Panasonic G9 vs G9M2 Normalised PDR Chart

knickerhawk said:

Serguei Palto said:

DaveyB said:

This thread is the equivalent of Jesuits arguing endlessly about how many angels can fit on the head of a pin. I’ve taken hundreds of thousands of wildlife photos with the G9 and a few thousand now with the G9 II. The new sensor is a definite step forward over the old 20 mp sensor. It produces more flexible files, you can raise shadows cleanly, crop a good bit more and there is also something undefinably more pleasing about the images when paired with the PL 100-400. The latter is something a photographer would recognise but a scientist might struggle to measure. On my social media, people who don’t know I’ve changed cameras have commented on the “wow” factor of some recent images and they aren’t tainted by some of the biases which inevitably taint all photography forums. Still, I guess this interminable measurebating keeps some people occupied and out of trouble instead of going out and taking photographs.

Click to expand...

Sure, with the G9II the Panasonic made a significant step ahead. I have said about this just after my first measurements of the G9II-DR many months ago.

The problem is in convincing "experts" like from "Photons to Photos" that there is the problem with their results showing very moderate improvement of G9II-DR compared to G9-DR. There is the problem with their methodology. The correct measurements definitely should correlate with real-life observations of photographers.

Click to expand...

OK. The "correct measurements" per IWE indicates about a +2.5 Ev advantage for the G9ii vs. the G9. Thanks to jrsforums, we have a set of well-controlled lowlight shots for the G9ii and G9. While these shots are of the test chart variety rather than real scenes that photographers typically shoot, they're more "real" than black frames and better controlled than you often get with scenes outside of a studio set-up. They definitely suffice for purposes of comparing the deep shadow behavior of the two cameras, which is where the purported +2.5 Ev advantage for the G9ii should be found. There's one challenge presented by these shots: the aperture setting and lens are constant, but the use of matrix metering resulted in non-matching shutter speeds (and, hence, non-matching exposures) between the two cameras. However, as you'll see from the range of displayed exposures, it's still possible to perform a good comparison and draw your own conclusions about the extent of the "real life" advantage of the G9ii.

In this post I'm using screen grabs from RawDigger. This is done for two reasons. First, RawDigger has been set to treat both sets of images in a very neutral and consistent way: As Shot white balance, no camera profile applied, and simple 2x2 pixel interpolation. RawDigger doesn't apply sharpening or noise reduction. Any "black box" processing differences that might arise later when we compare raws converted in ACR are minimized by my use of identical explicit presets in RawDigger. Second, RawDigger makes it very easy to see how black point settings impact the image display. As previously discussed and shown in this thread, even the smallest possible change to the applied black level can have a significant visible effect on deep shadows and associated read noise, especially when lightened in processing. One final point here about neutrality and consistency of the images presented: the screen grabs were made on my Mac Studio Monitor with display size set to 100% in RawDigger and on my Mac Studio. The screen grabs were copied to Photoshop, and output to sRGB JPEGs for upload here. No adjustments were applied to any of the screen grabs in Photoshop.

The RawDigger header portion of the screen grabs conveniently shows useful Exif info such as which camera, shutter speed, aperture, ISO, and lens was used. Note that only the camera body and shutter speed differs in any of these screen grabs. Also pay attention to the footer portion, where Black Level info is shown. The following exposures and corresponding Ev differences relative to the "base" G9 -7 Ev exposures (bolded) are shown below:

1. G9ii 1/1000 (auto black point) +2.67 Ev compared to the base G9 exposures
2. G9 1/1600 (auto black point) +2.00 Ev compared to the base G9 exposures
3. G9ii 1/4000 (auto black point) +0.67 Ev compared to the base G9 exposures
4. G9 1/6400 (auto black point) +0.00 Ev Base G9 exposure
5. G9 1/6400 (manual black point) +0.00 Ev Base G9 exposure
6. G9ii 1/8000 (auto black point) -0.33 E compared to the base G9 exposures

According to the Scientific DR metric used in IWE, which predicts about a +2.5 Ev advantage for the G9ii, #3 should appear to be less noisy than #2. Likewise, IWE predicts there should be a large and very visible advantage for #3 (over +2.00 Ev) compared to #4, and #6 should also have a large noise advantage over #4 and #5.

