Corsair
Yes, this, again. ASIAir had the promise of working like an appliance - plug it all together, run it, and go home with results. Which was pretty much the case upto version 2.3.3. It is incredibly frustrating to find they've gone backwards since.
Unlike most here I am an electrical engineer, with a pedigree in digital control systems (my thesis) and a 40 year career in large software systems - air traffic control, power stations, and railway signaling. Things that can go BANG in milliseconds and kill lots people.. And in my spare time I am a software developer for macOS and Windows.
I even volunteered to go to ZWO and have a look at sorting out their crap because its quite clear they know nothing about software quality management and the people they have are barely above the level of amateur home-hackers, know nothing about testing, and are not using scopes, nor imaging with what they write.
I suggest I am well able to spot a bug when I see one, as well as engineering process failures an its petty clear something happened in May - my guess is a key person who knew left their software team, and was probably hired by ToupTek, as they are suddenly getting their pile of pooh into shape, fast.
The point is this: ASIAir 2.3.3 worked properly, but 2.4.1 onwards demonstrably do not, by direct tests.
Platesolving locates the centre of the field to the limit of the precision of the image, which for most purposes is a second of arc. GOTO uses this for closed-loop feedback, and (as an electrical engineering graduate) I am well aware closed-loop systems using feedback should settle exactly on the target. That was their whole point. While the mathematics of control systems are a bit of a backwater, it's not rocket science either, the maths are well understood from the 1950s. Even with time-sampled digital systems (which is what ASIAir is) – I did my honours thesis on this stuff.
In version 2.3.3 GOTO with platesolving was accurate. But in 2.4.1 onwards ZWO broke it.
There two simple tests you can do which prove this.
- In preview mode, manually select an RA and Dec and then start a GOTO that position. Compare the final position (from a platesolve) with the position you defined.
- Configure an AUTORUN session for a target east and close to the meridian, so that a meridian flip will occur. Assuming it works (it might not) extract the images afterwards and attempt to stack.
Test 1:
With version 2.3.3, the GOTO with platesolve will complete properly and will land on the target within 1 minute of arc. With version 2.4.X the platesolve does not finish and the socpe is likely to be off target by 10 arc minutes or more. This is close enough that users of small rigs think it's OK and won’t notice, but on mine (0.5 degree field of view) the target is often outside the frame of the camera. Which means the feedback loop is broken, and is a FAIL.
Test 2.
When stacking the images, a displacement is evident between the frames taken before the flip vs the images after the meridian flip. If the GOTO has worked properly the scope should recover its previous position within an arcminute or so and this displacement is negligible. However, with 2.4.1 onwards, after a meridian flip the previous position is not restored properly and the errors observed can be as large as 10..15 arc minutes. Again, those doing wide-field imaging with small rigs probably don’t notice and think it's OK, but at 3000mm the target is often outside the frame of the camera. That is a FAIL.
