The effect of ISO

[Mr niall]

Hello all

Probably a daft question (stupid boy!) but I've been pondering the effects of ISO on DSLR imaging.

Due to my roaring ineptitude I limit myself to relatively short exposures with a wide lens (recently bought the canon nifty 50 based on recommendations here and what a lens!) but I'm aware that wide field exposure in suburban / rural areas is often limited less by accuracy of setup and more by the blowout caused by light pollution. The advice I have always received is to lower the ISO to compensate for this and therefore still obtain longer exposures.

And it is this I have been pondering. I understand how ISO works (at a rudimentary level probably) and obviously that in its most simplistic terms its analogous to increasing the sensitivity of the sensor at the expense of introducing noise. I quite liked the way that somebody explained to me once that it was similar to the "gain" control on a guitar amp; the way you can dial it right up but eventually it just becomes a distorted smush (mine goes up to 11!!).

My question is (sorry took a while to get there) is that does taking a longer exposure picture at a lower ISO provide a fundamentally better picture than a short exposure picture with a higher ISO? I'm not talking extremes here - I normally work in the 400-1600 ISO range and with exposures from 20 to about 120 seconds. But what I'm wondering is; when I'm getting LP blowout at say 90 seconds with 1600 ISO should I be:

• lowering the ISO to preserve the exposure length?
• lowering the exposure length to preserve the better level of sensitivity. afforded by higher ISO?

How would those changes fundamentally impact the image?

I don't dither or bin, I just stack and stretch, and I always take darks (but I rarely take flats as I just can't seem to get them to work at really short focal lengths).

Any insights appreciated.

Many thanks all!

[Scooot]

These articles are very helpful.

http://dslr-astrophotography.com/iso-dslr-astrophotography/

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

In short, there is an ISO that is the most appropriate for your camera. If you increase the ISO above that level it is of no benefit because it increases the noise by a corresponding amount. It probably also reduces your dynamic range so makes it worse.

 

[vlaiv]

Disclaimer: I don't do DSLR AP, but understand what happens, so will try to explain.

Let's first talk about what ISO is.

ISO is in principle multiplicative constant. As such it does not impact your image in any way with regards to SNR - which is important thing in AP. If you have some sort of signal that is let's say 100 units, and associated noise is 10 units (photons / electrons, whatever), and you apply high iso - you will be multiplying with some number like 10, so resulting signal will be 1000 and noise will be 100. Signal to noise ratio remains the same. Image remains the same once you scale it to display range 0-255. So in principle nothing major happens with high / low ISO with regards to captured data.

On the other hand, there are things that happen when you change your ISO settings - read noise depends on ISO, and usually low ISO has higher read noise, and higher ISO has lower read noise. This will have some importance later in discussion. Another thing that ISO does is change your effective full well capacity.

Let's say you have 14 bit DSLR. This means that digital values produced by your camera can have values in range 0-16383 (integer values). Actual signal that pixels detect can and usually is higher than this. In order to map hole range that pixel can receive to limited range of 14 bits - you need to divide it with something. Let's say your actual full well capacity is 60000 electrons. You need to multiply that with 0.25 to get values in range of 0-15000. Such values fit 0-16383. But when you select higher ISO, you will be using only portion of your true full well capacity. Sometimes with very high ISO you might end up having less than 14bits effective full well capacity. For example if certain ISO has multiplicative constant of 2, then actual electron count from 0-8191 gets mapped to output of 0-16383, and you end up having smaller effective full well capacity.

Now back to your question about blow out.

Only thing that makes a difference between stacking a few longer subs vs many shorter subs (of the same total exposure) is read noise. Or to be precise, how "important" read noise is. There are a few sources of noise (shot, thermal, LP, read noise) and difference between few longer vs many shorter comes down to how "big" read noise is compared to other noise sources. Maybe best way to illustrate this is by using triangles, as noise adds as linearly independent vectors:

If read noise is close in magnitude to other noise sources like LP noise this happens:

Total noise is larger than both read noise and LP noise.

