ISO Settings

[nephilim]

As if AP isnt confusing enough, I'm reading this http://homepages.nildram.co.uk/~sbuniak/AstroPhotog/Astrophotography_Canon_350D.htm and as regards to Iso it recommends using 1600 all the time & if using the CHDK firmware hack, going as far as 3200. His/Her reason for this is that as you'll be taking dark frames then all the noise will be removed & its better to have signal+noise rather than no signal at all. Although to me (total newbie) this makes sense, is this really the case?? As I've also read that imaging objects such as M42 the higher iso will just burn out the core?

Steve

[M00NMonkey]

Hi Steve,

As you well know, I'm a newbie too. but here's my 10pence worth

Yes, your right. 3200 ISO will definitely blow out the core, this is why you would take the shorter exposures at a lower setting, as well, so you get the core detail.

From my short experience, and from what I have read, YOU CAN NEVER HAVE ENOUGH DATA!!

Another thing to remember is that there are no real rules and everyone has their own personal preferences to what they want out of the image, so their settings are different. Best thing to do is take as many subs (lights, darks, bias, and flats) as you have time for, experimenting with different mixtures of settings and see what works for you.

I will probably be told shortly, that this is all wrong, but it is the conclusion I have come to as most people use different settings.

Hope this helps, mate

Jez

[dph1nm]

Just remember that changing ISO does NOT change the sensitivity of your camera. So you still need the same exposure times whatever your ISO.

NigelM

[Astrosurf]

It also seems to depend on the camera. My Canon EOS 1000D is quite noisy so I perefer to use ISO 800. But I have used 1600, but with more noise, especially in the red end. I do as many lights and darks as I can but I still get some noise. I'm never sure about this subject so can't give you much more than that!

[M00NMonkey]

Just remember that changing ISO does NOT change the sensitivity of your camera. So you still need the same exposure times whatever your ISO.

NigelM

I thought ISO settings change how sensitive the sensor is to light, so it does change the sensitivity of the camera. You take longer exposures to capture more light.

[Knight of Clear Skies]

I thought ISO settings change how sensitive the sensor is to light, so it does change the sensitivity of the camera. You take longer exposures to capture more light.

I think it just changes how the camera interprets the signal it receives, the actual sensitivity of the sensor is fixed. So, at a higher ISO setting, it will require less photons before a pixel 'fires'. Please take this with a small pinch of salt, I am guessing a bit here. Hopefully someone with more knowledge of how the sensors actually work will drop in and clarify matters.

[RikM]

I think of the ISO setting as a mutiplier that is applied to the chip readout. e.g ISO100 = photons captured x 100, ISO800 = photons captured x 800. It makes the image brighter (including the noise!) but doesn't change the number of photons captured in the first place. That depends on exposure time.

I tend to use ISO1600 as I find it easier to process the final image than if I start with a lower ISO, but it almost doesn't matter and you should use whatever happens to work for you.

[nephilim]

Thanks all, yeah think i'll do as Jez mentions, try diiferent iso's & see which works best. With our poxy weather the less time spent 'messing about' the better, spose we have to make do with what we get

Steve

[M00NMonkey]

I think of the ISO setting as a mutiplier that is applied to the chip readout. e.g ISO100 = photons captured x 100, ISO800 = photons captured x 800. It makes the image brighter (including the noise!) but doesn't change the number of photons captured in the first place. That depends on exposure time.

Yeah this was kind of my thinking, except I think the other way round.

i.e the lower the ISO the more photons you need to trigger the pixel, the higher the ISO the less photons you need to trigger the pixel.

I understand that you can not create more photons (unless you are of transcendent nature) and you only get X photons for (T)X amount of time, regardless of the ISO. But the ISO setting does change the sensitivity of the camera. It becomes more sensitive to photons, the higher the ISO setting.

I was just a little confused by Nigel's post.

[StuartJPP]

This is how I understand ISO in relation to digital sensors specifically in the world of astro. This is my own interpretation and is subject to misunderstanding on my behalf. A lot of assumptions here and trying to explain in layman's terms.

To make things simple lets assume that we have a 1x1 pixel sensor.

Assume the sensor is sensitive enough to detect each and every photon that hits it which in turn increases the stored charge/voltage by 1uV (microvolt). The sensor saturates at 3 volts so it can cope with 3V/1uV = 3,000,000 photons hitting it before it is fully saturated.

Assume that exposing for 1 second on a faint nebula produces 1 photon, 100,000 photons for a magnitude 2 star and exposing on the sun for 1/8000th of a second produces 3,000,000 photons.

