Uncertain about ASI 2400MC vs 6200MM choice. Speed vs resolution vs read noise.

[pipnina]

I want to improve the speed of my setup. I worked out that if shooting only RGB then the ASI 2400MC would afford me effectively the same resolution image as my RisingCam 571 (6000x4000) but with a wider FOV due to the larger pixels, since I am speeding up my imaging with a normal camera lens I expect that would be best for star shapes anyway, while affording me a 1 stop boost to my speed (so 6.2x faster than my RisingCam+ f5 scope)

However it means narrowband (My main interest turned out to be Halpha only really, not worried about the loss of SII and OIII) becomes much less efficient as only one subpixel gathers data for it.

If I got the ASI 6200MM I can keep my narrowband capacity but with the 3.76 micron pixel size, the same as my RisingCam, I won't gain the speed increase as I would with the 2400MC.

Unless, I binned. A 3200x4800 image might be a little small for me but could be servicable. However this increases the effective read noise by a factor of four even though in theory is gives me a 2 stop light sampling boost right? So am I worse off with the 6200MM and should stick to RGB imaging with the 2400MC if I want speed, and simply forgoe narrowband for now?

Interested to hear thoughts.

Thanks

[Elp]

Even if using mono for RGB it's faster than OSC as you're utilising all pixels, not one in four or two in four. You'll be able to image though LP in narrowband and generally have more flexibility how and what you image.

You could add in a reducer if you wish to saturate your sensor pixels quicker.

[symmetal]

I would wait until you see what results you get with your Canon lens with your current camera as far as corner star shapes are concerned before buying a full frame sensor. It wasn't until I bought the RASA 11 that I could take full frame images with the 6200MM with good overall star shapes.

My Canon 'L' 100-400 zoom gave poor star shapes over the whole frame unless I stopped it down to f8 when it was tolerable. A prime lens should be significantly better than a zoom of course, so hope yours performs well.

I tend to software bin my 6200 images after stacking, as the full size images can be a bit slow to process, especially if doing a mosaic.

2x2 software binning gives 2x the noise and 4x the signal, so a 2:1 improvement in S/N. Noise adds by using the square root, so with a read noise of say 2e, the 2x2 binned read noise is sqrt( 4 * (2^2))  = 4e. 🙂

Alan

[pipnina]

13 minutes ago, symmetal said:

I would wait until you see what results you get with your Canon lens with your current camera as far as corner star shapes are concerned before buying a full frame sensor. It wasn't until I bought the RASA 11 that I could take full frame images with the 6200MM with good overall star shapes.

My Canon 'L' 100-400 zoom gave poor star shapes over the whole frame unless I stopped it down to f8 when it was tolerable. A prime lens should be significantly better than a zoom of course, so hope yours performs well.

I tend to software bin my 6200 images after stacking, as the full size images can be a bit slow to process, especially if doing a mosaic.

2x2 software binning gives 2x the noise and 4x the signal, so a 2:1 improvement in S/N. Noise adds by using the square root, so with a read noise of say 2e, the 2x2 binned read noise is sqrt( 4 * (2^2))  = 4e. 🙂

Alan

Wait noise adds with the square root?? How does this work?

[symmetal]

Noise is random so one sample has no correlation to the previous or next sample, unlike the signal which will add samples together linearly. If we had a noise free signal arriving that registered say 10 units arriving per second on the sensor we would register 10, 20, 30, 40, 50 units on the sensor over 5 seconds.

As there is noise present the actual values arriving wouldn't be 10, 10, 10, 10, 10 but something like 9.9, 10.2, 9.8, 10.0 10.1

On those five samples the noise is -0.1, 0.2, -0.2, 0.0, 0.1 units. In reality the noise will be a random +ve or -ve value away from the expected 10 units, with the magnitude of the variation dependent on the poisson  distribution curve so there will be many more small noise variations compared to large noise variations.

You can consider these noise samples as vectors pointing either in either a +ve or -ve direction, with the length of the vector being how far away it is from the noise free value of 10. Over many samples it's likely that the next noise sample will tend to have a value averageing about 0 so it's vector direction likely ends up near 90 degrees to the previous sample which was possibly slightly +ve or slightly -ve.

Over many samples these noise vectors can be considered to be at 90 degrees to each other so the final noise value is the vector addition of these vectors, which is the hypotenuse of the triangle formed by the two noise vectors. Pythagoras says the length of the hypotenuse (of a right-angled triangle) is the square root of the sum of the squares of the other two sides. Hence my initial formula adding the four noise vectors together using Pythagoras. 🙂

I'm sure that was as clear as mud. 😁

That's my best understanding as to why noise vectors are considered to be at 90 degrees to each other, but I'm sure @vlaiv will be able to give you a better answer if I'm wrong.😉

Alan

Edited March 17 by symmetal

[ONIKKINEN]

If you swamp read noise by some factor at bin1, then any level of binning further will also be swamped by the same amount so you dont really need to worry about extra effective read noise at bin2.

I'd go with the mono camera, if within budget. The price on that thing makes me a bit light headed though... Better test current camera performance first and then figure out if full frame is worth it.