But if read noise is far less in magnitude to some other noise source - in this case LP noise, than this happens:

So difference that read noise makes gets minimal and there is almost no difference between few longer subs vs many short subs.

With lower ISO you will effectively have larger full well capacity, and LP signal levels will not occupy significant portion of the range, so single sub will not look blown out, but lower ISO will have larger read noise - which means you need to again swamp your read noise with LP noise in order not to make difference.

What all of this means is: yes you can change ISO but there won't be any significant improvement in stacking result. If you have high background LP signal at 90 seconds - just use those exposures - going longer simply won't produce better results.

 

[wxsatuser]

On 12/08/2019 at 22:07, Mr niall said:

Hello all

Probably a daft question (stupid boy!) but I've been pondering the effects of ISO on DSLR imaging.

My question is (sorry took a while to get there) is that does taking a longer exposure picture at a lower ISO provide a fundamentally better picture than a short exposure picture with a higher ISO? I'm not talking extremes here - I normally work in the 400-1600 ISO range and with exposures from 20 to about 120 seconds. But what I'm wondering is; when I'm getting LP blowout at say 90 seconds with 1600 ISO should I be:

• lowering the ISO to preserve the exposure length?
• lowering the exposure length to preserve the better level of sensitivity. afforded by higher ISO?

How would those changes fundamentally impact the image?

I don't dither or bin, I just stack and stretch, and I always take darks (but I rarely take flats as I just can't seem to get them to work at really short focal lengths).

Any insights appreciated.

Many thanks all!

ISO is not really a measure of sensitivity of the camera, QE and the construction of the sensor is.

Compared to the lower ISO we should not decrease the exposure at the higher ISO, if you do the SNR will go down.
We do this in normal daylight photography but should not do it for AP.

[Stargazer33]

I believe this video has been linked to before but I can't find it!

 

[JamesF]

It has.  There's a second link somewhere to the bit he ran out of time for at the end, too.

James

[Scooot]

I had watched the Dr Robin Glover video before but have just rewatched it.

Does anyone know how I can work out my light pollution in (e) at my location for his formula In my screenshot below? Apologies if I’ve missed it in the presentation.

[vlaiv]

17 minutes ago, Scooot said:

I had watched the Dr Robin Glover video before but have just rewatched it.

Does anyone know how I can work out my light pollution in (e) at my location for his formula In my screenshot below? Apologies if I’ve missed it in the presentation.

Rather simple - you take a sub of "empty" piece of the sky (there really is no empty piece of the sky, but aim to miss milky way, major galaxy or nebula, also aim for sparse star field) and you calibrate your sub with dark (you can also do flat, but in principle it's not needed if you measure in central region to avoid vignetting and you also avoid any major dust shadows).

Then you measure average signal in each channel - it will give you sky flux when you divide it with exposure length and number of pixels used to measure. You also need to know your e/ADU gain and convert measured ADU values to electrons.

[Scooot]

6 minutes ago, vlaiv said:

Rather simple - you take a sub of "empty" piece of the sky (there really is no empty piece of the sky, but aim to miss milky way, major galaxy or nebula, also aim for sparse star field) and you calibrate your sub with dark (you can also do flat, but in principle it's not needed if you measure in central region to avoid vignetting and you also avoid any major dust shadows).

Then you measure average signal in each channel - it will give you sky flux when you divide it with exposure length and number of pixels used to measure. You also need to know your e/ADU gain and convert measured ADU values to electrons.

Thanks Vlaiv,

So in Pixinsight the readout of a section of the background, the R,G & B in the image is the signal?

Edited August 19, 2019 by Scooot

[vlaiv]

3 minutes ago, Scooot said:

Thanks Vlaiv,

So in Pixinsight the readout of a section of the background, the R,G & B in the image is the signal?

Yes, it should be - just be careful if it is ADU value and you might need to convert it to electrons for above to work.