Mono or Bayer - sensitivity comparison

[Shiraz]

Just tested the relative sensitivities of my QHY5 cameras, one colour and the other mono. Might help quantify the difference in sensitivity of a Bayer matrix camera when compared to a mono camera.

test conditions:

source - white paper illuminated primarily by skylight from overcast sky, viewed through a refractor and defocused.
cameras tested alternately and results checked by one repeat - identical gain and exposure settings were used.
an IR/UV cut filter (luminance from an LRGB set) was fitted to the cameras when tested (same filter used on both)
images were analysed in Nebulosity, with 2x2 software binning to average out Bayer filter effects and reduce fixed pattern and read noise.


Results:

the mono camera detected 2.3x as many photons as the Bayer camera.
the signal to noise ratios were in proportion to the square roots of the signal, so camera noise/offset was probably not a major factor.


Discussion: the Bayer matrix does not divide the light up into three equal and precise spectral regions, but does so with some overlap. It appears that this, in concert with the green bias inherent in the 1:2:1 matrix results in higher sensitivity than would be expected from a simplistic division of the input into equal thirds. This test assumes that the underlying MT9M001 chips are the same for the two cameras and that the QHY software sets them up identically - I have no way to check this. The spectral distribution of the test source was not known, but it should have been fairly representative of real targets. I also have no way of knowing if the Bayer filters used by Aptina are similar to those from other makers, but assume that they are not too different and that these results may be more generally applicable.

had previously assumed that a Bayer camera would be about 0.33 times as sensitive as a mono one. That is incorrect - in this test the Bayer QHY5 was better than that, being about 0.43 times as sensitive as the mono counterpart.

thanks for reading. Regards Ray

[RickS]

Interesting experiment. Thanks for that, Ray!

Cheers,
Rick.

[MrB]

Nice work Ray.
Don't forget there are two green for each blue and red in a Bayer matrix.
I couldn't find a colour data sheet for the MT9M001, but another Aptina MT9 sensor data sheet shows a QE of ~38% for red, ~43% for blue and ~50% for green.

Oops, I didn't see you had addressed this in your OP, I may have missed your update.

[Dennis]

Good work and analysis Ray, thanks for this interesting and useful experiment.:)

Science at home eh, who needs a certified laboratory and calibrated equipment! Top stuff.:)

Cheers

Denis

[nebulosity.]

Cool experiment Ray, I reckon you would be pretty correct, about the same as what I roughly measured with my colour and mono 350D sensors.

Jo

[Astro_Bot]

Thanks for that comparison, Ray. Very handy. :thumbsup:

[bert]

I think you may overstating the sensitivity of the mono chip.

You didn't take into account light loss when shooting through colour filters (rgb) as opposed to the constant loss via the Bayer matrix on the colour filters.

There would be a few ways to test this.

Brett

[Shiraz]

thanks Rick - obviously cloudy down this way. have also been messing around with star shapes, trying to find a way to get around the halo problem from your post - interesting problem in PI.


Thanks Simon - yep, added that as an update/

Thanks Dennis. Nowhere near a controlled experiment, but probably worth doing.


Thanks for the confirmation Jo - where did you get a mono 350D from?


Thanks - was interesting to do


Thanks Brett. Agree, this was just a test of the relative sensitivities in detecting photons from a bright broadband source - ie, obtaining luminance (and colour in the case of the OSC). Colour data will require extra exposure for the mono system, but mono allows binning to be used, so it is difficult to compare directly with Bayer. It is probably reasonable to say that a cooled Bayer system will require something like 1.5x as long to produce a similar overall SNR result as a cooled mono system with filters etc. And then there is the resolution reduction inherent in a Bayer matrix, but that is a different issue.

One often sees very high quantum efficiencies quoted for Bayer cameras and I wanted to confirm that these do not translate into broadband sensitivity. However, it was interesting to find that OSCs are actually a bit better than might be expected from a cursory understanding of how they work - must get my QHY8 back under the stars.

Regards Ray

[nebulosity.]

Just a standard 350D sensor from which I had removed the Bayer CFA, quite a risky process but possible if your careful.

Jo

[MrB]

I had planned to do this to an 1100D, bought two bodies but still haven't tried as it is apparently impossible with the 1100D's sensor, at least, no one has been successful yet to the best of my knowledge. There are however many stories for dead 1100D sensors. Shame really as the 1100 is a great body.

Sadly I missed out by minutes on Forgey's bargain sale : http://www.iceinspace.com.au/forum/showthread.php?t=118192
Those sensors would have been my CFA removal guinea pigs :(

[nebulosity.]

Have you seen the monoing thread over at SGL? I believe a fella recently managed to monofy a 1100D, the only problem appears to be getting the cover glass off the sensor, not sure what he did.

Jo

[MrB]

Hi Jo, yes the SGL threads are where all the action is, I've even bought a hot air rework station for removing the cover glass, but the results are not great when compared to earlier model sensors and the failure rate is too high for my liking. That's why I was super excited when I saw Paula's ad and I PM'd as quickly as I could, just not quick enough unfortunately :(
I hope whoever bought them does something useful with them, hate to think they're just going to gather dust as a good-intentions project that festers.

For now I'll be modding and cooling one of the bodies, but debayering will have to wait a while.