Getting the best out of your equipment

[MadMarco]

I'm always trying to improve my technique and resultant photos, so I am running a series of tests on my M43 cameras to see which settings and methods provide the best results with regards to astro photography and particularly for long exposure and multi exposure stacking.

ISO Invariance
I read an interesting article claiming that in most cases (for most cameras) the ISO setting has no effect on noise, it's the exposure settings that effect noise. At first that sounds completely wrong and counter to all my experience, but testing can prove that this is the case. If you want to reduce the amount of noise in your images, you either need to open up the aperture (decrease the f-stop number) or decrease the shutter speed (longer exposure time).

Most digital cameras exhibit some degree of ISO invariance i.e. the ISO setting won't affect the noise in the image. There is an experiment that you can run to test how invariant to ISO the camera is. Take a series of images at a fixed shutter speed and aperture and then change only the ISO setting. In post processing you adjust the images to correct for the ISO settings by boosting the exposure by the required EV. This works so long as you aren't blowing highlights or burying shadow detail.

I ran the tests with both my M43 cameras, the E-M1 II and a full spectrum IR modified E-PL5. I used my WO ZS71 telescope which has an aperture of f/5.9 and a focal length of 418mm. The test images are below, the the EM1M2 on the left and the E-PL5 on the right. The images are cropped and then resampled to 2x pixel density to show all the horrors:

ISO 6400 0ev adjustment

Subscribe to see EXIF info for this image (if available)

ISO 3200 +1ev adjustment

Subscribe to see EXIF info for this image (if available)

ISO 1600 +2ev adjustment

Subscribe to see EXIF info for this image (if available)

ISO 800 +3ev adjustment

Subscribe to see EXIF info for this image (if available)

ISO 400 +4ev adjustment

Subscribe to see EXIF info for this image (if available)

ISO 200 +5ev adjustment

Subscribe to see EXIF info for this image (if available)

Looking at the images, I'm struggling to see any difference from ISO 6400 to ISO400. At ISO 200 it goes a bit wonky (the full size image shows very slight banding), but I think that this is caused by trying to push for detail that just hasn't been recorded (run out of dynamic range in the shadows).

My conclusions from all this are that both the E-M1 II and E-PL5 are largely ISO invariant and that the noise from both the cameras is all but identical in quantity and quality. The E-PL5 looks a bit fuzzier, but this is mainly due to the full spectrum modification. IR light and visible light focus at slightly different points. I did not do dark frame subtraction for any of these images and I noted that the E-PL5 had more stuck and hot pixels than the E-M1 II.

I'm going to run some more tests and will post below. I want to show how increasing the exposure decreases noise (fairly certain that this will be the case) and to see if stacking many images with short exposures improves noise over shooting a single long exposure (I have no idea which way this will go).

This has really set me thinking and I will be changing my approach to setting shutter speed, aperture and ISO in the future. In order to get the cleanest image possible it is preferable to capture as much light as possible, so setting the shutter speed and aperture (where possible) are the priority and ISO will be used to correct for the exposure. Where ISO becomes an issue is in relation to dynamic range. Increasing the ISO will decrease the dynamic range of the sensor, so there is going to be an upper limit depending on the subject.

 

[MadMarco]

Reserved

 

[MadMarco]

Reserved

 

[archaeopteryx]

(I have no idea which way this will go)

The mechanisms are well understood technically. Specifically, I think you might be looking for the distinctions between FPN, PRNU, and random noise. See, for example, Martinec 2008 or Hornsey 2008.

Related factors to consider might be green pixel averaging and lack of chroma demosaicing errors in high resolution modes.

 

[TheMenWhoDrawSheeps]

i think you are refferingto base ISO - it´s the iso which camera is calibrated to, any other iso setting is just postprocessing, like pushing ev in lightroom.
so your exposure is just a math setting, i doubt you will get any kind of different resuts - making and merging two pictures with iso 400, will give you same noise as 1 with 200, ev values are just different aproach describing it.

and by getting same amount of light should give you same picture - same on dofferentsensor formats(if same technolog(i know, it´s not(but close)))

 

[PakkyT]

base ISO - it´s the iso which camera is calibrated to, any other iso setting is just postprocessing, like pushing ev in lightroom.

