What binning you use most with your ASI294MM pro | |
---|---|
Bin 1x1 | |
Bin 2x2 | |
Login to vote and view results. |
1.20
#...
·
1
like
|
---|
A test I keep meaning to do is to collect data at bin1, then collect data at bin2, but do a 2x drizzle integration. In both cases, I would end up with a final image at 8288x5644. I've done both independently in the past, but I haven't shot the same target with both methods. Thanks. My drizzled image looks like one on the left. I think drizzle is being not applied. Might have to watch some tutorials for drizzle in APP. |
1.20
#... |
---|
Amrinder Singh:A test I keep meaning to do is to collect data at bin1, then collect data at bin2, but do a 2x drizzle integration. In both cases, I would end up with a final image at 8288x5644. I've done both independently in the past, but I haven't shot the same target with both methods. That’s a huge difference! The right has superb stars. I would love to see bin 1 data compared to the right image. Then drizzle bin 1 data, 😂. File would be humongous! |
1.81
#...
·
1
like
|
---|
I normally use bin 2x2 with my Stellarvue SV102 and .8 reducer, which gives me a 1.67 sec resolution. I can, but haven't yet, use it bin 1x1 on my AT72EDII with a .8 reducer to give me 1.37 sec resolution. |
7.83
#...
·
2
likes
|
---|
Brian Meyerberg: I posted earlier that if I were to try drizzling the bin1 data, my MacBook Pro would curl into the fetal position and just whimper uncontrollably LOL. I really don't think I'd see much improvement if I ever did it, though. For drizzling to be effective, you need three things:
I can get the second two bullet points, but the first one... with my current setup, I'm at 1.25"/px when using bin1. I suppose if I had a night of above average seeing, I would be a bit under-sampled... but I fall back to the real limiting factor: my poor MacBook Pro trying to deal with drizzling the 47MP files |
1.20
#... |
---|
My most recent Bin1 stack with drizzle was 730MB!! |
1.20
#...
·
1
like
|
---|
Let's just hope computers get much faster, quickly and cheaply. I'm already thinking of building a 16 core Rizen computer. AMD is up to 64 cores, it can work on a lot of files simultaneously ($$$$$). Large SSD hard drives and fast memory cards are also coming down in price. Is there no end to the craziness of our hobby ?? |
10.16
#...
·
2
likes
|
---|
Drizzel is a mathematical emergency solution for cases with undersampled pictures. It is not a substitute for real data. |
1.20
#...
·
1
like
|
---|
Finally figured out how to drizzle and sharing the comparison. Bin 2 with 2x drizzle comes pretty close to bin 1. When you do extreme zoom you can see some artifacts around stars but there is huge improvement from normal bin 2. Bin 1 Full 2.jpg |
1.20
#...
·
1
like
|
---|
Did a quick re stack of North America Nebula shot at 150mm with bin 2. There is no improvement I see with drizzle on this one but it caused some banding on the sides of the picture. I guess drizzle only works to fix undersampled stars? |
1.20
#...
·
1
like
|
---|
I posted earlier that if I were to try drizzling the bin1 data, my MacBook Pro would curl into the fetal position and just whimper uncontrollably LOL. I really don't think I'd see much improvement if I ever did it, though. For drizzling to be effective, you need three things: The 3 requirements listed above are in fact requirements. |
1.20
#... |
---|
Funny how this poll thread changed for “Bin 1 or Bin 2” to “dither or not to dither” |
1.20
#... |
---|
Brian Meyerberg: |
10.79
#...
·
|
---|
I am strictly using 2x2 since with that mode, the sensor can make use of the high conversion gain (HCG) which isn’t available at 1x1. The advantage: I can use a higher gain without losing dynamic range and keep exposure times pretty low. I‘m running the camera always at gain 120. |
10.79
#...
·
·
2
likes
|
---|
I am strictly using 2x2 since with that mode, the sensor can make use of the high conversion gain (HCG) which isn’t available at 1x1. I should correct myself. As I have learned, the ZWO ASI294MM has the high conversion gain enabled for all gain settings at bin 1x1. Hence, it might be worth considering shooting everything in 1x1 and do a downsample later on. |
12.28
#...
·
|
---|
Yes, bin 2 in the ASI294 is purely an after the fact firmware implementation and isn't done at the hardware level. You can shoot with bin 1, and I have done that on a couple of images. The file sizes are huge. It would be more useful if they could allow a field of interest defined so the image could be cropped prior to downloading. |
10.79
#...
·
|
---|
Yes, bin 2 in the ASI294 is purely an after the fact firmware implementation and isn't done at the hardware level. I'm aware that most CMOS sensors just do a mathematical binning. Since I always believed that the high-conversion gain is only activated with 2x2 and gain 120, I never considered 1x1. Now, knowing more about the camera, I'm considering shooting at 1x1 and gain 0 to maximize dynamic range. |
12.28
#...
·
|
---|
I’d be interested in peoples experiences with Gain 0. I have heard second hand about banding issues at low gain when imaging from dark sites or in narrow band images with low sky background. I also would like to image at low gains to take advantage of the higher full well capacity available. |
1.20
#...
·
|
---|
Based on my processing workflow (Using Ha as luminance layer), I shoot Ha data at bin 1 and Oiii and Sii at bin 2. Gain for both is 120 to keep it simple with managing calibration files. |
4.92
#...
·
·
1
like
|
---|
I use Bin 1 on L and Bin 2 on the rest of the filters. Cuts down on my file sizes and on RGB it really doesn't make a difference. I scale up the Bin2's before integration but leave them in Bin2 for calibration. |
0.00
#...
