Enhanced UVB sensitivity in monochrome converted EOS 5DSR

[JMC]

I've been playing around with UVA and B imaging as part of my recent sunscreen work. I have a Hamamatsu LC8 light source with a 200W HgXe lamp, and an interesting internal filter setup. The end result is it produces a lot of UVA B and even some C, with virtually no visible and IR. This is a spectra of the output taken using my Ocean Optics FX spectrometer.

 

For UVA and UVB imaging I used a Baader U an Invisible Vision 308nm filter respectively. For my initial tests I used my Canon 5DSR monochrome conversion from MaxMax, and a Rayfact 105mm UV lens. First of all I wanted to know the relative sensitivity of the camera when images were taken with the 2 filters, so I kept ISO and f stop the same (f11 and ISO800) and collected a set of images of a Labsphere 40% diffuse reflectance standard. I then plotted the average channel response for vs exposure time for the 2 filters and got the data below;

 

This was telling me there was about 2 stop difference in response when using the Invisible Vision 308nm filter compared to the Baader U - I needed 2 stops more exposure for the UVB images vs the UVA ones with my monochrome camera and Rayfact lens and this light source.

 

The sunscreen images from the intiial work are a bit boring, so I've done some images with a Snowdrop from the garden. Firstly with UVA (Baader U). Settings for this one was ISO800, f16, 1/15s, Raw file loaded into Darktable and simply exported);

 

And secondly with the Invisible Vision filter (ISO800, f16, and 1/4s, treated the same as the Baader U image).

 

These look similarly exposed to me and there was 2 stops between them as expected from the diffuse reflectance standard test (1/15s for the Baader U and 1/4s for the Invisible vision). While the are some slight differences in the Snowdrop, the biggest change is in the shadow from the glass vase. In UVA it is letting some light through to the backdrop. In UVB it is opaque - no light getting through.

 

I got to thinking, given there is a 2 stop difference between the UVA and UVB images using the monochrome camera conversion, how do they compare for the standard multispectral conversion I have with the Bayer filter array retained? So I repeated the experiment, keeping ISO and f stop the same but now with my multispectral EOS 5DSR instead of the monochrome one. This time though I increased the exposure slightly. Firstly with the Baader U UVA image (no white balancing, RAW file loaded into Darktable and exported, 1/3s);

 

The image is slightly over exposed, but the aim here was to capture an image in UVA and compare it to the UVB one with 2 stops more exposure. Increasing exposure 2 stops, and replaced the Baader U with the Invisible Vision 308nm filter, I got this at 1.3s exposure time;

 

Blackness now in the UVB image with the same 2 stops additional exposure used for the monochrome camera. I then kept increasing the exposure, and ISO, and re-imaging, and eventually got to ISO3200 and 30s (a huge increase) and ended up with this;

 

Even at this exposure the Snowdrop is barely visible in the darkness, and this was about an extra 6.5 stops of light, above and beyond the 2 stops I thought I would need.

 

I knew the Bayer filter was highly absorbing in the UVB region, but hadn't quite realised it was this extreme. It looks therefore like the monochrome conversion is an extremely handy one for UVB imaging. Not sure whether this is Canon specific, unfortunately I don't have loads of different monochrome converted cameras to test....

 

As always, sharing in case it is of interest.

[Andy Perrin]

That is pretty convincing evidence that it’s not worth even trying UVB imaging without a monochrome camera.

 

When you get the time, do some more objects for us! Wonder what dandelions do...

[JMC]

Andy, the plan is definitely to look at more things in UVB. Not much growing in the garden just yet though.

 

I may be able to use it as a source for UV induced fluorescence too - by filtering the light source I can get UVA and UVB independently and look for differences.

 

It'd be interesting to revisit the Sigma cameras with their foveon sensors with this. Without the Bayer filters they may behave more like the monochrome camera. We shall see.

[dabateman]

Wow,

Not at all what I see. You must have glass on your converted Canon. I have 2 full spectrum converted cameras. My first Olympus E510 and recently converted Em1.

