What's the easiest and most affordable way for obtaining UV reflectance images?

I've been tempted to buy a UV pass filter, which cost approximately two hundred dollars for my camera, which I've sent to be made into full spectrum. I've wanted to combine UV with IR and/or visible images. I've thought of an idea where I could use a ZWB2 filter, which can pass infrared also. I could maybe mount the filter somehow to my camera, then at night I could use a UV light source (possible a full spectrum lamp might work?) which does emit a small amount of IR. Next, I could take two photos, one with the ZWB2 filter, next with a UV/IR Block filter. Thirdly, a photo taken using an Infrared pass filter could be taken. I could later open all three images in Photoshop, where I could combine the images, subtracting the Infrared and visible light from the composite. But on second thought, wouldn't there still be a problem because of the initial IR that can pass through the ZWB2? I'm not sure if this is the proper technique also.

Edit: *if I'd want UV-only image

UV imaging is neither easy or cheap. For digital photography you need three non standard things for reasonable results. As with so many things the easiest is usually the most expensive:

A camera converted so it can see UV - your full spectrum conversion SHOULD do this, but some converters think replacing the hot mirror with glass makes it full spectrum, while others know glass is fairly effective at absorbing UV & silica must be used instead.

A filter (or filters) that block visual & NIR but transmit UV, as you say the best of these are £200+. Cheapskates like me should be able to get results with a pair of carefully selected cheap er filters - this route reduces the UV range possible.

The third factor you need is a lens that transmits UV well - most don't! The best here are lenses specially made for UV imaging, but these are rare & very expensize. I went for a nikon enlarging lens that was recommended by Dr Klaus Schmitt it only transmits about 30% at best in the UV but manages to transmit a bit further than any of the other lenses I've looked at (using the UV spectrometer at work).

Despite having a full spectrum camera, a combination of filters that transmit a short band of UV very well, & this better than usual lens, I've only managed some poor images that faintly show UV features - proof of concept at best. Sometime I'll have another go with some more powerful UV lighting.

Using film rather than digital changes the challenges - film is typically very sensitive to UV and insensitive to IR, so the filter only has to block visual light - there are many cheap filters that can manage this (all leak NIR). Film throws in some other problems - the viewfinder shows nothing with the filter mounted, you can't see to correct focus (like IR lenses typically focus UV differently than visual light. then exposures are complete guesswork with no feedback till the films processed

UV induced fluorescence is very much easier, than either approach to reflected UV photography. Which is probably why so many on-line sources on UV photography seem to only cover fluorescence 

petrochemist wrote:

UV imaging is neither easy or cheap.

You said it all there... but....

I'll just mention that decoding multispectral information can be cheaper and easier than you'd think. If you look at the spectral profiles for typical RGB Bayer filters, they're really not narrow-bandpass filters, yet we get decent RGB extracted from captures with them. Basically, as long as the filters being used have different levels of response to the bands you care about, being able to extract the bands is primarily a matter of calibration and a bit of math.

For example, if you have a filter that does almost nothing beyond blocking UV, relatively simple differencing with an unfiltered view can give a decent UV image. For NIR, even the spectral response differences of Bayer RGB can allow extraction of a noisy NIR channel. However, I don't think the Bayer RGB colors differ much in NUV.  There are cheap gel filters that differ significantly in NUV response -- see the samplers from Rosco ; each (designed for theatrical lighting) filter comes with a response curve going from 360-740nm in 20nm steps.

BTW, you'll want to do the band extraction math on a linear representation of the pixel values, which generally means raws -- the log encoding used in JPEGs makes add/subtract behave more like multiply/divide, which is not what you want.

Interesting. I don’t know much about this.

ProfHankD wrote:

petrochemist wrote:

UV imaging is neither easy or cheap.

You said it all there... but....

I'll just mention that decoding multispectral information can be cheaper and easier than you'd think. If you look at the spectral profiles for typical RGB Bayer filters, they're really not narrow-bandpass filters, yet we get decent RGB extracted from captures with them. Basically, as long as the filters being used have different levels of response to the bands you care about, being able to extract the bands is primarily a matter of calibration and a bit of math.

