Quantum Efficiency of Mono DSLR?

[glend]

I am having some problems finding a way to get a handle on the quantum efficiency gains of my debayered Canon 450D. I can find references to the stock colour camera's QE (which is listed as 33% by Sensorgen) but how do I translate the QE for a mono sensor in which every pixel is gathering photons without that bayer matrix blocking some? I read that a colour CMOS sensor is about 1/5 as efficient as a mono sensor. I would like to be able to plug my camera sensor details into something to work this QE improvement out. Pixel pitch is 5.2, APS-C sensor, 12.2 mega pixels. Sensor test data is here:

http://www.sensorgen.info/CanonEOS-450D.html

This sensor is cooled (cold finger behind the sensor itself) and runs at below 0C - where Darks and Bias Frames are equal when stretched. So thermal noise is not an issue.

The mod process removed the microlense layer so I know that has some negative impact on QE but how much and isn't this simply overcome by running longer subs.


:question:Any ideas on working this out?

[Merlin66]

Glen,
The method used by spectroscopists to measure the QE instrument curve, I'm not sure is easily available to you....it involves using a spectrometer measuring a known source and then by dividing the results getting a response curve for the camera (including the rest of the optical train -lenses etc.)
I'm not sure what other rigorous methods are available.....
http://www.astrosurf.com/buil/50d/test.htm

Re resolution... that can also be done very easily with a spectrograph - measuring the FWHM of nominated emission reference lines - the FWHM is a measure of the pixel size/ sampling etc.
(It should be in your case close to the actual pixel size - after the mono conversion)

[Shiraz]

agreed, in order to determine what proportion of photons you are detecting, you need to know how many are being made available to the detector. A standard lamp and calibrated monchromator could be used, but such a setup would cost maybe $10k+. You may be able to use the moon as a standard source (doesn't cost much :). If you wish, will try to work out a method for doing so? - it won't be super accurate, but will give some idea of what you have achieved. What filters do you have available?

[glend]

Thanks Ray and Merlin, I didn't realise it was so complicated. I have three narrowband filters I am using: Ha, OIII, SII. I just ran across some old threads on CN that postulate QE reductions due to removal of the microlenses offsets any gain in going mono. I think more testing is required.

[Shiraz]

I would have thought that going mono is definitely well worth the effort. Removing the microlenses may possibly reduce sensitivity by a factor of maybe 2, but the Bayer filter does worse than that for broadband sources - and if you use narrowband, only a subset of the pixels see any photons at all under Bayer filters (eg with Ha, only 1 in 4 of the pixels get any light at all). mono is much more effective overall.

[glend]

That is certainly what I am seeing Ray. I had thought that while the microlenses are good at funneling more photons into a pixel well, simply running longer subs and collecting more photons coming straight into the pixel would negate the absence of the microlenses. This was my test stack in Ha with the mono camera, just 10 * 360" subs at f5. I am building longer subs now and they look better.

http://www.astrobin.com/full/244755/0/

It's a start.

[luka]

If you had another 450D you could use the same light source for both and compare the total intensities. You could even use your narrow band filters to compare different wavelengths.

[rmuhlack]

Similar to luka's suggestion, could a comparison be done with another mono camera that has a known QE response . eg using a sub length that is within the linear range of both cameras,take a series of test shots, compare the resulting signal? if you combine with the narrowband filters, you would get relative QE at different wavelengths. Then use the known QE values for those wavelengths from the known camera to calculate the QE of the new mono DSLR.

Edit - if the pixel size of each camera is different, I suspect that this would also need to be taken into account.

[rustigsmed]

if pixel size matters you probably want a camera with the pixels as close in size to a 450d (5.4 microns) are;
- Basler acA2040-25 with 5.5 micron pixels
- 8300 chip with 5.5 micron pixels.

[luka]

I think that you can integrate the counts over the whole sensor size and not worry about the individual pixels for the measurements. Then you can work back (divide by pixel count) to get the per-pixel efficiencies.

1. Make sure that the lighting conditions are identical for the 2 cameras.
2. If the sensors don't have the same physical size the image from the camera with larger sensor must be cropped so that the two resulting images come from identical physical dimensions of the sensors. This will ensure that optical conditions are similar and hopefully identical.

And then repeat the measurement several times. With several cameras :-)