My E510 tested poorly for range, able just to see UVa. Since I have the Em1, I thought I would take it apart to see if I could convert it to monochrome. In doing so I discovered how it was converted. The BG glass was just removed from the sensor, the dust shaker was left in place and the sensor was adjusted to get infinity focus. So I cut the dust shaker off. Now my E510 has just the sensor, CFA, microlenses and coverglass. So it will not focus to infinity in IR with four thirds lenses. But I now can see to 300nm. I was able to get a flower shot at 6 minutes at ISO 800 using a reptile light during its warm up time. After 15 minutes I can't see any 302nm light from the bulb. I do now get a strong 313nm image using this camera.

Similarly the Em1 with no dust shaker glass can easily see 313nm.

My Sigma Sd14 can only see to 335nm. So you may not get anything better.

 

So I would say either the Canon is really poor or that you discovered that there is glass on your camera blocking your uvb.

[dabateman]

Something else it could be, is the coverglass on the sensor. Olympus sensor coverglass do not have an AR coating. The first Olympus camera to have a sensor with AR coating was the Em1mk2 and the new Em1x with same sensor.

When your Canon was converted to monochrome, most likely not the same coverglass was added back.

[JMC]

David I have the same Schott WG280 on both cameras (multispectral and monochrome) as cover glasses over the sensors. This shouldn't be impacting the imaging at 310nm. Differences in the dyes for the Bayer filter for the Canon and Olympus could also account for this.

 

I do have a Canon 450d with half the Bayer filter removed, and no cover glass at all. I'll try and repeat this experiment with that. Of course they may have used different dyes between the 450d and 5DSR, but at least it's the same manufacturer.

 

EDIT 19/2/19 - it seems I still have lots to learn. I thought the cover glass was something that is easily removed and replaced during conversions. Apparently it is not. My current understanding is that the cover glass on my multispectral 5dsr doesn't get removed during the conversion, while the monochrome conversion does remove it. This was something I had not considered until now when comparing the cameras.

[JMC]

Well, that is interesting. My 450d half and half (with half the Bayer filter removed, and nothing at all over the sensor) behaves differently to my EOS 5DSRs, and I can see something at 308nm even on the Bayer filter side with the modified 450d. David, what I'm seeing is more in keeping with what you mentioned about your Olympus cameras.

 

Given I have been assured that the same Schott WG280 was put back over bot my monochrome and multispectral conversions, that leads me to think that is something different in the Bayer filter layer between the 450d and 5DSR (adhesives, coatings, microlens material, Bayer filters themselves etc). Not sure how I would get to the bottom of that though. It does however make it hard to make generalisations about the suitability of cameras especially for UVB. All I can say is that it looks like while my 5DSR monochrome version is pretty good for UVB, my normal multispectral one isn't.

 

EDIT - to be a bit systematic with the 450d, I used a Labsphere 40% diffuse reflectance standard. Images opened in RawDigger as RAW Composite files and then screen-shotted for sharing. Firstly with Baader U filter (monochrome side on left obviously). This was at ISO800, Rayfact UV lens, f16 and 1/15s).

 

And the Invisible Vision 308nm filter (at ISO800, f16, Rayfact UV lens and 1/3s).

 

Looking at parts of the image where the illumination is the same, I get the following.

 

Baader U image, Bayer filter side about 1.5 stops darker than monochrome side.

Invisible vision 308nm side, Bayer filter side about 2.5 stops darker than monochrome side.

 

This is very different to my 5DSRs, where the multispectral conversion had basic zero UVB sensitivity with the Invisible Vision 308nm filter. I shall get to the bottom of why at some point, but not at the moment.

 

So, don't go giving up on the Bayer filter cameras for UVB just yet. As David says they don't all behave the same way.

[Andy Perrin]

Huh, well until we have a sticky for UVB-capable Bayer cameras, I think I would not feel comfortable using one for that purpose.