On the other hand, since PhotonsToPhotos' Photographic Dynamic Range (PDR) predicts just under +0.33 advantage for the G9ii, it predicts that #6 should show almost the same amount of noise as #4 and #5. Likewise, PDR predicts that there should be about a +1.00 Ev noise advantage for #3 compared to #4 and #5.

Considering the wide disparity in the predictions, the screen grabs should make things pretty clear as to which DR metric is more useful for "real life" shots like this test scene (and similarly low-lit shots that require a lot of lightening in conversion).

Here are the shots in the order listed above:


#1 g9ii 1/1000


#2 G9 1/1600 (Auto black point set to 142,143,142,143)


#3 g9ii 1/4000


#4 G9 1/6400 (Auto black point - 142, 142,142,142)


#5 G9 1/6400 (Black point manually set to 143,143,143,143)


#6 g9ii 1/8000

So, what are you seeing? A 2.5 Ev advantage for the G9ii or something much less impressive? Let's get some feedback and observations about the comps shown above and then we can continue with a discussion of why there's an obvious confounding factor based on which black level is applied to some lowlight G9 shots, and how that "problem" (and, yes, it's a problem) recurs when these lowlight G9 images are converted at default settings in the Adobe tools (and, likely, most other raw converters).

Click to expand...

Serguei Palto said:

knickerhawk said:

Serguei Palto said:

DaveyB said:

This thread is the equivalent of Jesuits arguing endlessly about how many angels can fit on the head of a pin. I’ve taken hundreds of thousands of wildlife photos with the G9 and a few thousand now with the G9 II. The new sensor is a definite step forward over the old 20 mp sensor. It produces more flexible files, you can raise shadows cleanly, crop a good bit more and there is also something undefinably more pleasing about the images when paired with the PL 100-400. The latter is something a photographer would recognise but a scientist might struggle to measure. On my social media, people who don’t know I’ve changed cameras have commented on the “wow” factor of some recent images and they aren’t tainted by some of the biases which inevitably taint all photography forums. Still, I guess this interminable measurebating keeps some people occupied and out of trouble instead of going out and taking photographs.

Click to expand...

Sure, with the G9II the Panasonic made a significant step ahead. I have said about this just after my first measurements of the G9II-DR many months ago.

The problem is in convincing "experts" like from "Photons to Photos" that there is the problem with their results showing very moderate improvement of G9II-DR compared to G9-DR. There is the problem with their methodology. The correct measurements definitely should correlate with real-life observations of photographers.

Click to expand...

OK. The "correct measurements" per IWE indicates about a +2.5 Ev advantage for the G9ii vs. the G9. Thanks to jrsforums, we have a set of well-controlled lowlight shots for the G9ii and G9. While these shots are of the test chart variety rather than real scenes that photographers typically shoot, they're more "real" than black frames and better controlled than you often get with scenes outside of a studio set-up. They definitely suffice for purposes of comparing the deep shadow behavior of the two cameras, which is where the purported +2.5 Ev advantage for the G9ii should be found. There's one challenge presented by these shots: the aperture setting and lens are constant, but the use of matrix metering resulted in non-matching shutter speeds (and, hence, non-matching exposures) between the two cameras. However, as you'll see from the range of displayed exposures, it's still possible to perform a good comparison and draw your own conclusions about the extent of the "real life" advantage of the G9ii.

In this post I'm using screen grabs from RawDigger. This is done for two reasons. First, RawDigger has been set to treat both sets of images in a very neutral and consistent way: As Shot white balance, no camera profile applied, and simple 2x2 pixel interpolation. RawDigger doesn't apply sharpening or noise reduction. Any "black box" processing differences that might arise later when we compare raws converted in ACR are minimized by my use of identical explicit presets in RawDigger. Second, RawDigger makes it very easy to see how black point settings impact the image display. As previously discussed and shown in this thread, even the smallest possible change to the applied black level can have a significant visible effect on deep shadows and associated read noise, especially when lightened in processing. One final point here about neutrality and consistency of the images presented: the screen grabs were made on my Mac Studio Monitor with display size set to 100% in RawDigger and on my Mac Studio. The screen grabs were copied to Photoshop, and output to sRGB JPEGs for upload here. No adjustments were applied to any of the screen grabs in Photoshop.

