I have been using colour cameras since I began doing Astrophotography. This is the usual way for people to start in the hobby and many people never try the other type of camera. To astronomers they are known as "One shot colour" or OSC cameras. That other type of camera is the “Monochrome” camera, with which one applies different light filters to obtain colour images.
Back in December 2020 I purchased some used Astro junk from an Astro junkie. He’d been doing astrophotography since the days of film cameras and had stories to tell about nearly freezing to death trying to find clear dark skies in the depths of a Canadian winter. He dropped off the precious package and took a look at some images on my laptop. First he said “Beginners shouldn’t have such good images”, then “How did you keep Alnitak (a bright start in Orion) from blowing up?”. Finally he turned to me and said “It’s time for you to get a mono camera”.
Figure 1 shows the efficiency (percent of photons that should produce signal in the CMOS sensor) of the ZWO ASI6200MC colour camera. Sensitivity of the Red, Green and Blue channels as well as the total are shown. Blue is on the left, green is in the centre, and red is on the right. The yellow line is the sum.
My colour camera uses a standard “Bayer” matrix to create the colour sensitivity. The sensor is covered in a grid of different filters, with a set of four pixels having the pattern RGGB. If a red photon hits the G or B pixels it is rejected. If it hits a red pixel it will likely (a likelihood depending on the filter characteristics and sensor’s efficiency) register. A green photon is more likely to register in a colour camera than a red or blue – this roughly matches human perception and problems can be dealt with in later processing.
However we want a value for red, blue and green for every pixel, not just the ones with that particular filter over it. For a pixel with a green Bayer filter the blue and red intensities are determined by interpolating from the neighbor pixels which are sensitive to those colours. For my main camera each pixel is a square, 3.6 microns on a side. What resolution results from this interpolation process? We can certainly say that the pixels are not independent.
A ”monochrome” camera has no Bayer matrix and cannot separate colours on its own. To capture the red spectrum one uses a filter in front of the camera which rejects all wavelengths except red. One then has to separately deploy green and blue filters. Finally the three are combined to create a colour image. My fear is that if I were to try this I’d get an hour of “blue” colour and then the clouds would roll in for a month.
In spite of my fears, I decided to investigate…
My “super” camera is a ZWO ASI6200MC Pro colour camera. It is advertised as having a very high quantum efficiency of over 90% and produces excellent images from a full-frame CMOS sensor (Sony IMX455). The sensor also has a protective window on top of it, which cuts out some unwanted Ultraviolet and Infra-red light, but also cuts back on the visible light as well. The camera has a Bayer matrix, rejecting half the green photons, and three-quarters of the blue or red (RGGB = two greens, one red and one blue out of four pixels).
Finally the Bayer matrix light filters have imperfect passbands for these colours, causing addional loss of photons. There is one filter that passes red light, but that filter's response is imperfect and it may block some red light and pass some other frequencies (green or infra-red, most likely). This is shown below in figure 4. Similarly the two green filters are imperfectly passing green light. Of four pixels two will be covered by the imperfect green filter, one by the "red" filter and one by the "blue".
Conveniently there is a near identical Mono camera, the ZWO ASI6200MM Pro to compare this with. It has the same quantum efficiency but a slightly different protective window. There is no Bayer window and no equivalent loss.
Note that a colour camera produces three colours at once, while the mono will produce one colour at a time (or no colour, just luminosity).
For the colour camera we need to look at several factors:
- Sensor quantum efficiency as a function of wavelength.
- The protective window’s spectral response (percentage of light passed as a function of wavelength).
- The three passbands of the Bayer filters.
- The fact that for four pixels we have two with the green filter's passband and one each with the red and blue passbands. The overall efficiency would be the sum of the green efficiency multiplied by two, the red and the blue, all divided by four.
I started this work by manually digitizing the quantum efficiency curve of the sensor, which is the same for both cameras. This efficiency curve came as quite a shock. The sensor efficiency is highly variable across the visible spectrum, and is mostly far below 90%. I’d imagined a moderately flat curve across the frequencies of interest, but it is far from that. This curve is shown in figure 2.
Figure 2 shows the quantum efficiency of the sensor found in the ZWO ASI6200MM Pro monochrome and ZWO ASI6200MC Pro colour cameras. Visible light is approximately from 400nm to 700nm - wavelengths outside this range should be ignored (also see Figure 3) for the colour camera.
In these cameras there is a piece of glass sealing off the sensor. It keeps the sensor clean and is used to filter some unwanted light. For the colour camera the “protective window” excludes wavelengths in the UV and IR.
Figure 3 shows the spectral response of the protective window used in the colour camera. Wavelengths lower than 400nm or higher than 700nm are completely blocked.
Naively one would think that the RGB filters in a “super” camera would be nice and neat, with no photons getting ignored or potentially appearing as two colours. Not so! There is much overlap between these filters and the spectral response is certainly not flat anywhere. There is also a real drop-off in sensitivity in the lower wavelengths (i.e. Blue). Yikes!
Figure 4 shows the RGB frequency response of the Bayer filters used in the colour camera.
Armed with all this information we can determine the probability of a photon being registered by this camera. There is some overlap: Photons of these specific wavelengths might be registered as green OR red. The quantum efficiency varies widely across the visible spectrum: Lots of photons are lost and don’t register at all. This image is shown at the top and again below.
Figure 5 shows the efficiency (percent of photons that should produce signal in the CMOS sensor) of the ZWO ASI6200MC colour camera. Sensitivity of the Red, Green and Blue channels as well as the total are shown.
This graph was produced by MS Excel, and the line colours are not the spectral colours they represent. On the left is blue, green is in the middle and red is on the right. The yellow line is the sum of all three, for later comparison with the Mono camera. If we summed up all the colours and made a monochrome camera out of it we would have the response of the yellow line.
It’s a mess. The peak efficiency of the red and blue are well below 20%. The green peaks higher (RGGB has two greens) but is still below 40%. The red and blue are well separated, but the green washes into both colours. The peak efficiency, even summing the three colours is below 50% - most photons are lost.
A second posting will look at the monochrome version of this camera and see how it looks when used with no filters or with red, green and blue filters.
Comments