Jump to content
UltravioletPhotography

Split sensor image - monochrome/multispectral at the same time


JMC

Recommended Posts

A couple of weeks ago I found a guy on eBay selling a monochrome conversion of an EOS 5D for not much money (mainly because it had a line of dead pixels in it), and I reached out to him and asked him about the conversion. He'd done it himself and mentioned he had done a few, and was now just trying to recoup some of the money he had spent on sensors and cameras. I got to wondering whether he could do me a custom modification - half and half, remove the Bayer filter/microlenses from half the sensor and leave the other half intact. Well, unsurprisingly he had never done one before but was willing to have a go on an EOS 450d he had. A few days later he lets me know it's done, and send me a sample image, and we agree a price. Then this morning it arrives....

 

I took it out in the garden and got a few test shots, some plants, a sunflower and a PTFE tile. These are RAW composite images from RawDigger, screenshotted and saved as jpegs. Eos 450d half and half, Asahi UAT 85mm lens, f8, ISO800, Baader U filter and lens hood.

 

Hedge, 1s

post-148-0-68044900-1537185319.jpg

 

Hedge, 1/4s

post-148-0-13876500-1537185344.jpg

 

Sunflower, 1/3s

post-148-0-85187800-1537185321.jpg

 

Sunflower, 1/13s

post-148-0-32791000-1537185345.jpg

 

PTFE tile, 1/80s

post-148-0-79001500-1537185320.jpg

 

It's certainly a little rough around the edges, and with a few scratches, and the focussing is slightly off - there is no WG280 or spectrosil over the sensor, which I presume has an impact on the final focussing of the image. I must learn to use live view focussing again. The aim here was to look at the effect though, and not to use it as a fine art camera.

 

The PTFE tile shows the increased light hitting the sensor in the monochrome side quite well. I took the RAW image of the PTFE tile and put that in Rawdigger, and measured the RGB values in 2 regions; monochrome, Bayer filter + microlenses, and then in between where the microlense have been removed but the Bayer filter remained. Plotting them out I got this;

post-148-0-88132800-1537185844.jpg

 

The average response for the monochrome side is about 4.5x that of the multispectral side in this case. Interestingly the area between the 2 sides, is where the top microlens layer has been removed and the Bayer filter remains. The sensitivity in that region for the Bayer filter alone, is about half of that of the Bayer filter and microlenses, so it looks like even in the UV the microlenses are doing their job and are not absorbing the UV. The Bayer filter is very highly absorbing in the UV, even more so than we thought before.

 

So, camera manufacturers please make us a monochrome sensor with microlenses on for our UV work :)

 

Oh, and a final picture, the sensor itself.....

post-148-0-64211200-1537185843.jpg

Link to comment
Oh this is fascinating. A camera with a split personality! What I would really like is a Bayer that is designed for UV. It's pretty clear from your and others' prior work that two of the channels basically follow almost the same curve, so we are not getting the full range of colors possible in the 350-400nm region we use most often. A properly designed UV Bayer would fix that.
Link to comment

That is very interesting. Its amazing that you have monochrome, CFA, CFA+Micro lenses all on the same sensor for comparison!

 

The SD14, does not have a color filter array, has micro lenses, but clearly has different type of silicone sensor. I tested it with my mercury vapor lamp and 313bp25 and I see nothing. Its all noise and black even at 2 minutes shutter speed. My converted EM1 has much better response. So can not easily compare with and without CFA as you can here. I do get signal though with the 340bp10 with the sigma. So that seems to be the lower limit for that sensor.

 

I have a feeling Sony (whom made the sensor in my EM1) is doping there sensors with sulfur to improve response. Sigma, I don't think is at all. Canon may be doing the same.

