Debayering camera or making sensor sensitive to UV

[Avalon]

So as you probably already know first barriers to ultraviolet photography are IR/UV cut dichroic mirror and color filter array in front of sensor which in most normal cameras blocks out UV starting from about 400nm. Strangely CFA even blocks out near infrared spectrum which seems to have excellent transmission through semitransparent dyed plastics but infrared converted cameras appear to have this pink tint.

http://www.fen-net.de/walter.preiss/bilder/spectrum.gif

 

That’s why necessary step is removal of Bayer CFA filter what is called also demosaicing or debayering. Although there are different camera and sensor designs, color filter arrays such as Foveon or even prismatic ones used in 3CCD cameras as well already monochrome sensors which can have greater sensitivity but they are exotic and expensive. There is solution also of swapping sensor for monochrome but they are hard to find too. Hot mirrors also might differ and I was lucky to find JVC camera IR-cut filter which passes 365nm UV LED spectrum but most IR-cut filters block UV that’s why for UV and IR photography it must be removed.

 

UV also is blocked/reflected by other components such as lens, optical cement, AR coatings, microlens (used in older front lit sensors) and silicon nitride coating on sensors which blocks ultraviolet and blue light: https://www.photonic..._sensors/a47007

 

Backlit CMOS sensors can actually detect even UV-C spectrum even down to 200nm but that is limited due to silicon nitride coating and it would be interesting to know if something could be done about it?

 

IR-cut filter removal, unlike debayering, is much simpler, reversible and less risky operation. I was attempting many years to find way to make UV sensitive monochrome camera and have ruined several cameras doing this due to breaking tiny golden wires or damaging sensor structure. Here is my topic where I explained how I tried unsuccessfully debayering Samsung 5230 and JVC camcorder sensors using acetone, small swabs and razor: https://www.physicsf...-filter.599701/

 

Now after gaining experience, getting microscope and access to many solvents I want to try again my luck this with modern BSI CMOS sensor based camera. Backlit or back-illuminated CMOS sensors are different from older generation of CMOS or CCDs in that they do not have anymore in front of sensors electrodes which not only block some light but also make it vulnerable to debayering. Front sensors due to these electrodes also require microlens so their removal would inevitably reduce light sensitivity. BSI CMOS are much more sensitive to overall spectrum and especially UV which does not penetrate deep into most materials. So for these reasons I strongly recommend using cameras of 2013 year or later when BSI CMOS become started becoming popular. They have many advantages such 1080p support and so on. For long time I believed CCD are superior to CMOS due to signal amplification using transistors but was wrong.

 

So camera I have now camera Fujifilm F500 which is not best camera due to fixed lens but what is affordable to risk destroying and as well compact. I already was able to remove IR/UV cut mirror so now I need to remove CFA as well. Here is photo of my camera sensor, luckily there are no golden wires exposed which would have to be first filled with resin to avoid breaking.

 

Does anybody here have experience debayering modern CMOS sensor? I have access now to various solvents used to dissolve old painting overpaints and varnishes. From what I read in other forums it’s best to use strongly polar solvents such ethanol, MEK or acetone which is used myself because they easily dissolve plastics, acrylics and so on. I do have 99% ethanol with 1% MEK as well acetone, not sure which first to use. If this would fail I could try using nonpolar solvents such toluene or xylene. We use as last resort in oil painting overpaint removal or it’s softening solvent such as dimethylformamide or even toxic chlorinated solvents so I could test these too if CFA would be hard to remove with cotton micro swabs. Longer exposure to solvent also can help soften CFA, I actually submerged JVC gz mc100e camcorder sensor in acetone for day and that helped. It is enough to soften or „swell“ CFA so that it could scraped off gently with wooden or plastic needle or softer metal blade. Sodium hydroxide seems like too extreme but is very effective at stripping away organic matter such as paint. Microscope would came in here handy especially for smaller sensors. It’s very import to avoid damaging delicate circuitry, so camera could even be left turned on to monitor pixel health (unless there is risk of short circuit or static electricity), I think with backlit CMOS this risk should be lower. I seen also one person describing removal of CFA and microlens using air blasting with sodium bicarbonate which is softer than silicon dioxide sensor layer.