Right, my understanding of ISO on digital cameras is that when you increase the ISO you are not increasing the sensitivity of the sensor (which is what a lot of people state) but instead you are simply amplifying what the sensor presents and this includes the noise. Basically the noise present is the same through all ISO settings, it just becomes a matter of how much you amplify that noise along with your signal when picking an ISO level. Does that sound correct?

 

[archaeopteryx]

Does that sound correct?

That's the definition of ISO invariance, yes. There's body to body variation and technology evolution but, broadly speaking, ISO invariance tends not to be a good approximation at extended ISO settings. For relatively current ILC bodies it's generally an effective model from ISO 200 up to somewhere in the 1600-6400 range depending on the specifics.

 

[MadMarco]

i think you are refferingto base ISO - it´s the iso which camera is calibrated to, any other iso setting is just postprocessing, like pushing ev in lightroom.
so your exposure is just a math setting, i doubt you will get any kind of different resuts - making and merging two pictures with iso 400, will give you same noise as 1 with 200, ev values are just different aproach describing it.

and by getting same amount of light should give you same picture - same on dofferentsensor formats(if same technolog(i know, it´s not(but close)))

Not all cameras and sensors are ISO invariant, so this cannot be assumed and really needs to be tested for each camera. Most cameras have a point at which this falls down, so for optimum results you need to know the specific equipment you are using and hence the point of my tests. If you try this with a Canon 6D for instance, you will get very different results and find that best results are obtained from ISO3200-6400. The Cannon is an example of a camera that is extremely ISO variant.

As for the stacking vs. long experiment I suspect that the results will be similar, but I am wondering if there is a benefit to stacking with regards to averaging the noise out between frames.

 

[Machi]

I'm always trying to improve my technique and resultant photos, so I am running a series of tests on my M43 cameras to see which settings and methods provide the best results with regards to astro photography and particularly for long exposure and multi exposure stacking.

ISO Invariance
I read an interesting article claiming that in most cases (for most cameras) the ISO setting has no effect on noise, it's the exposure settings that effect noise. At first that sounds completely wrong and counter to all my experience, but testing can prove that this is the case. If you want to reduce the amount of noise in your images, you either need to open up the aperture (decrease the f-stop number) or decrease the shutter speed (longer exposure time).

Most digital cameras exhibit some degree of ISO invariance i.e. the ISO setting won't affect the noise in the image. There is an experiment that you can run to test how invariant to ISO the camera is. Take a series of images at a fixed shutter speed and aperture and then change only the ISO setting. In post processing you adjust the images to correct for the ISO settings by boosting the exposure by the required EV. This works so long as you aren't blowing highlights or burying shadow detail.

I ran the tests with both my M43 cameras, the E-M1 II and a full spectrum IR modified E-PL5. I used my WO ZS71 telescope which has an aperture of f/5.9 and a focal length of 418mm. The test images are below, the the EM1M2 on the left and the E-PL5 on the right. The images are cropped and then resampled to 2x pixel density to show all the horrors:

ISO 6400 0ev adjustment
View attachment 691118
ISO 3200 +1ev adjustment
View attachment 691117
ISO 1600 +2ev adjustment
View attachment 691116
ISO 800 +3ev adjustment
View attachment 691115
ISO 400 +4ev adjustment
View attachment 691114
ISO 200 +5ev adjustment
View attachment 691113

Looking at the images, I'm struggling to see any difference from ISO 6400 to ISO400. At ISO 200 it goes a bit wonky (the full size image shows very slight banding), but I think that this is caused by trying to push for detail that just hasn't been recorded (run out of dynamic range in the shadows).

My conclusions from all this are that both the E-M1 II and E-PL5 are largely ISO invariant and that the noise from both the cameras is all but identical in quantity and quality. The E-PL5 looks a bit fuzzier, but this is mainly due to the full spectrum modification. IR light and visible light focus at slightly different points. I did not do dark frame subtraction for any of these images and I noted that the E-PL5 had more stuck and hot pixels than the E-M1 II.

