The ASI6200MM Pro - First Light
In an earlier blog, first impressions were shared on the observatory’s latest camera, the ASI6200MM Pro. This is one of the latest cameras from ZWO. It has a full frame, backside-illuminated sensor with a resolution of 62MP. This Sony sensor represents some of the latest sensor technology. Of particular interest to astrophotography are the native 16-bit resolution, the very low noise level and the 14-stop dynamic range. That dynamic range can be achieved at both gain 0 and 100, thanks to its native dual-gain design. In real life this promises more detailed low-light images with less noise and with a great latitude in post-processing.
IC1805 - Heart Nebula
The weather in this part of the world has been very poor over the past few months, so it took a while before the camera could be properly used for its first image. For details on capture and processing, see here. During two sessions, one in November and one in January, almost 7h of data was collected. The dataset contains 82 subframes of 5 mins each, divided over the narrow-band channels H-alpha, OIII and SII. The final image has a resolution of 9316 x 6214px, or 58MP. That is slightly less than the 62MP from the camera, due to a small crop.
Image quality
The most important aspect of any camera is of course the image quality. And this camera does not disappoint. As described below, the final image was processed under suboptimal conditions, but the images are a pleasure to work with. The files are clean and have a lot of flexibility to play around with. The high resolution images responded well to any of the PixInsight processes thrown at them. Fine-tuning the settings for each of the processes worked well with often satisfying results. For a full resolution version of the image, see here. Below are details of the corners of the image. Stars are well-shaped, even into the far corners of the full frame field, no halo’s around them, and due to the lack of clipping, kept their colours quite well. The setup does not account for a tilting-adjustment option. The tilt-adjustment ring of the camera is sacrificed when adding the EFW and using an EF-mount connection. But judging by this image, that wont be necessary either. The nebulous regions with their dark areas of dust show a lot of detail at pretty decent sharpness. This sharpness could probably have been further increased if it was not for some heavy noise reduction that had to be applied because of the calibration issues mentioned below.
Dynamic range
Without doing some fundamental testing it is difficult to say anything quantitatively, but certainly qualitatively it felt that the camera had a lot more room to play with when it comes to dynamic range. Thanks to the deeper full-well capacity, stars were clipping less. In the below picture some 3 min luminance shots are compared to 3 min luminance shots of the ASI1600 under similar circumstances (although comparable, but different part of the sky). Any clipped areas are made black here. While subtle, the overall impression is that there is more latitude in stretching the images with less risk of bloated stars.
On the other side of the histogram, the background signal of the ASI6200 is with approximately 0.042 about half that of the ASI1600 which has a background signal of approximately 0.085. All numbers normalized to a 0-1 range. It must be noted that these measurements have not been made under identical conditions, but the difference is large enough to draw the conclusion that overall there is more dynamic range between the deepest blacks and the clipped highlights.
Calibration issues
As you can see on the processing page, calibrating the images turned out to be difficult. This was most likely a user error. During the first observation session on 07 November 2020, the background levels of the narro-band images were in the range of 0.002. But dark and bias-images that were captured the following day had a base-level of 0.003. Background levels during the second session on 13 January 2021 were just a bit above 0.003. So something had happened to the first set of images that made it impossible for them to be corrected for bias and darks. In an attempt to reproduce the background levels of the first session, many parameters were checked for their effect. This includes parameters like USB bandwidth, hardwarebin mode, highspeed readout mode and on camera dew heater. But none of these had a significant effect on base noise level. In fact, looking at the results, none of these settings had a significant effect on the noise level. The highspeed mode however did increase the variability by quite a bit. This is probably due to the fact that in highspeed mode, the camera switches to a 12-bit mode, rather than the 16-bit mode. In the original ‘First Impressions’ blog, it was suggested that turning on HardwareBin would also switch the camera to 12-bit mode, but this turned out to be wrong. With HardwareBin on, the camera is still in 16-bit mode.
BandWidth HardwareBin HSMode DewHeater Median avgDev
80 0 0 0 0.00322 0.00007
80 1 0 0 0.00322 0.00007
80 0 1 0 0.00312 0.00011
80 1 1 0 0.00312 0.00011
80 0 0 1 0.00323 0.00008*
80 1 0 1 0.00323 0.00008
80 0 1 1 0.00313 0.00011
80 1 1 1 0.00313 0.00011
40 0 0 1 0.00323 0.00008
*Default settings going forward
The conclusion here is that the issues encountered during first light must be related to user error and will likely not happen in subsequent sessions. The conclusion also is that without any calibration files applied, the final result is actually pretty good. The camera does not have ‘bad columns’, banding, amp glow or other sensor-related artefacts that would create an uneven effect on the image for which bias and dark frames would normally correct. Also vignetting is very minimal, so flats may not even be needed, unless some dust-bunnies start to interfere. All in all these files are very clean and very robust to work with.
Gain 100
The camera has a so-called dual-gain setting, meaning that the amplification during readout from the sensor-plane can actually switch between two states. The result is that with this second gain setting you can get nearly the same dynamic range (probably half a stop less) as with base gain setting, but with higher sensitivity. For this camera that gain setting is 100. In one of the YouTube videos on the internet, someone mentioned that he was not quite sure whether the high sensitivity mode would kick in at 100, or above 100. Below three examples of background noise, all stretched in identical manner, at three gain settings: 99, 100 and 101. It is clear that the high sensitivity mode kicks in at 100, so no need to set the gain on 101 or higher to get the benefit.
File-size
One of the downsides of higher resolutions is the increased file-size. The 62MP sensor produces 122.4MB files, and that is a lot. You need enough hard-disk space to store all these files. But with the continuous drop of price per MB for hard-disks, this will probably not cause too much of a problem. What did appear to be more of an issue than anticipated was the actual transfer-time of the files. Files move quickly over USB 3.0 connection from camera to the Fitlet2 (scope-based computer). But transferring those files over WiFi to the remote laptop took longer than expected. In original testing, the total time to get images in was about 20s. But in real use, due to different relative positions of scope-router-laptop, the transfer time is more often in the range of 23-25s. While only a few seconds longer, in total it gets to a pretty long time. For a 5min exposure this is not a big deal, but if scope and target combination would require 1 min exposures, this seems not the right setup. There are various ways to solve this issue. The most logical one is to store files locally on the Fitlet2, and bring them over using FTP. This would be working well for automated capturing during the night, but it would lack the direct feedback you would want during the early parts of the session. Binning or capturing subframes are other options, but would not make maximum use of the resolution of the camera. The most ideal solution would be for KStars/Ekos to asynchronously capture frames to server and download them to client. Lets hope this solution will become available at some point.
Processing the larger files obviously demands more of your computer. So experiences here very much depend on the setup used. For AstroWorldCreations, a 16-core Mac is used, which handled the files well. It was reassuring to see PixInsight using all 16 cores for most processes. So investing in higher core processors will probably benefit processing these large files.
Conclusion
While the first image was not without its challenges, most of the issues are probably related to user error of not knowing the ins and outs of the camera yet. The camera behaved very well in the setup, and from a practical point of view, the handling was quite comparable to the ASI1600 setup. The extra size and weight were never an issue. The images seem to have a nicely improved dynamic range, leaving a lot of latitude during processing. The clean nature of the frames means less correcting needed in post-processing. The larger full-frame sensor size allows for a wider field of view, which can be used for just that, more object in the image. Alternatively, crops of the increased FoV, could be used so that the object can be positioned and rotated in the most optimal way. Because the pixels are of a similar size to the ASI1600, one could also record subframes the size of an ASI1600 and have similar images to the ASI1600 but with less noise and more dynamic range.
If the next images will follow a similar experience, this camera is likely to become the default go-to camera in the observatory.