Jump to content
UltravioletPhotography

Build thread - at home measurement of camera UV spectral response


JMC

Recommended Posts

But I don't like it that those two cameras have an internal filter about which we know nothing? With that internal filter we really aren't learning anything about the sensors! We are learning instead about that internal filter. What the heck did ACS use for a filter there anyway? It does seem to pass quite a lot of blue.

 

I worry that viewers might be mislead because the charts are not labeled to indicate that the cameras are filtered. Perhaps you could add such a label in the legend? "Canon 7D with ACS Filter", or something similar? What do you think about this, Jonathan??

 

Sure I can re label the charts Andrea.

 

With regards to them letting some blue through according to my charts, without seeing the filter itself, I cannot say whether it is an artifact of my measurement or a real effect.

 

Edit - on the previous page there is a plot of the monochrome 5DSR with Baader U filter vs 7D with ACS filter. They overlap at 420nm which suggests it may be something about the data processing, as I pretty sure the Baader U doesn't let anything through at 420nm. This is new territory for me and I'm learning something new every time I run an experiment. I know that each of my '20nm buckets' does collect over that full 20nm so could the 420nm be picking up some data from just over 400nm? Perhaps, I just don't know at this stage. I hope to be getting a StraightEdge U in a couple of months - the straight drop on the high wavelength side of that will be interesting to test, and see what I can see with my device.

 

With hindsight, I don't think modelling the data using the normal distribution in the graph above is the best idea - the baseline is non-zero, so that will automatically raise the edges of the curve. This could also be exaggerating any signal at 420nm. As I say, I'm learning on fly with all this - as far as I'm aware, it's not been done before.

 

I realised today I really need to get an EOS 5DSR with the UV/IR filter removed but the Bayer filter retained to better understand what is going on. Better start saving up...

Link to comment

pretty sure the Baader U doesn't let anything through at 420nm.

 

The substrate is UG11. Looking at the Schott data, I don't see anything coming thru there in a 2mm filter. Not sure what thickness is used for the Baader. But even 1mm thick UG11 is OD5 by 420.

 

Thx for the nice labels. Sometimes readers don't read the surroundings carefully (where you of course have nicely mentioned that these are filtered cams) and just run with what they think a chart is telling them. So I always think charts in papers, posts, wherever should be somewhat self-contained, if possible.

Link to comment
  • 2 weeks later...

What to do with a rainy, cold day in the UK? For me, it was take apart my device for measuring spectral output from my cameras and redo it for measuring between 400nm and 800nm. Someone asked me the other day whether I could look at the output between 500nm and 600nm, and it got me wondering if I could use the monochromator I have for this, which is mainly a UV one. To cut a long story short it worked, and I created a new set of calibration data for looking at data between 400nm and 800nm, by just using the halogen lamp of my light source. In the end the 400nm to 440nm data just using the halogen was a bit noisy as the lamp doesn't produce much light down there, but the data from 440nm to 800nm is usable.

 

First experiment. Repeat the test of my two main cameras - Standard unmodified EOS 5DSR, and Monochrome EOS 5DSR (with Bayer filter, microlenses, and UV/IR cut filter removed). I used the same settings for the image capture and analysis as previously (ISO 1600, 30s exposure, Rayfact 105mm lens at f4.5, captured in RAW and analysed in RawDigger for response on each channel). This is how the two cameras compare for relative sensitivity to light between 440nm and 800nm.

post-148-0-90434400-1523204927.jpg

 

I was told by MaxMax that the sensitivity of their monochrome cameras would be slightly higher than a standard cameras in the visible part of the spectra, and that looks to be the case here, on average anyway between 400nm and 700nm. The Standard camera has the UV/IR cut filter over the sensor, and this can be seen as it's sensitivity drops dramatically as you approach 700nm. It is possible to combine this set of data with the data from 280nm to 480nm so look at the whole 280nm to 800nm region;

post-148-0-77767900-1523205210.jpg

 

The two curves here are each made up of 2 sets of data - one from 280nm to 480nm and the other from 440nm to 800nm. In essence the line for the monochrome camera should be the same shape as the sensitivity of the sensor itself, without the Bayer filter, microlenses or factory UV/IR cut filter. Of course Canon wont share that, so it's impossible to check for certain, although the shape is typical for these types of sensors from what I have been able to see online. The line for Standard camera takes into account the Bayer filter, microlenses and UV/IR cut filter.

