While I have been working on overcoming limitations of my mount, I decided to test whether short (within reason) narrowband subs will yield a similar result to longer ones, given the same total integration time.
The attached shows a section of a stack of 22 five-minute subs (left) and 9 fifteen-minute subs. Total integration time is a tad less for the shorter subs, but anyway the difference between the two is not as astonishing as I was expecting.
Data captured with 3nm H-alpha and ICX 814 in Paddy in Brisbane.
Hopefully I will be able to produce at least a semi-decent image in the coming weeks - it has been ages since my last one...
The biggest difference is always going to be in the darker areas. I understand that my test from way back isn't exactly comparable in many respects but.
There are all three just single subs but even if I combine the 1200s and 2400s they don't even come close to the 3600s in contrast and signal. These were taken with my QHY9 from Melbourne on a full moon imaging towards the heaviest part of Melbournes sky glow with 3nm filters
It depends on the camera and the f ratio, in my option. I shoot all my narrowband at 300" with the ASI1600MM-C. All my scopes are f5 to 5.3. I have Gain and Offset control on this camera and use what is becoming a standard, of Gain 139 and Offset 20.
Shorter subs just shoot more of them. In my testing there was very little noticable difference between 20 x 300" subs and 10 x 600". As you suggest sky conditions are a factor.
Definitely a dramatic difference with your QHY9 Colin. Are these single subs or stacked images with equal amount of integration time? If these are single subs, then obviously the longer subs with have better SNR.
I agree with Glen. With my gear (f/5.6, ICX814, imperfect mount and heavy light pollution-I live very close to the CBD) there is very little difference between images integrated from 5-minute and 15-minute Ha subs, even in the dark areas.
If I had a better mount, I would certainly go for fewer but longer subs, even if only to simplify stacking, but given that my mount does not deliver consistently good guiding, I may try living with shorter subs.
The dominant factor is the read noise of the camera. QHY9 read noise is quite high. Sony sensors are probably less than half the read noise.
These older Kodak sensors can be made less noisy. QSI extracts less noise out of their 8300s.
I now have a FLI MIcroline 16200 I have been using now for about 6 weeks.
One difference I have noticed is how well subs stack together compared to other Kodak based sensors I use/have used. The read noise is quite low similar to the Sony sensors (not quite but its a large sensor at 27 x 22mm).
6 subs with fairly light backgrounds suddenly get nice dark backgrounds when stacked that I would normally expect from way more subexposures.
So improvements are being made by OnSemi on these old Kodak designs which seem to have been very static in their developments. Perhaps all the money is going to CMOS for mobile phone cameras and digital cameras and industrial vision.
It might be an academic test anyway if your mount won't do 20 minute subs or longer without eggy stars.
But generally speaking with ex-Kodak sensors longer subs are better than shorter subs due to the high noise levels which reduce the effectiveness of stacking.
Suavi, personally I reckon you have a really good niche, imaging from so close to the city, I'd be emphasising that, showing the effectiveness of your filters. you get some world class results.
Even with my excellent, over the top mount for my scope (MyT!) and wide F ratio, I'll be doing testing like Glen and hopefully using 5 min NB subs. My theory is the more quick subs I get, the more individual subs I achieve and don't get burned by errors or bad subs. We'll see.
Glen, where have you seen the standard of 139 gain and 20 offset? Hmm, I need to google what on earth offset even means!
Suavi, personally I reckon you have a really good niche, imaging from so close to the city, I'd be emphasising that, showing the effectiveness of your filters. you get some world class results.
Even with my excellent, over the top mount for my scope (MyT!) and wide F ratio, I'll be doing testing like Glen and hopefully using 5 min NB subs. My theory is the more quick subs I get, the more individual subs I achieve and don't get burned by errors or bad subs. We'll see.
Glen, where have you seen the standard of 139 gain and 20 offset? Hmm, I need to google what on earth offset even means!
