Jump to content
UltravioletPhotography

Camera sensitivity - chasing ghosts in spectral sensitivity measurements


Recommended Posts

This is going to be a bit on an ongoing post as I add more data to it.

 

As a bit of background I've been trying to get a handle on spectral sensitivity of cameras, and ended up building a device consisting of an Ocean Optics light source (DH-2000-BAL), a little monochromator I bought from eBay which is a older version of one of the Mini-monochromators sold by Edmund optics (https://www.edmundoptics.com/testing-detection/spectroscopy/200---800nm-manual-mini-chrom-monochromator/), and a 2" Spectraflect coated integrating sphere (again another eBay buy). The build is outlined here (http://www.ultravioletphotography.com/content/index.php/topic/2580-build-thread-at-home-measurement-of-camera-uv-spectral-response/page__hl__%20spectral%20%20sensitivity).

 

I've been using this a lot now, and recently tested a Sigma SD14 camera, and got a strange result - a bump in the red channel I wasn't expecting to see at 420nm (http://www.ultravioletphotography.com/content/index.php/topic/2770-foveon-sensor-transmission-charts/page__st__20__p__21821#entry21821).

 

As a result of that and subsequent discussion, I've gone back to try and understand what this bump was, and it turns out that there is a problem with stray light in my setup. Given what has been discussed a few times on this forum, this is not a huge surprise as the monochromator is only a single grating one (although the specifications for it do state stray light is <0.003%).

 

This thread is intended to be a chance to share what I find as I dig into this further, and try and understand these ghosts in the measurements.

Link to comment

Experiment 1 - Monochrome 5DSR with different filters (Hoya R72, Baader UVIR cut, GG455, GG495, OG550, OG570)

 

The aim here was to look at the signal from my monochrome 5DSR camera between 280nm and 480nm, with different filters attached to the lens and compare this with the unfiltered setup. I used a Rayfact 105mm UV lens at f4.5, ISO1600 and 30s exposure for all images. RAW files were imported into Rawdigger as raw composite files and red, green and blue channel data measured for the exit port of the integrating sphere. This was then averaged to give an average response for the camera either unfiltered, or with the different filters attached. I collected images from 480nm down to 280nm in 20nm intervals. Then at 280nm I also closed the shutter on the light source and collected another image - a dark image with no light being let into the integrating sphere.

 

The filters tried were Hoya R72, Baader UV/IR cut, Mason Vactron GG455, GG495, OG550, and OG570 (custom Schott glass filters made for the forensics firm Mason Vactron). The aim with these filters was to block UV and certain parts of the visible (or in the case of the Baader UVIR cut filter the IR) and see what effect this had on the camera sensitivity measurement in the UV. These are the transmission spectra of the filters used;

post-148-0-42341600-1528465296.jpg

 

Here is the data from the unfiltered camera, and then with the different filters presented in table form;

post-148-0-53997500-1528465376.jpg

 

The table shows sensor sensitivity values as a function of wavelength set on the monochromator. These have been corrected for light source intensity as a function of wavelength as mentioned in the original build thread. As expected with the unfiltered setup, the sensitivity drops as the wavelength gets shorter. Even down at 280nm there is a signal observed, above what is seen when the fiber shutter is closed and all light entering the integrating sphere stopped.

 

When I add in the different filters, which should be blocking either part or all of the UV, the signal doesn't drop to the same score as when the light source is blocked off. Now this could be due to the filters not being perfect (not a bad assumption, no filter is perfect and blocks all the light) or it could be due to stray light from the monochromator (more likely as it a single grating monochromator).

