I'm one of the astronomers who put the thing together - thanks for the interest!
We use standard SBIG off the shelf astronomical cameras as detectors. The lenses are fast enough, and the integration times long enough, that read noise is negligible even with only modest cooling of the detectors.
We've published a paper that describes the thing; this is the publicly available version, for those who are interested:
As you may notice we're mostly concerned with scattered light; we measured the point spread function out to ~1 degree (Fig 6) and found that it is amazingly well controlled (this was recently confirmed by another group, who compared our results to a wide range of other telescopes).
We've had some science results out this year, too - and we just put out a press release on the discovery of seven very faint galaxies (which might be why CR posted it today!). The reason why we went with Canon is that I'm a Canon shooter (with a special interest in dragonflies..) and I was aware of the quality of the updated lenses. I know lenses and Bob knows telescopes, so it all worked really well.
Other groups have used Canon lenses for astronomical purposes, but typically just to cover a wide area of sky - not to detect very low surface brightness emission, beyond the reach of reflecting telescopes.
Anyway, sorry for rambling on - it's a fun project!
Your results evaluating the lenses are quite interesting. I am not sure they are correct, however. I have measured 6 of these lenses under photopic light at an aperture of f/4, and even then they are nowhere near the diffraction limit. I have little to no doubt they are designed this well - but realizing such a performance in large scale manufacture, when I have measured to the contrary, leaves me skeptical.
I am not familiar with the technique you have used to measure the wavefront error, but judging by the use of defocused images, it seems to measure only wavefront curvature, not tip/tilt, yes? I am then confused how your synthesized interferogram is dominated by tilt, if the test would be unable to see tilt.
Further, your defocus is quite enormous, and I would suspect that the very large defocus term you have added to the zernike polynomial may be masking other aberrations. However, I am not familiar with the method, and perhaps it is robust with respect to that parameter. It seems that by increasing the size of the defocused spot, you will increase the spatial resolution of the reconstructed wavefront, but decrease the sensitivity of the test.
Are you able to describe in more detail the workings of this method? It intrigues me, but I have my doubts as to its accuracy.