer filter in various stages of removal.

Is that correct?

 

Thank you Alex for this presentation.

I have some questions/thoughts, but I need to think things over first. "-)

 

I'm trying to figure out if I have the conclusions correct.

In a converted full-spectrum Sony NEX-5N:

  • There is no UV-induced UV fluorescence.
  • There is no UV-induced IR fluorescence.
  • There is some UV-induced Visible response. And it is blue unless you further remove some layers?
    Added: blue response in area where bayer filter is intact.

Hmm....I've never seen any blue.

OK, thinking.

Link to comment

Andrea,

 

Just to be clear:

 

1) It is a picture taken with de-bayered monochrome converted NEX-5N.

 

2) The sensor imaged here is my busted sensor from NEX-5N that I used to test different de-bayering techniques. The cover glass of that sensor was removed, and also some of the layers (microlenses, bayer filter array, protective transparent layer) were removed in different parts of it. I marked them with the numbers and labelled them. I will repeat myself again:

 

area 1 - undamaged sensor with microlenses, bayer filter array and protective layer;

area 2 - microlenses removed, bayer filter array is exposed;

area 3 - microlenses and bayer filter array are removed, thin transparent protective layer located on top of photodetectors is exposed;

area 4 - microlenses, bayer filter array and protective layer are removed, photodetectors are exposed.

Link to comment
OK, I went back and corrected some misprints, and also replaced the word "fluorescence" with "response". I will let everyone judge for himself what it is.
Link to comment

Intriguing, especially figure 2, panel 2. Evidently, on this sensor, the microlenses themselves fluoresce blue. I would expect this to show up as enhanced UV response in the blue channel. I do not know if my old Canon had microlenses. If whatever we see in areas 3 and 4 is fluorescence, the implication is that the excitation wavelength does not penetrate through the Bayer dyes; area 2 is black or nearly so.

 

Certainly some of the channel imbalance is caused by differing channel gains; however, one would expect that component to be constant, not variable.

Link to comment

Certainly some of the channel imbalance is caused by differing channel gains; however, one would expect that component to be constant, not variable.

 

There is a lot of information online stating that the red channel needs to be amplified considerably during demosaicing in order to represent true colors. Unfortunately, I am not able to locate any credible original sources for this information.

 

When considering the graph that Andrea posted, the transmission of UV below 370nm is different between different channels, with red having several times higher transmission (at least in that narrrow region). If you add the standard gain that is given to red channel during normal demosaicing, the finel image will appear red.

Link to comment

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×
×
  • Create New...