 

All the filters with a cutoff above 440nm, show that strange bump at 420nm seen with the Sigma SD14 data which started this all off. This is easier to see if the data is plotted out. Below are 3 graphs, all the same data, just plotted with 3 different scales on the y axis, showing sensor response as a function of wavelength;

post-148-0-94907900-1528465801.jpg

 

post-148-0-89579000-1528465816.jpg

 

post-148-0-73821300-1528465827.jpg

 

It looks to me as though the peak I see at 420nm is mainly driven by IR as the Hoya R72 filter which blocks the visible and UV light, still has a big peak here. The filters which let through the visible and IR gradually increase this size of the peak at 420nm as more visible light is included. So it looks to me as though it is mainly IR with some visible (and potentially UV) contamination. Then there is the general background signal between 280nm and 400nm and at 440nm and above. This looks to be more of a more even combination of IR and visible stray light to me.

 

Ideally I'd go lower with my cutoff, and I will be getting some additional filters as John Dowdy mentioned in one of the other threads. I suppose the question then becomes if it is stray light, how to deal with it? Do I do various measures when collecting data changing the filters so that the cutoff is always not far above where I am measuring? Not sure. I'll have suck it and see, as they say.

 

How big a problem is this? Below is a graph of the monochrome 5DSR camera data unfiltered, the camera with the GG495 filter attached and also one line where the GG495 data has been subtracted from the original data - the aim being to remove the stray light contamination. I chose the GG495 here as I plot between 280nm and 480nm;

post-148-0-67552800-1528466337.jpg

 

The light leak as shown the by the GG495 filter line is generally low compared to the signal from the unfiltered camera. However that bump at 420nm is about 10% of the overall signal. As the wavelength gets shorter the camera sensor sensitivity data gets closer and closer to the stray light signal. However with the GG495 the stray light seems to account for about 20% of the overall signal at 280nm where the sensor sensitivity is the least.

 

This is the best case scenario - the Monochrome camera has the highest sensitivity to UV. With the 5DSR with the Bayer filter still present, it will have much lower UV sensitivity, and I'd expect the stray light to have a bigger impact there. I will be testing that at some point.

 

I also want to check in the visible region using the R72 to see whether there are more 'bumps' like I saw at 420nm. More experiments to come.

Link to comment

This title might be attracting those who chase "ghosts" in old houses with IR and UV cameras. But so far only one comment about that.

 

Added later: I should have added a "humor alert" tag to that comment. It made me chuckle to get asked about the other kind of "ghost hunting". :D :D :D

Link to comment

Jonathan,

 

I assume you have been using both the deuterium lamp and the halogen lamp for these tests.

Is that a correct guess?

For investigating the UV response it might be a good idea to not switch on the halogen lamp.

Then you'll avoid much IR contamination.

Link to comment

Jonathan,

 

I assume you have men using both the deuterium lamp and the halogen lamp for these tests.

Is that a correct guess?

For investigating the UV response it might be a good idea to not switch on the halogen lamp.

Then you'll avoid much IR contamination.

Yes Ulf, it's with both lamps on. When I first built it I decided to measure in two parts 280nm to 480nm and 440nm to 800nm, to give me some overlap. So the 280nm to 480nm part needs both lamps. Unfortunately to go back to just the deuterium lamp for the UV part means taking it all to bits and recalibrating, which at the moment I'd rather not do. Although to be honest I may have to go down that route, depending on what I find.

Link to comment

I was going to ask at what wavelength you are switching over from Du to Halogen, normally in a UV/Vis spectrophotometer that is around 340nm without overlap. The Du spectrum starts to get rough above 400nm and normally the relative power of the two lamps is scaled so crossover to higher relative halogen irradiance is shorter than that.

 

Your second figure is interesting in that it separates stray radiation contributed by Vis wavelengths (Baader UVIR Cut) from the total unfiltered. The fact that the your 420 peak remains with both the 455 and R72 blocking filters implies that peak is either real or is caused by stray radiation >R72 cutoff or is from shorter wavelengths (less likely). The monochromator lacks 2nd order blocking filters so at 420nm there would be a a 840nm grating harmonic not blocked by R72 but why then would harmonics show up only there? With the wavelength range of your source overlap the next point at 440nm is also the first point in the 440-800 segment, is it and the longer points an average?