Link to comment

Steve, white balancing based on the colour images does some very odd things to the monochrome image. I balance in Darktable, and if I balance a whole image using the PTFE tile information for the multispectral image, I get as an example, the following (monochrome on the left and multispectral on the right);

post-148-0-37724800-1537256053.jpg

 

Or for the sunflower;

post-148-0-92017400-1537256195.jpg

 

So, maybe I need a split white-balance instead of one for the whole image :)

Link to comment

The monochrome White balance makes sense based on your graph. The red channel is almost 4 fold more responsive.

 

What filter did you have on the camera for the images?

 

This may actually be a new style in photography. Picking up its roots from spot filters and half macro lenses. With the right composition you could do some interesting things.

Link to comment

I think it is really interesting. I thing you are the only person on the planet that has done this maybe.

Yes, the white balance doesn't work for both at the same time. Guess you would need to white balance each side separately.

Also, the monochrome side is more overexposed than the Bayer side. Not sure how to expose the overall image optimally for a post adjustment to each side,

I guess maybe just half way between each side's best exposure?

I may have missed something above, I do that... but how are you setting exposure for these so far?

Link to comment
I set the exposure by guesstimate Steve. I was thinking portraits using direct light on one side of the face and slight shadow on the monochrome side. We shall see, away with business at the moment so it'll have to wait for now.
Link to comment
Turn on spot exposure. Measure exp for relevant area on left, then on right, and then average the two. Old fashioned way of determining exposure. "-)
Link to comment

Jonathan,

You asked for a Monochrome camera with microlenses for UV photography. It might exist and be affordable. I came across the ZWO 1280mm monochrome camera. Its intended for Astrophotography, but I don't think any thing will stop you from using it for regular UV work. There seem to be 3 versions. Regular 1200mm, a cooled version with fan to cool the sensor down and a cooled version called Pro which adds a 256MB internal ram chip for faster movies. The thing uses a Panasonic 43rds monochrome sensor.

https://www.highpointscientific.com/zwo-asi1600mm-pro-cooled-monochrome-astronomy-camera-asi1600mm-p

 

And old UV photography forum user has one, Boon, and has written a couple reviews:

 

http://myphotojourney.co.uk/zwo-asi1600mm-cool-uv-imaging/

 

It also has accessories to add dead m43rds mount or canon EF mounts. Although the m43 to tmount accessory doesn't make sense to me, the flange with it is 6.5mm, where as the m43 system uses 19.25mm. So I think a tube would still be needed. I also see that RAF camera has a female tmount to female M42 mount which may work better with this camera. Then with a helicoid any m42 lens could be used.

If I could figure out how to directly run it corded from my Android phone, I may be interested.

Link to comment

Thanks David.

 

As an aside, I have been doing some experiments using this 'half and half' camera on my spectral sensitivity rig. Because there is an area in the middle of the sensor without microlenses, I have been able to measure the spectral response of just the Bayer filter (with no microlenses) all the way down to 280nm. It looks like the microlenses are doing their job in the UV all the way down to 280nm, and are not blocking the UV.

Link to comment
  • 2 weeks later...

Yes, I was wondering if you could use your fiber to target the three regions of the sensor to get an idea of the response. Seemed mentally tricky to me. Also if your sparticle individual filters may be better, lined up over the region to get an idea.

 

I was reading that the quantum efficiency of CMOS is low and that microlenses are required to boost the sensor to a usable level. Thus why CCD was preferred for so long, until only recently when CMOS overcame these limitations. There does also seem to be major differences between the Canon and Sony sensors. Canon seems to use a glass or ceramic CFA layer and glass microlenses. Sony sensors seem to use plastics.

 

I found interesting that the ZWO ASI 1600mm sensor, I mean to be talking about above. It costs $1280 cooled, which I have switched in the name or $900 not cooled. Its the Exact same sensor as in my Olympus Em1, the Panasonic 34231 (MN34230). So I could, if I purchased one be looking at with or without just the CFA layer. Also it can be controlled directly from an Android phone using ASICAP app from the play store. You just need an OTG phone, tablet, or you can use a computer.