·
·
1
like
|
---|
I used to image through my 600mm Celestron 80ED at 1x1 but I lost all my data due to snow getting in the laptop. I felt I needed to make up for lost time so switched to 2x2 which is 4 times more sensitive. I haven't noticed the loss of resolution and faster processing of the smaller files is great. |
2.97
#...
·
·
1
like
|
---|
Jonny Bravo: Exactly. You're very unlikely to get good enough estimates of all of the parameters to calculate the combined impact of your scope, seeing, guiding, etc., so just measure it. And measure it as FWHM in pixels. If it's between 4-8 pixels, you pretty good to go. At the high end, you might consider dropping to the smaller pixel or Drizzling but it might not be worth it for the SNR hit (no free lunches in astro; you will get finer detail but lose some of the faint stuff entirely). At less than 4, you should really just bite the bullet and drop to the smaller pixel (natively or through Drizzle, if that's not an option) if you're not satisfied with the resolution. You'll still lose SNR (speed), but the detail is likely worth it. Depends on what you value in the image: faint stuff or sharp details/round stars. Above 8 and you're just taking the SNR hit without getting much if any additional detail information, i.e. you're over-sampled, so either increase your pixel size or integer downsample your data. That's the worst outcome but the easiest to fix. |
1.20
#...
·
|
---|
I shoot at Bin2 with FL 1200mm newt (0.8"/px) and with my SVX80T (1.99"/px) when I'm shooting broadband, which are short (<=60sec) subframes. This is just due to long term storage considerations as hundreds of Bin1 subs are easier to store than thousands and short subframes lead to thousands. I don't shoot Luminance... just pure RGB so I don't need/want/use the Bin1 for Lum. I shoot at Bin1 with my SVX80T (1"/px) when narrowband with 3-5 min subframes. Difference in file size is compensated for by the fewer files with the longer exposures in the individual files. My computer/hard drive space react approximately the same to 1500x 25mb files as it does with 300x 93mb files and I only calibrate the raw subframes after shooting. I don't register/measure/normalize/integrate until after I have a sufficient number of hours. Even then, I often delete the intermediary registration/normalized images after the channels are integrated as I can re-generate them from the calibrated frames at any time and if I add data, I'm going to re-register/normalize/integrate anyway. |
2.97
#...
·
·
1
like
|
---|
Arun H: I don't have any experience with it on this camera to comment on the banding, but there are some clear principles to follow for single exposure length approaches depending on the Bin mode. For Bin 2, gain 0 would not be worth it because you'd need more than a 20x increase in FWC to overcome the more than 4x increase in read noise over gain 120, assuming that you take the sky-limited sub-exposure necessary at each respective read noise level. You only get somewhere between 6-7x. Bin 2 is basically set-it-and-forget it at gain 120 under all conditions. Maybe a different story if you're composing different exposure times or shooting your stars separately, since stars are alone bright enough and you don't care about the noise of the background you'll just replace with other data; FWC is all that matters. In Bin 1 however, the lowest gain possible (either 0 or the lowest gain that doesn't exhibit idiosyncrasies like banding) would always be a win under the same circumstances (minimum sky-limited sub times) unless you have headroom to spare. That's because you only need 2.25x more FWC to cover the 2.25x longer exposure required over unity gain (to pick gain) when you get 3x. I'm not intimate with the inner workings of the mode switch, but Bin 1 appears to be effectively working in HCG mode over the whole range of gains. That would lead me to believe that if there is banding at gain 0, that would be at gain 0 in Bin 2, where the sensor is truly working at its highest conversion rate, and we've already eliminated that option from serious contention. But if it does happen at gain 0, Bin 1, then just increase it as minimally as necessary. As for the general idea that you should lower gain in dark skies or with narrowband, they've got it exactly backwards. When you reduce the shot noise -- by moving to a dark site, using a filter, or both -- read noise threatens to become more and more of a factor and running out of FWC becomes less and less of a factor. Those are the moments where you would favor higher gains. Under light pollution or when unfiltered is when read noise more easily gets swamped by the shot noise, and your primary concern shifts to FWC. Either way, you should be looking to maximize DR, subject to practical considerations. |
2.97
#...
·
|
---|
Björn: That's mostly an illusion. The high conversion is essentially always already there in Bin1. Look at where the read noise is for both modes at similar e-/ADU levels while imagining what the read noise of Bin2 would have been if HCG applied to the whole range (extend the line back from 120). True the FWC is lower at the same read noises, but that's to be expected with smaller pixels. HCG is a boon on Bin2 because it's starting at an atrocious 8e- of read noise. But Bin1 isn't and so lack of an HCG mode switch isn't really a reason to avoid it. It's a matter of whether one wants the smaller pixel and what comes with it. |
12.28
#...
·
|
---|
Roman Pearah: Most sub exposure calculations I have seen assume that you want the sky background shot noise to drown out the read noise by some huge factor like 50x. In reality, you can get away with a lot less - the point being that raw read noise is low enough in modern cameras that it isn't really worth spending much time on. There is an exception and that is in the case of where the read noise has a pattern, since the human eye is much more sensitive to patterns than randomness. I understand the theory quite well. My question was for people who had actual experience using the camera at Gain 0 so that I can make the appropriate decision on whether to prioritize FWC or read noise for my particular imaging needs. For reference, I also own a 2600 MC Pro. I'll always use that camera at Gain 0 to take advantage of the FWC. That sensor has no fixed pattern noise to speak of. |