Stock Canon cameras - internal filters leaking UV

[JMC]

I had an odd result in a test recently. While doing some fluorescence imaging, I got something which looked like red specular reflection (the camera was a stock Canon with the internal filters still in place). I realised that I'd forgotten to put a long pass filter on the camera lens, but thought it may have been down to some red contamination from the light source.

 

However this was where it got a bit weird. If I put a 420nm long pass filter on the lens, the red specular reflection disappeared. If is was red light just reflecting from the subject this would not be removed using the long pass filter.

 

It turns out that this red reflection is due to the UV. The internal filter setup in my Canon 5DSR doesn't block all the UV - there is a double peak with quite a large transmission at around 370nm. Given the sensitivity of the sensor at that wavelength, this looks red in the image.

 

I've done quite a lot of work on this now to try and understand it a bit more, and have seen the same effect with my Canon 6D, which makes me think that it's not just an issue with the 5DSR. I'll share my findings on here, when I've made the graphs and pictures a bit more neat and tidy.

 

Has anyone seen this type of effect before?

[JMC]

Here's the transmission spectra of the 3 filters from the stock Canon 5DSR. These are the ICF and hot mirror which is normally removed during a multispectral conversion, and the OLPF filter from the sensor (which I had removed from one of my 5DSR's when I have it converted to multispectral). These were measured using my Ocean Optics FX spectrometer. I've also included a line which is all 3 multiplied together, to show their combined behaviour.

 

 

There is a significant UV leak in the ICF filter, which overlaps with where the other 2 filters still transmit UV. The end result is quite a sizable leak in the UV, with a small peak at 365nm (about 1%) and 377nm (about 5%). The Bayer sensitivity is quite high for the red channel in this region, hence the peak looks red.

[dabateman]

I don't have that sensitivity to lnow. But my Nikon DF, stock looks horrible in UV. So I don't think that is happening with its filter set.

The Panasonic GM5 I have also is very poor at transmitting UV. But slightly better than the Df. I don't think it has a peak like that as its mostly blue (upper 380's and 390's) getting through.

The Olympus cameras leak a whole bunch of UV and can be used without conversion at and above 370nm.

 

Canon have their own sensors and stacks. So maybe doesn't affect Sony based sensor cameras.

[Cadmium]

Which seems to compare to some of the camera hot mirror graphs shown here:

https://kolarivision.com/articles/internal-cut-filter-transmission/

[dabateman]

The lack of blocking filter on the Pentax 6Mpixel *stD and *stDL look great for 2003 and 2005 cameras. I wonder how that performed.

There seems to be slight bump on the Panasonic GF1 which I might be similarly seeing on the GM5. Its not good for UV, but leaks just a little.

 

Graphs like these today would help selecting a good camera for just UV induced fluorescence.

[Cadmium]

My first UV/IR converted camera was the Nikon D90, which looks like it would leak a lot of UV in stock version.

However, it was probably like when I use BG38 on my UV/IR converted cameras as a UV/IR-Cut filter for visual, even though BG38 doesn't cut or block UV,

the UV doesn't show up in visual because the visual is so overpowering, not permitting enough exposure time for the UV to show up in the mix.

I still prefer BG38 on all 3 of my UV/IR cameras for visual, it has the perfect blue/red balance for me. I prefer it to Baader UV/IR-Cut which looks too red for me.

However, doing UVIVF, with a stock camera that leaks more UV than others may be a problem unless you have the right barrier lens on the lens.

I like the UVIVF shots I get with my stock Nikon D610. I really like the D610, although mine is just stock, not converted.

[JMC]

Just as a quick example of the effect, here's some images of a small glass vase.

 

Lighting was my Hamamatsu LC8 light source, with a xenon lamp, and filtered using a Baader U filter (I just realised I do not have a scan for this, and must rectify that). The vase was sat in a cardboard box which was painted with Semple Black 2.0 paint.

 

I imaged this with a Eos 6D, using 2 different lenses (Rayfact 105mm UV lens, and Nikon 105mm f4 macro lens). Images done at ISO6400, and 0.5s exposure, f11 and saved as Raw and jpeg in the camera with daylight white balance. The images shown here are from the jpegs and have been resized for sharing.

 

Firstly, the images with the Rayfact 105mm UV lens.

 

1. 420nm long pass filter from Klaus

 

2. Firecrest 400UV filter

 

3. No filter on the lens

 

4. LaLa U filter from UVIR optics

 

The 420nm long pass filter from Klaus is a bit of an unknown, as there isn't much information on it. It does however appear to have a coating on it, it does not fluoresce under UV, and it has very good blocking of the UV. The Firecrest 400UV filter leaks slightly at around 370nm, and you can see a slight red cast on the image. With no filter at all on the lens, the red is obvious. The light source is very powerful, hence the small amount of reflection from the Semple Black paint can be seen. With the LaLa U on the lens, the fluorescence is now blocked, and all that is left is the UV reflection.

 

Ok, next set of images same exposure, but this time using a Nikon 105mm f4 macro lens instead of the Rayfact. All other settings the same.

 

5.420nm long pass filter from Klaus

 

6. Firecrest 400UV filter

 

7. No filter on the lens

 

8. LaLa U filter from UVIR optics

 

With the Nikon 105mm f4 lens the fluorescence can still be seen (as expected) and is similar to the Rayfact. The red cast from the Firecrest 400UV is still there, but less obvious. Removing the filter from the lens, still gives the red, but is less than with the Rayfact. The LaLa U still gives an image corresponding to the red areas in the shot with no filters, but the brightness of this image is lower than with the Rayfact (as expected as the Nikon 105mm will let less UV through).

[Cadmium]

Jonathan, Great tests.

[Andrea B.]

I'm not particularly surprised that some stock cameras can pass some UV.

 

If you look at filters sold as UV-blockers (used to help with distant "haze"), they typically are not blocking everything in the high UV range. I suspect some of those were used initially in older digital cameras because a bit of UV passage is not particularly damaging to a visible photograph. So I suppose that is also why more expensive, steep-shouldered, hard-cutoff at 400nm filters are still not always used internally in stock cameras.

 

Jonathan's experiments reinforce the need to always filter the lens in UViVF work. Internal UV/IR blocking is still not "perfect".