I'm going to run some more tests and will post below. I want to show how increasing the exposure decreases noise (fairly certain that this will be the case) and to see if stacking many images with short exposures improves noise over shooting a single long exposure (I have no idea which way this will go).

This has really set me thinking and I will be changing my approach to setting shutter speed, aperture and ISO in the future. In order to get the cleanest image possible it is preferable to capture as much light as possible, so setting the shutter speed and aperture (where possible) are the priority and ISO will be used to correct for the exposure. Where ISO becomes an issue is in relation to dynamic range. Increasing the ISO will decrease the dynamic range of the sensor, so there is going to be an upper limit depending on the subject.

M43 cameras with Sony's 16 Mpix sensor (as is EPL5) have ~1 to 1.5 EV better read out noise at higher ISO settings. Biggest difference is between ISO200 and ISO400 (almost 1 EV gain in shadows). Difference between ISO400 and ISO3200 is then only up to 0.5 EV.
So those cameras aren't entirely ISO invariant.
But in your tests this effect is hidden by dark noise (from long exposure) and shot noise (from the sky). Small ISO variance is best to be seen at short exposures (up to ~1 second) of very dark target (close to SNR 1) as read out noise affects visibly only shadows.
In practice difference looks like this:
(It's from E-PM2 which has the same sensor as EPL5)

Subscribe to see EXIF info for this image (if available)

 

[0000]

Yeah, I've definitely noticed significant gains (so to speak) using 400 vs. pushing 200 a stop in post, so I generally bump it up at least that high in low light. Above that, the difference is negligible within the range I generally consider usable (up to 1600 or equivalent post-boost). This is with the 16MP sensors on my E-PL7s. So... it's not so much about "invariant or not" as about doing some empircal testing to see what happens where along the graph.

 

[ToxicTabasco]

Right, my understanding of ISO on digital cameras is that when you increase the ISO you are not increasing the sensitivity of the sensor (which is what a lot of people state) but instead you are simply amplifying what the sensor presents and this includes the noise. Basically the noise present is the same through all ISO settings, it just becomes a matter of how much you amplify that noise along with your signal when picking an ISO level. Does that sound correct?

I think you're correct. That's my understanding also. There are some online info on the invariance, and that is what they conclude. That's why dynamic range is a huge variable when it comes to invariance.

 

[MadMarco]

M43 cameras with Sony's 16 Mpix sensor (as is EPL5) have ~1 to 1.5 EV better read out noise at higher ISO settings. Biggest difference is between ISO200 and ISO400 (almost 1 EV gain in shadows). Difference between ISO400 and ISO3200 is then only up to 0.5 EV.
So those cameras aren't entirely ISO invariant.
But in your tests this effect is hidden by dark noise (from long exposure) and shot noise (from the sky). Small ISO variance is best to be seen at short exposures (up to ~1 second) of very dark target (close to SNR 1) as read out noise affects visibly only shadows.
In practice difference looks like this:
(It's from E-PM2 which has the same sensor as EPL5)

Using long exposures such as those used in astrophotography is pretty much the whole point of the testing I'm doing, that's why this is posted in the astrophotography forum.

Without a cooled camera, the dark noise is by far the largest component. I can't do anything about shot noise and I can tackle readout noise to some extent with bias frames. Knowing the cameras are in every practical sense ISO invariant between ISO 400-6400 at long exposures is extremely useful information for me, I can ignore the ISO when calculating the exposure (RESULT!).

My goal is to increase SNR to the highest value that I can achieve, so I will be using this information to gather the maximum exposure I can within the limitations of sky-glow, exposure limits (time limit of the tracking I can achieve) and dynamic range.

 

[Machi]

Using long exposures such as those used in astrophotography is pretty much the whole point of the testing I'm doing, that's why this is posted in the astrophotography forum.

Without a cooled camera, the dark noise is by far the largest component. I can't do anything about shot noise and I can tackle readout noise to some extent with bias frames. Knowing the cameras are in every practical sense ISO invariant between ISO 400-6400 at long exposures is extremely useful information for me, I can ignore the ISO when calculating the exposure (RESULT!).