 

Jonathan, I do wonder a tiny bit about that replacement WG280 cover glass. Is it real Schott? Could someone have used a substitute glass?

[cancington42]

This is amazing! Thanks for taking the time to do this! I’m curious to know how it is you converted your camera to monochrome? Was this something you did yourself?

[JMC]

Andy, it's early days to explain my results. The 'half and half' camera shows it is definitely possible for a Bayer filtered camera to see in the UVB, and I think I saw David has started a thread showing his work in the UVB with his Olympus cameras. As for the nature of the cover, I have no reason to doubt it not being the same on both 5dsr cameras. At this stage the result is what it is, and I shall continue looking into the possible reasons why it is like that. Early days :)

 

Cancington, no the monochrome conversion isn't done by me. It was done by MaxMax.

[cancington42]

Is that a member in the forum? I’m very curious about this!!

[JMC]

Is that a member in the forum? I’m very curious about this!!

Sorry, no. Just search for 'MaxMax monochrome conversion' on Google.

[Andy Perrin]

One thing that would be nice is if you and David could image the same subjects so that we can start to get a feel for what a UVB image “should” look like. It would make it easier to catch things like contamination with UVA etc.

 

 

[dabateman]

I have thought about what would be a good standard. I was hoping the color checker passport would work. But I am not sure now. And good ideas for a standard that would work on both sides of the pond?

I am in USA, Jonathan is in UK.

[Andy Perrin]

Well, when the dandelions bloom, that might be a thought! Only another month or so, I think.

But the best thing would be stuff that shows a significant difference from UVA photos, which only you guys can determine at present.

[JMC]

Well, a bit of an update here. I've had some very enlightening talks with Dan Llewellyn at MaxMax about my cameras. Here's where we came out - I'm sharing as it has some pretty important information about anyone trying to image UVB.

 

The normal Canon sensors have a cover glass on them to protect the sensor. This isn't normally removed during a UV-VIS-IR conversion, as it is an extremely delicate operation, and can lead to more broken sensors..... This material is lithium niobate, is referred to as the OLPF, and it has some pretty serious UV absorption especially at the shorter wavelengths. With the three cameras I tested;

 

EOS 5DSR monochrome conversion, lithium niobate cover glass gets removed.

EOS 5DSR UV-VIS-IR multispectral, lithium niobate cover glass still there.

EOS 450d, half and half, with the half the Bayer filter removed, lithium niobate cover glass removed.

 

This explains why my multispectral 5DSR had such poor UVB performance. I've seen the transmission curve for the Li niobate filter, and it is basically zero down around 320nm and below. Scary thing is it starts dropping in the UVA too, so this is not good for any UV imaging. The removal of this filter no doubt improves the UV sensitivity, and it is not just the Bayer filter that is absorbing the UV. This was a huge revelation to me, and I'm probably not the only one who this is news to. This is hugely important for UVB imaging, and probably explains the problems people have when trying to do it - I believe Nikons also have OLPF filters, which could be holding their performance back too.

 

David, apparently the Olympus has the OLPF filter as part of the ICF/AA stack which gets removed as part of your conversion. This would explain your great UVB sensitivity - no OLPF, no UV absorption by it, and good UVB sensitivity.

 

I want to thank Dan at MaxMax, for some pretty in depth conversations on this, and for the testing he went off and did for me. It's explained my results, which were really confusing me. I am hoping to get my multispectral 5DSR back to him in a month or 2 to have the OLPF removed, which should get me down the UVB with it. Very exciting.

[dabateman]

That is very interesting information. I actually think the OLPF if present in an Olympus camera, may actually be part of the dust shaker and not on the ICF AA stack. I assume this as when I removed the dust shaker from my E510, which is only 0.5mm thick, I magically could see below 350nm, which I couldn't before with that camera.

 

Thank you for digging into this Jonathan. Do you still have your Multispectral Nikon? You could do a quick UVB test with it to confirm your suspicion.

