Jump to content
UltravioletPhotography

Thought expt - correcting filter transmission with camera sensitivity and light intensity


Recommended Posts

This is a bit of a thought experiment and please feel free to comment.

 

I got to thinking, filter transmission charts are all well and good, but the lighting we use is not the same intensity at different wavelengths, and the camera sensor sensitivity also drops as the wavelength decreases. So what would be the real life relative contribution of different wavelengths to a filtered image, and how would this be different to the transmission curves for different filters?

 

I've assumed sunlight as a light source for this, and the camera is my Monochrome converted Eos 5DSR. To understand this I needed 2 additional pieces of information (in addition to the filter transmission curves). Specifically, sensor sensitivity, and sunlight absolute irradiance curves. Thankfully I have both. Firstly the sensor sensitivity between 300nm and 400nm - y axis scale is relative;

post-148-0-31600400-1522061701.jpg

 

Secondly, absolute irradiance curve for sunlight between 300nm and 400nm;

post-148-0-96082100-1522061718.jpg

 

I then take these 2 curves, and rescale so they go from 0-1. Then multiply these 2 together, to give a total response factor for sunlight and sensor between 0 and 1 between 300nm and 400nm;

post-148-0-73559100-1522061736.jpg

 

I fitted a line of best fit through this in Excel, as I have nothing more complicated than that. It's ok, but I'm not 100% happy with it to be honest. I then took the equation from that and applied the correction factor to 2 different filters - Baader U and U340 (4mm). I then plotted the original filter transmission curves, along with their curves when corrected for sensor response and sunlight intensity;

post-148-0-00904000-1522061748.jpg

 

Note the corrected curves are plotted against the right hand y axis, which has a different scale to the left one. The scale on the right hand side is not % transmission, but relative contribution of a specific wavelength. As the correction factor drops quickly as a function of wavelength, the Baader U filter effective contribution line looks very different now - its biggest contribution comes in at about 380nm and then contribution drops as the wavelength shortens, which is different to its transmission chart which is at its maximum in the 340 to 360nm region.

 

The corrected U-340 peak contribution is now at around 360nm and is around half of the Baader U. This is very different to the transmission chart, but happens because the U-340 peak transmission for the filter is at a shorter wavelength than the Baader U. Also the ability of the U-340 to transmit at shorter wavelengths is hugely attenuated in the model here, as the sunlight and sensor correction factor approaches 0 as the wavelength reaches 300nm. It doesn't matter what the filter can transmit if there is no suitable light, and no sensitivity in the sensor.

 

This correction factor suggests that our images, in sunlight at least, are predominantly made up of the longer wavelength parts of the UVA region - 340nm and upwards. It'll be interesting to see what the StraightEdge filter looks like when corrected for this given the shape of its curve.

 

Anyway, as I say, this was a bit of a thought experiment, and I wanted to share in case it's of interest.

Link to comment
For color imaging, the white balance chosen is also a factor in determining the spectral distribution of an image.
Link to comment

Interesting effort, Jonathan. I'm not altogether sure about seeing that BaaderU peak move from 350 to 380 nm because that filter just does not image like you would expect to get around 380 when used under broad UV illumination. I have to think about this later when I have more time to read in depth.

The question, however, is a very important one to answer.

 

A thought occurs -- in the Schott transmission program you can add information about illumination. We should check that out. Do they have any sunlight data?

Link to comment

I don't have a copy of that Andrea, so I don't know.

 

The other thing of course is the reflectance of what is being imaged. My thought experiment ignores that completely.

Link to comment
Your thought experiment makes perfect sense to me. I often think of this, except with an added layer... IR contamination (notably, IR leak through "UV" filters). In such a scenario there is competition at the sensor with wavelengths (IR) to which the sensor is very, um, sensitive. Each photosite can only fill up with a certain amount of charge before its 'full'. Many of us here know just how easily IR can contaminate a UV image - and we go to great lengths (or depths [filter thickness ;)]) to avoid it. I don't know much about how optical filters are made, so I can't speak to it - I just assume it must be very difficult to make an IR-leak-free UV filter (without getting overly expensive).
Link to comment

'Out of band', or red/IR, or IR suppression usually works good under OD3, and very good under OD4, and extremely good under OD5. OD6 or more would be overdone, and would simply be cutting down more on UV.

