Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

understanding sensorgen.info

  • Please log in to reply
22 replies to this topic

#1 joelin

joelin

    Viking 1

  • -----
  • topic starter
  • Posts: 584
  • Joined: 14 Jan 2008
  • Loc: Saratoga, CA

Posted 13 March 2017 - 03:42 AM

I recently discovered sensorgen.info. I think I need a little tutorial to understand everything. I understand the basics of read noise, saturation and dynamic range. However, I have some questions beyond that. 

 

1) If you're mainly focused on collecting data that is very faint, would the camera/ISO setting with the lowest read noise always be preferable (usually the highest ISO)? This way you don't need a lot of data for it to rise above the read noise. I know at these settings the sensor is more easily saturated (lower DR), but really the goals are the data in the shadows and I'm ok with blown out stars or bright areas.

 

2) If two sensors have the same ISO and the same DR but a different read noise + saturation pairing, which one is better. I'm interpreting this to just mean that the sensor with the higher read noise + saturation would just require longer exposures but give an equivalent image in the end.

 

3) What is the difference between measured ISO and ISO.

 

4) Some sensors have a greater than 100% QE. Is this a bug?

 

5) What makes a sensor so great? I know the Nikon D5000/7000s have a very high DR at low ISOs. Would this mean I would need to use ISO 100 and expose for very long periods of time to maximize it?


Edited by joelin, 13 March 2017 - 03:52 AM.


#2 whwang

whwang

    Surveyor 1

  • *****
  • Posts: 1764
  • Joined: 20 Mar 2013

Posted 13 March 2017 - 04:41 AM

1. yes.  For low-light, low read noise is preferred.  DR is less important, but you don't want very very low DR either.

 

2. the low read noise one is preferred.

 

3. There is strict definition of camera's ISO sensitivity, defined by ISO, not surprisingly.  I believe the measured ISO is based on this.  I read somewhere that ISO is defined at the saturation level, and I don't think this is practical.  Camera manufactures set the "ISO" based on the apparent brightness of the images, which is not determined by the saturated highlight in most practical situations.  So it is very understandable (at least to me) that the sensitivity adopted by most cameras is different from ISO's definition. Ironically, it's still called ISO on the camera.   

 

4. should be a bug.  There is also substantial noise in sensorgen's data.

 

5. low read noise, high QE.

 

Cheers,

Wei-Hao



#3 Jerry Lodriguss

Jerry Lodriguss

    Vendor

  • *****
  • Vendors
  • Posts: 5649
  • Joined: 19 Jul 2008
  • Loc: Voorhees, NJ

Posted 13 March 2017 - 09:55 AM

The other thing you need to think about is the noise associated with thermal signal, which, unfortunately, is not listed at sensorgen.

 

If you're shooting at high ambient temperatures, the thermal noise can be a greater contributor to the noise than readout noise.

 

And, you can always overcome readout noise by just exposing so you are sky-noise limited and not read-noise limited.

 

Once you reach this point, it's all about gathering more signal to improve the signal-to-noise ratio.

 

Jerry


Edited by Jerry Lodriguss, 13 March 2017 - 10:02 AM.

  • Jon Rista likes this

#4 Midnight Dan

Midnight Dan

    Voyager 1

  • *****
  • Posts: 12891
  • Joined: 23 Jan 2008
  • Loc: Hilton, NY, Yellow Zone (Bortle 4.5)

Posted 13 March 2017 - 10:05 AM

1. It depends on the camera and the curve.  In the case of my camera, the Canon 600D, the read noise curve flattens out above ISO 800, while the DR curve flattens out below ISO 800.  So for my camera, the sweet spot is 800 ISO.  On the other hand, for the Nikon D7000, the read noise curve is pretty much flat.  In that case, the lower your ISO, the more dynamic range you'll get.

 

2. I think I'd go with lower read noise, but Jerry brings up a good point about thermal noise potentially being the overriding factor.

 

3. On Sensorgen, the "ISO" is the setting on the camera dial.  The "Measured ISO" is the true ISO as measured by Sensorgen's testing.

 

4. I Agree with Wei-Hao's comment.

 

5. Yes, if you want to maximize the dynamic range, you'd need to use low ISO.

 

However, to build on Jerry's comments about thermal noise - in a DSLR, I found that the rated dynamic range is significantly limited by the thermal noise.  You might reach those DR ratings at short exposures for normal daytime photography, but for long exposure astrophotography the dim end of the dynamic range will be dominated by thermal noise which will limit its usefulness.

