A question about noise

[StuartT]

I was watching a nice youtube video about imaging M31 and stacking with a DSLR. The guy said something which surprised me. He said that noise decreases with increasing ISO. This is the opposite to what I always understood - that the higher the ISO, the more grainy your image.

Can anyone explain this for me?

[DAVE AMENDALL]

Hi , Certainly with film the higher the ISO the more grain in the image therefore you are right. Also true of digital. Sounds like he made an Oops! in his video presentation. .........Dave

[shropshire lad]

1 hour ago, StuartT said:

I was watching a nice youtube video about imaging M31 and stacking with a DSLR. The guy said something which surprised me. He said that noise decreases with increasing ISO. This is the opposite to what I always understood - that the higher the ISO, the more grainy your image.

Can anyone explain this for me?

Can you put a link to the video please.

 

[The Lazy Astronomer]

Assuming ISO on DSLRs behaves the same as the gain value on astro cams, then the video is right: sensor read noise decreases with increasing gain. 

However, sensor read noise is not the only source of noise when imaging. The actual image grain/noise does increase, because at high ISO/gain you are not capturing any more photons, just amplifying the signal more, and in amplifying that signal you are also amplifying any noise associated with it as well. 

That is my understanding anyway, happy for someone much more knowledgeable than me to come along and tell me I'm an idiot and don't know what I'm talking about! 😁

 

[davew]

I first became aware that all is not well in ISO land in the summer of 2017. I read this article from a guy named Ian Norman -

https://www.lonelyspeck.com/how-to-find-the-best-iso-for-astrophotography-dynamic-range-and-noise/  

I tried a few experiments of my own and it turned out he was telling the truth ! It must be stressed that this ISO result depends to some extent on the camera and if it's ISO variant or invariant. It also only applies where the read noise is the dominant source of noise. In other words, if light pollution is the dominant noise then you won't see the read noise but you will at a dark site. If you need persuading that the read noise reduces when you raise the ISO then have a look here - 

https://www.photonstophotos.net/Charts/RN_e.htm

That's the good news 😎 The bad news is that dynamic range also decreases - 

https://www.photonstophotos.net/Charts/PDR.htm

I've also tried that experiment and in real life Dynamic range reduction isn't as bad as some might suggest especially when stacking. 

So, should we raise the ISO to the max ( In some cameras to almost half a million ISO ) and have no noise ? No ! There comes a point where there is no Dynamic range worth talking about. I would suggest in most ISO variant cameras that you push up to 1600 or 3200 and stop there.

Alyn Wallace gives a good explanation on the latest cameras with invariance and the older cameras with variance -

https://www.youtube.com/watch?v=d8QV00mkJW4

The video gets interesting around the 4.30 mark and shows what happens in most cameras.

So why can so many people prove beyond all reasonable doubt that raising the ISO totally wipes out the image with noise ? Simple. If you follow the exposure triangle you'll know that for one stop more of ISO you need one stop less of exposure. So if I take an image at 20 seconds at ISO 100 that equates to 10 seconds at ISO 200, 5 seconds at 400, 2.5 seconds at 800 etc. No one EVER mentions the fact that the light is being reduced in half every time the exposure is reduced ! So 2.5 seconds at 800 is 12.5% the amount of light that was gathered at 20 seconds ISO 100. Who would have thought the image would have been noisier ?

It turns out that some cameras have twin gain circuits where a second stage kicks in. In my camera, the z6, and the Canon R5, this occurs around ISO 800 so at that setting the read noise drops significantly.

Right, lunch calls and my fingers are falling off 😀

Dave.

Edited March 28, 2021 by davew

[StuartT]

9 hours ago, DAVE AMENDALL said:

Hi , Certainly with film the higher the ISO the more grain in the image therefore you are right. Also true of digital. Sounds like he made an Oops! in his video presentation. .........Dave

I don't think he did. He was explaining that you have to balance ISO and noise. Dynamic range decreases with ISO, but noise increases. So you want the sweet spot between the two. He pointed to a rather nice site called photons to photos which allows you to see this noise vs ISO profile for any camera. Here is mine (screenshot)

8 hours ago, shropshire lad said:

Can you put a link to the video please.

 

 

Edited March 28, 2021 by StuartT

[StuartT]

6 hours ago, davew said:

I first became aware that all is not well in ISO land in the summer of 2017. I read this article from a guy named Ian Norman -

https://www.lonelyspeck.com/how-to-find-the-best-iso-for-astrophotography-dynamic-range-and-noise/  

I tried a few experiments of my own and it turned out he was telling the truth ! It must be stressed that this ISO result depends to some extent on the camera and if it's ISO variant or invariant. It also only applies where the read noise is the dominant source of noise. In other words, if light pollution is the dominant noise then you won't see the read noise but you will at a dark site. If you need persuading that the read noise reduces when you raise the ISO then have a look here - 

https://www.photonstophotos.net/Charts/RN_e.htm

That's the good news 😎 The bad news is that dynamic range also decreases - 

https://www.photonstophotos.net/Charts/PDR.htm

I've also tried that experiment and in real life Dynamic range reduction isn't as bad as some might suggest especially when stacking. 

So, should we raise the ISO to the max ( In some cameras to almost half a million ISO ) and have no noise ? No ! There comes a point where there is no Dynamic range worth talking about. I would suggest in most ISO variant cameras that you push up to 1600 or 3200 and stop there.

Alyn Wallace gives a good explanation on the latest cameras with invariance and the older cameras with variance -

https://www.youtube.com/watch?v=d8QV00mkJW4

The video gets interesting around the 4.30 mark and shows what happens in most cameras.

So why can so many people prove beyond all reasonable doubt that raising the ISO totally wipes out the image with noise ? Simple. If you follow the exposure triangle you'll know that for one stop more of ISO you need one stop less of exposure. So if I take an image at 20 seconds at ISO 100 that equates to 10 seconds at ISO 200, 5 seconds at 400, 2.5 seconds at 800 etc. No one EVER mentions the fact that the light is being reduced in half every time the exposure is reduced ! So 2.5 seconds at 800 is 12.5% the amount of light that was gathered at 20 seconds ISO 100. Who would have thought the image would have been noisier ?

It turns out that some cameras have twin gain circuits where a second stage kicks in. In my camera, the z6, and the Canon R5, this occurs around ISO 800 so at that setting the read noise drops significantly.

Right, lunch calls and my fingers are falling off 😀

Dave.

this sounds like exactly what he was saying. Thanks!