 

Second experiment. How do the 4 colour channels look for the colour camera, and what can they tell us? I plotted the 4 colour channel data (Red, Green 1, Blue an Green 2) from RawDigger, and compared this with the overall signal from the Standard colour camera and the Monochrome camera (Monochrome and Standard camera data is now shown as the SUM of the 4 colour channels rather than the AVERAGE as in the previous graphs, hence the numbers are higher);

post-148-0-47646300-1523205683.jpg

 

The colour channels split out quite nicely, and look to behave as expected. The 2 green channels are almost identical which is good. However it's possible to take this further. By directly comparing the channel data from the Standard Camera and the Monochrome Camera it is possible to see what the effect of the Bayer Filter/Microlenses and UV/IR cut filter has on the light hitting the sensor. Dividing the channels data of the Standard Colour camera by the Monochrome one and multiplying up to give % transmission of the Bayer Filter + Microlenses + the UV/IR cut filter between 440nm and 800nm I get the following;

post-148-0-85390600-1523205972.jpg

 

At first glance this looks strange, how can I get >100% transmission for the Standard camera data? But keep in mind the standard camera has the microlenses over the sensor which the monochrome one doesn't have. So values of >100% are possible as ore light will be focussed on the sensor by these. This graph then is a transmission graph for the "Bayer Filter + Microlenses + UV/IR cut filter" as a function of wavelength, as measured across the 4 channels.

 

All I need now is a Multispectral EOS 5DSR with the Bayer filter retained but the UV/IR cut filter removed, and then I can look at the effect without the pesky internal UV and IR cut. MaxMax said the sensitivity of the monochrome converted cameras in the IR drops a bit compared to just removing the UV/IR cut filter, so that's something I want to look at as well as the UV behaviour. I'll add it to the list to test when I can afford another camera.....

 

Anyway, sharing because it was an exciting days experiment, and it may be of interest.

Link to comment
Andy Perrin
Wow! This is phenomenal. I want to know what the last graph does in UV, though, because that would explain our false colors.
Link to comment

Wow! This is phenomenal. I want to know what the last graph does in UV, though, because that would explain our false colors.

Thanks Andy. As mentioned I will look at that as soon as I get a multispectral EOS 5DSR. I hope to do that later in the year, or once I can get a spare camera to have converted. The good thing is that from what I've seen so far, it'll be an easy experiment to do, once I have a suitable camera.

 

In the mean time here is a screen shot of what the colour camera is seeing when I take the photos for later analysis;

post-148-0-67535000-1523259542.jpg

 

These are cropped (not very well I hasten to add, as I cropped them in a hurry earlier) from the originals, and they are taken at 20nm intervals. Along the top they go from 400nm to 520nm (left to right). The middle row is 540nm to 660nm and the bottom row is 680nm to 800nm. These were done from JPEGS taken at the same time as the RAW files I used for the analysis, so obviously they have been processed, but I suppose to me it looks like a 'virtual Sparticle' board and is useful for visualising what is going on.

Link to comment

Interesting. Well done.

 

The general behaviour look like these graphs:

https://photo.stacke...s-only-by-looki

Came across this one too while looking around yesterday;

 

http://astrosurf.com/buil/50d/test.htm

 

The shape of the curves looks I'm getting looks pretty similar to those seen for the 5D mkII. The wheels have been set in motion to get a multispectral 5DSR with the Bayer filter left in, to complete the set. I hope to get the camera in the next few days, and then have it modded next month. Fingers crossed I'll be able to do the work assessing it towards the end of May. Then I'll have data on all 3 (unmodified, multispectral with Bayer filter present, and monochrome multispectral no Bayer filter), all between 280nm and 800nm.

