This is just a brief placeholder review. Gary Walker loaned me his ASI2600mm-pro camera, but it has a non-functional sensor temperature readout and so can't regulate the sensor temperature. He is returning this camera for repair, but will be sending me a second ASI2600 for testing. When I get that one, I'll do a more extensive test.
The ZWO ASI2600mm-pro uses the same IMX571 sensor as does the QHY268M. The differences are how the unit is packaged, and what parameters are available for the user to control. The 2600 has a very similar look and feel as their earlier 183/1600/etc. cameras - red anodizing, cylindrical (90mm diameter), with a 2-port USB2 hub for accessories. It uses USB3 to communicate with the host computer, and requires a 12V 3A power supply (not supplied) for the TEC. Like with the QHY camera, there is a tilt ring on front to make sure the sensor is perpendicular to the optic axis, and dew heater resistors near the entrance window. The back focus is 17.5mm, including the tilt ring. A female M42x0.75 thread on the faceplate attaches the camera to the telescope or filter wheel.
The driver only gives access to one of the multitude of readout modes for this sensor (QHY gives 4). However, the mode ZWO enables appears to be the same as the "High Gain Mode 1" of QHY. It enables selection of gain and offset for that mode. This readout mode is what I prefer in any case. For largest dynamic range, setting the gain to 0 and an offset in the 10 range should be chosen. Like on the QHY268M, there is a changeover from the low-gain amplifier to a high-gain amplifier, in this case around a gain setting of 100. I only tested this camera at a gain of 0.
At a gain of 0 and offset 10, the system yields 0.77e-/ADU of system gain, and a read noise of 2.80e-.
When binning, QHY sums the native pixels to give the superpixel value. Since pixels are limited to 16-bit size, this means that each native pixel of a 2x2 bin on the QHY would be limited to a count of 16383 before saturating the superpixel. This means that is a 1x1 bias level is 500, a 2x2 bias level is 2000. For ZWO, native pixels are (I think) put in 12bit readout mode, summed, and then scaled to fit the 0-65535 range. This means that if a 1x1 bias level is 500, a 2x2 bias level is also 500. This preserves the full well depth of each native pixel, at the loss of some dynamic range. It may be that the 2x2 mode is done in true 16bit mode and then divided by 4. The documentation is not clear, but I should be able to figure this out before the updated post. A better solution for both QHY and ZWO is to image in 1x1 mode and then sum images after the fact in floating point before storage. However, this complicates the acquisition step, so there is no simple solution.
Given that, binning the ASI2600 2x2 yields system gain of 3.08e-/ADU and read noise of 6.63e-. The gain is exactly 4x the 1x1 value. However, the system noise should be sqrt(4) = 2x the 1x1 value, or 5.60e-. Instead, I measure a read noise of 6.63e- in 2x2 binning mode. This is still very good (and less than almost all CCD read noise values), but larger than theory. I need to study this difference!
I'll expand this post when the replacement camera is tested. I'll also do a sky test with the replacement camera. In general, the ZWO ASI2600mm-pro appears to be a good one to consider when upgrading from a smaller CCD setup.
Here is my linearity testing for this camera at the high dynamic range setting (0 gain, 50 offset), 1x1 binning, temp -10C
Linearity results look very promissing for the ZWO ASI2600MM.
Looks like a good candidate for my first cooled CMOS Monochrome camera to do Photometry.
Thanks for sharing.
I purchased one recently and like it a lot.
Something recently (like today) occurred to me about camera gains when doing VS and Spectroscopy work. You can find my post in the Spectroscopy section. Unfortunately I don't know how to attach images but if you have a copy of the 2600MM manual you can see what I"m talking about.
Barbara (and/or others),
I ran into an odd issue when I tried to do a linearity test with my ASI2600mm-pro last night. My light box is fairly bright so I have to keep exposures less than 2 seconds even with an H-alpha filter in-line. I used ASIImg and entered custom exposure durations at 0.05 second intervals using the ms units (vs seconds). The exposures that are recorded in the FITS headers are expected but it looks like the actual exposures were not. Has anyone else run into this? I suspect it's something dumb that I'm doing.
P.S. I tried to include a plot of the data (ADU vs exposure) and failed. Can someone tell me how to do that on this forum?
Thanks for the review of the ZWO ASI 2600mm Pro, a camera which I am seriously considering to purchase this year to multi-task between Astrophotography, Spectroscopy and Photometry, where I would like to have my first cooled monochrome camera to upgrade from DSLR Photometry, which is the only type that I have practiced up until the present time.
So for Photometry/Exoplanet work would you guys say that the ASI2600M pro is superior to my old STF-8300? (it's getting old so I'm thinking about selling it).
My 1st thought is that if the STF-8300 is working for you, keep it.
That said, the ASI offers higher DQE, especially in the blue. And larger field.
Would you need new larger filters as well?? I don't know.
Hi Peter, thanks for the response. That's the problem, the poor thing is getting old and beginning to fail, first the STi guider now the FW-8 filter wheel all easily fixed BUT expensive. I just would like to know strategies with the ZWO2600 with it's new chip as far as gain levels (as you know with the CCD gain was not a factor) etc.
I already have 36mm unmounted filters and I just order a new set of UVBRI form Baader..
Arne, have really enjoyed your reports on your experience with both the QHY600 and ASI cameras as I prepare to build my own instrument.
Regarding binning on these IMX571 series chips (the 455 uses the same architecture in a larger sensor) - my understanding is that if 2x2 hardware binning is selected your comment about reducing bit depth of the individual elements is correct, however if software (driver based) binning is utilised, no such reduction occurs. Given that CMOS doesn’t offer the read out noise advantages of a CCD, my understanding is that there is no functional difference between hardware and software binning other than download time with regards to SNR? (Therefor favouring the use of software binning and the full resolution of the 16 bit sub-pixels?)
Of course if a pixel saturates then the linearity of the super pixel is compromised, but I imagine that if a 3.6u pixel is saturating then it’s an indication i’m probably under sampling? Thanks again.