It is not any more expensive to use thicker IR suppression, if anything thinner suppression glass is more expensive.

Link to comment
Andy Perrin
I've actually wondered about the opposite of this —what would happen if we designed a filter that was weighted to give "equal" sensor response across the UV (say, 320nm-400nm)? This is essentially what our white balance is trying to accomplish, but suppose instead of reweighting R, G, and B, we reweighted the incoming light with an appropriate filter? It would obviously have to strongly attenuate the longer wavelengths, which means the overall transmission would be pretty awful, but it might bring out some interesting false colors.
Link to comment

I don't have a copy of that Andrea, so I don't know.

 

The other thing of course is the reflectance of what is being imaged. My thought experiment ignores that completely.

 

Here is the latest version of the Scott Optical Filter Glass Program:

http://www.schott.co...sh-16022017.zip

 

You can add your own Filter- an Light Source-data. :)

 

The data needs to have an integer wavelength step length. One row in the spreadsheet equals one nm.

If your data has less resolution, as from the Hoya filter data sheets, you can enter just that data.

The program interpolate the missing wavelengths.

 

I can contribute with data that fit the Schott program, from my spectrometer measurements, but have not found any way to upload such text files to the forum. :(

Link to comment

Andy,

UV gain flattening filters for Si detectors exist and I have actually tried one. You can find them surplus as "Optical Filter UV Gain Flattening for Silicon Photodiode or Linear Array 305BP40" and "Optical Filter UV Gain Flattening for Silicon Photodiode or Linear Array 290BP40". I do not recall which one I tried, but it was a 25mm x ~2-3mm thick. Not an easy filter to get anything through as you thought, and the thickness also made for nasty reflections. With a fully UV capable lens I was able to get images with long exposures and yes the false colors were weird. Of important note these filters are for flattening the response of Si radiometric detectors, not Bayer CFA imaging sensors, so the response is very likely not flat. One of the de-Bayered monochrome cameras would no doubt make better use of such a filter.

Link to comment
Andy Perrin

John, that's very interesting, and I may try to get my hands on one. I've been meaning to try out a quartz singlet or something too, so maybe that would make a good combined experiment.

 

One of the de-Bayered monochrome cameras would no doubt make better use of such a filter.

Although since the attraction (for me anyway) is the false color profile, that would defeat the point for me a bit?

Link to comment

Jonathan, you should toss a U-360 2mm into that assortment, or actually a U-360 2mm + S8612 1.5mm+ stack.

U-340 4mm is a poor comparison. It doesn't even have enough IR suppression. Alone, it is not a viable or functional UV-Only filter. I know it has an extended lower UV transmission, because it is not a stack,

and therefore it is so tempting, but I discourage people about the U-340 4mm all the time, it is a dream, not a reality.

U-340 4mm leaks Red/IR at 0.5%! That is a most definite Red/IR leak, it pollutes the UV shot, unless you keep your eyes closed to reality.

It is not like the Baader U, it transmits Red/IR. It leaks. It is not a UV-only filter. Sorry, but true.

You only show the UV part of the U-340 4mm graph.

I don't have any data to include the sensor into the mix, but short of a $7,000.00 UV-Nikkor, the biggest bottle neck is the lens transmission, for most of us, that is a reality.

Most other lenses will truncate the UV transmission more than any other factor in your mix/stack.

Of course, lighting is always a factor, type and amount, but strong sunlight is standard, and not that hard to come by part of the year anyway.

I think the Schott program has a D65 illumination selection, which is:

https://en.wikipedia.../Illuminant_D65

 

"D65 corresponds roughly to the average midday light in Western Europe / Northern Europe (comprising both direct sunlight and the light diffused by a clear sky),

hence it is also called a daylight illuminant. As any standard illuminant is represented as a table of averaged spectrophotometric data,

any light source which statistically has the same relative spectral power distribution (SPD) can be considered a D65 light source.

There are no actual D65 light sources, only simulators. The quality of a simulator can be assessed with the CIE Metamerism Index The CIE positions D65 as the standard daylight illuminant."

 

And I think the D65 seems to be the default in the Schott program. And indeed, natural light is what I use mostly, and is the most 'natural' choice.

Way to go Sun! :-)

 

You are smarter than me, and you know all of this already, but my main point is the lens.

 

Your scenario leaves the lens data out of the mix, yet it is the biggest truncation of all.