 

I added a thermo-electric "cold-finger" cooler to my DSLR to cool the chip and it helped dramatically, especially in the warm summer months.  But when I moved to a cooled astro camera (the ZWO ASI071) it was another huge improvement due to the better cooling capabilities.  I can pull details out of the dim areas that I never could before, even with the cooled DSLR, due to the thermal noise.

 

-Dan



#5 joelin

joelin

    Viking 1

  • -----
  • topic starter
  • Posts: 584
  • Joined: 14 Jan 2008
  • Loc: Saratoga, CA

Posted 13 March 2017 - 03:14 PM

The other thing you need to think about is the noise associated with thermal signal, which, unfortunately, is not listed at sensorgen.

 

If you're shooting at high ambient temperatures, the thermal noise can be a greater contributor to the noise than readout noise.

 

And, you can always overcome readout noise by just exposing so you are sky-noise limited and not read-noise limited.

 

Once you reach this point, it's all about gathering more signal to improve the signal-to-noise ratio.

 

Jerry

Can you provide more detail about being sky-noise limited? 

 

I'm guessing read noise constant (irregardless of exposure length) so I can expose longer and make it insignificant. Exposing longer will increase S/N by exposing the stars long enough to bring out the noise over the background (sky noise)?



#6 joelin

joelin

    Viking 1

  • -----
  • topic starter
  • Posts: 584
  • Joined: 14 Jan 2008
  • Loc: Saratoga, CA

Posted 13 March 2017 - 03:15 PM

1. It depends on the camera and the curve.  In the case of my camera, the Canon 600D, the read noise curve flattens out above ISO 800, while the DR curve flattens out below ISO 800.  So for my camera, the sweet spot is 800 ISO.  On the other hand, for the Nikon D7000, the read noise curve is pretty much flat.  In that case, the lower your ISO, the more dynamic range you'll get.

 

2. I think I'd go with lower read noise, but Jerry brings up a good point about thermal noise potentially being the overriding factor.

 

3. On Sensorgen, the "ISO" is the setting on the camera dial.  The "Measured ISO" is the true ISO as measured by Sensorgen's testing.

 

4. I Agree with Wei-Hao's comment.

 

5. Yes, if you want to maximize the dynamic range, you'd need to use low ISO.

 

However, to build on Jerry's comments about thermal noise - in a DSLR, I found that the rated dynamic range is significantly limited by the thermal noise.  You might reach those DR ratings at short exposures for normal daytime photography, but for long exposure astrophotography the dim end of the dynamic range will be dominated by thermal noise which will limit its usefulness.

 

I added a thermo-electric "cold-finger" cooler to my DSLR to cool the chip and it helped dramatically, especially in the warm summer months.  But when I moved to a cooled astro camera (the ZWO ASI071) it was another huge improvement due to the better cooling capabilities.  I can pull details out of the dim areas that I never could before, even with the cooled DSLR, due to the thermal noise.

 

-Dan

So true dynamic range also depends on the exposure length.

 

Sensorgen doesn't take in account exposure length right?

 

What is the typical point where thermal noise stars to overwhelm read noise?



#7 Jerry Lodriguss

Jerry Lodriguss

    Vendor

  • *****
  • Vendors
  • Posts: 5649
  • Joined: 19 Jul 2008
  • Loc: Voorhees, NJ

Posted 13 March 2017 - 03:46 PM

 

The other thing you need to think about is the noise associated with thermal signal, which, unfortunately, is not listed at sensorgen.

 

If you're shooting at high ambient temperatures, the thermal noise can be a greater contributor to the noise than readout noise.

 

And, you can always overcome readout noise by just exposing so you are sky-noise limited and not read-noise limited.

 

Once you reach this point, it's all about gathering more signal to improve the signal-to-noise ratio.

 

Jerry

Can you provide more detail about being sky-noise limited? 

 

I'm guessing read noise constant (irregardless of exposure length) so I can expose longer and make it insignificant. Exposing longer will increase S/N by exposing the stars long enough to bring out the noise over the background (sky noise)?

 

It's not really about stars. It's about the signal from the object you are interested in. For long-exposure deep-sky, it's usually a deep-sky object.

 

Sky noise is photon shot noise. Anytime you have signal, you have noise associated with it due to the quantum nature of light. The noise is basically the square root of the signal.

 

If you plug all of the noise sources into the equations, at some point the read noise becomes trivial and either the thermal noise or shot noise is the dominant term. 

 

At that point, it's all about gathering more signal as it is the only way to improve the s/n ratio.

 

Jerry



#8 joelin

joelin

    Viking 1

  • -----
  • topic starter
  • Posts: 584
  • Joined: 14 Jan 2008
  • Loc: Saratoga, CA

Posted 13 March 2017 - 06:40 PM

 

 

The other thing you need to think about is the noise associated with thermal signal, which, unfortunately, is not listed at sensorgen.