The RawDigger header portion of the screen grabs conveniently shows useful Exif info such as which camera, shutter speed, aperture, ISO, and lens was used. Note that only the camera body and shutter speed differs in any of these screen grabs. Also pay attention to the footer portion, where Black Level info is shown. The following exposures and corresponding Ev differences relative to the "base" G9 -7 Ev exposures (bolded) are shown below:

1. G9ii 1/1000 (auto black point) +2.67 Ev compared to the base G9 exposures
2. G9 1/1600 (auto black point) +2.00 Ev compared to the base G9 exposures
3. G9ii 1/4000 (auto black point) +0.67 Ev compared to the base G9 exposures
4. G9 1/6400 (auto black point) +0.00 Ev Base G9 exposure
5. G9 1/6400 (manual black point) +0.00 Ev Base G9 exposure
6. G9ii 1/8000 (auto black point) -0.33 E compared to the base G9 exposures

According to the Scientific DR metric used in IWE, which predicts about a +2.5 Ev advantage for the G9ii, #3 should appear to be less noisy than #2. Likewise, IWE predicts there should be a large and very visible advantage for #3 (over +2.00 Ev) compared to #4, and #6 should also have a large noise advantage over #4 and #5.

On the other hand, since PhotonsToPhotos' Photographic Dynamic Range (PDR) predicts just under +0.33 advantage for the G9ii, it predicts that #6 should show almost the same amount of noise as #4 and #5. Likewise, PDR predicts that there should be about a +1.00 Ev noise advantage for #3 compared to #4 and #5.

Considering the wide disparity in the predictions, the screen grabs should make things pretty clear as to which DR metric is more useful for "real life" shots like this test scene (and similarly low-lit shots that require a lot of lightening in conversion).

Here are the shots in the order listed above:


#1 g9ii 1/1000


#2 G9 1/1600 (Auto black point set to 142,143,142,143)


#3 g9ii 1/4000


#4 G9 1/6400 (Auto black point - 142, 142,142,142)


#5 G9 1/6400 (Black point manually set to 143,143,143,143)


#6 g9ii 1/8000

So, what are you seeing? A 2.5 Ev advantage for the G9ii or something much less impressive? Let's get some feedback and observations about the comps shown above and then we can continue with a discussion of why there's an obvious confounding factor based on which black level is applied to some lowlight G9 shots, and how that "problem" (and, yes, it's a problem) recurs when these lowlight G9 images are converted at default settings in the Adobe tools (and, likely, most other raw converters).

Click to expand...

The sigma has completely different meaning in presence of the image-signal like in all of your screenshot examples. I am not familiar with the RawDigger, but if the RawDigger has tools for the local analysis of the image regions, where the photon noise is absent or minimal then one can discuss the results for these regions. Up to now, there is nothing to discuss.

Click to expand...

knickerhawk said:

Serguei Palto said:

knickerhawk said:

Serguei Palto said:

DaveyB said:

This thread is the equivalent of Jesuits arguing endlessly about how many angels can fit on the head of a pin. I’ve taken hundreds of thousands of wildlife photos with the G9 and a few thousand now with the G9 II. The new sensor is a definite step forward over the old 20 mp sensor. It produces more flexible files, you can raise shadows cleanly, crop a good bit more and there is also something undefinably more pleasing about the images when paired with the PL 100-400. The latter is something a photographer would recognise but a scientist might struggle to measure. On my social media, people who don’t know I’ve changed cameras have commented on the “wow” factor of some recent images and they aren’t tainted by some of the biases which inevitably taint all photography forums. Still, I guess this interminable measurebating keeps some people occupied and out of trouble instead of going out and taking photographs.

Click to expand...

Sure, with the G9II the Panasonic made a significant step ahead. I have said about this just after my first measurements of the G9II-DR many months ago.

The problem is in convincing "experts" like from "Photons to Photos" that there is the problem with their results showing very moderate improvement of G9II-DR compared to G9-DR. There is the problem with their methodology. The correct measurements definitely should correlate with real-life observations of photographers.

