Jump to content
UltravioletPhotography

Microscopy of a monochrome converted camera sensor


JMC

Recommended Posts

I've been doing some microscopy recently, and a few days ago took a look at a camera sensor I had. This one had been partially converted to monochrome and was damaged, so regions of the sensor has different degrees of removal of the microlenses and Bayer filters.

 

There's a full write-up here showing images with different stages of removal of the filer and lens layers - https://jmcscientifi...-camera-sensor/

 

Here's a couple of images from the writeup. Firstly, part of the sensor which had not been touched (microlenses and Bayer filter still present).

post-148-0-14139000-1599297003.jpg

 

And now part which had had the microlenses and Bayer filter removed.

post-148-0-83207000-1599297007.jpg

Link to comment

Nice. Can you do that in UV or IR, to show how much the Bayer filters absorb?

 

In IR the green subpixels should absorb some, and in UV the least absorbing ones should be the red ones, followed by the blues.

Link to comment

Thanks Stefano. Not with the light source setup I have at the moment. This was done with a cool white LED. However I do have a tungsten filament bulb which would give out plenty of IR so could try that at some point. Above about 850nm all three Bayer filters should be essentially transparent.

 

In theory a HgXe lamp I have could be used for UV imaging, but there would be a number of issues there. The internal optics in the microscope setup would need to transmit/reflect the UV effectively in order for me to get the image. It may work for 365nm but not any lower (and certainly not below 320-330nm). I'd also need a different adapter to be made to be able to mount the light source properly and safely.

 

I theory a 365nm LED would also do the job, albeit with much longer exposure times.

 

It'd be an interesting experiment to try at some point though.

Link to comment

Thanks Stefano. Not with the light source setup I have at the moment. This was done with a cool white LED. However I do have a tungsten filament bulb which would give out plenty of IR so could try that at some point. Above about 850nm all three Bayer filters should be essentially transparent.

 

In theory a HgXe lamp I have could be used for UV imaging, but there would be a number of issues there. The internal optics in the microscope setup would need to transmit/reflect the UV effectively in order for me to get the image. It may work for 365nm but not any lower (and certainly not below 320-330nm). I'd also need a different adapter to be made to be able to mount the light source properly and safely.

 

I theory a 365nm LED would also do the job, albeit with much longer exposure times.

 

It'd be an interesting experiment to try at some point though.

 

That would be an interesting thing to see....

Link to comment

Above about 850nm all three Bayer filters should be essentially transparent.

Are you sure? Above 850 nm images are monochromatic, but the pink/magenta color is still there.
Link to comment
Stefano, yes, he is sure, he measured the responses awhile ago with his spectrometer! The writeup is on the board somewhere. The pink color is NOT there with raw images.
Link to comment

Thanks all. Hopefully more experiments to come.

 

Stefano, to help explain the IR behavior, I'll write a bit more later or tomorrow - I'm currently out and writing on the phone isn't easy.

Link to comment

Stefano, with regards to the IR images, I'll try and explain.

 

Cameras (at least the ones I've tested) tend to boost the red channel when creating the RAW file. This is because the internal filters over the sensor cut into the red channel quite a bit and reduce the red sensitivity. Therefore the camera boosts the red to try and get some sensitivity back. Hence the unprocessed images have that strong red cast. White balance would normally take care of this. Anyone - please if I am wrong on this please correct me, but this is my understanding.

 

EDIT - please see Andy's post below for the correct explanation of the cause of the red colour.

 

The filter transmission at those wavelengths is a different matter though. Above 700nm the transmission of the red, green and blue filters in the Bayer filter start to increase. By the time you get to 800nm you can see the rise of the green and blue ones here - https://www.ultravio...dpost__p__21122

 

Literature work has shown that at > about 850nm red, green and blue are all pretty much transparent. I expect that if I take an image of the Bayer filter at >850nm, and I white balance the image, it would come out grey, and the individual filters should look transparent. I doubt it would look exactly the same as the monochrome sensor image in the my first post though, due to the presence of the microlenses.

 

These sorts of images are quite complex to set up though, and I suspect that for IR and UV that level of complexity will increase to another level. When I have a few spare hours though, I shall see what I can do to try and look at the sensor in IR and UV.

Link to comment

Stefano, with regards to the IR images, I'll try and explain.post-94-0-54718400-1599333993.jpg

 

Cameras (at least the ones I've tested) tend to boost the red channel when creating the RAW file. This is because the internal filters over the sensor cut into the red channel quite a bit and reduce the red sensitivity. Therefore the camera boosts the red to try and get some sensitivity back. Hence the unprocessed images have that strong red cast. White balance would normally take care of this. Anyone - please if I am wrong on this please correct me, but this is my understanding.