 

If you're shooting at high ambient temperatures, the thermal noise can be a greater contributor to the noise than readout noise.

 

And, you can always overcome readout noise by just exposing so you are sky-noise limited and not read-noise limited.

 

Once you reach this point, it's all about gathering more signal to improve the signal-to-noise ratio.

 

Jerry

Can you provide more detail about being sky-noise limited? 

 

I'm guessing read noise constant (irregardless of exposure length) so I can expose longer and make it insignificant. Exposing longer will increase S/N by exposing the stars long enough to bring out the noise over the background (sky noise)?

 

It's not really about stars. It's about the signal from the object you are interested in. For long-exposure deep-sky, it's usually a deep-sky object.

 

Sky noise is photon shot noise. Anytime you have signal, you have noise associated with it due to the quantum nature of light. The noise is basically the square root of the signal.

 

If you plug all of the noise sources into the equations, at some point the read noise becomes trivial and either the thermal noise or shot noise is the dominant term. 

 

At that point, it's all about gathering more signal as it is the only way to improve the s/n ratio.

 

Jerry

 

what is this "at some point the read noise becomes trivial"?

 

does it happen with longer exposures? more stacking?



#9 Midnight Dan

Midnight Dan

    Voyager 1

  • *****
  • Posts: 12891
  • Joined: 23 Jan 2008
  • Loc: Hilton, NY, Yellow Zone (Bortle 4.5)

Posted 13 March 2017 - 07:11 PM

 what is this "at some point the read noise becomes trivial"?

 

Jon Rista has done a lot of posting in threads about the ASI1600 and ASI071 cameras regarding this issue. His take is that you should try to expose so that the sky background is 20 times the level of the read noise to properly "swamp" the read noise.  

 

I've been trying to follow that advice with my ASI071 and it seems do a good job.  But, to do that, you have to calculate what the read noise level will be in 16-bit values after gain and offset are applied.  You then use software like SGP to get a reading of your image's mean pixel value, which mostly represents the sky background, to see if it is 20x the read noise.  Problem with a DSLR is I don't know how you'd calculate the values.  You know what the ISO is, but not the gain and offset that the ISO represents.

 

There may be other ways to evaluate whether your exposures are long enough.  Jerry might be able to help out here.

 

-Dan



#10 Jerry Lodriguss

Jerry Lodriguss

    Vendor

  • *****
  • Vendors
  • Posts: 5649
  • Joined: 19 Jul 2008
  • Loc: Voorhees, NJ

Posted 13 March 2017 - 07:26 PM

 

 

 

The other thing you need to think about is the noise associated with thermal signal, which, unfortunately, is not listed at sensorgen.

 

If you're shooting at high ambient temperatures, the thermal noise can be a greater contributor to the noise than readout noise.

 

And, you can always overcome readout noise by just exposing so you are sky-noise limited and not read-noise limited.

 

Once you reach this point, it's all about gathering more signal to improve the signal-to-noise ratio.

 

Jerry

Can you provide more detail about being sky-noise limited? 

 

I'm guessing read noise constant (irregardless of exposure length) so I can expose longer and make it insignificant. Exposing longer will increase S/N by exposing the stars long enough to bring out the noise over the background (sky noise)?

 

It's not really about stars. It's about the signal from the object you are interested in. For long-exposure deep-sky, it's usually a deep-sky object.

 

Sky noise is photon shot noise. Anytime you have signal, you have noise associated with it due to the quantum nature of light. The noise is basically the square root of the signal.

 

If you plug all of the noise sources into the equations, at some point the read noise becomes trivial and either the thermal noise or shot noise is the dominant term. 

 

At that point, it's all about gathering more signal as it is the only way to improve the s/n ratio.

 

Jerry

 

what is this "at some point the read noise becomes trivial"?

 

does it happen with longer exposures? more stacking?

 

It depends on what camera you have and how much readout noise it has, and how much thermal signal it has.

 

The old rule of thumb was expose until the histogram is 1/3 of the way over from the left and they you are sky-noise limited.

 

That rule is a little different for the so-called ISOless cameras.

 

Of course, that doesn't include any crazy thermal signal.

 

It's really complicated. If you want to learn all this stuff, you can. Then you need to measure your camera for all of the various parameters you need to input in the the equations, and then measure your sky, and you can calculate it.

 

Or you can go out and take pictures  using the 1/3 histogram rule.