Link to comment

OK Jonathan I have a question about those colors in Post #55.

 

Are those raw colors? Or white-balanced colors? What converter was used to produce these files? What color space was assigned to the JPGs? What illumination was used to make the photos? These and many other factors all affect what "color" is produced and viewed. (Yes we all know this.)

 

I do not intend to be "picky" here, but I do want to see some accurate description of the various factors which produce these colors. And I do understand some of these factors were discussed in previous posts, but when you post a chart then a short recap of these factors is quite useful if not necessary.

Thanks!!

 

If I'm being harsh, I apologize in advance. I don't think I am and I don't want to be, but I can't always determine how printed conversation comes across. :rolleyes:

Link to comment

Andrea, you're not being picky. As mentioned in the post the JPEGs were created at the same time as the RAW file. The JPEGs were not used for any of the analysis and are purely for illustrative purposes.

 

They were captured in the camera when I took the photo for the RAW files, and processed by the camera. The white balance was AWB (Auto), and the colour space sRGB. The illumination is the light coming from the integrating sphere - I focus on the exit port of the integrating sphere, and image what's coming through the monochromator and illuminating the integrating sphere.

 

I posted the JPEGs purely for illustrative purposes and they should be taken as such. I had limited time to get this posted hence I used the JPEGs (as they were available). I could go through each RAW file individually, and make TIFF files using RawDigger. I may well do that as I do more with the technique, however this time, as mentioned I wanted to share them as a way of visualising what I was imaging, hence the quick post.

 

EDIT, 14th April 2018 - I have just done a quick test on RawDigger to look at how the RAW files appearance compares with the JPEG files. There are some striking differences especially on the brightness (the JPEGs are much brighter). Therefore as mentioned above the JPEGs showing the coloured disk should only be seen as being illustrative of what the method can produce, not as absolute representations of the colours direct from the RAW files. When I get my spare EOS 5DSR converted to UV-VIS-IR I shall come back to this in more detail, including if I have time a more detailed comparison of RAW vs JPEG image appearance at the different wavelengths.

Link to comment

Thanks for the additional information! You are kind to satisfy my color curiosity on this most interesting output. For years now I've been interested in this (along with many others)!!

 

Anyway, I would love to see the Raw Composite for just one of the photos.

pleeezzzzeeeeee.........

 

When exporting the raw composite (as a TIF), may I suggest the following settings for the raw composite?

 

Camera color profile settings in Raw Digger Preferences for viewing your raw composite.

  • No profile (raw color)
  • 2x2 pixels (this is the setting for demosaicing the raw file)
  • Gamma 2.2 (curves the output for human visibility)
  • Autoscale (forget what this does)

When exporting the raw composite via a Control-E make sure these corresponding boxes are checked:

  • Raw composite
  • 3-channel (RGB) output
  • Autoscale to use full 16-bit range (there is it!!)
  • Apply gamma 2.2 curve

 

All those settings will give you the human-visible version of the raw file while viewing in RD and when exporting the TIF. (Jonathan, I'm quite sure you know this stuff already, but for the sake of other readers, I list it.)

 

Now the raw composite output for one of your color samples is going to look VERY dull and low contrasty as you have no doubt noticed. This would be when viewed in either RD or in the exported TIF. That's OK. Drag the output into some converter and measure the color.

  • What is its hue on the color wheel (the "degree" between 0-359°)?
  • What does that color look like at full saturation and brightness?

You might have seen me add a dab of the fully saturated, fully bright color to a raw comp in order to get a feel for what the raw colors are. I'll quick make a small example to show you.

 

Here is a raw composite UV example of a Kalmia flower showing that the D200-mod + UV-Nikkor + UV-Baader + Sunlight records mostly in the Red channel for this particular background and flower. Anything new there? No, we know this. But still, this is information about how that particular system works. I took 5x5 samples on the raw composite TIF in PS Elements. (I'd like to take a larger averaged sample, but was just being quick today.) All this was smashed into an sRGB color-space JPG so there will be some deterioration to measurements which already have some flaws. But it is all we can do.

kalmiaAngustifolia_uvNikkor_uvBaader_sun_20080715swhME_23770rawCompLabel.jpg

 

 

 

Jonathan, do you concur or not that the raw colour output might actually tell us something about how the camera is recording our *UV* photographs? I think that it does in some fashion. Although just how useful that information is or will be, well, who knows? :D Depends on one's own purpose I suppose.

 

And, finally, Jonathan, what I'm really, eagerly looking forward to seeing is what raw colours you are going to get from the UV spectral analysis at various nm outputs. There has been a long standing interesting in trying to correlate output colour to UV wavelengths. (No discussion needed here in Jonathan's topic about all the perils inherent in that quest. We have done it to absolute bits elsewhere.)

Now I will stop hijacking your thread!!!!!

Link to comment

Andrea, it's not thread hijacking at all, so don't worry about it. You're far more experienced with RAW images than I am, so I'll send you through a few to have a play with.

 

As for your question;

 

"Jonathan, do you concur or not that the raw colour output might actually tell us something about how the camera is recording our *UV* photographs? I think that it does in some fashion. Although just how useful that information is or will be, well, who knows? :D Depends on one's own purpose I suppose."

 

To me the RAW file does tell us about how the camera is recording the UV photographs. However whether anything useful can be derived from that colour remains to be seen. I've read a few papers which describe how a camera can be 'calibrated' to provide spectral information, so from what I have seen I believe it can be done. I think it does require knowing a few things though - the Bayer filter transmission curves in the UV, and the sensor sensitivity. Knowing the Bayer filter curves allows the way each channel responds to the different wavelengths to be determined. The combination of these three will determine the colour produced, as it does in the visible spectrum. Knowing the sensor response curve allows you to determine something about the intensity of the light hitting the sensor at each wavelength.

 

Anyone know any millionaires with an interest in science that want to fund some fundamental research :)

Link to comment

Anyone know any millionaires with an interest in science that want to fund some fundamental research :)

 

I wish!!! :lol: :lol: :lol:


 

Thanks for your response about raw colors. I think the raw color sensor response might give us some small, informal insights about using the converted camera as a shortcut to determining reflected UV wavelengths. But I think the converted camera can only give us a crude measurement of reflected UV wavelengths. By "crude", I mean in the sense of such measurements not being totally precise and not being totally accurate. Too many variables at play.

 

THE ONE THING I want to see resolved is about the many-to-1 problem. For example: how does the converted camera sensor respond to a spectral yellow input in contrast to how does the camera sensor respond to a mixed red+green input. If the converted camera responds the same way to both, well, then what? There is no way to determine what actual wavelengths we are recording.

Link to comment
For example: how does the converted camera sensor respond to a spectral yellow input in contrast to how does the camera sensor respond to a mixed red+green input. If the converted camera responds the same way to both, well, then what? There is no way to determine what actual wavelengths we are recording.

 

If you are shooting UV....there is no spectral yellow input. It's just UV input being recorded through the RGB filters. It is there the combination of R,G and/or B may produce a yellow as an output.

Link to comment

Yes, I get that. Clearly obvious. Never not known.

But I want to still see the results from aiming a spectral yellow at the sensor and aiming a red+green mix at the sensor. Is there anything, anything at all which will permit untangling the mixed signal from the spectral signal? I've always said no. But, hey, there is so much that I don't know.

 

BTW, has there been any Bayer filtration with CMY?? I know that was proposed and there are patents for it.

 

Added: I have to go look up some physics of light. If red+green = yellow for light, then how do those photons mix themselves up when propagating through the aethers? :D

Link to comment
BTW, has there been any Bayer filtration with CMY

I believe the Kodak DCS420 might have used that, or one of those models.

 

Was the DCS620x.

 

Yes, I get that. Clearly obvious. Never not known.

I thought so but was unclear what you were asking.

Link to comment

I believe the Kodak DCS420 might have used that, or one of those models.

Interesting!!!

Wonder how it would have recorded if converted?

 

 

I thought so but was unclear what you were asking.

Winnowing away the haze and knowing what question to ask is key.

And of course I am not sure if I have done that. :)

 

This sensor to wavelength calibration has always always bothered me.

 

Try this thought experiment.

(If it is correctly posited? Again I'm not sure.)

You have spectral wavelength A and an equivalent mixed signal B+C. Both A and B+C pass through the same set of Bayer filtration. But the B portion of the mixed signal induces a fluorescence in a Bayer dye that the spectral signal A does not. Would you get the same recording of A and of B+C?

 

Lots of assumptions there. Like, for example, I have no idea of Bayer fluorescence, assuming it exists, can reach the sensor. Maybe not because of selsel spacing. But I don't know. Maybe it reaches the sensor but is accounted for in system software. And would using UV wavelengths up-end any assumptions about that or anything else?

 

My promise to stop hijacking must start now! I'm done. Let's await further results from our hardy spectrographic experimenters.

Link to comment
But I want to still see the results from aiming a spectral yellow at the sensor and aiming a red+green mix at the sensor. Is there anything, anything at all which will permit untangling the mixed signal from the spectral signal? I've always said no. But, hey, there is so much that I don't know.

 

This reminds me of these days ( the link isn't any good anymore, sure it must be there somewhere).

http://nikongear.com/live/index.php?/topic/34851-uv-gh1-uvir-standardized-uv-colors/

Link to comment

I think I have another link like that. I will go look for it.

*****

 

We have not even discussed gamut. Not only might equivalent spectral and mixed signals get recorded as the same same color, but a lot of "nearby" signals, in certain cases, will get recorded as the same color. This is because the camera software has a color space into which it is stuffing the colors. For example, various violet hues were recorded as blue or blue-violet in some cameras. This has been improving over the years. But who knows how the false UV colors (in raw form) are being stuffed into the red/orange areas of the profile gamut.

Link to comment
  • 2 months later...

When I originally built this monochromator device it was to try and get some idea of response as function of wavelength for different camera, lens and filter combinations. What it doesn't allow me to do though is capture multiple wavelengths at once. There is something that does though - the Sparticle, that Steve and others have made. I wanted to build my own, even if it was just to see whether I got similar colours to what I got with my monochromator. So I sourced some filters from eBay and Omega. These were all seconds as I didn't have a lot to throw at this and it was just to check my results anyway. Eventually these arrived in the UK and I set about building a Sparticle.

 

First, hats off to you Steve and anyone else who has built one. I thought it would be a case of simply putting loads of filters together. Making it taxed me a bit. I original bought a block of plastic to use to mount the filters in. After about 30mins working on it I managed to tear that up while drilling the holes, so in the bin it went. I went back to wood, as it's easier to work with. So I built everything from wood from old pallets laying around the man cave. I also made a hole to mount some PTFE tape in addition to the filters - this amused me, I needed a tube 12-12.5mm external diameter, so I cut down a 45ACP shell case as it was the right diameter (I'm one of the few people in the UK that still shoots so have a whole box full of these). The insides were then painted with Culture Hustle Black 2.0, which I have used before for this type of thing. Anyway, some pictures from the build.

post-148-0-24674200-1531656897.jpg

 

post-148-0-28912400-1531656924.jpg

 

post-148-0-38736200-1531656925.jpg

 

post-148-0-53227500-1531656926.jpg

 

post-148-0-27516100-1531656929.jpg

 

And this is the setup for using it - point it at the sky away from the sun and put the camera in the back end. Not very complex.

post-148-0-80275700-1531656991.jpg

 

post-148-0-87400300-1531656995.jpg

 