 

It would be helpful to actually know what exactly materials are utilized in sensor production at least at surface to choose best CFA removal technique minimizing damage since trial and error option doesn’t work out so good.

[Cadmium]

You might want to ask Alex H, he has some experience with that.

[dabateman]

I have not done a chemical or physical CFA removal yet. But have tried to research it. You may find more answers on Astronomy sites. It was more common in the past to see people there try more risky attempts to gain very little increase in sensitivity. However since the release of the ZWO 1600mm camera, most who were successful stopped as it is far cheaper and off the shelf better. It came out in 2016 with the exact same sensor as the Olympus Em1. It is a front iluminated monochrome sensor with microlenses and the non cooled version is only $900.

 

For a UV camera you will need to be able to change lenses. If not you are throwing away too much by just the lens on the front. BSI sensors are still quite new. As far as I know all of them also have microlenses. The number of cameras with those sensors is very few. I count only about 13. And on that list some have other problems like IR shutter monitors that also make them useless for uv like the Sony ones. So you are limited to the new Fuji Xt3, Nikon D850, samsung NX500, NX1, pentax q7, Q-s1, and Nikon J5.

 

Of those I wonder if anyone has full spectrum converted a Samsung NX1 and if they saw a real improvement over a front side illuminated sensor. That camera has some popularity, but I haven't found an answer to that yet.

Maybe one day Andrea will convert her D850 and we may also know.

 

Jonathan's recent work with microlenses suggest they will work well down to 300nm helping to improve sensitivity and not harm it. So they still maybe are needed even on a BSI sensor.

This too is why I have switched my attention to the ZWO. May just be the best cost effective solution and I hope they release a newer BSI monochrome camera as the 1600mm is over 2years old now.

[Andrea B.]

...have ruined several cameras doing this due to breaking tiny golden wires or damaging sensor structure.

 

Avalon, given the toxicity of several of the chemicals mentioned above and your having already ruined several cameras, I am thinking that the best all round solution for you is to save up the money and let a professional debayer your camera!

 

Although I do grant that there is a certain satisfaction in learning how to do things like this yourself. :D

[Avalon]

BSI sensors are still quite new. As far as I know all of them also have microlenses. The number of cameras with those sensors is very few. I count only about 13. And on that list some have other problems like IR shutter monitors that also make them useless for uv like the Sony ones. So you are limited to the new Fuji Xt3, Nikon D850, samsung NX500, NX1, pentax q7, Q-s1, and Nikon J5.

 

Of those I wonder if anyone has full spectrum converted a Samsung NX1 and if they saw a real improvement over a front side illuminated sensor. That camera has some popularity, but I haven't found an answer to that yet.

Maybe one day Andrea will convert her D850 and we may also know.

 

Jonathan's recent work with microlenses suggest they will work well down to 300nm helping to improve sensitivity and not harm it. So they still maybe are needed even on a BSI sensor.

This too is why I have switched my attention to the ZWO. May just be the best cost effective solution and I hope they release a newer BSI monochrome camera as the 1600mm is over 2years old now.

 

I was actually thinking getting in future Sony camera since they have such great sensitivity such as A7s. Could you explain what is IR shutter monitor?

Dont know how many BSI sensors still utilize microlens but for debayered camera's it's best to have sensors that need microlens since they are going to removed. The larger individual pixel sensor are the better is sensitivity, so maybe some normal camera's might have so big pixel photodetector surface that microlens no longer needed? Special purpose monochrome camera's are just too expensive, as well they are not so portable. So my best option is just learn to debayer camera's and get maybe even full frame camera so that even with microlens removed quantum efficiency drop should be compensated.

What about digital night vision monoculars? They are engineered for maximum sensitivity to operate even at 0.0001lux conditions so they have monochrome sensors only problem they might have fixed lens and not support high resolution to increase pixel surface area. Some security camera's also utilize monochrome sensors for low lux night lighting but are there any portable?