Maybe not with a UV-Nikkor, but this site spends countless topics, and years pondering the best lenses to use for UV, and that is why, because the lens is #1 in the mix.

Link to comment

Wow, a lot for me to read and digest this morning, and apologies if I missed anything.

 

Steve, I used the 340 as I wanted something with a very different peak position to the Baader U. This was purely aimed at understanding the effect of reduction in the sensitivity of the sensor and the reduction in short wavelength UV in sunlight, not be say that the 4mm U340 was suitable on its own for UV imaging. I only show UV data, as my sensor sensitivity measurement only goes fro 280nm to 480nm, I cannot measure higher than that with my setup. One thing the data looks to say to me is that the majority of the contribution of the UV image is coming from the higher wavelength parts of the UVA - the sensor is more sensitive there, and there is more UV there in sunlight too, therefore more contribution to the image. So really this actually favours the use of a lot of the 'accidental' UV lenses. The extra ability of lenses such as UV NIkkors to go down to 200nm is sort of lost with our type of imaging. With different light sources with more shorter wavelength UV then that would change of course. Obviously there are benefits to reduced chromatic abberation and focus shift from such dedicated UV lenses, but this was never aimed at looking at that. I've used D65 it for skin colour measurement for years, but only in the visible not the UV. I'm not sure how well it matches data in the UV to be honest - never looked at it, so I'll add it to the list. Finally, I doubt very very much whether I am smarter than you.

 

John, those filters are interesting, I'd not heard of those before. I wonder if I can back calculate from my sensor sensitivity graph to give me an idea of an 'ideal' correction transmission profile.

 

Ulf, thanks, I'll have a look at that. EDIT - had a quick look. To me it looks as though the illuminant type is only used for calculating the colour of the filter. As far as I can see it doesn't alter the transmission data at all.

 

Mark, absolutely, I did not include red/IR in this. I can only measure sensor sensitivity from 280nm to 480nm with my setup. Only day I'll make a wider band measurement device, but not today unfortunately :)

Link to comment

{{{{ Off Topic: We're all intelligent. We simply have different areas of expertise. :D B) }}}}

 

Jonathan: To me it looks as though the illuminant type is only used for calculating the colour of the filter. As far as I can see it doesn't alter the transmission data at all.

 

Read some more myself and I think that is also what I've gotten from the illuminant notes. They help you figure out what color cast you will get with, say, a BG filter and where that fits into a standard color space. Quite important if colour is an integral part of an experiment or of an industrial need (paint manufacture for example?)

 

Jonathan: The extra ability of lenses such as UV Nikkors to go down to 200nm is sort of lost with our type of imaging.

 

Yes, no surprise there. The UV-Nikkor, UAT, Coastal 60 - all are scientific or industrial lenses and are/were marketed as such. It might be quite interesting to get a CCD camera with a sensor sensitive below 300 nm and see what things look like way down there in the shortwaves. B)

 

Andrea has a question for John or Jonathan -- what means "UV Gain Flattening" ?? Does that mean just suppressing some UV in the longer wavelengths? "Gain" I thought meant "amplification"? What does gain mean in a filter? Or does the gain refer to the sensor sensitivity? That seems to make the most sense. This UV gain flattener is meant to supress some of the UV to which the sensor is most sensitive??? Thanks.

Link to comment

Ulf, thanks, I'll have a look at that. EDIT - had a quick look. To me it looks as though the illuminant type is only used for calculating the colour of the filter. As far as I can see it doesn't alter the transmission data at all.

 

 

I think that would work to treat the light source data as transmission data and enter it as a custom filter in the Schott program instead.

See your light source-information as the result of a constant intensity light source filtered by an imaginary filter.

Link to comment

Jonathan, I don't agree with your corrected BaaderU curve yet.

 

Where do you account for filling the photosites?

 

My though experiment........

 

For the BaaderU 350CWL UV-pass filter having a 325-369 nm half-maximum region, there is an approximate transmission of 50% at 325 nm and 369 nm.

 

Sunlight at 325 nm and 369 nm has a relative intensity of about .22 and .60. So there is about 2.7 times more 369 nm Sunlight than 325 nm Sunlight.

Assume you are at sea level, high noon, at the equator during summer.

Assume you have a Mojito at hand.

