Sub length vs Sub count?

[Christopher Davenport]

So been at the this imaging malarchy for a couple of months and sometimes things just don't seem rational.

Why do I get better results from 30 x 300s subs than from 15 x 600s?

I suspect that it is the signal to noise ratio of the longer exposures. i.e DSLR noise increase with exposure time.

But maybe I am missing something?

[Pint Of Stellar]

I was just reading about this in Bracken's Deep Sky Imaging Primer the other day and yes, to cut a long story short (which you can also read about in the book) it is a signal/noise ratio thing.

[nmoushon]

It boils down to signal/noise ratio. But lots of things factor into that. Such as your local sky conditions, weather, the camera you are using, the target you are imaging, ect.

DSLRs are much noiser than CCD and unless your DSLR is modded to be cooled it will heat up much more the longer the subs are. Usually darks will take care of this. So make sure you are using seperate darks for those two exposure lengths and taking enough darks as well. I alsways went with at least half as many darks as lights. Also with an unmodded DSLR your  darks are also temperature dependant. So you will need plenty of sets of darks at set exposure lengths at different temperatures.

Local LP can make a 600sec sub look like complete trash while the 300sec sub looks excellent even though normal thinking would claim the 600sec sub should be better. Especially when stacked. This is a common mistake that effects everyone starting out. Myself included. I even fell for it again after 2 years of imaging with a dslr when I switched to CCD. Yes, it even effects CCD cameras. Understanding the limits of the location you're imaging from is very important and can be tricking to find out what it is. I have still trying to figure mine out since moving 9months ago. (Mainly due to weather and my laptop dieing.)

[lensman57]

So been at the this imaging malarchy for a couple of months and sometimes things just don't seem rational.

Why do I get better results from 30 x 300s subs than from 15 x 600s?

I suspect that it is the signal to noise ratio of the longer exposures. i.e DSLR noise increase with exposure time.

But maybe I am missing something?

You need both in fact, the longer the sub ( subject to your  sky condition ) the stronger the collected signal, the more the subs then the less the noise as the result of stacking. DSLRs do need in excess of 35 subs if noise is to be kept away and CCDs usually have the edge here but by no means it is a done and dusted matter as a lot of factors can affect the outcome.

A.G

[nmoushon]

You need both in fact, the longer the sub ( subject to your  sky condition ) the stronger the collected signal, the more the subs then the less the noise as the result of stacking. DSLRs do need in excess of 35 subs if noise is to be kept away and CCDs usually have the edge here but by no means it is a done and dusted matter as a lot of factors can affect the outcome.

A.G

This is a good point I forgot to mention. When imaging with DSLR you will need HOURS of data to keep the noise down. I took 8 hrs of 2min sub on M42 to keep the noise down to be able to really pull out the data that is in the 2min subs. Thats 240 subs plus darks, bias and flats. Now if I was to shoot the same 2min subs but with my CCD I'm sure I could achieve the same desired data quality in about 4hrs or so worth of data. When you get longer subs the amount needed goes down because you are increasing the signal you are getting. With DSLR you are also increasing the noise but as i said above darks take care of most of it but not all. So go back and take another 2-3 hrs of data for each set then compare them. I'm sure you will notice an improvement. Unless your sky is your limiting factor here and ruining your 600sec subs. You might be limited to 300sec. I know that is my limit in my western sky because of LP, even with a LP filter too.

[lensman57]

This is a good point I forgot to mention. When imaging with DSLR you will need HOURS of data to keep the noise down. I took 8 hrs of 2min sub on M42 to keep the noise down to be able to really pull out the data that is in the 2min subs. Thats 240 subs plus darks, bias and flats. Now if I was to shoot the same 2min subs but with my CCD I'm sure I could achieve the same desired data quality in about 4hrs or so worth of data. When you get longer subs the amount needed goes down because you are increasing the signal you are getting. With DSLR you are also increasing the noise but as i said above darks take care of most of it but not all. So go back and take another 2-3 hrs of data for each set then compare them. I'm sure you will notice an improvement. Unless your sky is your limiting factor here and ruining your 600sec subs. You might be limited to 300sec. I know that is my limit in my western sky because of LP, even with a LP filter too.

You have got it right 100%. Another problem  with DSLR Darks are due to the lack of set point cooling, imagine taking 30X600s subs and then another 25x600s Darks, the chances are that due to  significant temperature variation the Darks will be mismatched and you'd need a sophiscated software to scale the Darks, the situation is more managable during long cold winter nights but in a hot summer eveing there is no chance. Mismatched Darks create more problems that they solve.

