UVC (254nm) imaging with monochrome Nikon d850

[JMC]

Ok, disclaimer first - UVC is extremely dangerous, please do not attempt to use it without fully understanding the risks involved.

 

Background to this work. I've been seeing everyone playing around with UVC, and had assumed that there was little point even trying it with the cameras I've got. A few weeks ago I bought a couple of 254nm 8w tubes to fit my UVP lamp, and today thought I'd just give it a go with my monochrome Nikon d850 (which has a quartz coverglass instead of the usual WG280) and Rayfact 105mm UV lens.

 

Setup. 2x8W 254nm tubes in a UVP lamp, no filter on the lamp. Subject a vase with a couple of feathers in it, and a 20% diffuse reflectance standard. Filters, a 254nm bandpass filter from the Sirchie forensics camera, WG305 2mm and WG295 3mm. Camera, monochrome converted Nikon d850 with fused silica window from MaxMax. Lens, Rayfact 105mm UV lens. Settings, ISO400, f8, 30s exposure for the UV images. Whitebalanced in Darktable and reduced in size in XnConvert. Images shown with reduced resolution, but full frame.

 

Firstly, image with visible light.

 

Now with 254nm lamp and the 254nm bandpass filter only.

 

So, with the 254nm lamp and bandpass filter I get an image (which I wasn't expecting to be honest), and because the glass of the vase is opaque at least some of this is UVB or even UVC.

 

Next I tried putting a couple of different longpass filters in front of the 254nm filter. The aim here to try and block what the 254nm filter should be letting through, while letting through the out of band areas of the 254nm filter. I did this to check for leaks.

 

With the WG305 2mm in front of the 254nm filter.

 

And with the WG295nm 3mm filter in front of the 254nm filter.

 

Both the WG305 and WG295 filters resulted in a reduction in the image brightness, but they did not eliminate the image, which suggests to me that there are some leaks in the 254nm filter in the longer wavelength regions. But it also tells me that most of the image is coming from the UVC region, which I did not expect. I must admit, I hadn't realised when I chose these two long pass filters that the 254nm one had such a long tail on it. Perhaps in hindsight a longer long pass filter would have been a better option.

 

EDIT - Although looking as closely as I can at those WG images, the vase looks dark which would suggest leakage in the UVB/C region, so perhaps it is coming from the overlap region? More work needed there I think.

 

Something to consider is the filter transmission spectra and the light irradiance spectra, which are given below (full range and then zoomed in on the UV region).

 

 

Yes, there is some overlap between the 254nm filters and the WG ones in that 280nm to 300nm region, but it occurs at a region where the light is not really emitting anything, which makes me think that the 254nm images are really driven by that big 254nm line in the lamp. It also tells me that even good filters for UVC run the risk of letting enough light through in the out of band regions to contaminate the image - OD4 is no longer good enough, even with a light source like this with a really strong 254nm peak.

 

I learned something new today - don't make assumptions about cameras capabilities before testing them. Sorry for any typo's, I'll go back and proof read when I have a minute.

[ulf]

I suspect that the last toe of the WG-filters comes from internal crosstalk in the spectrometer, originating from longer wavelengths.

However, as you say, there is still some actual overlap and those filters cannot block the 254nm-peak completely.

 

Interesting lesson.

[Stefano]

Very nice test. Silicon sensors, theoretically, should see down to 190 nm (source).

 

Your camera can really go deep into UV. I would love to have a camera like yours one day.

[dabateman]

Wow this great to see. The D850 isn't a Sony sensor, but its similar to the one in the Z7. The Z7 just has PDAF, which would be stripped off with monochrome conversion.

If I had the cash a couple z7 cameras would be fun, monochrome, and normal. I think it too suffers from the banding issues as the Z6 and wouldn't be recommended for full spectrum. Amazing to see a regular camera could be used for UVC.

[JMC]

Thanks all. Imaging this far down brings a whole new level of challenges, and is not something I plan on doing routinely.

 

David, surprised to hear that the d850 sensor isn't a Sony one, I thought it was - https://petapixel.com/2018/06/15/the-nikon-d850s-sensor-is-made-by-sony-report/

 

I'm thinking of trying my Sony A7III too. My d850 conversion has a problem with internal light (I think from the LED) which is why I ran this test at 400ISO. However the converted A7III has no issues with that and I can turn the ISO right up. Be interesting to see if I can see anything with the Bayer filter still in place - see whether the RAW file is still 'green' that far down.

[Stefano]

One day you could even try a tri-color UVC (blue), UVB (green) and UVA (red). You can do a lot of interesting things.

[JMC]

Little bit of a follow up after yesterdays work.

 

When I did the original imaging yesterday, I put the WG305 filter in front of the 254nm filter (WG305 closer to the subject). I did this because when I put it behind the 254nm, the image looked foggier. I repeated that today, to try and show it. Three images with the monochrome d850, same settings as yesterday - 254nm only, WG305 in front of the 254nm one, WG305 behind the 254nm one.