 

Something to consider is to filter the input energy with a high OD bandpass filter centered on the monochromator setting. If you have a Sparticle filter array those filters positioned between the lamp and monochromator would greatly reduce both short and long wave stray radiant power but grating harmonics would remain. Placed between the monochromator and the sphere those filters would clean up harmonics as well if the filter has high OD blocking > limit of sensor detection. I am pretty sure OO has fiber coupled filter holders.

 

The rule of thumb for the ASTM opaque filter method is the sharp cut long pass filter should have transmittance <1x10-4 at and below the wavelength setting of the monochromator. The auxiliary test for stray NIR recommends a Schott RG-665 3mm which passes shorter than your R72.

 

The rabbit hole is deep and winding, the better to keep the fox a bay.

Link to comment

This is great work chasing down that odd 420nm light you thought was in the Baader filter long ago. The curve looks good and linear. There seems based on my eye fit to maybe have a small impact on the 440nm point on the curve.

Things like this can be fun. You are not doing a PhD. So have fun and take it were you want. Too many PhD students see stuff like this and chase it for years. Thus not getting any real data and only learn about the oddities in their specialized equipment. So, time and money for your studies are at your discretion and not a committee.

I have seen many ways people get caught through the years. So you can chase it, or scrap it and rebuild. I used to use tungsten lamps with cut off at 330nm for measurements. But Halogen bulbs seem better, even with my home grating I can see UV-A in commercial Hallogen bulb but not a commercial tungsten. I dont have the money to go more specialized and get real equipment. So you may want to add your filters, as John said which would be easy modification. Or lower the lamp switch point, which may be more involved modification, but cheaper with less filters needed. What may come down to is how much impact it has on your full spec Canon. You knows its there, its limited. Hopefully there aren't others that impact what you want to do. This is how you fall into a hole.

Link to comment
Food for thought, and thanks all. I tend to get consumed by problems - I spend too much time focussing on them, chasing down every little issue, rather than on realising what it is I've got. So for me I will continue to look at trying to understand the extent of the issue rather than redesigning and rebuilding everything at the moment. I can do the understanding work with the equipment I have and a few new filters (which can be used for other things in addition to this problem solving). If I get into redesigning and rebuilding chances are I'll end up with a whole new raft of issues to address.
Link to comment

Experiment 2 - Colour Multispectral 5DSR from 280nm to 480nm, with and without GG495

 

I've taken the approach above and collected images with my colour multispectral Eos 5DSR camera, with and without the GG495 filter, between 480nm down to 280nm. The aim being to subtract the data from the scan with the GG495 filter from the scan with no filter to try and minimise the effect of the stray light. I'm not going to rehash the method as I've covered that above. The only difference here is that these were run at ISO6400, unlike the monochrome camera scans above, as with te Bayer filter on the camera is much less sensitive to UV light. Hence the y axis score are significantly different to the monochrome data above.

 

Firstly the relative response of the difference colour channels with no GG495 filter;

post-148-0-58194200-1528562953.jpg

 

And magnified to see the details better in the UV;

post-148-0-85096800-1528562988.jpg

 

The double humped red channel is a bit weird, but I'll come to that in a minute. Now, add in the GG495 filter and re-run;

post-148-0-81833200-1528563052.jpg

 

As before there is a huge peak around 420nm, which I'm putting down stray light from the setup. Interesting to see it does not impact each channel exactly the same, but it does give signal from red, green and blue. Could be IR - this gives a signal in each channel, especially as the wavelength increases.

 

I can now subtract this scan with the GG495 filter from the one with no filter, to try and remove the effect of stray light. This is what that looks like;

post-148-0-00974200-1528563178.jpg

 

And again, with a rescaled y axis to make UV part easier to see;

post-148-0-50314400-1528563217.jpg

 

Now the strange double hump in the red channel has disappeared, and the overall sensitivity across the range has dropped slightly. In fact at 340nm and below the sensitivity is now very very low, with slightly more from the blue than red and green.

 

This final plot is now my best attempt so far at measuring the sensitivity of the multispectral 5DSR camera (with Bayer filter present) in the UV, as it has effect of the stray light removed as best I can at the moment. It does change my initial measurements and conclusions slightly, especially in the relative contributions of red, green and blue at the different wavelengths.