Click to expand...

OK. The "correct measurements" per IWE indicates about a +2.5 Ev advantage for the G9ii vs. the G9. Thanks to jrsforums, we have a set of well-controlled lowlight shots for the G9ii and G9. While these shots are of the test chart variety rather than real scenes that photographers typically shoot, they're more "real" than black frames and better controlled than you often get with scenes outside of a studio set-up. They definitely suffice for purposes of comparing the deep shadow behavior of the two cameras, which is where the purported +2.5 Ev advantage for the G9ii should be found. There's one challenge presented by these shots: the aperture setting and lens are constant, but the use of matrix metering resulted in non-matching shutter speeds (and, hence, non-matching exposures) between the two cameras. However, as you'll see from the range of displayed exposures, it's still possible to perform a good comparison and draw your own conclusions about the extent of the "real life" advantage of the G9ii.

In this post I'm using screen grabs from RawDigger. This is done for two reasons. First, RawDigger has been set to treat both sets of images in a very neutral and consistent way: As Shot white balance, no camera profile applied, and simple 2x2 pixel interpolation. RawDigger doesn't apply sharpening or noise reduction. Any "black box" processing differences that might arise later when we compare raws converted in ACR are minimized by my use of identical explicit presets in RawDigger. Second, RawDigger makes it very easy to see how black point settings impact the image display. As previously discussed and shown in this thread, even the smallest possible change to the applied black level can have a significant visible effect on deep shadows and associated read noise, especially when lightened in processing. One final point here about neutrality and consistency of the images presented: the screen grabs were made on my Mac Studio Monitor with display size set to 100% in RawDigger and on my Mac Studio. The screen grabs were copied to Photoshop, and output to sRGB JPEGs for upload here. No adjustments were applied to any of the screen grabs in Photoshop.

The RawDigger header portion of the screen grabs conveniently shows useful Exif info such as which camera, shutter speed, aperture, ISO, and lens was used. Note that only the camera body and shutter speed differs in any of these screen grabs. Also pay attention to the footer portion, where Black Level info is shown. The following exposures and corresponding Ev differences relative to the "base" G9 -7 Ev exposures (bolded) are shown below:

1. G9ii 1/1000 (auto black point) +2.67 Ev compared to the base G9 exposures
2. G9 1/1600 (auto black point) +2.00 Ev compared to the base G9 exposures
3. G9ii 1/4000 (auto black point) +0.67 Ev compared to the base G9 exposures
4. G9 1/6400 (auto black point) +0.00 Ev Base G9 exposure
5. G9 1/6400 (manual black point) +0.00 Ev Base G9 exposure
6. G9ii 1/8000 (auto black point) -0.33 E compared to the base G9 exposures

According to the Scientific DR metric used in IWE, which predicts about a +2.5 Ev advantage for the G9ii, #3 should appear to be less noisy than #2. Likewise, IWE predicts there should be a large and very visible advantage for #3 (over +2.00 Ev) compared to #4, and #6 should also have a large noise advantage over #4 and #5.

On the other hand, since PhotonsToPhotos' Photographic Dynamic Range (PDR) predicts just under +0.33 advantage for the G9ii, it predicts that #6 should show almost the same amount of noise as #4 and #5. Likewise, PDR predicts that there should be about a +1.00 Ev noise advantage for #3 compared to #4 and #5.

Considering the wide disparity in the predictions, the screen grabs should make things pretty clear as to which DR metric is more useful for "real life" shots like this test scene (and similarly low-lit shots that require a lot of lightening in conversion).

Here are the shots in the order listed above:


#1 g9ii 1/1000


#2 G9 1/1600 (Auto black point set to 142,143,142,143)


#3 g9ii 1/4000


#4 G9 1/6400 (Auto black point - 142, 142,142,142)


#5 G9 1/6400 (Black point manually set to 143,143,143,143)


#6 g9ii 1/8000

So, what are you seeing? A 2.5 Ev advantage for the G9ii or something much less impressive? Let's get some feedback and observations about the comps shown above and then we can continue with a discussion of why there's an obvious confounding factor based on which black level is applied to some lowlight G9 shots, and how that "problem" (and, yes, it's a problem) recurs when these lowlight G9 images are converted at default settings in the Adobe tools (and, likely, most other raw converters).

