• Ultraviolet Photography
  •  

Sigma camera for UV reflectance imaging

30 replies to this topic

#21 Shane

    Member

  • Members
  • 324 posts

Posted 17 October 2018 - 14:45

It used to be that the Foveon blue photodiode layer junction depth was 0.2micron and designed for 420nm blue light but the overlying structures severely attenuated any UV and some blue light. Maybe the Quattro sensor has reduced overlying structures in a bid to enhance capture of blue light and in doing so has improved its UV capability.

#22 dabateman

    Da Bateman

  • Members(+)
  • 335 posts
  • Location: Maryland

Posted 18 October 2018 - 07:27

Jonathan, I don't think you will ever see the color that we get with bayer cameras. The Sigma is really just like monochrome camera.

However, your comment does offer an interesting idea that I will try to test. You may be able to use the Sigma for large channel separation. Using a UG1 or U330 filter which have nice UV band and a large IR band, it may be possible to illuminate a subject and get both UV reflectance in the the blue channel and UV induced IR fluorescence in the Red channel. I tried to quickly test this using ug1 1mm filter and two 365nm LED lights, but my subject was not very fluorescent in the IR. Using just a 720nm IR filter I needed 15 second exposure, where the uv reflectance was 1 second. I will have to find a better subject, but testing this out might be interesting. Getting both types of images for the same exposure. The U330 filter may give some green information, in the mid wavelengths. I also have a 2.5mm UG5 which might work, as the green band is very low at that thickness.
Something I will also have to test with the Quanttro when it arrives.

#23 Shane

    Member

  • Members
  • 324 posts

Posted 18 October 2018 - 16:07

When the Foveon sensor first came out it theoretically offered great potential for UV and IR work due to its stacked photodiode channels. Although some pushed it into service for these purposes, it never really reached its theoretical potential. I tried it for both UV and IR, and quickly determined that even the unconverted Nikon D40 outperformed it for UV work, although perhaps due to better in-camera firmware processing. Foveon UV was very noisy and IR, although not bad, had some strange in-camera processing going on. Theoretically, UV should be in the blue channel which it was primarily, and IR in the red channel which certainly wasn't the case. Sigma processing showed IR data in the blue and red channel but none in the green......not possible! Further investigation bypassing Sigma processing revealed that it occurred in all 3 channels increasing in intensity from blue through to the red channel but certainly not confined to the red channel only.

The presentation of the Quattro UV is certainly promising and looks to far exceed what I have seen in the past with regards to Foveon UV capture. I'm certainly interested to see more but I would also like to understand what has changed to produce such an outcome. Certainly in 10 years there must have been some improvements in firmware noise reduction but what else?

Ideally we need 5 stacked photodiodes (and decent firmware), for UV, B, G, R and IR, although the physical limitations for producing a true UV channel are extremely challenging.

The Foveon is similar to a monochrome camera in that it does not have a Bayer array and so demosaicing is not really part of its processing, but it still requires colour interpolation of the three stacked photodiodes.

On a side note - I tried to find links to postings I had made about this in 2009 on various UV and IR forums but those forums and important postings made by many have long gone. This often leads to trying to reinvent the wheel. This site houses some extremely useful postings and I hope that these also don't eventually disappear when the site is longer active.

#24 Andrea B.

    Desert Dancer

  • Owner-Administrator
  • 6,700 posts
  • Location: USA

Posted 18 October 2018 - 17:27

[OFF TOPIC] How would we preserve this website so that the data would not be lost?
Andrea G. Blum
Often found hanging out with flowers & bees.

#25 Shane

    Member

  • Members
  • 324 posts

Posted 18 October 2018 - 19:29

PM sent....don't want to hijack thread.

#26 Shane

    Member

  • Members
  • 324 posts

Posted 19 October 2018 - 13:36

Found the old Sigma links at the now Fotozones
https://www.fotozone...&comment=119134
top histogram is unprocessed linear RAW and lower is standard Sigma processed RAW histogram.
https://www.fotozone...on-cmos-sensor/
unfortunately in the archiving of these posts some text seems to have gone missing..but the top image is the B channel, middle the G channel and bottom the R channel. As you can see most UV fell into the B channel as expected, with a little falling into the G channel and none in the R channel.

#27 OlDoinyo

    Member

  • Members(+)
  • 631 posts
  • Location: North Carolina

Posted 20 October 2018 - 01:31

View PostShane, on 18 October 2018 - 16:07, said:

Theoretically, UV should be in the blue channel which it was primarily, and IR in the red channel which certainly wasn't the case. Sigma processing showed IR data in the blue and red channel but none in the green......not possible! Further investigation bypassing Sigma processing revealed that it occurred in all 3 channels increasing in intensity from blue through to the red channel but certainly not confined to the red channel only.

Welcome to the wacky world of Sigma firmware! I wonder how you managed to bypass the processing. I was never able to do that on the SD14 I used to have.

#28 dabateman

    Da Bateman

  • Members(+)
  • 335 posts
  • Location: Maryland

Posted 20 October 2018 - 02:58

Shane,
Thank you for those historical links! The first one is very similar to my thoughts. Reading through it took me a long time as there seemed to be a bunch of off topic discussions with posts missing, so following it was fun. Also since the main players seemed to have fallen here now.

I think software has gotten better but the leakage through layers will be the problem. Will be like specral unmixing of fluorophor channel signals, like in a 6 color flow cytometer. But may be possible, just need lots of control images, which may elimate the point really. I am still thinking about it and will see if the Quattro files are better.

The terminology we use may also be a problem. I do find Sigma cameras good for uv. They are really sensitive. However their range is really limited. I found that my SD14 is great from 370 to 400nm section of the UV spectrum. Equal to a well converted Bayer camera. But the Range is limited to a maximum of just 350nm. Whereas my well converted Olympus Em1 can quite easily see to 300nm. But I purchased my SD14 for the cost of converting my Em1. So if you were just interested in great uv monochrome signal response, the early Sigma cameras are still good. If you want broad range or mixed color uv with various filters, then No don't buy an earlier Sigma. Testing the new Quattro should be fun. But I think the range will be similar.

#29 Shane

    Member

  • Members
  • 324 posts

Posted 20 October 2018 - 05:18

Quote

I wonder how you managed to bypass the processing
Almost 10 years ago.....not sure anymore.

#30 dabateman

    Da Bateman

  • Members(+)
  • 335 posts
  • Location: Maryland

Posted 20 October 2018 - 05:35

The major advantage of the Sigma SD14 is it was the last camera model fully supported by almost all software. You can open Raw files in dcraw, lightroom, and convert to dng using Adobe free dng converter software. Using the DNG converter is how you can make a linear uncompressed DNG file, which maybe what you used.

The files can also be opened in ImageJ. The Sigma photo pro software is a pain to use, but does have the best artistic output.

The revived Photivo, may be the best free software to fully manipulate Sigma files though now.
https://photivo.bouc...s/Download.html

#31 Shane

    Member

  • Members
  • 324 posts

Posted 20 October 2018 - 05:56

I might have used a custom version of dcraw because back then the Sigma files weren't so well supported if I remember rightly.