Canon 1300D ISO - Read Noise - Dynamic Range... = Exposure Time

[RolandKol]

Hi Guys, 

Long story short:

Is it worth doing 4 min exposures with ISO 100 using Canon 1300D (not yet moded..) on DSO.

Or it's better to stick with default ISO 800 and 1min? 

@ISO800 1min is the top exposure I can use from my back garden in London even with a light pollution filter on....

 

Long Story:

I "recently" managed to get all beginner's equipment and started to experiment with long exposures (I had DOB previously).

"Recently" - almost 2 months ago, and during that period I managed to play with it 3 times only... London is like the Adams Family house, - constantly under the clouds...

I have Canon 1300D and SW 130PDS guided via ZWO224 (not completely guided yet... as still learning ).

Initially, I thought I will figure out the best ISO/exposure settings while experimenting...

But 1 experiment per month... VERrrrY painful... and it would be the top rate masochism to enjoy experiments with such a frequency...  

So I have been googling, reading forums and even going through the astrobin/flickr pictures and checking ISO/Exposure parameters, checking all kind of data sheets with sensor data, but managed to find only for Canon  1200D like for example http://sensorgen.info/CanonEOS-1200D.html

or http://www.photonstophotos.net/Charts/DXOPDR.htm and nothing for 1300D. If you have any link, you are welcome to share.

From all data I have read for my range CANON ... usually, - ISO800 is recommended or even ISO1600!!! - Wow!!! Seriously???

I used ISO1600 with my DOB for 10sec exposures

As far as I learned, - All DSLR cameras have the best Dynamic Range with the lowest ISO possible, with 1300D it's ISO100, so I would expect to get the best DSO colours and shadows using ISO 100,

However, I was amazed founding out that Canons have the Highest Read Noise at low ISO settings... (I was not able to find any data for 1300D, but I guess data from 1200D sensor is more or less the same).

It looks like a sweet spot would be ISO 400 which would be 1.5 - 2min exposures maximum...

Please correct me if I am wrong.

P.S.

Last time I tried ISO800 and 1min exposures, - after stacking with all darks, flats and bias the result was very noisy.

I even afraid thinking about the amount of noise with ISO100, but I will try it next night anyway...

Curiosity ...

 

[michael.h.f.wilkinson]

This website has a lot of information, but the 1300D seems to be missing

http://dslr-astrophotography.com/iso-values-canon-cameras/

[Handy_Andy]

I came across this link in the forum some time ago http://dslr-astrophotography.com/iso-values-canon-cameras/ unfortunately no info on the 1300d but it does show you the best iso vs dynamic range for subs. Noise is reduced with increasing numbers of subs.

Quite a few people on here don't rate darks but use bias as a master dark and dithering to eliminate hot pixels during stacking, something for you to look into as I'm only really starting down that track myself.

With the light pollution LED are the bane of astronomy now as they are broadband wavelengths instead of a nice easily filtered emission line. As I understand it the only ways around it are to travel to darker skies or look into the deep dark abyss that is narrow-band imaging.

 

hope this helps you on your way

[ChrisEll]

Trouble with using lower ISOs you end up losing a lot of imaging time as it takes longer to collect the photons. Choosing the ISO is a compromise... I've been using the standard 800 for quite a while but I'm not getting the total time needed and I will experiment with 1600.

[drjolo]

2 hours ago, RolandKol said:

As far as I learned, - All DSLR cameras have the best Dynamic Range with the lowest ISO possible, with 1300D it's ISO100, so I would expect to get the best DSO colours and shadows using ISO 100,

However, I was amazed founding out that Canons have the Highest Read Noise at low ISO settings... (I was not able to find any data for 1300D, but I guess data from 1200D sensor is more or less the same)

Yes, most DSLR and also CMOS cameras has lower read out noise at higher gain/ISO. The reason is that read out noise contains two components - first one comes before signal is amplified, and second after amplification. When you increase gain/ISO, then first component is also increased, but second one is not. That's why read noise (defined in electrons) can be lowered for higher gain. 

This is however not the case for ISO-less cameras (like Canon 80D and many Nikons), where read noise is proportional to gain/ISO even for low settings. For these cameras you can do astroimages even at low ISO 100-200 and stretch image later. It was not confirmed yet for Canon 1300D, but there is 99% probable, that this model is not ISO-less, so you should use higher ISO 400-800, shorter times, and then stack more images. This way you can limit read noise, and dynamic range can be still increased by stacking many frames. However under heavy LP sky the difference between low and high ISO images will not be large, because read noise will be only small part of total noise. 

You may check if your calibration workflow is correct - try to stack uncalibrated images and compare result to calibrated ones. Also use many calibration frames - for 1 minute subs use at least 30-40 darks, 20-30 flats and you can use many bias frames 50 or even 100. Also ambient temperature is important, it is worth to do darks in similar temperature as lights were done. Another thing is stacking algorithm - if you have more than 10-15 light frames make sure you use some pixel-rejection algorithm, not just simple sum or average. Dithering also helps.

[vlaiv]

Yes, ISO800 - 1600 is usually recommended and for a reason.