Link to comment

Experiment 3 - Using the stray light corrected multispectral data to look at Baader U transmission

 

Taking the data from Experiment 2 above, I can multiply the final sensor sensitivity data by the transmission curve for the Baader U filter, to get an idea of the channel response as a function of wavelength. If I multiply this up, the curves look like this;

post-148-0-28319600-1528565092.jpg

 

This now looks different to my previously measured curves, such as here;

http://www.ultraviol...dpost__p__21270

 

In particular the new data corrected by the stray light goes to zero when it should (at 320nm and below, and 420nm and above). There is a small signal at 400nm, which I am putting down to the FWHM of my measurement setup (I reckon at 400nm its seeing light from 390nm to 410nm). But as the sensor is much more sensitive at longer wavelength this could still be important to any final image.

 

As a bit of a stretch I can also take the R : G : B ratios at 400nm down to 340nm, and try and predict the 'colour' that would bee seen at each of these wavelengths. For this I worked out the RGB ratio and then normalised it so the highest value in each set of three was 255. I then put this into Excel 'custom colour' selection for the cells, and it gave the following;

post-148-0-03346300-1528565419.jpg

 

This would be for colour from Raw composite image in Rawdigger - not in RGB rendered mode.

 

I now want to use this new Baader U curve to repeat what I did here, looking at reflection from the different flowers in my garden;

http://www.ultraviol...dpost__p__21414

 

In the original work on this the correlation between measured and predicted colours from the flowers was not as good as I'd have liked. As the original data was based on sensitivity curves with stray light from how they were measured, in hindight, that isn't surprising. So, next experiment, repeat this prediction and see what happens. That's a job for tomorrow though, I can feel the beer calling :)

Link to comment

There's one thing I don't understand. (Probably many things!). If the BaaderU peaks around 350 nm and tapers off around that peak, then why is the channel response so strong at 380 nm under the BaaderU? I get that you are plotting channel response and not a transmission chart. But I'm not getting why the BaaderU doesn't supress some channel response at 380 nm.

(I hope I phrased this question properly.)

Link to comment

There's one thing I don't understand. (Probably many things!). If the BaaderU peaks around 350 nm and tapers off around that peak, then why is the channel response so strong at 380 nm under the BaaderU? I get that you are plotting channel response and not a transmission chart. But I'm not getting why the BaaderU doesn't supress some channel response at 380 nm.

(I hope I phrased this question properly.)

 

I get what you're saying Andrea. It's driven by two things, the underlaying sensor sensitivity and the transmission of the Bayer filter/microlense array. It's sensitivity drops steadily but quickly below 400nm. And the Bayer filters absorbs more are shorter wavelength. Multiplied together these outweigh the transmission of the Baader U, and skew the overall response to higher wavelength. Hope that makes sense, if not I'll try again tomorrow.

Link to comment

Jonathan, your new data is interesting. Your predicted colors are not what I expected or have seen. However, my color is biased by white balance. What I see is similar to what Cadmium has shown with the sparticle. 340nm region is green, 370nm region is yellow and 390nm region is blue.

I see this with E510 full spectrum camera in sunlight and biased with lights.

I recently got a 365nm led bulb from amazon and this one is around that. The light out is yellow on sensor and looks similar to the convoy S2+ I have.

Using baader venus filter on camera, white balance in sunlight and under the 405nm led light bulb, the image has hard light with blue tint, most likely only looking at 390nm. Using the 365nm bulb the light is softer with slight yellow tint. This bulb seems to not be so tight in spectrum. Using Convoy the light again is hard yellow tint. I don't have a solid 340nm source, but in sunlight at this range it seems green, using pinhole pro lens.

Its possible the Canon sensor is different than the typical Sony sensors in other cameras. Or my observations are biased due to selected white balance. I haven't looked at raw channel data using Raw digger yet.

I may pick up these specific filters to do 3 channel color, being 340, 370, 390. May be interesting.

Link to comment
Dabateman, keep in mind I'm looking at my images in Raw composite mode in Raw digger. This is before the images have had their usual RGB render and before any white balancing, as I wanted to see them with as little processing as I could. So it doesn't surprise me the colours look different to what's been seen before. I've not been able to get hold of raw files from a sparticle image, so have ended up buying some small filters with the aim of building one to test.
Link to comment

Jonathan, thank you for the raw colours, for which I've long been an advocate because that removes one step from the various approximations that go into making wavelength to color correlations. The 380nm pink/magenta and 360nm orange do correspond in a general way to the raw colours I've seen with a BaaderU on my Nikons and shown in many posts here on UVP.

 

However, I can't say I've ever seen that blue-violet at 400nm in my BaaderU raw photographs. That makes sense, I suppose, because transmission there by the BU is almost nil. And it is somewhat of a surprise to see that desaturated violet-blue at 340nm. With the BaaderU I don't recall seeing that shade either - although you would naturally think that the BaaderU would transmit enough at 340nm for that colour to show up. Perhaps the Nikons record slightly differently at 340? Just a speculation that we should probably let rest until someone provides a Bayer sensor transmission chart for some current converted Nikon.

 

For your readers who don't have something like Raw Digger, you could always add a white balanced false colour column next to the raw column in your modeled colours chart.


 

.....so have ended up buying some small filters with the aim of building one to test.

I'm eagerly awaiting this. I've wanted to get some narrowband UV filters for the longest time. My first foray into that resulted in disaster due to the hard coating on the purchased 340x10 which rendered it unusable for reflected UV photography. I'll try again eventually.


 

......Multiplied together these outweigh the transmission of the Baader U, and skew the overall response to higher wavelength. Hope that makes sense, if not I'll try again tomorrow.

 

Jonathan, thank you. It has all gotten much clearer. I think it is clear now. So let me pester you just a bit more to make sure the following facts are correct. :D

  • You are investigating how unfiltered sensor sensitivity at point X nm (last chart in Post #9) compares to filtered sensor sensitivity at point X nm (chart in Post #10).
  • The 5D sensor is very UV-sensitive at 380 nm (and its UV-sensitivity is rising from, say, 340 nm to 400 nm).
  • The BaaderU transmits much less at 380 nm than it transmits at its peak around 350nm.
  • So the height of the BU filtered sensor sensitivity curve at 380 nm would naturally be shorter than it would for the unfiltered sensor sensitivity curve at 380nm. The two charts bear this out.
  • In general, the strongly peaked unfiltered sensor sensitivity at 380 nm forces a strongly peaked filtered sensor sensitivity at 380 nm.

It would be very instructive to see an Experimental Conclusion which has four charts:

  • BaaderU transmission
  • 5D with BaaderU transmission
  • 5D sensor sensitivity
  • 5D with BaaderU sensor sensitivity


And then there is the effect of a lens on sensor sensitivity and filter transmission. Yikes!


 

I've made several comments over the years about thinking that we are all recording mostly from between 360-370 nm and 400 nm. I've made posts here and there about how it is tough to get much between 300 - 330 nm -- not that it cannot be done, it's just harder. So far this is obvious in your charts.

Link to comment

Andrea, see comments below;

 

However, I can't say I've ever seen that blue-violet at 400nm in my BaaderU raw photographs. That makes sense, I suppose, because transmission there by the BU is almost nil. And it is somewhat of a surprise to see that desaturated violet-blue at 340nm. With the BaaderU I don't recall seeing that shade either - although you would naturally think that the BaaderU would transmit enough at 340nm for that colour to show up. Perhaps the Nikons record slightly differently at 340? Just a speculation that we should probably let rest until someone provides a Bayer sensor transmission chart for some current converted Nikon.