 

In this strong sunlight, set up one 325 nm reflector and one 369 nm reflector against a UV absorbing background. Spot meter against the 325 nm reflector and make a BaaderU image with the UV-Nikkor* on a CCD camera that has equal sensitivity between 300 - 400 nm. (For now, we are assuming the 'perfect' camera.)

 

The intensity of the sunlight reflected off the reflectors, through the BaaderU, through the lens and then onto the CCD sensor would be approximately .5(.22) and .5(.60), or .11 and .30, respectively, for the 325 nm and 369 nm wavelengths. Still an almost 1-to-3 ratio. The resulting image would show the 325 nm as well-exposed, but the 369 nm reflector would be over-exposed. The excess 369 nm light has been "lost" in the over-exposure. The 369 nm wells are completely filled up (so-to-speak).

 

Now, if we had an 'imperfect' camera, like one of our modded DSLRs, for which the sensor is less sensitive to recording the 325 nm wavelengths than the 369 nm wavelengths, then the 369 nm reflector might be even more over-exposed in the image. Although again, there is a limit there -- once the photosite wells are full, well, they are full. Again, the excess 369 nm light doesn't get recorded.

 

The point being that you don't get a net relative contribution in the photo which looks like a BaaderU curve shifted to the right because any excess light is not being recorded and processed when you are metering on the less intense light. (If you meter on the more intense light, then the 325 nm reflector will be under-exposed. But there will never be more light in the photo than what the BaaderU/lens/sensor can pass & record.)

 

*UV-Nikkor: flat transmission of about 70%. But I haven't multiplied the intensities by that factor. Do so if you like. "-)

Link to comment
Andy Perrin
Andrea has a question for John or Jonathan -- what means "UV Gain Flattening" ?? Does that mean just suppressing some UV in the longer wavelengths? "Gain" I thought meant "amplification"? What does gain mean in a filter? Or does the gain refer to the sensor sensitivity? That seems to make the most sense. This UV gain flattener is meant to supress some of the UV to which the sensor is most sensitive??? Thanks.

Did you mean me? Jonathan wasn't discussing that.

 

Gain here is the ratio of the output to the input of the whole system (not just sensor) at each wavelength. It's not really "amplification" unless it happens to be bigger than 1, which is not possible in this application. Gain FLATTENING means you get equal response at each wavelength. Since the apparatus is more sensitive to the longer wavelengths, that means a gain flattening filter must suppress the longer ones more than the shorter ones.

Link to comment

Hi Andrea. Sorry I don't follow, and I've tried reading that a few times (you mentioned we were all intelligent, I'm now beginning to doubt mine). I'm wondering whether the scale on my graph is causing confusion. Perhaps I can explain my reasoning a different way. Firstly, I'll replot the comparison graph, but making the 2 y axes the same scale;

post-148-0-50924500-1522170625.jpg

 

So the results for the 'corrected' spectra are never letting more light through at a specific wavelength than the original filters.

 

The way I think of it is this.

 

Assume that the light source is the same intensity at each wavelength. Also assume the camera sensor is equally sensitive to all wavelengths.

 

Then consider this to be a 3 filter system. One filter which is put over the sensor to drop the amount of light reaching the sensor at shorter wavelengths. A second filter over the light source to drop its intensity at shorter wavelengths. And finally the actual physical filter - Baader U or U-340. The final effect would a product of the 3 filters, the sensor filter * the light source filter * the actual physical filter. In effect this is what I have plotted.

 

Does that make more sense - anyone else, please feel free to chip in?

Link to comment

Yes I think that was the problem. I was interpreting the corrected Baader U relative composition graph the wrong way. I had its curve in my head the way you just now showed it (approximately). Don't know why I didn't mentally adjust the right-hand scale, but I didn't. Couldn't for the life of me figure out why that light blue line was tall and right shifted!!

 

In my example I looked at at two filter transmission points of equal transmission and accounted for sunlight intensity at those two points to describe the same thing. I don't know if I described what I was trying to say well enough to make it clear. You have the approximate 1-to-3 ratio there at 335 & 380 nm in this new chart that I referred to also. My endpoints would have moved if I'd included specific sensor sensitivities (and lens transmission?), but I didn't know what to use for that.

 

Always fun to think about these things even when a step behind. :D B)

 

It seems rather surprising to think that the System as described only captures between a 10 - 30 % rate. Makes you wonder how we ever capture any UV images at all.