[dabateman]

An IR shutter monitor is an IR led placed near the shutter mechanism that fires IR light within the camera to detect the condition of the shutter. For a normal camera this is ok as the sensor has blocking filter to prevent the light from interfering with the sensor capture. However for a full spectrum conversion camera the IR monitor will wash out your UV image.

The IR monitor is required. So you cannot remove it. A work around is to use electronic shutter only. But this has other limits. There is a site where I saw someone who replaced the IR light with a higher wavelength IR led and this worked out ok for astrophotography, but I don't know if there used it for uv imaging.

 

However, when you are thinking about extreme modification. You need to place the gain into perspective. A monochrome camera with CFA removed has only a maximum of 2 stops advantage over a regular full spectrum conversion. This really isn't a whole lot. Similarly BSI sensors only have 1/2 to a maximum of 1 stop advantage over front illuminated sensors. Useing the maximum, 1 stop is really not a lot.

So you need to know what your base expectations are first.

With a quartz lens, a high pressure mercury vapour lamp, and a 300bp10 filter on my EM1, my exposure was ISO200, f8 and 9 minutes. So a monochrome camera would knock that down to 2.5 to 3 minutes. Not really a major boost. If BSI, then would be at 2 minutes. So there is a cost to benefit ratio to consider. Also there wasn't much excitement down there. Things are starting to absorb, so getting really dark. Plastics remained quite reflective.

 

I also own a SD14 camera and its limit of detection is 340nm. It can not see UVb at all. One day I still may rent the Sigma SdQ to see if it can see at least some lower wavelengths. But I doubt it, I don't think the silicon wafers Sigma uses are as sensitive.

[dabateman]

However, thecost to benefit is significant for a quartz lens. I see about 4 stops advantage at 340nm. Imaging at 370nm, there is about 1 stop advantage. But I just get signal with the best known UV lenses at 313nm. I am getting at least a 6 stop advantage with quartz lens, but more likely an 8 to 10 extra stops at 313nm. The image from the Steinheil 50mm f2.8 is really not usable.

So its worth it to try and get a quartz lens.

 

Don't forget all the variables for an image. You seem fixated on the sensor, but it is only one of the four parts.

For deep uv imaging you will need a quartz lens. UKA optics sell the cheapest, and that may work for you with narrow band pass filters.

Second you need good filters. Ideally you want at least 80% transmission at your desired wavelength and ideally no IR leakage. An 80% transmission filter is 2 stops better than a cheap 20% transmission filter. Also due to camera sensitivity, the camera will lean on the upper portion of the filter range. So a 370bp15 filter, the camera is really seeing that 375nm point in the spike more so than the 365nm point. However this will depend on lighting. Which is the third part of the equation. The sun is not a good consistent uv source. It has very high IR an only usable signal to 350nm. You will not see UVb well with only using the sun as your light source. Mercury vapour lamps are good for getting 313nm, 334nm and 365nm lines in the uv region. LED now seem to be the best. No IR, low temperatures and LEDs can be tight at the wavelengths of interest. They are however still quite expensive.

So you need to think about each part of the equation. Locking in an a good lens I think is the first part. Then getting good filters and lights. Then see if a boost in sensor will help you. Spending $35 to add lights make more sense than $1000 for a sensor. So does spending $600 on better filter if you only have 10% transmission, as you will gain 3 stops.

[Avalon]

 

However, when you are thinking about extreme modification. You need to place the gain into perspective. A monochrome camera with CFA removed has only a maximum of 2 stops advantage over a regular full spectrum conversion. This really isn't a whole lot. Similarly BSI sensors only have 1/2 to a maximum of 1 stop advantage over front illuminated sensors. Useing the maximum, 1 stop is really not a lot.

 

 

Between color and monochrome camera sensitivity difference is much higher. I fact CFA blocks out UV almost completely, maybe blue pixels let in some UV-A - that's it. Unless you have in mind somekind different color detection method that allows UV to reach photodetector. I noticed huge difference in sensitivity between older CMOS, CCD and BSI CMOS camera's. My FSI CMOS smartphone could not even show night while with modern BSI CMOS phone it can even show stars. Maybe there other factors ilvolved but gettting rid of electrodes in front of photodetectors made significant difference in quantum efficiency.