That is actually not correct - the RAW file has the unprocessed sensor values and if you use a program like PhotoNinja and just uncheck the "color correction" box (and every other box) on a 850+nm IR photo you will see it is monochrome grayscale. It's the auto white balance that gives the red cast for the reason you mentioned, but with no white balance it's gray.

 

Here is a demo:

950nm filter ("NEEWER" but good enough for this purpose). With the RAW and color correction on the default, it looks pink:

post-94-0-54718400-1599333993.jpg

 

Without any white balance it's grayscale:

post-94-0-06350900-1599334042.jpg

Link to comment
Ah right, thanks Andy. I will correct my piece then. I hold my hand up to being a bit hazy on 'what was applied and when' by the camera with regards to creating the RAW files, so thanks for explaining it.
Link to comment

Fun times Jonathan.

I also do some reflective microscopy. But my microscope is just for transmission. The trick is the objective working distance. This also helps for better images.

Just align a bright UV flashlight, like the 365nm Nemo (I think you have one) on the side of the scope at an angle to your subject with a support. Or channel the light through a fiber and direct it very close to the subject surface.

Then if your working distance is great enough you can catch the surface reflection through the objective.

Try to keep some movement flexibility in the light source as you may need to optimize the angle. But this works really well.

Thus why I try to get objectives with great working distance.

Link to comment

Blimey, I've just re-read what I wrote yesterday evening about the red colour cast. It doesn't even make sense to me now. I have no idea where I was going with that - I've looked at hundreds of monochrome camera RAW files in Rawdigger, and every one is monochrome before doing any whitebalance. I must learn not to write forum posts after going out for the evening.....

 

I've just done some IR experiments, and need to process the images now. Will post something later.

 

David, thanks for the advice, I'm still learning about microscopy as I'm going along, and the images above were my 2nd attempt at reflection microscopy after building the setup. I still have plenty to learn.

 

A 365nm torch. Hmm, that is a bit of a sore point at the moment. I currently have 2 torches, but thanks to either eBay or a dodgy postal service (or a combination of both), no batteries and no charger. So, no 365nm torches for me for now.

Link to comment

I have 3 images to share. These were taken with 630nm, 800nm and 900nm bp10 Newport filters, on a multispectral EOS 5DSR camera. The filters were placed on top of the photoeyepiece. The same setup was used as for the previous images, expect I used a Tungsten light instead of the LED one (to give me some IR to play with). Images shown as monochrome, given the narrow band imaging.

 

630nm.

post-148-0-55151000-1599391751.jpg

 

800nm.

post-148-0-55270400-1599391756.jpg

 

900nm.

post-148-0-20056000-1599391760.jpg

 

On the 630nm and 800nm ones I have labelled where the different coloured filters are.

 

For the 900nm one I cannot identify the individual coloured filter as they all look the same (as expected). I suspect the 900nm image is much too dark (all the filters are transparent at this wavelength), but I was struggling a bit with exposure given the light source and camera sensitivity. Also, for some reason live view wasn't working for these, so there was an element of guesswork for alignment and focusing the IR - I had to remove the camera from the microscope for each filter change. Alignment was pretty good for the 630nm and 800nm ones but a bit off for the 900nm image.

 

UV ones will have to wait for a while, until I can figure out how to light them.

post-148-0-55151000-1599391751.jpg

post-148-0-55270400-1599391756.jpg

post-148-0-20056000-1599391760.jpg

Link to comment
Nice Jonathan, thank you for doing this. The Bayer filters definitely behave the same way at 900 nm. The 800 nm image is a bit odd to me: the reds are transparent (as expected), but the greens are more transparent than the blues, so the color should be a salmon-orange? I would have expected the opposite, with the greens being the least transparent, to give a pink-magenta. I may be wrong again.
Link to comment

Colin, yes the light comes in from above and then is either reflected from the surface of the subject, before being either transmitted or absorbed by any of the filters, reflected or absorbed by the sensor underneath, then passing back through any filters before going back through the objective and to the camera.

 

Stefano, according to the work I have done on Bayer transmission, I'd expect the greens to be more transparent than the blues at 800nm, which was what I observed. The final colour you are seeing in a photo is after any white balance has been imposed on the image, as Andy described above. This is likely why it looks more magenta than 'salmon orange'.

Link to comment

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×
×
  • Create New...