A.G

[wxsatuser]

If you have'nt seen Samir Kharusi's site, take a look, good explanation of sub lengths and noise.

http://www.samirkharusi.net/sub-exposures.html

[wxsatuser]

 Re darks.....

I am in the process of making darks at various subs and temps, doing this on cloudy nights.

If a master dark is made then you really want at least 64 darks or may be 100 at the most.

64 frames will only add 1/8 of the noise that a single dark adds and 1/10 for 100 frames.

So your master dark really wants to be a good quality one.

[Pete Presland]

 Re darks.....

I am in the process of making darks at various subs and temps, doing this on cloudy nights.

If a master dark is made then you really want at least 64 darks or may be 100 at the most.

64 frames will only add 1/8 of the noise that a single dark adds and 1/10 for 100 frames.

So your master dark really wants to be a good quality one.

when i was told this by Olly on here, i was very surprised.  i cant believe how much time i wasted taken them very night i was out! when you say "good quality" what do you mean by good quality?

[Christopher Davenport]

Thanks everybody. Starting to make sense now.

I do use darks created in my fridge for 15c sensor temp.

But never realised that I need same amount of lights for longer exposures.

[sharkmelley]

In theory there should be very little to choose between 30x300 and 60x150 on a DSLR.  If anything the longer subs might be very marginally better.  Almost certainly sky conditions or a difference in ambient temperature played a major part.

During a long DSLR imaging session, the sensor temperature will be increasing for at least the first hour of the session before it stabilises.  So do the darks at the end of the session.  If you are shooting library darks during a diffferent session then throw away the first hour's worth since each one will have a different thermal characteristic. 

If possible, during processing, use sofware that "matches the dark" individually to each light (since each light will have a different thermal characteristic) - the software does this by scaling the dark to best fit the noise found in the light frame.

Mark

[sharkmelley]

One thing I should have added is that if your software matches the dark to the lights then you don't need a big library of darks - in fact I use just one set of darks taken at a relatively warm 15C ambient (15C is a warm nighttime tmeperature in the UK).  They get scaled down appropriately for imaging sessions done on colder nights.  I reshoot this set of darks once a year, just in case the sensor changes in the meantime.

Mark

[wxsatuser]

when i was told this by Olly on here, i was very surprised.  i cant believe how much time i wasted taken them very night i was out! when you say "good quality" what do you mean by good quality?

The signal-to-noise ratio is improved in light frames by the square root of the total number of frames.

Noise is reduced in the master dark by the square root of the total number of dark subs.

By good quality, just take a lot, 64 is very good at a pinch 100 is ok but after that the returns are limited.

[lensman57]

The signal-to-noise ratio is improved in light frames by the square root of the total number of frames.

Noise is reduced in the master dark by the square root of the total number of dark subs.

By good quality, just take a lot, 64 is very good at a pinch 100 is ok but after that the returns are limited.

I don't do much DSLR imaging now but for both the 428EX and the 314L+ with subs up to 1200s and @ -15c no darks are necessary but I do shoot 101 Bias frames and atleast 51 Flats at the end of each session. On a warm night,  with regret both Atiks fail to reach -20c  so the necessaity to shoot the Bias frames at the end of each session, this takes about two minutes.

Regards,

A.G

[wxsatuser]

In theory there should be very little to choose between 30x300 and 60x150 on a DSLR.  If anything the longer subs might be very marginally better.  Almost certainly sky conditions or a difference in ambient temperature played a major part.

During a long DSLR imaging session, the sensor temperature will be increasing for at least the first hour of the session before it stabilises.  So do the darks at the end of the session.  If you are shooting library darks during a diffferent session then throw away the first hour's worth since each one will have a different thermal characteristic. 

If possible, during processing, use sofware that "matches the dark" individually to each light (since each light will have a different thermal characteristic) - the software does this by scaling the dark to best fit the noise found in the light frame.

Mark

I found my Canon 60Da  takes about 30mins to stabilise.

Here is a run of 180sec darks at ISO800, 10sec gap between each frame for writing.

Ambient was 7°C, outside was 5°C, this stayed steady thru the whole session.