 

Don't forget all the variables for an image. You seem fixated on the sensor, but it is only one of the four parts.

 

 

Currently discussion is about making camera sensor more sensitive in UV region. I do understand that other component have big effect on ultraviolet transmission and for my future project I plan to get fused silica or other UV pass material lens. For now I will use F500 and UV spectrum which fixed lens can pass. For outdoor photography and as well painting ultraviolet reflectography 350-400nm should be sufficient.

[dabateman]

I think your mixing up the uv/ir block filter and the color filter array.

See Jonathan post her comparing a full spectrum camera with a Monochrome camera in post #20.

 

http://www.ultravioletphotography.com/content/index.php/topic/2813-camera-sensitivity-chasing-ghosts-in-spectral-sensitivity-measurements/page__view__findpost__p__22045

 

Then look at next page at post #24 with sunlight factored in.

Jonathan also has a post comparing a monochrome, full spectrum conversion camera and a normal camera. The monochrome is 6x greater response than the full spectrum and the normal has no signal.

 

[Andrea B.]

CFA blocks out UV almost completely, maybe blue pixels let in some UV-A - that's it.

 

No, this is not correct!!

We are all using cameras which have had the internal UV/IR-blocking filter removed -- but which retain the Color Filter Array (Bayer Array) -- to successfully make reflected ultraviolet photographs.

 

Removing the CFA does improve the passage of UV light to the sensor.

 

The reason that older digital cameras appeared to be more sensitive to UV was because their internal UV/IR blocking filters were very weak. For example, one could successfully make reflected ultraviolet photographs with an unmodified Nikon D70 because it had such a weak internal UV/IR-blocking filter.

 

Please use correct terminology! B)

[Andy Perrin]

The many color flower photos on this forum show that the CFA lets through more than enough UVA to make nice color photos in that range. It does block UVB mostly, but there are many other obstacles there too.

[Cadmium]

"The reason that older digital cameras appeared to be more sensitive to UV was because their internal UV/IR blocking filters were very weak"

 

I think basically all cameras have a blue/green filter (BG type filter), which passes UV and suppresses IR.

I think It is the coatings that then remove the UV, and not all stock cameras were or are created equal with identical internal filters or internal coatings.

Edited: Unfortunately, sensor and Bayer CFA manufactures don't log on and participate in these discussions. :)

 

Whoops! Just noticed the picture I linked to is copyright, so I will post a link to the page containing it, darn... it would have looked cooler with the pic in my post. :(

https://wolfcrow.com...camera-filters/

[Andrea B.]

Who are "the people who make this stuff" who "don't log on here"?

Make what stuff?

[Cadmium]

I meant the makers of image sensors and Bayer CFA's, like:

Top 10 manufacturers of CMOS Image Sensors (2012 market share)

1. Sony - 21%
   
2. Omnivision - 19%
   
3. Samsung - 18%
   
4. Canon - 9%
   
5. Aptina Imaging - 8%
   
6. Toshiba - 6%
   
7. STMicroelectronics - 4%
   
8. Nikon - 4%
   
9. GalaxyCore - 4%
   
10. SiliconFile - 2%
    

Of course they don't participate in these such discussions here... :)

I will edit the above post to make it clear..

[dabateman]

Cadmium,

That might be the list of sensor manufacturers. But really wonder how long the list is for the dye manufacturers that are used to source the blue, red and different green dyes. It maybe all internal like the way kodak did it or there maybe just one or 2 companies that provide the CFA dyes.

[Avalon]

 

No, this is not correct!!

We are all using cameras which have had the internal UV/IR-blocking filter removed -- but which retain the Color Filter Array (Bayer Array) -- to successfully make reflected ultraviolet photographs.

 

Removing the CFA does improve the passage of UV light to the sensor.

