How Efficient Are You?

[jimjam11]

Clear sky is a rare beast so I have spent a decent amount of time studying my imaging efficiency and trying to improve it. I want to get as much acquisition time as possible per clear dark hour but I wonder if I am alone in pursuing this?

Products like SGPro make automation relatively easy but in my experience this automation can easily lead to poor efficiency especially if you are using a cmos camera which works best with shorter exposures.

When I first got my ASI1600 and used it with SGPro I measured my interframe delay to be approximately 16s. Combined with 30s L subs that equated to a wastage of circa 50% meaning each hour would yield on 30mins of data. With 60s RGB subs it was slightly better but still leading to a wastage of around 27%.

Studying the SGPro logs I noted the following:

1. I had image history turned on, this added approx 5s per frame and SGPro did not perform this analysis asynchronously.
2. Despite my CEM60 only having a USB2 hub, image download was taking 2s.
3. I was using an older i5 laptop for image aquisition. This was taking approx 12s to download, save and start the next exposure (this was not a slow laptop by any means and had plenty of ram and an ssd).

I made the following changes to try and improve my efficiency:

1. I upgraded my acquisition laptop to a decent (and brand new) i7 - This dramatically cut interframe delay from 12s to 3s (without image history).
2. SGPro changed the analysis; it is now an async process. Nina was always async and considerably faster.
3. I switched to using 60s gain 0 L subs. I still use 60s RGB subs @ gain 76.
4. I measured image download as 2s through my cem60 usb2 hub, 1s without.

 

This gets me to approx 48-50mins of acquisition per hour in broadband) but there are still some things I am looking at:

1. I normally interleave my frames (LLRGB-dither) to minimise dithers and balance any sky changes. However, this costs time (approx 2-3s per filter change). I may stop doing this to save the additional 2-3s. Dithers are pretty fast especially if they are only performed every few frames (approx 5m worth of frames), but the filter changes are costly.
2. I use autofocus typically set hourly or every 2C. Looking at my FWHM measurements I think 2C is correct, but the time interval can probably be stretched to 2 hours (maybe more). An autofocus run typically takes a few mins.

 

Has anybody else looked into this, if so what kind of things have you changed to make improvements?

N.B. For narrowband I use 5m subs so typically get 50-55mins worth of acquisition per hour.

[DaveS]

I'm currently running 600 sec NB and RGB subs, and 300 sec l subs with my ASI`1600 set to Max DR Download is through a USB2 hub, but am considering running a dedicated USB3 lead.

[SamAndrew]

Just looking at the time stamps on my saved files I can see it takes 5m 20s ~ 5m 30s for each 5 min fame inlcluding auto focus runs every hour; so > 55m per hour which is better than i was expecting.

I think I have dithering still set to once every 5 frames although i was thinking of removing it entirely. Camera is connected directly to PC via usb3 (ASI 1600mm). Auto focus probably the most time consuming, I'd like to get tempratre compensation setup at some point soon and see if that removes the need to refocus.

[jimjam11]

I have been doing some more testing to see where I am losing time and what I can do about it.

5min subs significantly reduces wastage but at the cost of dynamic range (more stars become saturated because the histogram is pushed right). For my sky conditions (bortle 5) 60s is still slightly above optimal sub length but I am keen to reduce interframe wastage. FWHM is also lower with 60s subs compared to longer. My FWHM was not lower with 30s subs so 60s seems optimal for my setup.

To do some testing I compared SGPro and Nina. I also tested using USB2 and USB3. Imaging laptop was my usual i7 with fast SSD.

I programmed a sequence to capture 60x5s Lum frames to see what the wastage was like. The results surprised me a lot:

I always knew nina was faster than SGPro but I didnt realise the difference was so large. SGPro wasted 3.3s per frame on USB2 and it only improved to 2.58s with USB3. Nina had a wastage of just 1s with USB2 and a staggering 0.07s with USB3!

I then wanted to test interleaving of frames (the way I normally image) to see what the impact was there. I changed the test so that 15x each filter were captured. I knew swapping filter after each exposure was costly but this really highlighted it:

Once again it looks like Nina is miles ahead of SGPro, but interleaving adds approx 2s per exposure in Nina and 3.5s in SGPro. I cannot fathom why a filter change is faster in Nina than SGPro!

I definitely need to test Nina some more, I just dont have the confidence in it yet for unattended imaging overnight...

[andrew s]

With my ASI1600 bined 3x3 I get a 4s delay using CCD Autopilot. I take very short exposures typically 10 to 25s as I am looking for fast transient events. 

The reported download time is 1s the rest is overhead.

Regards Andrew 

[jimjam11]

I managed to test Nina a couple of nights ago and it performed very well. My SGPro routine yields about 48 subs per hour with interleaving, this increased to approx 51 without. Nina without interleaving yielded an average of 55 subs per hour, a significant improvement. The lowest value I got from Nina was during the meridian flip hour which appeared somewhat inefficient and wasted approx 4 mins. Autofocus after the flip took another couple of minutes so I ended up with 49 subs for that hour.

SGPro is still the functionally richer option but its performance is so poor compared to Nina. Clear sky (especially moonless) is rare and wasting valuable time just waiting for stuff to happen in SGPro is becoming increasingly frustrating... 

[jimjam11]

On 15/04/2020 at 20:17, andrew s said:

With my ASI1600 bined 3x3 I get a 4s delay using CCD Autopilot. I take very short exposures typically 10 to 25s as I am looking for fast transient events. 

The reported download time is 1s the rest is overhead.

Regards Andrew 

Are you able to use something faster like Nina for acquisition and then CCD Autopilot for analysis so you can significantly increase your efficiency?

[andrew s]

I think it would be fastest if I implemented it in Python and directly ran the camera via The Sky X Jarva script to minimize the delay. I have done this before but now I run remotely I wanted to start with a robust automation application. 

I analyse the results using astroimagej.

Regards Andrew 

 

[Ags]

I can get 99% imaging time utilization efficiency doing lucky imaging, but in processing will discard 50% of the data...

If you you have a dual scope rig, can you claim 200% efficiency? 

Edited April 17, 2020 by Ags

[Skipper Billy]

I dont keep stats but I would guess that on average I actually image for about 35 mins an hour - the other 25 mins being taken up with download time, recovering from dithering, cloud knocking out PHD2 and particularly electronic focusing routines. My scope is very temperature sensitive and refocuses every 0.5 degree change as well as every filter change which is quite time consuming especially with narrowband filters.

[JamesF]

I'm still at the stage of losing a bit of time to stuff that doesn't work.  That's mostly down to not being as familiar with the software as I need to be however.  On the positive side, I can be running three imaging OTAs at the same time.

However, a quick look at the timestamps for my ten minute luminance subs of M101 on 14th (just one of the rigs) suggest that there was just over one minute over the course of three hours when I wasn't actually capturing photons, so when things are going well efficiency looks pretty good.  Filter-swapping and dithering will eat a little more time out of that, but I'd guess that on a good night I'd still be capturing photons for better than fifty minutes out of every hour, perhaps better than fifty-five.

James

Edited April 17, 2020 by JamesF