 

Yes, the Baader U transmission is almost nil up at 400nm, but if the subject emitted/reflected a large amount of light in that region then this outweighs the small transmission. For instance, take a picture of a Daisy with using Baader U, and load it as a Raw composite in Rawdigger. The white part of the daisy strongly reflects at 400nm and higher (based on the info in the FReD database). When I tried it it looked purple when loaded as a RAW composite, although very dull of course given the extremely low transmission. It comes back to how much light is being reflected as a function of wavelength. I would expect Nikons and Canons to be different, although not sure by how much. Again remember that at 340nm there is less UV in sunlight, and the sensor is less sensitive. So if 360nm and 380nm are well exposed the 340nm would look much darker - my colours are all normalised to the the highest Red, Green or Blue value at each wavelength being 255.

 

.....so have ended up buying some small filters with the aim of building one to test.

I'm eagerly awaiting this. I've wanted to get some narrowband UV filters for the longest time. My first foray into that resulted in disaster due to the hard coating on the purchased 340x10 which rendered it unusable for reflected UV photography. I'll try again eventually.

 

Filters have been ordered, and are currently with a family friend in the US, to be brought to Europe later this month. To be honest I expect it to give the same 'colours' as my setup for given wavelengths. After all, all I'm doing is taking pictures of small circles of light from the output of an integrating sphere at different wavelengths. The only difference is I'll be able to do a few at once, depending on the final size of the array.

 

......Multiplied together these outweigh the transmission of the Baader U, and skew the overall response to higher wavelength. Hope that makes sense, if not I'll try again tomorrow.

 

Jonathan, thank you. It has all gotten much clearer. I think it is clear now. So let me pester you just a bit more to make sure the following facts are correct. :D

  • You are investigating how unfiltered sensor sensitivity at point X nm (last chart in Post #9) compares to filtered sensor sensitivity at point X nm (chart in Post #10).

I took the data shown in point #9 and then multiplied it by the Baader U transmission spectra data to get the graph in point #10.

  • The 5D sensor is very UV-sensitive at 380 nm (and its UV-sensitivity is rising from, say, 340 nm to 400 nm).

It is more UV sensitive at 380nm than it is below that, and the sensor sensitivity drops steadily as wavelength gets shorter down to 300nm when it is virtually zero, atleast in my system with WG280 glass over the sensor. To be honest given how the sensitivity curve is dropping I'd be surprised if there is any sensitivity below 300nm even with something with a lower cutoff than the WG280 window I have.

  • The BaaderU transmits much less at 380 nm than it transmits at its peak around 350nm.
  • So the height of the BU filtered sensor sensitivity curve at 380 nm would naturally be shorter than it would for the unfiltered sensor sensitivity curve at 380nm. The two charts bear this out.
  • In general, the strongly peaked unfiltered sensor sensitivity at 380 nm forces a strongly peaked filtered sensor sensitivity at 380 nm.

All the factors multiply together, so yes, although the main effect is the drop in transmission through the Bayer filter. The 5th graph in point #9 above shows how quickly the response of the sensor with the Bayer filter drops.

 

It would be very instructive to see an Experimental Conclusion which has four charts:

  • BaaderU transmission
  • 5D with BaaderU transmission
  • 5D sensor sensitivity
  • 5D with BaaderU sensor sensitivity

OK, when I get time to pull them together, I guess you mean to compare the Monochrome (bare sensor) vs Multispectral conversion (with Bayer filter retained).

 

And then there is the effect of a lens on sensor sensitivity and filter transmission. Yikes!

 

Yes, and the light source too. Non dedicated UV lenses and sunlight will skew that Baader U curve to higher wavelengths.

 

I've made several comments over the years about thinking that we are all recording mostly from between 360-370 nm and 400 nm. I've made posts here and there about how it is tough to get much between 300 - 330 nm -- not that it cannot be done, it's just harder. So far this is obvious in your charts.

 

Yes, especially under sunlight, most of what we see will be >360nm. As you say, it becomes harder and harder. You'd need to block longer wavelengths more effectively the lower you go.