For example, if you have a filter that does almost nothing beyond blocking UV, relatively simple differencing with an unfiltered view can give a decent UV image. For NIR, even the spectral response differences of Bayer RGB can allow extraction of a noisy NIR channel. However, I don't think the Bayer RGB colors differ much in NUV. There are cheap gel filters that differ significantly in NUV response -- see the samplers from Rosco ; each (designed for theatrical lighting) filter comes with a response curve going from 360-740nm in 20nm steps.

BTW, you'll want to do the band extraction math on a linear representation of the pixel values, which generally means raws -- the log encoding used in JPEGs makes add/subtract behave more like multiply/divide, which is not what you want.

I know it can be cheaper than I described, The OP want cheap & easy, so I thought suggesting a self modified compact & combining filters which might be the cheapest way wouldn't fit.

Your post does suggest I'd could get much better results from my poor mans UV set up if I learnt to process RAW better. Differencing is not something I'd ever considered I can see several potential applications for it in IR photography (where i already have a wide range of cut offs) I'm not a fan of extensive processing so even with that I'll probably never try it. I don't think any of my gel sample books have the wide range of long pass filters you mention - I'll have to keep my eyes open for other types...

Most of my experience with UV is in organic spectroscopy where I can clearly see how much UV is blocked by glass. The glass cells used for visual light have walls about 1mm thick, 2 of these walls effectively blocks all light below about 320nm, while UV quartz cells of the same dimension go right down to about 190nm about the point where air itself becomes opaque to UV..

I've played around with the works spectrometer for some of my photographic gear & have found a few of my lenses actually block all UV (blocking everything below 410nm) more effective than some UV filters. Without careful research the RAW files you try processing could have no significant UV data even in the 'unfiltered' shot.

petrochemist wrote:

Just to be clear, I was in no way indicating anything you said seemed wrong. I have a reasonable bit of experience with multispectral imaging, but very little with UV; everything you said sounded right for as much as I know.

ProfHankD wrote:

petrochemist wrote:

UV imaging is neither easy or cheap.

You said it all there... but....

I'll just mention that decoding multispectral information can be cheaper and easier than you'd think. ... extract the bands is primarily a matter of calibration and a bit of math.

I know it can be cheaper than I described, The OP want cheap & easy, so I thought suggesting a self modified compact & combining filters which might be the cheapest way wouldn't fit.

Your post does suggest I'd could get much better results from my poor mans UV set up if I learnt to process RAW better. Differencing is not something I'd ever considered I can see several potential applications for it in IR photography (where i already have a wide range of cut offs) I'm not a fan of extensive processing so even with that I'll probably never try it. I don't think any of my gel sample books have the wide range of long pass filters you mention - I'll have to keep my eyes open for other types...

It can help, but SNR is the big issue. If some channels are lots weaker than others, you get some channels with really poor DR. On the other hand, you can use filters to balance strength over a sampled spectrum, thus increasing SNR and DR. It just gets to be a pain.

Most of my experience with UV is in organic spectroscopy where I can clearly see how much UV is blocked by glass. The glass cells used for visual light have walls about 1mm thick, 2 of these walls effectively blocks all light below about 320nm, while UV quartz cells of the same dimension go right down to about 190nm about the point where air itself becomes opaque to UV..

Yup, quartz lenses are the "right answer." That's why old EPROM chips used to be so expensive: to get enough UV to erase 'em, they needed a little quartz window and a ceramic package to mount it in. UV is also scary in that bright UV lights are a serious health hazzard, so compensating for poor response by boosting light levels is a really bad idea.

I've played around with the works spectrometer for some of my photographic gear & have found a few of my lenses actually block all UV (blocking everything below 410nm) more effective than some UV filters. Without careful research the RAW files you try processing could have no significant UV data even in the 'unfiltered' shot.

Absolutely can't recover what's not there... and calibration can be a huge pain. Here's one example of what we did.

Of course, multiple shots are not suitable for some photography no matter what, so expensive filters that directly select the band you want can be the only option.

ProfHankD wrote:

Yup, quartz lenses are the "right answer." That's why old EPROM chips used to be so expensive: to get enough UV to erase 'em, they needed a little quartz window and a ceramic package to mount it in. UV is also scary in that bright UV lights are a serious health hazzard, so compensating for poor response by boosting light levels is a really bad idea.