[dabateman]

Jonathan and others here are some images of the Stacks for the my cameras. I don't have the E510 stack as I bought it converted. But this is an image of the Dust shaker I cut out from the Olympus E510. The thickness of the glass is just 0.5mm, the outer shaking ring adds an other 0.5mm making it a total of 1mm that fits in the ring that has an o seal on the camera.

 

As you can see it has a blue color to it and whatever coating is on it is most likely why I could not see below 350nm until I removed it.

 

This is the Shaker glass that was cut out from my EM1, it is 0.5mm thick and has a different color.

 

This has a green color to it.

 

This is the "Top" not really sure what side but on direction looking down on the EM1 filter stack that was removed from the sensor.

 

This one has a strong red color on this side.

 

This is the " Bottom", looking down on the EM1 filter stack:

 

This has a faint blue color looking at it.

 

This is the side, you can see the whole thing is 3mm thick with the blue layer being just under 1mm thick.

[Andy Perrin]

The blue-colored one may be the IR blocker?

[JMC]

David. I bow to your thoughts on where it might be on the Olympus. I've never taken one to bits to check. Unfortunately the Nikon is UV only with an internal filter which is more UVA so I doubt UVB would get through that anyway. If that filter ever needs to come out I will be checking it to see if it can see UVB.

[dabateman]

Sorry maybe not clear. The E510 Sees UV and IR. After removing the thing in the first picture, It now sees UV-A better and UVB. Did not see any UVB before.

 

The Thing in the second picture is the Dust shaker removed from the EM1. It has a green rather than blue colour when held up to the light.

 

The last 3 pictures are of the same thing, just looked at from different angles. This was the glass stack that was on the EM1 to block UV and IR. I thought it was just BG glass, but now that I look at it from different angles, I looks more like a UV/IR interference type coated glass sandwiched under optical glass.

[JMC]

Bit of an update on this one. Just come back from a trip to the US, and while I was there I took my multispectral 5DSR back to Dan at MaxMax. During its stay it had the OLPF which is attached to front of the sensor removed. As a reminder, the 5DSR has 2 x OLPF, one of which is removed during a normal UV-Vis-IR conversion as it is part of the IR cut filter assembly. The other one though is glued to the front of the sensor and is not normally removed during one of these types of conversions. This second one has now been removed.

 

I have taken some images with it, and I can now see down into the UVB with it. Yay. Will put some pictures up when I have had time to process them. I also have this OLPF from Dan, and can measure a transmission spectra of it to see how much UVB it was absorbing.

 

This was quite an eye-opener for me - that some cameras have an OLPF filter on the front of the sensor which is not normally removed during a UV conversion. While I doubt this will have much impact on the predominately UVA images most of us do, most of time, it does have a huge impact on trying to look at shorter wavelengths.

 

David - my UVB pictures (un-white balanced) look green like yours. As I say, pictures to follow when I have a bit of time to go through and process them.

[JMC]

Some flower pictures to show the enhanced UVB sensitivity now the OLPF has been removed from the sensor.

 

Been a while since my initial images and the snowdrops have long gone, so Buttercup images. All taken with the modified 5DSR (now with the second OLPF removed) and Rayfact 105mm f4.5 UV lens (ISO 800, f16). Images captured as RAW and then imported into Darktable and exported without whitebalancing (the light source has some sharp peaks).

 

Firstly in UVA using a Baader U. Exposure time 1/3s.

 

And in UVB using the Inivisible Vision 308nm filter. Exposure time 5s.

 

Before removal of the second OLPF I could see nothing in the UVB so this has definitely improved things. At some point I will get out the diffuse reflectance standard again and do a proper recalibration of the difference.

 

I also need to redo my work on camera sensitivity in the UV with this camera, as I expect the results <350nm to now change quite a bit as a result of removing this filter.

[dabateman]

Interesting,

What did Dan place on the sensor for infinity correction, was it the standard 280 glass? Or is this a macro/ close conversion without corrective glass.

[JMC]

WG280 David, as was used on my monochrome one.