My goal is to increase SNR to the highest value that I can achieve, so I will be using this information to gather the maximum exposure I can within the limitations of sky-glow, exposure limits (time limit of the tracking I can achieve) and dynamic range.

I understand. My point was just about ISO invariance.
I did also tests for longer exposures and general rule is that higher ISO always reduces noise if one works with physical definition (RMS electrons). Longer the exposure, less pronounced this effect is.
But because higher ISO also means less and less space for highlights it practically means reduced SNR.
So my take on it is that for long single exposures (Bulb or Time, >1 min) I always use ISO200.
For shorter exposures (10 - 60 sec) I'm using ISO200 if I want best SNR or ISO400 if I want a lowest noise in the shadows.

 

[georgian82]

I'm always trying to improve my technique and resultant photos, so I am running a series of tests on my M43 cameras to see which settings and methods provide the best results with regards to astro photography and particularly for long exposure and multi exposure stacking.

ISO Invariance
I read an interesting article claiming that in most cases (for most cameras) the ISO setting has no effect on noise, it's the exposure settings that effect noise. At first that sounds completely wrong and counter to all my experience, but testing can prove that this is the case. If you want to reduce the amount of noise in your images, you either need to open up the aperture (decrease the f-stop number) or decrease the shutter speed (longer exposure time).

Most digital cameras exhibit some degree of ISO invariance i.e. the ISO setting won't affect the noise in the image. There is an experiment that you can run to test how invariant to ISO the camera is. Take a series of images at a fixed shutter speed and aperture and then change only the ISO setting. In post processing you adjust the images to correct for the ISO settings by boosting the exposure by the required EV. This works so long as you aren't blowing highlights or burying shadow detail.

I ran the tests with both my M43 cameras, the E-M1 II and a full spectrum IR modified E-PL5. I used my WO ZS71 telescope which has an aperture of f/5.9 and a focal length of 418mm. The test images are below, the the EM1M2 on the left and the E-PL5 on the right. The images are cropped and then resampled to 2x pixel density to show all the horrors:

ISO 6400 0ev adjustment
View attachment 691118
ISO 3200 +1ev adjustment
View attachment 691117
ISO 1600 +2ev adjustment
View attachment 691116
ISO 800 +3ev adjustment
View attachment 691115
ISO 400 +4ev adjustment
View attachment 691114
ISO 200 +5ev adjustment
View attachment 691113

Looking at the images, I'm struggling to see any difference from ISO 6400 to ISO400. At ISO 200 it goes a bit wonky (the full size image shows very slight banding), but I think that this is caused by trying to push for detail that just hasn't been recorded (run out of dynamic range in the shadows).

My conclusions from all this are that both the E-M1 II and E-PL5 are largely ISO invariant and that the noise from both the cameras is all but identical in quantity and quality. The E-PL5 looks a bit fuzzier, but this is mainly due to the full spectrum modification. IR light and visible light focus at slightly different points. I did not do dark frame subtraction for any of these images and I noted that the E-PL5 had more stuck and hot pixels than the E-M1 II.

I'm going to run some more tests and will post below. I want to show how increasing the exposure decreases noise (fairly certain that this will be the case) and to see if stacking many images with short exposures improves noise over shooting a single long exposure (I have no idea which way this will go).

This has really set me thinking and I will be changing my approach to setting shutter speed, aperture and ISO in the future. In order to get the cleanest image possible it is preferable to capture as much light as possible, so setting the shutter speed and aperture (where possible) are the priority and ISO will be used to correct for the exposure. Where ISO becomes an issue is in relation to dynamic range. Increasing the ISO will decrease the dynamic range of the sensor, so there is going to be an upper limit depending on the subject.

Have you read this article? This helped me a lot when I was trying to determine the best ISO to use for my modified EPM2 for long exposure astrophotography.

What is the best ISO for your DSLR for astrophotography?