The price of photographic lenses designed for UV is at least 10 times what I feel I can justify. I've been wondering what the results might be like if I just use a single quartz element (available for around £20). Spherical aberration will probably be an issue, but the increased light transmission might allow an aperture to be small enough to compensate.

Has anyone thought about a pin-hole setup? Also, I wonder what the UV cut-off for the glass on the sensor is?

Ed

Ed Constable wrote:

Has anyone thought about a pin-hole setup? Also, I wonder what the UV cut-off for the glass on the sensor is?

Ed

Yes I've considered pinholes. Mine range from around f/128 to f/180 which in itself reduces the available light considerably.

Which glass on the sensor are you referring to - The hot mirror & AA filter (parts of the sensor stack) have both been removed as part of the conversion of my camera. The Bayer filter & sensor itself will have more glass, that will reduce UV but the thickness of these is minimal (reducing their absorbance) & getting rid of them introduces complications.

It is quite possible that some models of camera will work considerably better than others in this respect.

AFIK the glass on the Bayer filter is borosilicate rather than spectrosil.

Ed Constable wrote:

AFIK the glass on the Bayer filter is borosilicate rather than spectrosil.

Quite probably not either, there are hundreds of types of glass, I think we can be pretty confident it won't be quartz (spectrosil or any other brand). The filtering dyes will absorb some UV, but removing the bayer layer is making for a very specialist camera so most will be happy to live with it's effect.

It means switching to very old technology but I guess I should give my SD14 a go. With the dust protector.hot mirror removed (single screw DIY job) it's a converted camera with no bayer filter. The foveon design also means most IR is seen in the red channel & UV should be mainly in the blue so getting rid of IR leaks from cheap filters is just a case of selecting the blue channel

I was surprised when I checked the UV transmission of a normal microscope cover glass that I had surlyned over an OLED material. It was a very effective UV filter (cutoff about 290 if I remember). You don't even want to imagine what thin quartz sheets cost. In the end I pirated a CaF2 window from an old spectrometer and cut it into sheets.

Ed Constable wrote:

I was surprised when I checked the UV transmission of a normal microscope cover glass that I had surlyned over an OLED material. It was a very effective UV filter (cutoff about 290 if I remember). You don't even want to imagine what thin quartz sheets cost. In the end I pirated a CaF2 window from an old spectrometer and cut it into sheets.

290nm isn't that bad, few if any of my lenses transmit anything near so low. I only recorded transmission down to 300nm and only two that I've checked give more than 0.05% transmission. Photographing down to 300nm would be quite enough for me.

I think my boss would be upset if I sliced up our only CaF2 cell. The only DIY conversion I've tried has been on a very cheap Kodak compact where I didn't replace the hotmirror with anything. Not surprisingly focus afterwards was way off. Quartz sheets would clearly be many times the cost of the camera here (£3 IIRC).

I could easily get hold of quartz cuvettes but they'd only be 10mm wide & 1mm thick (not wide enough for many & far too thick) I can buy 1mm thick quartz windows for under £20 but sadly thinner than that isn't a stock item.

I could possibly get away with cleaving & polishing some old NaCl windows, which should get down to 250nm, but I think atmospheric moisture would fog them quickly & they'd corrode the camera internals.

I tried NaCl before CaF2. It did the usual thing that IR cells do ... went cloudy and had to be republished. You know, way back in pre-history we used to make KBr disks. The KBr flows under high pressure and gives an optically clear disk. You might have some of the presses around still. But they were usually on 1cm or so diameter.

petrochemist wrote:

Not surprisingly focus afterwards was way off.

This is actually my #1 issue with full-spectrum conversions. Generally, if you remove a 1mm thick piece of glass typically with index around 1.5, you've basically pushed the sensor about 0.5mm closer to the flange. MFT uses a 4mm cover glass, so we're talking about a potential shift of 2mm!

For DSLRs, that's well out of the adjustment range for aligning the focus sensor (or screen). Fortunately, most lenses just loose a bit of close focus with a shift like that. However, I own some lenses (ultrawide/fisheye) where the entire focus range is a fraction of a mm. So, the take-away message is that you might want to make a custom adapter that compensates for the missing cover glass.  That's easy if you 3D-print adapters, it can be a bit harder to change a commercial adapter's length that much, but it is do-able.

