You are fairly new to this endeavor it seems to me, so you missed my rationale for the S90 choice which I had peppered over two huge threads on the SX110 and S90. The choice for my application took months of analytical work with these and S50 and S70 cameras. That was followed by massive support and effort to make S90 camera to work well and give predictable error-free performance for what I need, from the gurus of this forum to whom I am extremely grateful: CHDKPTP remote control, which is essential to my specific application, works perfectly due their intense dedicated effort, in addition to programmatic capability that I had described already in the INTRODUCTION on the first page here.
When I describe something I choose my words very judiciously. In my recent posts here, you'll find I refer to "dark performance," not "dark current." They are not all synonymous. Google lectures on the numerous sources of sensor noise such as 1/f noise, shot noise, fixed pattern noise, readout noise, among others like cosmic rays and pixel volume (size & depth), in terms of photoelectron-to-current transformers such as floating diffusion amplifiers and how they relate to signal-to-noise-ratio (SNR) in an imaging system, and try to understand well how correlated double samplers work, and how their own different architectures have an impact on SNR of the final product. Dark current is a measure of photoelectrons/pixel/s, it is temperature sensitive and is only part of the total noise in an imaging system.
Quote "The cmos cameras i've checked were better at dark current than the ccd cameras that i've looked at"
If by "dark current" you mean SNR under long exposures, you're trying to compare apples and spaghetti with the only similarity is that they are both foods. When analyzing two different cameras, electrical light collection conditions (exposure time & VGA amplifier gain) have to be *exactly* the same, CCD sizes must be exactly the same (if not, you have to compensate by modifying the flux), and the flux falling on the CCD must be identical, otherwise your comparison is of no use. When done right, even one CCD camera can have very different dark performance from the next CCD camera.
Quote "I'm curious why you think your camera has a low dark current? Can you post a ~5 minute darkframe (or better 2 to get an idea of the readout noise) at a low iso setting in DNG format?"
You cannot measure "dark current" independently of "readout noise." Both are (and others) always present and combined. On the PowerShots I've tested, you can't set a 5 minute exposure, only 15s. CHDK extends it 64s, but you run into other limitations characteristic of small-format CCDs (1/1.7" in the S90) that I have covered in my other threads.