 

 

Well I posted spectrum graph showing quantum efficiency of each Bayer color filters, most color sensors spectrum graphs show sharp drop in quantum efficiency in 400nm region. Too bad I can't find much images showing more ultraviolet region. Here is another spectrum chart of Exmor R (or backlit) CMOS camera showing both monochrome and RGB filter quantum efficiencies. Color camera picks up some UV-A but monochrome camera is significantly more sensitive. And this is with IR-cut filter on:

 

[Andy Perrin]

Avalon- don’t believe everything you find on the internet. Jonathan has done his own measurements here on this site. Look at his data.

[ulf]

The RGB curves in the diagram above look like a sensitivity of a non-modified camera.

 

The full-spectrum modified cameras, most of us use, has the IR-and UV-blocking filter(s) removed, but the Bayer-matrix intact.

That extends the cameras sensitivity into both ends of the spectrum.

 

It is true that a monochrome-modified camera ha an even higher sensitivity, especially into UV.

Search for Jonathans measurement results on this forum.

 

I measured the IR-and UV-blocking I removed from my Canon 60D:

http://www.ultraviol...dpost__p__17625

[Andrea B.]

Avalon, the proof is in the photos. We have easily captured reflected UV photos between 330 - 350 nm.

 

It looks to me like the chart you have referenced was measured with spectrometer which could not go below 360 nm. There are other charts and measurements made with UV-capable spectrometers which do show that there is sensor response below 360 nm.

 

While the response below 360 nm is not large, there is most certainly enough sensitivity to make good reflected UV photos.Below 330 nm it becomes more difficult, of course, because the amount of UV between 300-330 nm in sunlight has decreased so much. There are two examples linked below.

 

Several members have measured the sensor response and have posted results here on UVP. Here is one such example:

http://www.ultraviol...dpost__p__21140

 

 

Examples:

 

Floral signature captured with CopperU having peak at 339nm. The CopperU is a liquid copper sulphate filter.

http://www.ultraviol...th-the-copperu/

 

Floral signature captured with Edmund 340/10, a narrow bandpass filter.

http://www.ultraviol...-sunflower-sun/

 

A 30" photo made with a 293bp10 filter: http://www.ultraviol...dpost__p__17538

This is not a particularly convincing photo, but it does show that there is some very small response around 300nm.

[Avalon]

Important to note to those who plan to debayer camera is to check before buying or attempting to debayer camera. Some sensors might be damaged even when removing just protective glass plate because they are glued with opaque mass that can be softened only solder iron. Beware of golden contacts, if they are not visible they might be hidden under glue so removing it would destroy sensor. If however contacts are visible you can carefully attempt to remove glass and right after that I recommend filling contacts with transparent resin.

I attempted to remove CFA from Finepix F500 camera and damaged sensor because I removed black grey glue holding glass plate which had microscopic wires in them. Also be very careful around thin wires such as flexiboards, better detach them if possible because repairing them will be hard as hell even under zoom. I used 99% ethanol with 1% MEK and later acetone which was best solvent to use with micro cotton swab to peel away CFA.

So now I;m looking to get interchangeable and bigger sensor camera. I'm curious if Olympus PEN E-PM1 could good candidate? Need to see close up picture of sensor before buying to determine if debayering could be possible.

[Dmitry]

My e-pm1 without hotmirror. Each corner.

No golden wires, sensor is connected using PCB tracks.

Sensor cover is glued using transparent glue.

 

[Avalon]

Thanks. Protective glass can be removed without damaging wires? I'm looking to get as well APS sensor SONY NEX 3N or 5N camera since one guy already did debayer them succesfully, only few pixels were dead: http://danielmorrisonimaging.com/projects/sony-nex-3-monochrome-conve.html

[Andrea B.]

Avalon, be sure to check with Daniel about Sony cameras which have an infrared internal shutter monitor. Those should not be de-bayered or converted by removing internal filter glass.

[Andy Perrin]

Anyone going to the trouble of debayering a sensor on their own is probably also up to the job of replacing the IR monitor with a SWIR monitor.

[dabateman]

What Andy is referring to is on step two of this modification for Sony cameras:

 

https://www.nickspiker.com/blog/search-digital-kodak-aerochrome/