I'll also note that focus plane for NIR or NUV is often pretty far off from visible, so that alone might make you want a customized adapter if you're shooting only with one of those bands.

Ed Constable wrote:

I tried NaCl before CaF2. It did the usual thing that IR cells do ... went cloudy and had to be republished. You know, way back in pre-history we used to make KBr disks. The KBr flows under high pressure and gives an optically clear disk. You might have some of the presses around still. But they were usually on 1cm or so diameter.

An old story

When I started in the lab we used NaCl for all our IR cells & had to replace then every year or so. For one application I eventually switched to CaF2 & over 20 years later we're still using the same cell. Unfortunately its transmission doesn't go down far enough for some of our applications (no transmission below about 1500cm-1 while measuring benzene required us to use 675cm-1)

After 5-10 years of using CaF2 I had a good look at other options that could be used at 600cm-1 & ended up getting a set of KRS5 windows - they also have no issue with water, but they have the twin disadvantages of high refractive index (restricting the path length options that are practical) & high toxicity (not safe to polish yourself). These are also lasting well

We've never used KBr cells but yes I've used a press to make KBr discs, but never enough to be good at making them. I think our 10 ton press has now gone (we still have the die) ATR sampling has rather taken over for the things that needed it.

I certainly wouldn't want NaCl in my camera.

ProfHankD wrote:

petrochemist wrote:

Not surprisingly focus afterwards was way off.

This is actually my #1 issue with full-spectrum conversions. Generally, if you remove a 1mm thick piece of glass typically with index around 1.5, you've basically pushed the sensor about 0.5mm closer to the flange. MFT uses a 4mm cover glass, so we're talking about a potential shift of 2mm!

Could you explain it to me please?

What I have learned from the internet is that the filter stack increases the focal distance by about 1/3rd of its thickness. Leaving the sensor at its position would mean to lose the ability to focus to infinite.

My question: Must the sensor without the filter stack be moved nearer to the lens mount, or further away in order to be able to focus to infinite für visible light?

For DSLRs, that's well out of the adjustment range for aligning the focus sensor (or screen). Fortunately, most lenses just loose a bit of close focus with a shift like that. However, I own some lenses (ultrawide/fisheye) where the entire focus range is a fraction of a mm. So, the take-away message is that you might want to make a custom adapter that compensates for the missing cover glass. That's easy if you 3D-print adapters, it can be a bit harder to change a commercial adapter's length that much, but it is do-able.

I'll also note that focus plane for NIR or NUV is often pretty far off from visible, so that alone might make you want a customized adapter if you're shooting only with one of those bands.

Johannes Zander wrote:

ProfHankD wrote:

petrochemist wrote:

Not surprisingly focus afterwards was way off.

This is actually my #1 issue with full-spectrum conversions. Generally, if you remove a 1mm thick piece of glass typically with index around 1.5, you've basically pushed the sensor about 0.5mm closer to the flange. MFT uses a 4mm cover glass, so we're talking about a potential shift of 2mm!

Could you explain it to me please?

What I have learned from the internet is that the filter stack increases the focal distance by about 1/3rd of its thickness. Leaving the sensor at its position would mean to lose the ability to focus to infinite.

My question: Must the sensor without the filter stack be moved nearer to the lens mount, or further away in order to be able to focus to infinite für visible light?

The precise distance depends on the index of refraction. Basically, the index is a ratio of the speed of light in a vacuum divided by the speed of light in the material... and optical distance to the focus plane is really determined by how long it takes light to get there. Thus, a 1.5 index 4mm glass behaves like 1.5*4mm=6mm of vacuum... and speed of light in air is essentially same as in a vacuum (ok, air's index is actually about 1.000293). Thus, removing the glass effectively makes the sensor closer to the rear of the lens by 2mm.

To correct for removing the cover glass, you need the sensor to be farther from the rear of the lens (by 2mm in my example). Thus, you'd need to add 2mm of shims to the lens mount.