 

Added: After going thru all this, I'm not sure it is useful at all to even have an "absolute" filter transmission chart. I wonder why we can't have transmission charts fitted to the light we are going to use? Oh well.

 

Added added: The moral of the story might be to always use a UV-flash to boost the UV in sunlight?

Link to comment

{{{{ Off Topic: We're all intelligent. We simply have different areas of expertise. :D B) }}}}

 

Andrea has a question for John or Jonathan -- what means "UV Gain Flattening" ?? Does that mean just suppressing some UV in the longer wavelengths? "Gain" I thought meant "amplification"? What does gain mean in a filter? Or does the gain refer to the sensor sensitivity? That seems to make the most sense. This UV gain flattener is meant to suppress some of the UV to which the sensor is most sensitive??? Thanks.

 

Sorry, for running off topic.

 

I answer to your question, here is a fair description from an ~2012 Omega press release about their recent product in this category, which I suspect may be what is now showing up in surplus.

 

"...
designed to achieve OD 5 average attenuation at the red and NIR wavelengths while transmitting nearly 50% of the prominent UV A, B, and C regions.

The filter OD 1.0 points are 220 and 320 nm, with attenuation reaching OD 4.0 by 450 nm. The XUVGFF helps to resolve common issues with excessive system sensitivity in the red region (600 – 1000 nm) for silicon based detectors. The extreme red sensitivity of silicon detectors limits the dynamic range in the deep UV region."

 

It is very similar to the problem of red/NIR leaks, or secondary transmittance bands, in filters like the UG11 which can overwhelm the weaker UV we want to image.

 

Apparently custom filters can be made to flatten out diverse spectral profiles as the maker impressively illustrates.

Link to comment

It seems rather surprising to think that the System as described only captures between a 10 - 30 % rate. Makes you wonder how we ever capture any UV images at all.

Good, glad it made sense. The more I research this area the more I am amazed that these cameras can image successfully in sunlight. Keep in mind, my thought experiment was based on a camera with the Bayer filter removed, giving me 2 to 2.5 stops more light than a standard multispectral conversion.

 

John, I don't think you have any reason at all to apologise. Thanks for mentioning about the gain filters, I'd not heard of those before. I was reading on the Omega site earlier about their custom filters. Amazing what they can do. I dread to think how much custom ones are though.

Link to comment

Jonathan,

 

What you have presented is known in photobiological nomenclature as a convolution spectrum. An example would be a solar erythemal effectiveness spectrum with which you are doubtless familiar. Here is a plot from a Wikipedia article on sunburn, because I am feeling to lazy to plot it myself (and it is dinner time and I want to go home).

post-24-0-39812800-1522190416.png

Image Attribution:

By Hankwang (Own work) [GFDL (http://www.gnu.org/copyleft/fdl.html) or CC-BY-SA-3.0 (http://creativecommo...nses/by-sa/3.0/)], via Wikimedia Commons

 

What this tells us is that even though 300nm is roughly 2 orders of magnitude more potent than 320nm for causing sunburn, under this standard solar spectrum there is significantly more sunburn risk at 320nm. Adding in the transmittance of a sunscreen makes a three curve model, analogous to your plot, which is the basis for in-vitro SPF calculations.

 

How's that for going off topic..... :lol:

 

PS, Are you using a standard solar spectrum or one you measured with your spectrometer?

Link to comment

I was reading on the Omega site earlier about their custom filters. Amazing what they can do. I dread to think how much custom ones are though.

 

Hence we play with their eBay castoffs, but it is nice to know such things exist and possible applications.

Link to comment

I'm using one I measured myself John. Measured using my little OO spectrometer, but seems to correlate well with others I found. Ah yes, the erythemal action spectrum plot - many a meeting trying to explain that to marketing.

 

Nothing wrong with off topic :)

Link to comment

{{{Also somewhat off topic.}}} And there's a word I was looking for --- the effective spectrum. B) Or the effective transmission of the system consisting of camera+lens+filter+illumination.

 

Jonathan, could you attempt an effective transmission chart for some camera+lens+filter + UV-Flash?

 

John, thanks for the link. I'm going to go add this one to the pinned Solar chart we have (somewhere).

 

http://www.ultravioletphotography.com/content/index.php/topic/2323-sol-solar-radiation-spectrum/

Link to comment

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×
×
  • Create New...