For Comparison M33 10min vs 5min

[Earl]

Here are 2 Combined Images which have had nothing other than DBE and Histogram Stretch.

One is 12 x 600s the other is 23 x 300s so pretty much the same.

I cant see much difference between them. Any comments?

[bamus]

Hmm, interesting. I would expect more signal in 10 minutes stack and less noise in 5 minutes stack.

[Earl]

Hmm, interesting. I would expect more signal in 10 minutes stack and less noise in 5 minutes stack.

That was exactly what I was also expecting to see. these however look identical.

[bamus]

What about single frames? Is there no difference as well? Perhaps there is limit for your camera and once this limit is reached, there in no more improvement in final picture? Wonder if this is same case for DSLR cameras?

[Earl]

Interesting point, ill have a look at that later.

[dph1nm]

Why should they not look exactly the same? They are the same exposure time.

NigelM

[bamus]

Why should they not look exactly the same? They are the same exposure time.

NigelM

True, but I have always been told to aim for longest subs possible in order to increase amount of signal. Apparently 24*300 sec will not equal 12*600 sec. I will try to find source for this assumption.

[dph1nm]

24x300 will have the same signal-noise ratio as 12*600 - so if you do a histogram equalization stretch or the like they should end up looking identical on the screen.

NigelM

[dph1nm]

but I have always been told to aim for longest subs possible in order to increase amount of signal. Apparently 24*300 sec will not equal 12*600 sec

This only applies if a single sub is dominated by read noise. Otherwise all that matters is the total exposure time. In the example given, I would be surprised if a 300sec sub were significantly affected by read noise. Now if this were 7200 * 1sec I would expect too see a difference.

NigelM

[riklaunim]

Cams with low read out noise can easily handle more subs at shorter exposures vs less subs at longer exposure time (for the same total time).

[bamus]

I always get confused and my brain hurts J when discussion comes down to low/high read noise, S/n ratios etc.

So in theory there is no point to aim for 10 minutes exposures (providing that LP will let us exposure for so long) and risk bad image due to guide errors, planes, satellites. Let’s all do many 2 minutes subs and make sure to have enough of them in order to have long stacked exposure in the end. Surely this can not be right.

Can someone who knows this subject very well explain it to us?

[Earl]

True, but I have always been told to aim for longest subs possible in order to increase amount of signal. Apparently 24*300 sec will not equal 12*600 sec. I will try to find source for this assumption.

This is also the impression I was under.

[riklaunim]

So in theory there is no point to aim for 10 minutes exposures (providing that LP will let us exposure for so long) and risk bad image due to guide errors, planes, satellites...

If I want the faint outer shell of M57 or M27 I need to use very long exposures in H-a to push the very weak signal above read out noise (+/- a bit of dark current). People do 10 - 30 - 60 min exposures because they are also going for very faint signals. (and some cams may have 4 or 15 e RN ).

Strong signals esily going above the noise floor may use shorter exposures, like 5 sec Ha/OIII x 30

[ollypenrice]

Interesting. I have a dark site so skyglow isn't a problem. If the signal is faint (and M33 isn't faint in the scheme of things) then long subs are well worth having because you want to get the faint signal above the noise. In this image that doesn't really apply. Where it nearly always does apply is in NB where there is usually so little signal. Long NB subs always look less noisy.

If you just want the Leo Triplet then short subs will do fine. If you want the tidal tail of the NGC member, they won't. You need long subs.

Long subs can be a disadvantage for star colour, which can get burned out. I often use shorter colour subs and avoid adding long lum subs to the stars where possible.

Olly

[dph1nm]

because you want to get the faint signal above the noise

It only has to be above the read noise - doesn't matter if it is buried in the sky noise, as in that case adding short subs together will give you the same S/N as one long one.

However, if indeed you have very dark skies/very small pixels (in arcsec)/narrow filters or any combination thereof then it is very likely that shorter subs will be read-noise dominated. Hence the need for longer exposures.

To give an example, with my professional hat on I am involved in the Pan_STARRS and VST ATLAS imaging surveys. These both involve covering large portions of the sky with sub-1min exposures on ~2m class telescopes (with ~1/4" pixels). In the bluer bands (particularly u and g filters) where the sky is quite faint, we have a significant read-noise contribution - we would have to exposure much longer to overcome this, but then we couldn't cover as much sky in the time available, so we take the hit in depth this causes. However at redder wavelengths (r,i,z), where the sky is brighter, we do not have this problem - here we are dominated by photon counting statistics.

NigelM

[BlueAstra]

How do you know if you are above the read noise? What numbers should I be looking for in a typical Maxim DL type exposure with my SX-H16 camera?

[riklaunim]

If you see the features you want to see - then you are above :]