Cheers

 

[MadMarco]

Have you read this article? This helped me a lot when I was trying to determine the best ISO to use for my modified EPM2 for long exposure astrophotography.

What is the best ISO for your DSLR for astrophotography?

Cheers

Thanks for the link to the article, it's an interesting read and confirms much of what I was thinking.

For the E-M1 II and E-PL5 I'm probably going to stick to 800 ISO for most of my astro imaging. This should keep noise to the minimum and the DR as high as necessary.

 

[georgian82]

Thanks for the link to the article, it's an interesting read and confirms much of what I was thinking.

For the E-M1 II and E-PL5 I'm probably going to stick to 800 ISO for most of my astro imaging. This should keep noise to the minimum and the DR as high as necessary.

You are welcome.

I also stick to ISO 800 with my EPM2 and it seems to be the best compromise. The other thing that is really important is to acquire a lot of data...the more light frames the better. I usually try to get at least 80 light frames but my goal is always 100. Finally, I make sure to take at least 35 dark frames in order to calibrate the light frames correctly.

Do you use bias frames?

Cheers

 

[MadMarco]

You are welcome.

I also stick to ISO 800 with my EPM2 and it seems to be the best compromise. The other thing that is really important is to acquire a lot of data...the more light frames the better. I usually try to get at least 80 light frames but my goal is always 100. Finally, I make sure to take at least 35 dark frames in order to calibrate the light frames correctly.

Do you use bias frames?

Cheers

Most of my imaging in the past has been using ISO 800, more out of common sense rather than any science.

I agree that the more data, the better the images are going to be. So far I've tried 5 minute subs with my new mount, the iOptron CEM25P which is giving nice round stars. It's probably not worth going much beyond this, due to the light pollution where I live.

I usually stack with separate dark, bias and flats in DSS which gives pretty good results. So far the weather has been pretty bad, so I've only managed to grab 8x 5min subs before the cloud killed the session. I've attached the image of Pleiades that I managed to get:

Subscribe to see EXIF info for this image (if available)

 

[Darin Berdinka]

Did an iso invariance test last evening on my e-m1 ii and found similar results. 1600 may have had slightly less noise than 800 but 800 seemed to retain a little more edge detail (this is in raw files). Not sure which to use.

I'm trying astro with a PL15 1.7 on the e-m1 ii which is obviously not a great combination due to purple fringing but I think it's actually forced me to experiment, research and learn a lot more than I may have otherwise. While ultimately I may have to get a different lense it's real educational.

I took some star (not milky way) shots last night at 9 combinations of apertures and iso

(1.7, 2.2, 2.8) and (800, 1600, 3200) all at 10 seconds

I was shooting smaller apertures and lower iso with the hopes of decreasing CA (which did in fact work, though it wasn't eliminated completely). The biggest surprise is, that once normalized for exposure, the images shot at f2.8 had just as many stars in them as the images shot at f1.7. I really couldn't distinguish a difference in even the faintest stars. f1.7 had less noise in the black but the stars at f2.8 were definitely more pinpoint and sharper (and like I said the CA issue was partially resolved). And maybe this makes sense as stars are by definition bright objects even the ones that appear very small.

The proof is in the pudding though and I need to try the same test on the milkyway, with it's dimmer cloud like formations I presume there will be a observable difference in f-stops. We shall see. Anyone else find similar results?

 

[MadMarco]

New to m43 and based on my rudimentary attempt at astro, that result seems unbelievable. 8x5min light exposure at what focal length and apeture? plus how many dark frames etc? Thanks so much.

Hi Darin, the images were taken with my William Optics ZenithStar 71 Telescope which is f/5.9 418mm. I used 20 each of dark, bias and flat. The seeing conditions weren't great and I would have got more exposures if it weren't for the clouds rolling in.

I've recently got a new mount for the telescope, an iOptron CEM25P which has GOTO and makes life much easier. Also you really need good polar alignment and guiding to push for longer exposures.

Post processing is Deep Sky Stacker and Photoshop.

 

[Darin Berdinka]

I rewrote my post to focus on invariance....but thanks! What a rabbit hole this astrophotography is........