Full well capacity is usually larger than bit depth with DSLRs. With 14bit precision one is able to record 16384 levels of signal. Full well capacity is usually >20K and can often go up to 50K for large pixel sizes.

With astro imaging, ideally one wants 1:1 mapping between electron count and ADU value, but that is not practical in day time photography. So lower ISO values are obtained by larger ADU/e- conversion factor (larger than 1). This means that there will be additional noise injected - and bad kind of noise, called quantization noise. In day time photography that does not matter, one uses low ISO settings when there is plenty of light, so SNR is not affected much. It helps take advantage of whole full well capacity (for example having 30K full well capacity, and having conversion factor of 2 will get you values in range of 0 - 15000, and that fits in 0 - 16384 provided with 14bit ADC).

Just as simple explanation of what kind of noise we are dealing with: let's say you capture 16e of signal and your conversion factor is 2, actual recorded ADU value will be 8. So far so good. But let's suppose that you captured 15e of signal, and with same conversion factor 2 you will still get 8 (you can't have 7.5 on integer ADC). So when you try to reconstruct original signal level in first case you will get right value, but in second case you will have some noise (1e worth of noise) injected. This kind of noise does not have "normal" (and in fact normal meaning gaussian) distribution, and does not "behave well".

So in addition to having smaller read noise in higher ISO settings, you avoid this kind of "value clipping" noise that comes when you have conversion factor bigger than 1. It can also happen for some values of conversion factor smaller than 1, but when it happens it is usually with values where signal is high, so SNR is still good.

[drjolo]

29 minutes ago, vlaiv said:

Just as simple explanation of what kind of noise we are dealing with: let's say you capture 16e of signal and your conversion factor is 2, actual recorded ADU value will be 8. So far so good. But let's suppose that you captured 15e of signal, and with same conversion factor 2 you will still get 8 (you can't have 7.5 on integer ADC). So when you try to reconstruct original signal level in first case you will get right value, but in second case you will have some noise (1e worth of noise) injected. This kind of noise does not have "normal" (and in fact normal meaning gaussian) distribution, and does not "behave well".

I think you considered here only one part of quantization noise - clipping error. In this case for both 15 and 16e signal you would get always 8ADU. But there is another component - approximation error that is +- 1/2ADU, so 15e signal can be both 7 or 8ADU. You cannot overcome quantization noise, when signal noise is zero*. But you can, when signal noise is not zero - and that is the case for photon signal, because it contains intrinsic photon noise. In this case quantization noise can be reduced with stacking subframes with the same or (better) different exposure time. 

* actually you can using different subframe exposure times

[vlaiv]

3 minutes ago, drjolo said:

I think you considered here only one part of quantization noise - clipping error. In this case for both 15 and 16e signal you would get always 8ADU. But there is another component - approximation error that is +- 1/2ADU, so 15e signal can be both 7 or 8ADU. You cannot overcome quantization noise, when signal noise is zero*. But you can, when signal noise is not zero - and that is the case for photon signal, because it contains intrinsic photon noise. In this case quantization noise can be reduced with stacking subframes with the same or (better) different exposure time. 

* actually you can using different subframe exposure times

Yes, indeed such noise is reduced by stacking, so there is no problem in that. It can be even "masked" off by existence of read noise (combination is closer to well behaved noise).

I'm just pointing out that it is additional source of noise that can be avoided by choosing different ADU/e- (or in case of DSLR ISO setting), and that this kind of noise is not "well behaved" noise - meaning that you can't control how much SNR increase there will be by using a stack of N subs (unlike poisson and gaussian noise), and that in calculations contribution of such noise is not straight forward to do.

[RolandKol]

Thanks guys, I have got the point  

While reading, - I felt like in the lecture of Quantum-Nanomechanics in the Chinese language and probably misunderstood some of the statements, but the conclusion.

My soul and body still rejects ISO1600.

And ISO800 seems like the option I will keep on going with and will try to master...

After short experiments with ISO100/400 (curiosity must be fed!   )

I still hope we will get at least one clear night within next week and I will manage to experiment on Leo Triplet again.

 

[newbie alert]

The histogram will tell you all you need to know,but basically the longer you expose the lower the iso..im  big fan of iso 400-800 ...but have dropped lower..

I only ever go higher if I'm using a HA filter or if I'm on a static tripod with real short 15 sec exposures

[SteveNickolls]

Hi Rolandkol,

Thanks for posting.

Why not experiment with lower ISO values and see what effect this has on the final images, you really have little to lose and a lifetime's experience to gain? Somewhere in all of this your local light pollution plays its part. You might also want to read this-https://www.cloudynights.com/topic/500991-is-dslr-unity-gain-useful/ And this-http://dslr-astrophotography.com/iso-values-canon-cameras/ including the comments at the bottom. 

I have two canon DSLR's but unfortunately neither are your 1300D model. I've come to view my DSLR's as black boxes that use proprietary software built by humans and contain circuitry made for specific (and non-astro) purposes. In that perspective they may or may not behave as they are supposed to according to theory for the purposes you want to put them to for astro-photography, so it makes it even more intriguing to experiment and draw your own conclusions.