 

254nm filter only

 

WG305 filter in front of the 254nm filter

 

WG305 filter behind the 254nm filter

 

Perhaps a bit hard to see after uploading, but putting the WG305 behind the 254nm filter produces a much more foggy image. I think what is happening is that with the WG305 filter behind the 254nm filter, the UV is getting through and hitting it, resulting in fluorescence which is then picked up by the camera. If the WG305 is in front of the 254nm filter this cannot happen, hence it is less foggy. Filter order has an impact here.

 

As as second experiment, I tried to get a UVC image using a camera with the Bayer filter still in place. I have a Sony A7III which has been converted to multispectral by MaxMax, and with the sensor cover glass replaced with fused silica.

 

Same settings as yesterday for the monochrome d850, same lens etc. With the Bayer filter there, I expected it to block most of the UVC and indeed it did. At ISO 400 and 30s, the image was pretty dark. These are whitebalanced in Darktable.

 

A7III, ISO400

 

However it wasn't as dark as I thought it would be. I increased the ISO to 1600 and it produced an image.

 

A7III, ISO 1600

 

And with the WG305 in front.

 

So at 254nm, it looks like the Bayer filter/microlense array is blocking about 2 stops of the light, which is less than I thought it would based on what I've seen at 302nm.

 

Funky thing with the A7III though is you can crank up the ISO. This was at ISO 102,000, f8 on the lens, and 1 second (!) exposure with the 254nm light.

 

Noisy as hell, but amazing that it produces an image.

 

However the key reason I wanted to use the A7III was because of the Bayer filter. If the images are opened as RAW composite images in RawDigger, I can look at the Red, Green and Blue channels for the Spectralon disk, to get an idea of the contribution of the different channels to the image.

 

Opening an image with just the 254nm give the following.

 

In the Spectralon disk the red, green and blue channels are fairly similar, with the green being slightly higher. This tells me that at 254nm the three Bayer filter dyes are pretty close in their absorption of the light. This is very different to 302nm where green dominates.

 

Next, with the WG305 filter and 254nm filter.

 

This is now a different balance between the channels. It's about 47% blue, 35% green, 18% red. Funnily enough from work in the past, this ratio is similar to the sensor response of the A7III in 280nm region, which suggests is the 'leak' may well be coming from that region of overlap between the two filters. However that is purely speculation at the moment.

 

So with an A7III with the Bayer filter and microlenses still in place, it is possible to capture UVC images, although because of the sensor response, it is rather grey down there.

[Stefano]

Nice. There are so many false UV colors, so many layers as you travel down. bluish-violet (becomes blue/violet after WB), magenta, reddish magenta (becomes yellow), then green, then bluish (around 280 mm, in your sensor) and then gray.

 

The CFA (Bayer filter) + microlenses is a very thin layer of material, probably a few microns total, so it doesn't completely absorb UVC. If it was even 0.1 mm thick then it would be completely opaque.

 

You can image lots of things. David said dandelions still have nectar guides in UVC, it would be nice to see that (you probably don't have any now), and most things should be dark there.

[Andy Perrin]

This makes me want to try with my A7S conversion. I will have to check what LifePixel used as a cover.

[Stefano]

This should be their full-spectrum filter: https://www.cloudynights.com/uploads/monthly_09_2015/post-230910-0-24312400-1441277825.png

 

Does transmit some 254 nm UVC, but not much.

 

Do you still have the coverglass?

[Andy Perrin]

Stefano - nope, they steal your original glass. The sensor still has the glass on it, they don't touch that.

[Stefano]

"They steal your original glass"? I don't understand what you mean here. I know they remove the UV/IR cut filters, and you just said they don't touch the sensor's coverglass (at least this is what I understood), so what do they "steal" exactly?

 

Or maybe they replace the coverglass?

[dabateman]

I was wrong and got confused again Jim Klasseon data also suggests its the exact same as the A7R3. I mixed up the conclusion, its a Sony Not a Toshiba. The D4/DF might be the last Toshiba sensor than.

 

Jonathan did you get the cover glass changed on the Sony A73?

 

Now I have to test my newly converted Em5mk2 as its a Sony sensor and had better response at 302nm than my Em1mk1.

I may have a 5th camera capable of UVC and not know it. My Em1mk1 could see 254nm but was very slow and noisy.

[Andy Perrin]

"They steal your original glass"? I don't understand what you mean here. I know they remove the UV/IR cut filters, and you just said they don't touch the sensor's coverglass (at least this is what I understood), so what do they "steal" exactly?

 

Or maybe they replace the coverglass?

They don't touch the cover glass. They steal the IR cut filter. Nobody removes cover glass ever unless by special request. (Jonathan makes a lot of those kind of special requests, but it's not a standard conversion thing.) *I* was confused by what you were asking because essentially only for monochrome conversions and Jonathan's experiments have I ever heard of cover glass removal.

[Stefano]

The cover glass probably doesn’t transmit a lot of UVC, if any. Still worth a try in my opinion.

[JMC]

David, that makes sense, just wanted to check about the sensor.

 

Both of the cameras mentioned in this thread have had the sensor cover glass replaced with fused silica. For the monochrome d850, the cover glass needs replacing as part of the conversion. For the multispectral conversion of the A7III, this cover glass would not normally be replaced, as it is a very delicate process. However I pestered Dan at MaxMax to do it for me, and use fused silica for both of them (specifically because I wanted to try them for UVC at somepoint).