Link to comment

[[Off Topic: Here is a helpful tech writing hint I learned in my Bell Labs tech writing tutorial. The phrase "to be honest" should not be used in tech writing. It tells yours readers that up until then, you have been supressing the truth. :lol: :lol: :lol: ]]

 

 

And thank you for your response to my question. I've got hold of it now. And wishing for you to someday have a double grated monochromator.

Link to comment

I need to correct an earlier statement, where I'd got the wrong end of the stick (or perhaps even the wrong stick). I've highlighted the piece below;

 

Andrea, see comments below;

 

However, I can't say I've ever seen that blue-violet at 400nm in my BaaderU raw photographs. That makes sense, I suppose, because transmission there by the BU is almost nil. And it is somewhat of a surprise to see that desaturated violet-blue at 340nm. With the BaaderU I don't recall seeing that shade either - although you would naturally think that the BaaderU would transmit enough at 340nm for that colour to show up. Perhaps the Nikons record slightly differently at 340? Just a speculation that we should probably let rest until someone provides a Bayer sensor transmission chart for some current converted Nikon.

 

Yes, the Baader U transmission is almost nil up at 400nm, but if the subject emitted/reflected a large amount of light in that region then this outweighs the small transmission. For instance, take a picture of a Daisy with using Baader U, and load it as a Raw composite in Rawdigger. The white part of the daisy strongly reflects at 400nm and higher (based on the info in the FReD database). When I tried it it looked purple when loaded as a RAW composite, although very dull of course given the extremely low transmission. It comes back to how much light is being reflected as a function of wavelength. I would expect Nikons and Canons to be different, although not sure by how much. Again remember that at 340nm there is less UV in sunlight, and the sensor is less sensitive. So if 360nm and 380nm are well exposed the 340nm would look much darker - my colours are all normalised to the the highest Red, Green or Blue value at each wavelength being 255.

 

I'd got myself confused. The Daisy shot only looks purple after white balancing. As a RAW composite it still looks pinky purple. Like most of the rest of the image. The reason we don't see that colour from 400nm in the Baader U image comes done to the extremely low transmission of the filter there. When the 360nm and 380nm would be properly exposed, then assuming a similar reflectivity for part of the image reflecting 400nm, that part would then be extremely dark. The Daisy is still presumably reflecting some light at 380nm and 360nm, even if it is much lower that at 400nm. Given the transmission curve of the baader U though, the effect of the 400nm part would be extremely small given the filter blocks nearly all the light there.

 

It is interesting when looking at a Daisy though - it is extremely UV dark. It does have a pretty steep cutoff in reflectivity as a function of wavelength according to the FReD database.

Link to comment
Thank you for this correction. That we all make mistakes is never an issue. That we might not correct a mistake is a big problem. :D
Link to comment
It would be very instructive to see an Experimental Conclusion which has four charts:
  • BaaderU transmission
  • 5D with BaaderU transmission
  • 5D sensor sensitivity
  • 5D with BaaderU sensor sensitivity

OK, when I get time to pull them together, I guess you mean to compare the Monochrome (bare sensor) vs Multispectral conversion (with Bayer filter retained).

 

Andrea, In response to your request. I haven't redone everything yet to remove the effect of the stray light, however looking at some old data is still relevant - the transmission through the Baader U at 420nm is essential 0, and this is where the effect of the stray light is most prominent. So the final curves of Camera PLUS Baader U are still worth looking at to see the comparison between the two.

 

I'm not going to show the Baader U spectra again - we have plenty of those on here. Firstly then the average response for the Monochrome camera and the Multispectral camera (with Bayer filter retained). These are both 5DSR cameras, tested the same way. The average response is the mean of the R, G, B and G2 channel values at each wavelength from Rawdigger, with them being opened as Raw composites;

post-148-0-23495100-1529001971.jpg

 

Now, multiplied by the Baader U;

post-148-0-58908200-1529001999.jpg

 

The distribution of relative contributions of the wavelengths shifts to be slightly lower with the monochrome conversion, as the sensitivity of the monochrome conversion drops slower with wavelength than the multispectral conversion (with Bayer filter retained). Interestingly the area under the curve for the monochrome camera is about 6 times that of the Multispectral camera with Bayer filter. MaxMax told me to expect about 6x the sensitivity to UV by removing the Bayer filter array.

