Monochromatic vs one shot color (OSC) CCDs
My thoughts and experiences in this interesting topic

I have seen many discussions about this topic in several web-forums and Yahoo groups (SXV, OSC). I am an enthusiastic Starlight Xpress user, being experienced with OSC CCDs like SXV H9C, SXV M25C and monochromatic CCD like SXVF H16. I have also had the opportunity to become familiar with and process SBIG CCD data coming from ST-2000XMC and ST 1301E cameras.
For better understanding of my point of view, here is what I use for deep sky imaging (DSI).
My current equipment used for DSI consists of:
9" f/9 TMB Apo and Starlight Xpress SXVF M25C one shot color camera as well as Starlight Xpress SXVF H16 monochromatic CCD .
(see my review of the M25C on my website)

(see my review of the H16 on my website)

After a couple of years using OSC I wanted to explore the alleged difference
in luminance resolution between an OSC and a monochromatic CCD and see if it
is truly as dramatic as so many people claimed it would (or should...?) be...
Quite frankly - it is not.
Terry Platt of Starlight Xpress was kind enough to loan me his demo SXVF H16 monochromatic CCD (that has really a very low dark signal) for this purpose

In my humble experience (I use the 9" f/9 TMB Apo mostly at around f/7) the difference in luminance resolution (pseudo-luminance in case of an OSC that is extracted in the channel mixer in PS CS2) is almost negligible, as I am working at an image scale of some 0.7 (sometimes 0.4) arc seconds per pixel under a night sky with average seeing and good transparency.

This is a comparison (click image to see full size for better evaluation) I made in april 2007 (info in the image - make sure to hit the enlarge button on the lower right).  The more over-sampling the less the difference – as a very rough rule of thumb.  Naturally this has certain limits – as you will require a very stable and reliable mount allowing guiding at this rate of resolution. A socalled pseudo-luminance layer that can easily be extracted from the OSC-data (see below) can then be processed quite aggressively and filters like deconvolution and wavelets can be applied like we are used to do that in a classical luminance frame taken with a mono CCD.

Well, what was missing from many threads was the discussion on the importance of the conditions of the observing site.  In my opinion, the advantage of monochromatic CCD lies mostly in its ability to achieve really deep color data with highest possible accuracy (besides the opportunity to acquire narrowband data). To take full advantage of this however, excellent conditions are required. When over-sampling comes into play, the OSC will easily catch up with the monochromatic CCD in terms of (pseudo)luminance signal, as long as you have average seeing conditions like I have most of the time
at my observatory’s place (northern regions of Austria). But it is harder to equalize the color data and in case of objects with a low signal it is almost impossible.  In case of very good seeing a monochromatic CCD will be able to outgun the OSC in luminance signal and therefore resolution, but never as much as so many people believe. In the cases of under-sampling (more than 2 arc seconds per pixel)  'Mono' will always have superior resolution, as the information of an object (say a star) being focused onto a couple of pixels will be summed up by 4 pixels when interpolating the Bayer Matrix data into the OSC's color.  Thus, you lose resolution as a consequence and stars will look less round; the starcolor will be false.

At my location, having 'back to back' nights to image an object is rare due to fickle weather. With an OSC I can "Make every frame count" for the final result, as it holds both pseudo luminance and color.

Another important point is the dark signal of the chip. In the case of severe noise (ST-1301e), faint detail will simply drown in the random and read-out noise of the chip.  This detail will therefore be lost (or take much more effort to bring it out with heavy processing), in comparison with a low-noise chip (lM25C’s Sony detector), even with the Bayer matrix. 
So, at least for the equipment used for this test (SBIG 1301e, SXVF-H16, and SXVF-M25C described), the "big advantage" of mono CCD in terms of superior resolution does not exist. The pseudo luminance (i.e. Extracted in Photoshop using the Channel Mixer to choose best s/n mix of r, g, and b components) will surely be superior to a (almost "loud" (-: ) luminance image taken by a noisier monochromatic chip.

So, simply buying a mono CCD is no guarantee of anything! Make sure to choose a CCD with good S/N numbers.  As stated, a large advantage of monochromatic CCD is the better color saturation achieved using 100% of pixels through a variety of color filters. OSC suffers a bit in color depth and color data of the stars. The way around this is longer total exposure time and longer subframes. In my case (f/7 and M25C) I achieve the best results with 10 minute subs (it was trial and error stuff to figure out the appropriate value) and at least some 3 hours total exp. time. (the more the better). OSC can do a decent job in objects with a strong signal, say bright clusters, very bright galaxies and bright planetary nebulae. In case of low signal OSC will always be significantly inferior in terms of proper color-information compared to a mono-CCD RGB data set.

Here is an example (click image for full size) that demonstrates how nicely star color comes up with these parameters. Certainly in narrow band imaging (Ha, OIII, etc.) an OSC will always fall behind.  Particularly in the case of an H-alpha filter, only every 4th pixel (Red) receives any signal from the filter’s narrow bandpass. That light is further attenuated by the red dye of the matrix over those pixels. I find that 6-hours total exposure with the OSC comes rather close to matching monochrome's 3-4 hours, but will never surpass it.
OSC will do much better with more total exp. time and will come rather close to an RGB, but never be equal or beat it; at least this is my personal experience, as RGB displays the colors in much higher accuracy.

L-OSC (please click HERE to see more)

is an alternative technique for combining monochromatic luminance and color data from one shot color CCD to save time...as OSC and monochromatic CCD do not necessarily have to be competitors...as you can combine the data of both.  I tried this and found the result of such a combination most exciting. Luminance data was taken with a H16 mono (to be "faster" and color data taken with M25C, to make sure to finish this object in one single night (as I indicated, I can not be certain to have another clear night with good conditions).

Finally, here is my recommendation for a person shopping for a CCD -- it would be wise to deal with some certain considerations before picking a CCD, mono or OSC.  Some of them are listed below:

- Conditions of weather and climate at the observing site (steady skies with good seeing and transparency - is the weather a reliable partner...? Will you be able to take full advantage of a monochromatic CCD)

- Photographic equipment (mount, focuser) stable enough for "heavy" CCDs like SBIGs?  S/N ratio of the chip used in the (monochromatic) CCD?  In cases of monochromatic chips with moderate s/n you might have to operate with dithered guiding mode -- will your mount support this?

- Focal length for Deep Sky Imaging – Nyquist sampling (arc seconds per pixel resolution).  Monochrome will produce good results at any resolution greater than say 2-4”. If choosing an OSC, I recommend your image-scale be at least 1 arc second per pixel, but better is to go for much more oversampling! (though many fine images by others have produced good results with as low as 2” per pixel). The lower the value the better the results in terms of resolution, star color and smooth star edges. A pseudoluminance can be processed very aggressively, like a true luminance layer taken with a mono CCD.

- Strong nerves and endurance for highly demanding pre/post processing in both types, but particularly monochromatic CCD, as completing an LRGB image takes more time, particularly if the s/n is not great...

- Budget...

Feel free to browse through my images and do not hesitate to contact if you have questions or comments.

Dietmar Hager