Assume that the ADC (analog to digital converter) in the sensor is 14-bits wide (16384 steps). The ADC runs from 3 Volts and can detect voltages from 0 to 3 Volts so the resolution for each individual ADC bit is 3V / 16384 = 183uV.

As you can see trying to fit 3 million discrete values from the sensor into 16384 discrete steps in the ADC is a struggle, something is going to get lost/clipped. What we can see is that the dynamic range we are trying to capture is immensely larger than what can actually be captured by the ADC. E.g. a bright star will saturate the ADC long before the faint bits of wispy nebulae are even registered.

At base ISO (ISO100 for Canon), there is no gain applied to the input of the ADC. So if 75 photons hit the sensor during exposure, it will produce a voltage of 75uV. Once digitised this will produce a value of 75uV/183uV = zero, however:

At ISO200, a 2x gain is applied to the input of the ADC, now the same exposure time as above will produce a digitised value of 150uV/183uV which is still zero.

At ISO400, a 4x gain is applied, now the digitised value is 300uV/183uV = 1

At ISO800, an 8x gain is applied, now the digitised value is 600uV/183uV = 3

At ISO1600, a 16x gain is applied, now the digitised value is 1200uV/183uV = 6

What a higher ISO is doing is allowing us to get more out of the sensor near the bottom end of the histogram/noise floor, this is especially useful when using short exposures, however it is not all good:

Each gain step adds more noise to the input signal, however this can be reduced by averaging/stacking multiple exposures.

On most modern Digital SLRs, ISO values higher than 1600 produce unacceptable levels of noise (tantamount to digital zoom) so it is not recommended, though that may change in future sensors.

What is important is to get away from the left hand side of the histogram, not too far that you are unnecessarily clipping the highlights, but far enough that you are collecting the data that is near to the noise floor which contains the subtle data that you are actually after.

In the ideal world, base ISO (ISO100) and long enough exposures would be the best option, however there is noise introduced by having very long exposures so in practice, you want to use the ISO setting for your camera which produces the least noise for the shortest exposure time to achieve a correct histogram. Each camera model is slightly different so for example a Canon 7D may be best at ISO640. If it is possible to do long enough exposures without trailing then use the "ideal" ISO value for your camera, if you can't then use a higher ISO to ensure you are capturing the data.

In conclusion, I'd say that strictly speaking the sensitivity of the sensor itself is not changed with ISO, however the value presented by the sensor to the outside world i.e. the ADC converter is. I am aware that newer sensors have ADC's on chip, but the same still principle applies.

[Jannis]

What i do is usually this: first, set a low ISO and see for how long i can track without startrails. Then, when i have found out that, i set the ISO acordingly so that the spike at the histogram starts roughly 1/3 on the way from the left. I've found this to work OK for me at least. Still, i never use any higher ISO then the max analog ISO of the cameras sensor, as it's simply no point.

[Pete Presland]

What i do is usually this: first, set a low ISO and see for how long i can track without startrails. Then, when i have found out that, i set the ISO acordingly so that the spike at the histogram starts roughly 1/3 on the way from the left. I've found this to work OK for me at least. Still, i never use any higher ISO then the max analog ISO of the cameras sensor, as it's simply no point.

another interesting thread, so does this mean the rosetta image, i took a couple of nights back which the histogram spike was over 3/4 of the way to the right, was caused by the ISO setting 1600 i selected ?

[RikM]

Since the rosette is a large area of nebulosity the peak will be spread out more than for a smaller object. I took mine to half-way because that was the point when I got a separation from the left hand end.

[Pete Presland]

Since the rosette is a large area of nebulosity the peak will be spread out more than for a smaller object. I took mine to half-way because that was the point when I got a separation from the left hand end.

i initially had quite a narrow peak over to the right, i suspect this relects the amount of light pollution in my image

[StuartJPP]

You can always take an exposure in a part of the sky with no nebulosity to see what the sky fog/LP is like and take it from there. There isn't a one size fits all solution I'm afraid, each target and sky condition is different. No point exposing for longer than necessary if all you are capturing is LP.

[RikM]

So long as your target is brighter than the light pollution to start with, more signal is better so long as you don't burn out the star colour. Yes it will look like an orange square but you can deal with that in post. If you don't have the signal to start with, you have nothing to work with.

[nephilim]

This is how I understand ISO in relation to digital sensors specifically in the world of astro. This is my own interpretation and is subject to misunderstanding on my behalf. A lot of assumptions here and trying to explain in layman's terms.

To make things simple lets assume that we have a 1x1 pixel sensor.

Assume the sensor is sensitive enough to detect each and every photon that hits it which in turn increases the stored charge/voltage by 1uV (microvolt). The sensor saturates at 3 volts so it can cope with 3V/1uV = 3,000,000 photons hitting it before it is fully saturated.

Assume that exposing for 1 second on a faint nebula produces 1 photon, 100,000 photons for a magnitude 2 star and exposing on the sun for 1/8000th of a second produces 3,000,000 photons.

Assume that the ADC (analog to digital converter) in the sensor is 14-bits wide (16384 steps). The ADC runs from 3 Volts and can detect voltages from 0 to 3 Volts so the resolution for each individual ADC bit is 3V / 16384 = 183uV.

As you can see trying to fit 3 million discrete values from the sensor into 16384 discrete steps in the ADC is a struggle, something is going to get lost/clipped. What we can see is that the dynamic range we are trying to capture is immensely larger than what can actually be captured by the ADC. E.g. a bright star will saturate the ADC long before the faint bits of wispy nebulae are even registered.

At base ISO (ISO100 for Canon), there is no gain applied to the input of the ADC. So if 75 photons hit the sensor during exposure, it will produce a voltage of 75uV. Once digitised this will produce a value of 75uV/183uV = zero, however:

At ISO200, a 2x gain is applied to the input of the ADC, now the same exposure time as above will produce a digitised value of 150uV/183uV which is still zero.

At ISO400, a 4x gain is applied, now the digitised value is 300uV/183uV = 1

At ISO800, an 8x gain is applied, now the digitised value is 600uV/183uV = 3

At ISO1600, a 16x gain is applied, now the digitised value is 1200uV/183uV = 6

What a higher ISO is doing is allowing us to get more out of the sensor near the bottom end of the histogram/noise floor, this is especially useful when using short exposures, however it is not all good:

Each gain step adds more noise to the input signal, however this can be reduced by averaging/stacking multiple exposures.

On most modern Digital SLRs, ISO values higher than 1600 produce unacceptable levels of noise (tantamount to digital zoom) so it is not recommended, though that may change in future sensors.

What is important is to get away from the left hand side of the histogram, not too far that you are unnecessarily clipping the highlights, but far enough that you are collecting the data that is near to the noise floor which contains the subtle data that you are actually after.

In the ideal world, base ISO (ISO100) and long enough exposures would be the best option, however there is noise introduced by having very long exposures so in practice, you want to use the ISO setting for your camera which produces the least noise for the shortest exposure time to achieve a correct histogram. Each camera model is slightly different so for example a Canon 7D may be best at ISO640. If it is possible to do long enough exposures without trailing then use the "ideal" ISO value for your camera, if you can't then use a higher ISO to ensure you are capturing the data.

In conclusion, I'd say that strictly speaking the sensitivity of the sensor itself is not changed with ISO, however the value presented by the sensor to the outside world i.e. the ADC converter is. I am aware that newer sensors have ADC's on chip, but the same still principle applies.

Thanks Stuart, for taking the time to explain, although i'm a little bit of a luddite when it comes to anything 'Tech' related it took me a couple of re-reads for it to sink in . From the little bit of research i've done so far ( No practical yet, just theory thanks to the poxy weather), I see that the 350D tends to really 'spike' noise wise at iso 800, so i'll be sticking with 400 & 1600 to start with & see how it goes.

Steve

[badgerchap]

I've always had good (for me) results with ISO 800. Having said that I am toying with the idea of trying out some more efforts with 1600.

[Dave 2112]

This is how I understand ISO in relation to digital sensors specifically in the world of astro. This is my own interpretation and is subject to misunderstanding on my behalf. A lot of assumptions here and trying to explain in layman's terms.

To make things simple lets assume that we have a 1x1 pixel sensor.

Assume the sensor is sensitive enough to detect each and every photon that hits it which in turn increases the stored charge/voltage by 1uV (microvolt). The sensor saturates at 3 volts so it can cope with 3V/1uV = 3,000,000 photons hitting it before it is fully saturated.

Assume that exposing for 1 second on a faint nebula produces 1 photon, 100,000 photons for a magnitude 2 star and exposing on the sun for 1/8000th of a second produces 3,000,000 photons.

Assume that the ADC (analog to digital converter) in the sensor is 14-bits wide (16384 steps). The ADC runs from 3 Volts and can detect voltages from 0 to 3 Volts so the resolution for each individual ADC bit is 3V / 16384 = 183uV.

As you can see trying to fit 3 million discrete values from the sensor into 16384 discrete steps in the ADC is a struggle, something is going to get lost/clipped. What we can see is that the dynamic range we are trying to capture is immensely larger than what can actually be captured by the ADC. E.g. a bright star will saturate the ADC long before the faint bits of wispy nebulae are even registered.

At base ISO (ISO100 for Canon), there is no gain applied to the input of the ADC. So if 75 photons hit the sensor during exposure, it will produce a voltage of 75uV. Once digitised this will produce a value of 75uV/183uV = zero, however:

At ISO200, a 2x gain is applied to the input of the ADC, now the same exposure time as above will produce a digitised value of 150uV/183uV which is still zero.

At ISO400, a 4x gain is applied, now the digitised value is 300uV/183uV = 1

At ISO800, an 8x gain is applied, now the digitised value is 600uV/183uV = 3

At ISO1600, a 16x gain is applied, now the digitised value is 1200uV/183uV = 6

What a higher ISO is doing is allowing us to get more out of the sensor near the bottom end of the histogram/noise floor, this is especially useful when using short exposures, however it is not all good:

Each gain step adds more noise to the input signal, however this can be reduced by averaging/stacking multiple exposures.

On most modern Digital SLRs, ISO values higher than 1600 produce unacceptable levels of noise (tantamount to digital zoom) so it is not recommended, though that may change in future sensors.

What is important is to get away from the left hand side of the histogram, not too far that you are unnecessarily clipping the highlights, but far enough that you are collecting the data that is near to the noise floor which contains the subtle data that you are actually after.

In the ideal world, base ISO (ISO100) and long enough exposures would be the best option, however there is noise introduced by having very long exposures so in practice, you want to use the ISO setting for your camera which produces the least noise for the shortest exposure time to achieve a correct histogram. Each camera model is slightly different so for example a Canon 7D may be best at ISO640. If it is possible to do long enough exposures without trailing then use the "ideal" ISO value for your camera, if you can't then use a higher ISO to ensure you are capturing the data.

In conclusion, I'd say that strictly speaking the sensitivity of the sensor itself is not changed with ISO, however the value presented by the sensor to the outside world i.e. the ADC converter is. I am aware that newer sensors have ADC's on chip, but the same still principle applies.

If that's layman's terms I think I need the Ladybird book version.

I can see a few nights of pressing all the buttons and seeing what comes out again, assuming the sky ever clears up.

[StuartJPP]

I've always had good (for me) results with ISO 800. Having said that I am toying with the idea of trying out some more efforts with 1600.

If you can image at ISO800 with long enough exposures and not get star trailing, I don't think there is much point upping the ISO.

[badgerchap]

Because I am getting star trails! But I'm nearly there with guiding now so there may indeed be no point

[dph1nm]

At base ISO (ISO100 for Canon), there is no gain applied to the input of the ADC. So if 75 photons hit the sensor during exposure, it will produce a voltage of 75uV. Once digitised this will produce a value of 75uV/183uV = zero, however:

At ISO200, a 2x gain is applied to the input of the ADC, now the same exposure time as above will produce a digitised value of 150uV/183uV which is still zero.

At ISO400, a 4x gain is applied, now the digitised value is 300uV/183uV = 1

At ISO800, an 8x gain is applied, now the digitised value is 600uV/183uV = 3

At ISO1600, a 16x gain is applied, now the digitised value is 1200uV/183uV = 6

Is this right? Unity gain (i.e. where 1 photon gives 1 ADU) for the Canon 1000D is somewhere between ISO 800 and 1600.

NigelM

[dph1nm]

What a higher ISO is doing is allowing us to get more out of the sensor near the bottom end of the histogram/noise floor, this is especially useful when using short exposures, however it is not all good: Each gain step adds more noise to the input signal

It doesn't really add noise, it simply scales the existing Shot noise. But is also scales the signal by the same amount, so the signal-to-noise ratio is unchanged.

NigelM

[Russe]

Well, I tried imaging with my 450D at f/11.8 with ISO800 for 300s...

Not an awful lot you can see on the pic... M31 nor M101.

I shall try ISO1600 with 400 or 500s. I'm interested. Haven't gotten any startrails yet, since I'm guiding with PHP.

I just need a better scope...

[lensman57]

As if AP isnt confusing enough, I'm reading this http://homepages.nil..._Canon_350D.htm and as regards to Iso it recommends using 1600 all the time & if using the CHDK firmware hack, going as far as 3200. His/Her reason for this is that as you'll be taking dark frames then all the noise will be removed & its better to have signal+noise rather than no signal at all. Although to me (total newbie) this makes sense, is this really the case?? As I've also read that imaging objects such as M42 the higher iso will just burn out the core?

Steve

Hi,

Use 400~800. The higher the ISO the higher the noise and lower dynamic range. ISO 800 should be a safe bet for up to 3 minutes of exposure. DSLR sensors are no different to CCDs, there is only one sensitivity, upping the ISO is just upping the gain in the Amps.

A.G