BTW, this is also one of the reasons MFT doesn't really do so well with legacy lenses (in addition to having rather small pixels that really want higher resolution). You see, the index of refraction is actually somewhat dependent on wavelength of the light: that's what dispersionis about. Thus, the 4mm glass actually has slightly different optical thicknesses for different wavelengths, causing CA -- unless the lens is designed taking that into account. MFT lenses can be, but adapted lenses aren't... although the Metabones MFT focal reducers do correct for it (Brian Caldwell, designer of these focal reducers, if you see this please feel free to jump in with a better explanation ).

The reason MFT went with such a thick glass isn't documented, but likely has something to do with the fact that a thicker glass keeps dust particles farther from the sensor (effectively, about 6mm for MFT), so they cast less obvious shadows. Basically, if the dust is close to the sensor, it can block light from all angles coming to the pixel behind it; if the dust is further away, some ray angles to the pixel behind it aren't blocked.

One more odd note: If you had an adapted lens that didn't quite reach infinity focus, you might be able to make it reach simply by installing a glass flat behind it. However, you'd really want that flat close to the sensor to minimize the loss of image quality, so this isn't a great way to solve the too-long-flange-distance adapting problem.

This might be useful

http://www.savazzi.net/photography/uvpassfilters.html

Ed

ProfHankD wrote:

Johannes Zander wrote:

ProfHankD wrote:

petrochemist wrote:

Not surprisingly focus afterwards was way off.

This is actually my #1 issue with full-spectrum conversions. Generally, if you remove a 1mm thick piece of glass typically with index around 1.5, you've basically pushed the sensor about 0.5mm closer to the flange. MFT uses a 4mm cover glass, so we're talking about a potential shift of 2mm!

Could you explain it to me please?

What I have learned from the internet is that the filter stack increases the focal distance by about 1/3rd of its thickness. Leaving the sensor at its position would mean to lose the ability to focus to infinite.

My question: Must the sensor without the filter stack be moved nearer to the lens mount, or further away in order to be able to focus to infinite für visible light?

The precise distance depends on the index of refraction. Basically, the index is a ratio of the speed of light in a vacuum divided by the speed of light in the material... and optical distance to the focus plane is really determined by how long it takes light to get there. Thus, a 1.5 index 4mm glass behaves like 1.5*4mm=6mm of vacuum... and speed of light in air is essentially same as in a vacuum (ok, air's index is actually about 1.000293). Thus, removing the glass effectively makes the sensor closer to the rear of the lens by 2mm.

To correct for removing the cover glass, you need the sensor to be farther from the rear of the lens (by 2mm in my example). Thus, you'd need to add 2mm of shims to the lens mount.

I think to correct for removing the filter stack in front of the sensor, you need to bring the sensor nearer to the lens.

from: http://www.ir-photo.net/ir_mod.html

From the formula given on the above mentioned website you would bring the sensor round about 1,3mm forward to the lens. No shimming of the lens needed.

Yes, the speed of light is reduced in glass, but the light is refracted. The angle of refraction in the glass is less than the angle of incidence resulting in a focus point further away from the exit papilla of the lens than without the glass filter.

That's how I understand Snell's law.

BTW, this is also one of the reasons MFT doesn't really do so well with legacy lenses (in addition to having rather small pixels that really want higher resolution). You see, the index of refraction is actually somewhat dependent on wavelength of the light: that's what dispersionis about. Thus, the 4mm glass actually has slightly different optical thicknesses for different wavelengths, causing CA -- unless the lens is designed taking that into account. MFT lenses can be, but adapted lenses aren't... although the Metabones MFT focal reducers do correct for it (Brian Caldwell, designer of these focal reducers, if you see this please feel free to jump in with a better explanation ).

The reason MFT went with such a thick glass isn't documented, but likely has something to do with the fact that a thicker glass keeps dust particles farther from the sensor (effectively, about 6mm for MFT), so they cast less obvious shadows. Basically, if the dust is close to the sensor, it can block light from all angles coming to the pixel behind it; if the dust is further away, some ray angles to the pixel behind it aren't blocked.

One more odd note: If you had an adapted lens that didn't quite reach infinity focus, you might be able to make it reach simply by installing a glass flat behind it. However, you'd really want that flat close to the sensor to minimize the loss of image quality, so this isn't a great way to solve the too-long-flange-distance adapting problem.