I did a quick comparison with my Multispectral 5DSR and Monochrome 5DSR, both ISO400 and 1s with the Rayfact lens at f5.6. The RAW files were then put into RawDigger and looked at as both Raw Composites and RGB renders. These were captured as screen grabs. The wavelengths are centre wavelengths as measured on my OO spectrometer, not what was written on the filters (these are seconds after all). Firstly the RAW composites;

 

Monochrome camera - RAW composite

post-148-0-47714400-1531657116.jpg

 

Multispectral cameras - RAW composite

post-148-0-98008900-1531657118.jpg

 

And the RGB renders from the same RAW files (note these were done using the 'auto' setting in RawDigger, based on the camera profile, which gives them a strong red tinge):

 

Monochrome camera - RGB render

post-148-0-70836400-1531657117.jpg

 

Multispectral camera - RGB render

post-148-0-06531200-1531657120.jpg

 

The monochrome camera is picking up signal from much shorter wavelength UV than the multspectral camera, as I saw with my monochromator based setup. Not sure why in the monochrome camera some of filters have a red tinge even in the RAW composite image, but it looks like I'm getting more signal from the red pixels than the green or blue, or the red pixels have some boost to the signal which happens before the data is recorded in the RAW file.

 

Some learnings from this. I had to block up the PTFE tape hole for these images - it was letting too much light through and causing weird reflections inside the box even with the black paint. Some of the filter coatings have I think reacted slightly with the adhesive I used to help mount them as they have a speckled appearance now. 'Seconds' filters often have wavelengths significantly different from what they are advertised as, and as I found the filter charts sent with the filters do not always correspond with the reality of their actual performance. This setup does not normalise for light intensity at each wavelength unlike my monochromator system.

 

Would I replace my monochromator setup with this for my measurements - in a word, no. Too many variables, and even with its quirks, my monochromator setup is more controllable. But it was a fun exercise anyway.

Link to comment
Andy Perrin
Do you get the same colors as Steve when you white balance on PTFE? Also how close were the filter labels to the actual center wavelengths you measured? I have been wondering for a while about how much variance there is in the Omega seconds. If you have a table of labeled vs. actual, that would be really super!
Link to comment
Why bother with a 404 and 405 filter in the same array? Surely the space could have been used for some different wavelength.
Link to comment

Andy, Let me pull together something and share. Only 4 of the filters were Omega ones though, 5 were from another ebay seller, that wasn't one of regular filter suppliers. I have yet to do the PTFE comparison.

 

OlDoinyo, I had bought them as one at 400nm and one at 405nm, with aim of looking at that 390-405nm range for the Baader U. However as mentioned the actual values were not always close to the stock information values - these were seconds so they vary in closeness of specification.

Link to comment

Do you get the same colors as Steve when you white balance on PTFE? Also how close were the filter labels to the actual center wavelengths you measured? I have been wondering for a while about how much variance there is in the Omega seconds. If you have a table of labeled vs. actual, that would be really super!

 

While the suns still out I grabbed a test shot with the Baader U and then whitebalanced it in Darktable the whitebalance from a PTFE disk. Lens was Rayfact 105mm, at f5.6, and camera at ISO400. This is what I got;

post-148-0-49946100-1531673149.jpg

 

Overall, I think the colours look similar. Keep in mind I have a Canon and Steve has a Nikon, so would not be surprised at some differences in final colour after the UV gets through the Bayer filter.

 

As for the filters. Here's the information from the Omega surplus ones (information on the graph/paperwork, any label on the filter provided, and my measured central wavelength);

 

325bp10 on the paperwork, 320bp10 on the filter, measured center wavelength 321nm

365bp10 on the paperwork, no label on the filter, measured center wavelength 364nm

390bp10 on the paperwork, 394bp14 on the filter, measured centre wavelength 396nm

405bp10 on the paperwork, no label on filter, measured centre wavelength 405nm

Link to comment

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×
×
  • Create New...