The CN ASI1600 Beta Test thread played around with various gain and offset settings. BTW Offset is the equivalent to brightness I believe, and default when using Unity is 20. Have a look at Ray's exposure chart and work out what suits you. My standard, is just my standard given my scope f ratios, the darkness of my observatory, being pier based, etc.
The were single subs so obviously the longer integration (3600s) will be cleaner than the other two but even when combining the 1200s and 2400s it still doesn't match the single 3600s. As Greg points out, this is all to do with the quite high read noise of the KAF-8300 sensor. The principle remains the same with the ICX-814, the closer you get to overcoming read noise in the dark areas the better contrast you'll get.
Basically, if you stack a lot of images where the fainter regions and buried in the read noise, no matter how many you stack together you'll not regain those faint regions.
Basically you just want to image to whatever your mount can handle repeatably. With my EQ6 that was 10 minutes.
The offset is just a histogram pusher. With both of my QHY's I have had to calculate the gain and offset manually, takes about 10 minutes. The offset changes your bias level, an offset of 15 may give a bias of 100ADU where an offset of 20 could give a bias of 1000ADU. Just plucking numbers out of the air
You only need to change your offset if your bias frames are either too high (1500 ADU+) or too low (300 ADU-). I tried to get mine as close to 800 as possible so that it doesn't chew into the dynamic range, the higher your bias the lower your dynamic range. If you have it too low however (your bias) you can sometimes end up with negative numbers if you have less than absolutely perfect flat frames (VERY difficult).
The Offset parameter is tuned to get the ADU level in the range of 500 -1000 ADU when taking a bias reading. This is to ensure that no divide by zero errors occur during stacking (when the bias is subtracted from the lights).
The were single subs so obviously the longer integration (3600s) will be cleaner than the other two but even when combining the 1200s and 2400s it still doesn't match the single 3600s. As Greg points out, this is all to do with the quite high read noise of the KAF-8300 sensor. The principle remains the same with the ICX-814, the closer you get to overcoming read noise in the dark areas the better contrast you'll get.
Basically, if you stack a lot of images where the fainter regions and buried in the read noise, no matter how many you stack together you'll not regain those faint regions.
Basically you just want to image to whatever your mount can handle repeatably. With my EQ6 that was 10 minutes.
Totally agree.
However, I was quite surprised how little difference shorter subs made for this object (NGC6188) and with my gear.
I visually inspected the darkest area in both images and visually there is not much difference between the two, even though shorter subs add up to less total integration time.
On a happy side of things, this morning I finally managed to reduce play in the DEC axis to near zero and the motors do not bind, so should be good for collecting better quality data, fingers crossed :-)
Your QSI690 is MUCH MUCH cleaner than the KAF-8300 sensor so that certainly helps. At F/5.7 I calculated an approximate 1.5-2 hours exposure needed to be sky limited with 3nm filters. So my QHY9 showed what happens with read noise significantly more than what your very low read noise Sony sensor does
Basically, if you stack a lot of images where the fainter regions and buried in the read noise, no matter how many you stack together you'll not regain those faint regions.
This seems somewhat intuitive, but it's not true We overcome the effects of read noise with very faint data every time we create a master bias or master dark frame.
Not sure whether I have utilised this script correctly, but if I choose 10% readout tolerance, the suggested time for a single Ha sub is about 18minutes, while Anstey limit is 8minutes for my usual total integration time.
I haven't really played with the PI script. The maths behind the method Ray proposed is easy to verify and the method is very simple to use so that's how I check if my subs are sky limited. There are a number of calculators around and they all give different answers
Thanks Rick, will definitely check the maths behind it
I know that ultimately longer subs would be better, only if I could accurately and reliably guide for longer periods of time. However, it was pleasing to see that even with 5-minute 3nm subs I can still get decent data and capture relatively faint fuzzies.
Now, if I could only teleport clouds down south to Mexico...