2014/02/09 17:32:42 10022014 006.CR2 B    180s ISO:800                Canon 60D Temp:10C

2014/02/09 17:35:52 10022014 007.CR2 B    180s ISO:800                Canon 60D Temp:13C

2014/02/09 17:39:02 10022014 008.CR2 B    180s ISO:800                Canon 60D Temp:14C

2014/02/09 17:42:12 10022014 009.CR2 B    180s ISO:800                Canon 60D Temp:14C

2014/02/09 17:45:22 10022014 010.CR2 B    180s ISO:800                Canon 60D Temp:15C

2014/02/09 17:48:32 10022014 011.CR2 B    180s ISO:800                Canon 60D Temp:15C

2014/02/09 17:51:42 10022014 012.CR2 B    180s ISO:800                Canon 60D Temp:15C

2014/02/09 17:54:52 10022014 013.CR2 B    180s ISO:800                Canon 60D Temp:15C

2014/02/09 17:58:02 10022014 014.CR2 B    180s ISO:800                Canon 60D Temp:15C

2014/02/09 18:01:12 10022014 015.CR2 B    180s ISO:800                Canon 60D Temp:15C

2014/02/09 18:04:22 10022014 016.CR2 B    180s ISO:800                Canon 60D Temp:16C

2014/02/09 18:07:32 10022014 017.CR2 B    180s ISO:800                Canon 60D Temp:16C

2014/02/09 18:10:42 10022014 018.CR2 B    180s ISO:800                Canon 60D Temp:16C

2014/02/09 18:13:52 10022014 019.CR2 B    180s ISO:800                Canon 60D Temp:16C

2014/02/09 18:17:02 10022014 020.CR2 B    180s ISO:800                Canon 60D Temp:16C

2014/02/09 18:20:12 10022014 021.CR2 B    180s ISO:800                Canon 60D Temp:15C

2014/02/09 18:23:22 10022014 022.CR2 B    180s ISO:800                Canon 60D Temp:15C

2014/02/09 18:26:32 10022014 023.CR2 B    180s ISO:800                Canon 60D Temp:16C

2014/02/09 18:29:42 10022014 024.CR2 B    180s ISO:800                Canon 60D Temp:16C

2014/02/09 18:32:52 10022014 025.CR2 B    180s ISO:800                Canon 60D Temp:15C

2014/02/09 18:36:02 10022014 026.CR2 B    180s ISO:800                Canon 60D Temp:16C

2014/02/09 18:39:12 10022014 027.CR2 B    180s ISO:800                Canon 60D Temp:16C

2014/02/09 18:42:22 10022014 028.CR2 B    180s ISO:800                Canon 60D Temp:16C

2014/02/09 18:45:32 10022014 029.CR2 B    180s ISO:800                Canon 60D Temp:16C

2014/02/09 18:48:42 10022014 030.CR2 B    180s ISO:800                Canon 60D Temp:16C

2014/02/09 18:51:52 10022014 031.CR2 B    180s ISO:800                Canon 60D Temp:16C

2014/02/09 18:55:02 10022014 032.CR2 B    180s ISO:800                Canon 60D Temp:16C

2014/02/09 18:58:12 10022014 033.CR2 B    180s ISO:800                Canon 60D Temp:16C

2014/02/09 19:01:22 10022014 034.CR2 B    180s ISO:800                Canon 60D Temp:16C

[Christopher Davenport]

One thing I should have added is that if your software matches the dark to the lights then you don't need a big library of darks - in fact I use just one set of darks taken at a relatively warm 15C ambient (15C is a warm nighttime tmeperature in the UK). They get scaled down appropriately for imaging sessions done on colder nights. I reshoot this set of darks once a year, just in case the sensor changes in the meantime.

Mark

Very interesting what software do you use to match darks to lights?

[sharkmelley]

Very interesting what software do you use to match darks to lights?

PixInsight does it best but it is also possible in IRIS.  There are probably other software applications that also do it.

A "quick and dirty" alternative is to perform 2 stacks - one with your usual dark and one with no dark (or use a dark scaled by zero).  Then perform a final blend of the 2 raw stacks that minimises the backgound noise.  This works well but not as well as scaling the dark individually to each light frame.

Also you should be a little wary of the temperature you see in the EXIF file header.  With other Canons (though I haven't used your model) I have found it is not necessarily a good predictor of the thermal noise found in the image.

Mark

[nmoushon]

Thanks everybody. Starting to make sense now.

I do use darks created in my fridge for 15c sensor temp.

But never realised that I need same amount of lights for longer exposures.