 

Jerry



#11 sharkmelley

sharkmelley

    Apollo

  • -----
  • Posts: 1241
  • Joined: 19 Feb 2013

Posted 14 March 2017 - 12:56 AM

 

 what is this "at some point the read noise becomes trivial"?

 

Jon Rista has done a lot of posting in threads about the ASI1600 and ASI071 cameras regarding this issue. His take is that you should try to expose so that the sky background is 20 times the level of the read noise to properly "swamp" the read noise.  

 

 

 

You've missed a critical part of the rule:

"Sky background is 20 times the level of the read noise when both are measured in electrons." 

 

So you must convert your digital numbers to electrons using the gain - assuming you know the gain of course. 

 

Mark



#12 Midnight Dan

Midnight Dan

    Voyager 1

  • *****
  • Posts: 12891
  • Joined: 23 Jan 2008
  • Loc: Hilton, NY, Yellow Zone (Bortle 4.5)

Posted 14 March 2017 - 09:38 AM

 

 

 what is this "at some point the read noise becomes trivial"?

 

Jon Rista has done a lot of posting in threads about the ASI1600 and ASI071 cameras regarding this issue. His take is that you should try to expose so that the sky background is 20 times the level of the read noise to properly "swamp" the read noise.  

 

 

 

You've missed a critical part of the rule:

"Sky background is 20 times the level of the read noise when both are measured in electrons." 

 

So you must convert your digital numbers to electrons using the gain - assuming you know the gain of course. 

 

Mark

 

True.  That's why I said you need to do the calculations.  I didn't bother to go into the details of those calculations, which requires both gain and offset (not just gain) because you don't have easy access to that information with a DSLR anyway.  The 1/3 histogram rule that Jerry mentions is probably the best bet.

 

-Dan



#13 joelin

joelin

    Viking 1

  • -----
  • topic starter
  • Posts: 584
  • Joined: 14 Jan 2008
  • Loc: Saratoga, CA

Posted 14 March 2017 - 11:01 AM

Given the 1/3 rule, would it be best to use the lowest possible ISO to maximize DR? (Since read noise no longer matters)



#14 Jerry Lodriguss

Jerry Lodriguss

    Vendor

  • *****
  • Vendors
  • Posts: 5649
  • Joined: 19 Jul 2008
  • Loc: Voorhees, NJ

Posted 14 March 2017 - 11:54 AM

Given the 1/3 rule, would it be best to use the lowest possible ISO to maximize DR? (Since read noise no longer matters)

If you can guide accurately for that long, yes.

 

If you are shooting with a filter at f/10 and try to shoot at ISO 100 at a true dark-sky site, you will be surprised at low incredibly long your exposure will need to be.

 

And then you can get into quantization errors. So it's best not to use the very lowest ISO.

 

It depends on the camera.

 

As I said, this is a very complicated subject.

 

It has been discussed at length, so there really is no need to repeat everything.  Do a search in the B&I forum.

 

Jerry



#15 bobzeq25

bobzeq25

    Cosmos

  • *****
  • Posts: 8445
  • Joined: 27 Oct 2014

Posted 14 March 2017 - 01:40 PM

I cannot believe I'm going to disagree with whang.

 

I like using low ISO (as long as the read noise is decently close to the minimum), because I find it easier to get better star color.  I'll concede a top imager going after really dim stuff might think otherwise.

 

ISO 100 is often a bit strange, I use 200 on my D5500.

 

You can tie yourself in knots on this stuff.  Or simply use the lowest ISO that gives you acceptable read noise, and expose to 1/4 to 1/3 (stretched) histogram.  If you're in your first year of imaging that's all you need to know, for now.  Lotsa more important stuff to worry about.  "Close enough".  Shooting more subs is more important than getting the exposure per sub exactly right.


Edited by bobzeq25, 14 March 2017 - 01:41 PM.


#16 whwang

whwang

    Surveyor 1

  • *****
  • Posts: 1764
  • Joined: 20 Mar 2013

Posted 14 March 2017 - 09:18 PM

I cannot believe I'm going to disagree with whang.

 

I like using low ISO (as long as the read noise is decently close to the minimum), because I find it easier to get better star color.  I'll concede a top imager going after really dim stuff might think otherwise.

 

I wouldn't say this is agreement or disagreement.  This is just personal interest/style.  My interest is faint nebulas/stars/galaxies, not bright stars. So I never pay attention to bright stars.  On the other hand, if bright stars are important to you, of course you don't want them saturated.  In such a case, lower ISO indeed helps, but makes you pay the price of noisier background.



#17 SnowSailor

SnowSailor

    Vostok 1

  • -----
  • Posts: 105
  • Joined: 15 Nov 2016

Posted 15 March 2017 - 02:08 AM

I'd like to piggyback on this question, as I've been meaning to post this as an entirely new question.