 

Remember too, this is without the effect of the sunlight spectra, and is with a UV lens (Rayfact 105mm). Obviously the sunlight skews this to higher wavelengths as well, as there is more longer wavelength than shorter, and any les other than a UV will also skew these curves to the higher end given their transmission properties.

Link to comment

Jonathan,

I have a stupid question for you. I know the channels in uv shift depending on white balance for a full spectrum camera. Do you need to white balance the monochrome camera? If you do try, does it change the channel sensitivity in any way?

 

Link to comment

 

Remember too, this is without the effect of the sunlight spectra, and is with a UV lens (Rayfact 105mm). Obviously the sunlight skews this to higher wavelengths as well, as there is more longer wavelength than shorter, and any les other than a UV will also skew these curves to the higher end given their transmission properties.

 

It would be quite interesting to see a similar graph, but also multiplied with solar spectrum levels around your data points.

That weighting will bring us closer to our real life UV-photos.

Link to comment

Jonathan,

I have a stupid question for you. I know the channels in uv shift depending on white balance for a full spectrum camera. Do you need to white balance the monochrome camera? If you do try, does it change the channel sensitivity in any way?

Not a stupid question at all. First of all, I wondered the same thing when I got the camera - if all the channels now show approximately the same sensitivity, what would white balancing do? Well the channels are not exactly the same, there are some minor differences between them on my Canon 5DSR when I open the file as a Raw composite in RawDigger (not sure about the channels in the Nikon Monochrome conversions as I haven't got one). So there are some small differences to be balanced out anyway.

 

Also I understand it, after the image is captured the internal camera software boosts the red channel signal, given the internal IR filter normally cuts down on red a lot. With the monochrome camera, that would result in red now being very different to the other 2 channels. By all means any people that know more about whitebalancing please chip in, I am by no means an expert.

 

So yes, white balance is still recommended, but the end results are nowhere near as dramatic as with a camera with the Bayer filter attached.

Link to comment

It would be quite interesting to see a similar graph, but also multiplied with solar spectrum levels around your data points.

That weighting will bring us closer to our real life UV-photos.

Here you go - this is the Baader U and camera graph, now multiplied by spectral distribution between 480nm and 280nm;

post-148-0-37110200-1529049152.jpg

 

As mentioned above it skews the distribution to longer wavelengths.

 

But please, always keep in mind this is based on multiplying up data which has been collected where I still have a stray light issue from the monochromator. So there will likely be some minor changes in the graphs when I redo all my experiments, most likely a reduction in the actual contribution at 340nm.

Link to comment
  • 10 months later...

New experiment - second OLPF removed from Multispectral EOS 5DSR

 

Bit of a follow up, now my Multispectral 5DSR has had the 2nd OLPF removed from the sensor. The comparison graphs for the ones shown here, are graphs 3 and 4 in the post linked below;

 

https://www.ultravioletphotography.com/content/index.php/topic/2813-camera-sensitivity-chasing-ghosts-in-spectral-sensitivity-measurements/page__view__findpost__p__21902

 

I did my sensitivity standard test with the camera (Rayfact 105mm lens). I used a set of images with the GG495 filter as a correction for any stray light, and these are the spectral sensitivity curves I got.

post-148-0-16166700-1557242475.jpg

 

And with a higher magnification of the y-axis, to better see the short wavelength data.

post-148-0-89797800-1557242472.jpg

 

Removal of the 2nd OLPF has improved the spectral sensitivity at 360nm and 340nm. I can't see much different at 320nm and below, as we are down in the noise with my test there. But my imaging work with the Sparticle has shown an improvement at 321nm, and with my HgXe lamp and 308nm filter, I have seen improvements down in the 310nm region too.

 

Might help with getting more of a range of tones in my UV images, by having a bit more sensitivity at and below 360nm.

Link to comment

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×
×
  • Create New...