Serious issue with ASI185MC and latest SDK and/or drivers

[vlaiv]

Yesterday I was out and imaging just to find later on that 2 hours worth of exposure were totally ruined.

At first it seemed that data is clipped somehow (it did turn out to be the case) but I had no idea what might be the reason. Calibration of frames failed. At one moment I suspected even that camera malfunctioned.

Then I retraced my steps between last week's shooting (everything was fine then) and last night. Only thing different was new version of SharpCap. As I understand it comes with new SDK for ASI cameras.

There is now different setting for brightness (0-500) that I found out to be ADU offset. But main problem is that it is not ADU offset from some base value but rather from some arbitrary value. If you set it to 0 for bias frame it looks like it takes 0 value to be median of image! So it clips all values below that if brightness is set to 0. You might think that moving brightness to some higher value solves the problem, but it depends on your gain settings - as gain increases so does histogram spread around median value, and at one point it again starts clipping at 0.

Even more disturbing is the following:

I took 2 frames with same settings: gain 0, gamma 50 (now it is from 0-100, don't know what it was like before, I usually don't touch gamma as I want my data to be linear) and brightness at 500 (to prevent any clipping), one being 30s dark exposure, other sub 1ms bias frame. Then I analyzed both frames and it turned out that 30s dark has lower median value than bias frame! 7568 vs 8000 (one might expect to see 8000 as median for bias if brightness is offset from median point of bias as it is 16*500 - perfect fit). No way that 30s exposure would produce smaller median value than the bias frame!

I did try to revert to older driver version but it did not help. I did not try earlier version of sharpcap as it has expiry date and I did not want to mess with system time.

I posted issue on zwo user board and awaiting reply and eventual resolution of issue.

[vlaiv]

I investigated a bit further and found out that indeed there is some unwanted automatic level scaling behind the scenes now.

Here are three screen shots from SharpCap. First being 1/8000 ms exposure, gain 0, brightness 50. Pay attention to histogram peak position. Second screenshot is first frame immediately after switching to 4s exposure with same settings (just selected lx + 4s). Histogram stays in expected position. Third screenshot is very next (and all that follow) frames at 4s. Histogram suddenly jumps to the left (something being subtracted from frames at this point). I suspect that first frame of longer exposure was used to calculate something that is later being subtracted from following frames at that exposure length.

I recommend the following until (and if) this is resolved:

For long exposure shooting use only dark frames for calibration (yep this means much more time to spend getting sufficient dark frames). For short exposure this is not an issue as one does not need to use bias frames (you can take enough dark frames for exposures below few seconds). Also keep an eye on histogram after selecting your wanted exposure and gain settings (do this in "dark" mode first) and adjust brightness so there is no clipping to the left (very important on higher gain settings as histogram spreads to wider configuration an can easily get clipped to the left).

One might think that there is no harm in histogram being clipped a bit to the left at noise level, but if stacking you need all the information (every photon to count) in order to get the best from your data.

[vlaiv]

Ok, after some further investigation following came up:

Due to very strange behavior in previous versions of drivers and SDK above issue did exist but was masked and less noticeable. I managed to acquire previous versions of drivers and SDK (thanks Robin for that) and did some more tests.

Earlier there was very strange ghost signal that appeared and masked above issue. Here are some pictures of it:

50ms exposure, gain set to 149, and brightness in order: 1, 100, 200, 300

One may notice that ghost signal shifts to the right twice as fast as main signal when applying brightness (??? maybe some multiplication involved instead of addition when applying brightness setting)

Conclusion so far: It would be advisable to upgrade to new SDK (just use latest release of SharpCap and FireCapture) and drivers, and carefully manage your settings (avoid bias frames, calibrate with darks only (use as much darks as possible) and apply brightness to avoid histogram clipping to the left on given gain and exposure settings) until (and if) this issue is resolved.

Vladimir

 

[vlaiv]

Just a quick update, Sam of ZWO confirmed the issue and promised to try to deal with it after their holidays.