 

On sensorgen, we can see that for the 7D Mk II (http://sensorgen.inf...7D-Mark-II.html) the read noise at 1600 is 2.8, but the read noise at 3200 is 2.1. After that it begins to go up before dipping again. I read on here before that it's best to shoot at the ISO where the read noise is low, but then rises up again. So in this case it seems that shooting at 3200 is the best option. I always see people wanting to shoot at low ISOs but for most cameras the higher ISO values seem to have a much lower read noise.

 

Is it better to shoot at 1600 or 3200 with the 7D Mk II and why do people want to shoot at lower ISO values when the higher ones have less read noise?


  • Traveler likes this

#18 whwang

whwang

    Surveyor 1

  • *****
  • Posts: 1764
  • Joined: 20 Mar 2013

Posted 15 March 2017 - 03:44 AM

It's a trade off.  You should use ISO 3200 when read noise is the only concern.  However, the dynamical range at ISO 3200 is kind of low.  If one can afford a longer sub exposure time or has fast optics, then going to ISO 1600 is not a bad idea, to gain some dynamical range.  



#19 NorbertG

NorbertG

    Explorer 1

  • -----
  • Posts: 73
  • Joined: 02 Mar 2015

Posted 15 March 2017 - 03:45 AM

 

Is it better to shoot at 1600 or 3200 with the 7D Mk II and why do people want to shoot at lower ISO values when the higher ones have less read noise?

it depends. Finally you have to consider the S/N for the whole stack. If your esposure time is limited by saturation (or if you take care for bright stars) ISO1600 allows you to expose twice as long before the brightest pixels saturate. You would need twice as many subs for the same total exposure with Iso3200, therefore your final readnoise adds up by factor of sqr(2) for ISO3200 when compared with ISO1600.

In this example you have to compare 2.8 with 2.1xsqr(2)=2.97    

So in this case it is nearly the same, then it is better to go with higher ISO (as long as readnoise drops of course) and more frames, because more frames allow more tricks (Sigma clipping etc) during stacking and avoid digitization effects.   


Edited by NorbertG, 15 March 2017 - 03:47 AM.


#20 SnowSailor

SnowSailor

    Vostok 1

  • -----
  • Posts: 105
  • Joined: 15 Nov 2016

Posted 15 March 2017 - 12:57 PM

It's a trade off.  You should use ISO 3200 when read noise is the only concern.  However, the dynamical range at ISO 3200 is kind of low.  If one can afford a longer sub exposure time or has fast optics, then going to ISO 1600 is not a bad idea, to gain some dynamical range.  

I see. I've got an 8" Astrograph Newtonian so I guess that might be good for 1600. I'll have to do some test shots with the same exposure time and number of images (only difference being 1600 and 3200 ISO) to see how the results turn out.



#21 sharkmelley

sharkmelley

    Apollo

  • -----
  • Posts: 1241
  • Joined: 19 Feb 2013

Posted 15 March 2017 - 02:11 PM

One other important thing that Sensorgen can't tell you is the amount of banding noise.  Canon cameras are quite susceptible to it.  Roger Clark ( a great fan of the 7D2) recommends using it at ISO 1600 or higher to avoid banding:  https://www.dpreview.../post/58897393 

 

Mark



#22 SnowSailor

SnowSailor

    Vostok 1

  • -----
  • Posts: 105
  • Joined: 15 Nov 2016

Posted 15 March 2017 - 04:31 PM

One other important thing that Sensorgen can't tell you is the amount of banding noise.  Canon cameras are quite susceptible to it.  Roger Clark ( a great fan of the 7D2) recommends using it at ISO 1600 or higher to avoid banding:  https://www.dpreview.../post/58897393 

 

Mark

Interesting. Thanks.



#23 bclaff

bclaff

    Explorer 1

  • -----
  • Posts: 65
  • Joined: 03 Jan 2015
  • Loc: Sherborn, MA USA

Posted 20 March 2017 - 03:08 PM



One other important thing that Sensorgen can't tell you is the amount of banding noise.  Canon cameras are quite susceptible to it.  Roger Clark ( a great fan of the 7D2) recommends using it at ISO 1600 or higher to avoid banding:  https://www.dpreview.../post/58897393 

 

Mark

In addition to pretty much the same content as sensorgen.info plus data on newer cameras, you can also see sensor heatmaps at my site that can help you judge pattern noise ("banding").

 

Regards,


Edited by bclaff, 20 March 2017 - 03:09 PM.

  • whwang likes this


CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.







Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics