I have recently started using SGPro and it has been a very nice experience.
I am writing about one problem that I can’t figure out.
I am using an ASI1600MM-Cool camera over USB3 with the USB Traffic setting at 85.
I have the latest ASCOM and ASI driver installed.
I am using SGPro 2.6.0.21 on a Windows 7 Pro laptop.
The laptop has the AMD A6-3610 APU @ 1.80 GHZ and 8GB of RAM.
The issue is it is taking a very long time between frames for SGPro to write out the frames.
For comparison - with Sharpcap 3.0 using the ASI ASCOM driver I can write bias frames to disk @ 0.9fps.
With Maxim 4.62 I can write them at 0.4 fps.
With SGPro it is about 0.2fps or 5 second per bias frame.
This is significantly slower than the other software running on the same laptop.
I had Image History and the HFD display disabled.
I disabled autostretch for the histogram.
Is there any other optimizations I can make?
I ask because with the ASI cameras it is advantageous at time to use lots of short exposures and taking 5-6 seconds per 30s exposure really wastes lots of potential integration time.
Here is a log of running a simple sequence to take 10 bias frames - it to 50s to execute:
I just measured how long each program took to take 10 bias frames.
Sharpcap was the fastest, then Maxim DL, then SGPro.
The log shows the time is takes per image in SGPro via the timestamps.
I’m looking for feedback on the expected time it takes to write a 32MB frame to disk - it should be nearly instantaneous, but there seems to be much more going on in between frames. Both Sharpcap and Maxim DL are able to write frames out at 2 times or faster than SGPro on the same hardware so I believe that rules out a hardware issue.
I’ve disabled all the analysis functions I could find in SGPro so something else seems to be slowing it down.
Any chance someone could at least review and comment on the timing information recording in the log file I provided? I am just trying to understand why SGPro is so much slower than Sharpcap and Maxim DL on the same laptop.
Hi did ask this long time ago. If you want to reduce overhead, the only way is to upgrade your hardware. SGP+ASCOM does have a lot of highlights, but the performance is not one of them.
With standard CCDs where image download is about 20-30 seconds this is not a problem, with new cameras where you can get a lot of short frames, it is.
That is unfortunate - many people use small headless computers on their setups and they generally don’t have the performance of this laptop. Hopefully there are some simple optimizations they can do to improve this. Throwing faster hardware at it to solve it is unsatisfying since clearly the hardware is more than adequate based on the Sharpcap and Maxim results.
The lack of subframe support (best I can tell) is also somewhat glaring.
SGPro does not do anything fancy (except image analysis, which you already mentioned you had turned off). When the image is reported as complete, we ask the driver for the data and stream to a disk-based file. I’m not sure if there is actually any part of this that presents itself for optimization. We are using redundant write cycles to ensure that the data is fully intact, but I’m not sure how much extra time that requires.
Does the subframe defined in the frame and focus get used on sequences? That isn’t obvious if that is how it works.
I want to use subframes as the ASI1600MM is a large sensor and for some targets I don’t need the entire sensor. Just wastes disk space and time stacking.
As for the write speed - it is writing out 32MB of data so it isn’t clear why it would take so long, robust or not. I have a fast SSD drive.
I had hoped supplying a simple use case with a log would allow more enlightenment on this problem.
Again it all comes back to at least two other applications using the same ASCOM driver on the same laptop are significantly faster so it can only be how SGPro handles image data.
On my i7 desktop SGPro actually writes files out quickly but I’m not going to get an i7 field computer due to cost and power usage compared to a energy efficient processor like I’m using.
If this is the expected behavior for SGPro I guess I can just use Maxim DL in these scenarios where I want a subframe or am using very short exposures for DSOs.
I think this is a significant problem with people using low power “compute sticks” that are not running i7s, and also people using the newer CMOS low read noise cameras that do not require long exposures. Being able to knock off 15-30s exposures quickly and efficiently is going to become most the norm.
No, SGPro does not support subframes within a sequence, just when using the frame and focus module. You are welcome to make feature requests in our Feature Request channel (Feature Requests - Main Sequence Software).