 

However, I find the dark noise results striking for the sensor and surprised they would use it.

 

http://www.wrotniak.net/photo/m43/em1.1-5.1-dark.html

Link to comment

David, that sentence of your, "Canon seems to use a glass or ceramic CFA layer and glass microlenses. Sony sensors seem to use plastics.", is really interesting. I had assumed that standard consumer cameras would be pretty much the same. Can you share where you got that from please.

 

Oh, and it was tricky, measuring all three areas. Very tricky..... The Sparticle approach would be hard, as the filters are fixed in place (glued in place) so it would need a big redesign of my setup to make the work.

Link to comment
Thanks David. I must admit, I do struggle with the idea that Canon has a glass or ceramic CFA though, just seems unnecessary. I wish, I wish, I had still had access to my PhD lab - a wonderful array of surface analysis equipment, and I could have a definitive answer in a day. Oh well....
Link to comment
  • 4 weeks later...

Hi Jonathan

 

The color corrected sunflower look amazing :-)

 

Can you also try to take some photos in VIS and IR and show them in BW, so we can see that impact CFA has on BW photos

and where the loss of the microlenses outweigh the removal of the CFA

Link to comment

Hi Jonathan

 

The color corrected sunflower look amazing :-)

 

Can you also try to take some photos in VIS and IR and show them in BW, so we can see that impact CFA has on BW photos

and where the loss of the microlenses outweigh the removal of the CFA

 

Hi Mark, Sure, I have taken an IR image for you to show the effect. Taken with a 28-135mm zoom at 28mm, with a Heliopan 715 filter. Typically the sun went off the Magnolia tree after I set up the camera, but the lighting was even across the width of the tree, and the sky. Monochrome side on the left, Bayer filter side on the right.

 

First ones are white balanced on the monochrome side;

 

White balanced mono side

post-148-0-83103400-1540568980.jpg

 

White balanced mono side, then image desaturated in photoshop

post-148-0-38071200-1540568978.jpg

 

Second pair, white balanced on the Bayer filter side

 

White balanced Bayer filter side

post-148-0-28939500-1540569038.jpg

 

White balanced Bayer filter side, then image desaturated in Photoshop

post-148-0-10312400-1540569036.jpg

 

Looking at the desaturated images the monochrome side is darker than the one with the Bayer filter/microlenses, which makes sense as the Bayer filter is becoming more transparent in the IR, and the microlenses have been removed along with the Bayer filter on the monochrome side.

 

If you look at the line down the middle, that is where the microlenses have been removed, but the Bayer filter remains. It is the darkest part, which again makes sense - the microlenses have been removed, and the Bayer filter will be absorbing some of the light even in the IR.

Link to comment

Thanks Jonathan

 

These photos really help me with my decision on where to convert my camera to B+W or not

for the IR it look like we lose 1 stop do the the missing micro lenses

and for UV we gain 2+ stop as CFA is blocking UV

so my guess for VIS that the loss of the micro lenses is compensated by the lack of CFA

I do wish that it was possible to re-apply the micro lenses :-

Link to comment
No problem Mark. Yes, it would be great to get those microlenses back - my recent work is telling me that even at 300nm they are still working and providing a benefit for UV imaging. In the IR the Bayer filters become more transparent at longer wavelengths, so losing the Bayer filter does not make the sensor much more sensitive, and losing the microlenses makes it worse, hence the drop in IR sensitivity. If you need as much UV sensitivity as possible it makes sense. If not, then the benefits of having independent 3 colour channels wins out in my mind. Of course, if you just like monochrome imaging it also make sense - no need to worry about white balance any more.
Link to comment

Jonathan, how were you setting the exposure in the magnolia photos? Thanks.

 

I would like to be able to compare the two sides by looking at two photos: one with exposure set for the left side and one with exposure set for the right side.

Link to comment

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×
×
  • Create New...