Are you sure it sensor temperature? Unless you have modded your camera I don't think there is a accurate way of checking the actual temp of the sensor itself. Sticking it in the fridge that is set to 15C will work but that is ambient temp not sensor temp. Unless it is cooled your sensor will never be the same temp as the air around the camera it will always be warmer. So taking Marks advice and just trashing the first hour or so of darks is your best option to get consistent same temp darks. But there is nothing wrong with your fridge darks. There is just a difference between sensor and ambient temp that shouldnt get confused together. This technique can be used to build a library of darks so you dont have to take darks the same night you image.

On this note though I have always be confused: If we trash the first hour or so of darks because of irregular darks due to the gradual heating of the sensor what do we do with the first hour of light subs then? The senesor is doing the exact same heating up so the first hr of lights will not have matching darks. For this reason I have never trown away any darks and just add them to the stack. Especially if I take them on another night.

[wxsatuser]

PixInsight does it best but it is also possible in IRIS.  There are probably other software applications that also do it.

A "quick and dirty" alternative is to perform 2 stacks - one with your usual dark and one with no dark (or use a dark scaled by zero).  Then perform a final blend of the 2 raw stacks that minimises the backgound noise.  This works well but not as well as scaling the dark individually to each light frame.

Also you should be a little wary of the temperature you see in the EXIF file header.  With other Canons (though I haven't used your model) I have found it is not necessarily a good predictor of the thermal noise found in the image.

Mark

I would'nt use temperature to predict thermal noise anyway.

Imagesplus does dark scaling.

You must subtract the bias from the the dark first and then the bias is subtracted from the lights separately.

Also the dark exposures need to be longer than the individual light frames.

[RikM]

...

Also the dark exposures need to be longer than the individual light frames.

Why? I always took Darks the exact same length as my lights. I used sets of 35 fridge darks at different exposure times stacked into masters to match to my lights and they always applied okay. I had a 3min, a 5min and a 7min set.

[Alien 13]

I have only taken darks once not sure it helped i have considered using the canon inbuilt darks i know it halves the imaging time but every light must get a pretty acurate matching dark with that method.

Alan

[Christopher Davenport]

Are you sure it sensor temperature? Unless you have modded your camera I don't think there is a accurate way of checking the actual temp of the sensor itself. Sticking it in the fridge that is set to 15C will work but that is ambient temp not sensor temp. Unless it is cooled your sensor will never be the same temp as the air around the camera it will always be warmer. So taking Marks advice and just trashing the first hour or so of darks is your best option to get consistent same temp darks. But there is nothing wrong with your fridge darks. There is just a difference between sensor and ambient temp that shouldnt get confused together. This technique can be used to build a library of darks so you dont have to take darks the same night you image.

On this note though I have always be confused: If we trash the first hour or so of darks because of irregular darks due to the gradual heating of the sensor what do we do with the first hour of light subs then? The senesor is doing the exact same heating up so the first hr of lights will not have matching darks. For this reason I have never trown away any darks and just add them to the stack. Especially if I take them on another night.

Set fridge temp to 5c then I leave live view running for a hour while cam cools. Then set off data capture and I get a stable set of darks at 15c.exif temp.

Plus my library is 3min 5min 7min. I always stack kappa sigma. Gets rid of excess noise.

[Pete Presland]

The signal-to-noise ratio is improved in light frames by the square root of the total number of frames.

Noise is reduced in the master dark by the square root of the total number of dark subs.

By good quality, just take a lot, 64 is very good at a pinch 100 is ok but after that the returns are limited.

are you letting DSS make your master dark during stacking or do you make a master prior to using DSS and then use that?

[sharkmelley]

For interest, here is a graph of the thermal noise of various cameras I have tested indoors under similar conditions at ambient 20C:

The horizontal axis is the exposure number of successive 5min exposures.  Note that it takes around 90min to stabilise.  In each case the rear LCD was switched off because this generates significant extra heat which is very noticeable on the exposures. The conclusion is that the master dark should be individually matched to each light because of the variation between frames.

Mark

[wxsatuser]

Why? I always took Darks the exact same length as my lights. I used sets of 35 fridge darks at different exposure times stacked into masters to match to my lights and they always applied okay. I had a 3min, a 5min and a 7min set.

I will explain this better than my rushed post at dinner time.

For a scalable master dark the subs want to be at the longest exposure you are liable to use for your lights.

The software can then scale the master for shorter sub lengths.

Your way is perfect ok.