Do let us know the outcomes of any lower ISO experimentation you try should you fancy exploring this field.

Best Regards,
Steve

[RolandKol]

3 hours ago, SteveNickolls said:

Hi Rolandkol,

Thanks for posting.

Why not experiment with lower ISO values and see what effect this has on the final images, you really have little to lose and a lifetime's experience to gain? Somewhere in all of this your local light pollution plays its part. You might also want to read this-https://www.cloudynights.com/topic/500991-is-dslr-unity-gain-useful/ And this-http://dslr-astrophotography.com/iso-values-canon-cameras/ including the comments at the bottom. 

I have two canon DSLR's but unfortunately neither are your 1300D model. I've come to view my DSLR's as black boxes that use proprietary software built by humans and contain circuitry made for specific (and non-astro) purposes. In that perspective they may or may not behave as they are supposed to according to theory for the purposes you want to put them to for astro-photography, so it makes it even more intriguing to experiment and draw your own conclusions.

Do let us know the outcomes of any lower ISO experimentation you try should you fancy exploring this field.

Best Regards,
Steve

Thanks Steve,

I read your links previously and even several times.

I will experiment with low ISO values for sure as my curiosity will not allow me to go blindly by data from the internet, however, as I mentioned above, cloudy nights are not allowing to do it properly

And I wanted to learn about the optimal option ISO for my camera, which, unfortunately, has no data on the internet even to compare.

I saw some articles, where people been experimenting with total exposure times, - I will try to do the same...

Only if the Gods Of Clear Skies will start to listen to me!  

[SteveNickolls]

7 minutes ago, RolandKol said:

Only if the Gods Of Clear Skies will start to listen to me!

Yes we could all do with some clear, dark skies. I wish you every success with your experimetation.

Best Regards,
Steve

[Redscouse]

Hmmm.. 

To the OP; I have the same setup as you although probably nowhere near the same level as LP as you do.
I usually take 300s at ISO400 and they 'seem' ok. (If I drop down to 90s I tend to move to ISO800) Disclaimer: I am no expert! 

I say seem because I after reading this thread I feel like I've just been told off and I must use a higher ISO so will give it a try at the next opportunity. 

Thanks for the posts guys, some knowledge you have there! 

[RolandKol]

Last week and yesterday, at LAST, I had some luck with the clear skies. (waited soooo long)

At the start, I struggled with guiding issues... lost few nights... fell into depression...

had some beer... BBQ...

at the end of the week sorted out the guiding issues,

Targeted on M51 and started from ISO200 - 300sec. 
The APT histogram was slight to the right side from the middle... 
Stack was really noisy, fought to get some galaxy detail with Photoshop masking, high pass filters and etc, spent around 3 hours to get something descent to look at.

Yesterday night, - started with ISO400 - 300sec. 
The APT histogram was in quite close to the right (I have not even tried 330 or 360sec - it would be obviously overexposed). And I have spoiled my experiment, as I have decided to test APT dithering, it worked perfect and I left it ON (distance 2).

The result: details of the galaxy and star colours are Much, Much better.  

I do not even want to go back and test ISO200 again with dithering, as the difference is enormous.

However, after next clear night, I will post the comparison between ISO400-300sec and  ISO800-150sec, both with Dithering enabled.  (I guess 150sec is the correct exposure for comparison, if not, - let me know).
 

[RolandKol]

Ok,

The night was quite bright... Moon was very shiny...

So went on experiment only, in APT set routine to make 3 subs of each ISO several times.

I thought Moon will spoil everything

However, the experiment was more or less successful, - RGB histogram has not touched the right side, but was extremely close.

Not a perfect experiment... but 

Long story short,

My First impression, - ISO800 wins at the moment....  but I am not happy, I think, I need to match a total exposure time to make a proper comparison.

If anyone needs source files to investigate, - just let me know, I will keep them for a while.

I have stacked all subs in DSS and outstretched in PI Demo.

ISO	Max DSS Score	Min DSS Score	Median DSS Score	Avarage DSS Score	Bias Subs	Dark Subs	Flats	Lights	Lights Exposure in sec	Stars Found on 2% (Median Filter Enabled)	Stars Found on 10% (Median Filter Enabled)	Total Integration
400	2,494.60	1,124.31	2,327.18	2,228.89	96	12	36	16	300	402	155	1h20min
800	2,620.48	1,480.78	2,533.33	2,380.46	100	14	55	16	150	373	150	40min
1600	1,849.42	889.56			100	15	51	15	75	401	161	18min 45s
3200	1,486.80	731.28			100	15	51	15	37	406	165	9min 15s

 

ISO400 (Background removed and outstretched by demo PI)

M51_ISO400_16_frames_300s_[1h20min]

 

 

ISO800 (Background removed and outstretched by demo PI)

M51_ISO800_16_frames_150s_[40min]

ISO1600 (Background removed and outstretched by demo PI)

M51_ISO1600_16_frames_75s_[18min_45s]

 

ISO3200 (Background removed and outstretched by demo PI)

M51_ISO3200_16_frames_37s_[9min_15s]