Click to expand...

The sigma has completely different meaning in presence of the image-signal like in all of your screenshot examples. I am not familiar with the RawDigger, but if the RawDigger has tools for the local analysis of the image regions, where the photon noise is absent or minimal then one can discuss the results for these regions. Up to now, there is nothing to discuss.

Click to expand...

A range of the jrsforums' -Ev test shots were included in order to provide some context for comparisons. However, as I noted the "base" shots for comparison purposes are the -7 Ev shots. These are extreme lowlight shots where read noise is certainly a significant contributor. And, yes, of course RawDigger allows you to target/select specific regions of an image for statistics and histogram display. See below showing 100x100 targeted boxes over the darkest patch in the ColorChecker. The statistics for the targeted box are in the header and outlined in red.


#4 G9 (1/6400) .


#6 G9ii (1/8000)

Click to expand...

JosephScha said:

I had a G9 and now have a G9 II. I really don't trust the PDR chart. G9 II's Dynamic Range Boost does not just pick up at ISO 3200. In fact, it works from ISO 100. And, I think the result is more than reflected in that chart. I don't know why.

Click to expand...

JosephScha said:

I had a G9 and now have a G9 II. I really don't trust the PDR chart. G9 II's Dynamic Range Boost does not just pick up at ISO 3200. In fact, it works from ISO 100. And, I think the result is more than reflected in that chart. I don't know why.

Click to expand...

Jens_ac said:

knickerhawk said:

Serguei Palto said:

knickerhawk said:

Serguei Palto said:

DaveyB said:

This thread is the equivalent of Jesuits arguing endlessly about how many angels can fit on the head of a pin. I’ve taken hundreds of thousands of wildlife photos with the G9 and a few thousand now with the G9 II. The new sensor is a definite step forward over the old 20 mp sensor. It produces more flexible files, you can raise shadows cleanly, crop a good bit more and there is also something undefinably more pleasing about the images when paired with the PL 100-400. The latter is something a photographer would recognise but a scientist might struggle to measure. On my social media, people who don’t know I’ve changed cameras have commented on the “wow” factor of some recent images and they aren’t tainted by some of the biases which inevitably taint all photography forums. Still, I guess this interminable measurebating keeps some people occupied and out of trouble instead of going out and taking photographs.

Click to expand...

Sure, with the G9II the Panasonic made a significant step ahead. I have said about this just after my first measurements of the G9II-DR many months ago.

The problem is in convincing "experts" like from "Photons to Photos" that there is the problem with their results showing very moderate improvement of G9II-DR compared to G9-DR. There is the problem with their methodology. The correct measurements definitely should correlate with real-life observations of photographers.

Click to expand...

OK. The "correct measurements" per IWE indicates about a +2.5 Ev advantage for the G9ii vs. the G9. Thanks to jrsforums, we have a set of well-controlled lowlight shots for the G9ii and G9. While these shots are of the test chart variety rather than real scenes that photographers typically shoot, they're more "real" than black frames and better controlled than you often get with scenes outside of a studio set-up. They definitely suffice for purposes of comparing the deep shadow behavior of the two cameras, which is where the purported +2.5 Ev advantage for the G9ii should be found. There's one challenge presented by these shots: the aperture setting and lens are constant, but the use of matrix metering resulted in non-matching shutter speeds (and, hence, non-matching exposures) between the two cameras. However, as you'll see from the range of displayed exposures, it's still possible to perform a good comparison and draw your own conclusions about the extent of the "real life" advantage of the G9ii.

In this post I'm using screen grabs from RawDigger. This is done for two reasons. First, RawDigger has been set to treat both sets of images in a very neutral and consistent way: As Shot white balance, no camera profile applied, and simple 2x2 pixel interpolation. RawDigger doesn't apply sharpening or noise reduction. Any "black box" processing differences that might arise later when we compare raws converted in ACR are minimized by my use of identical explicit presets in RawDigger. Second, RawDigger makes it very easy to see how black point settings impact the image display. As previously discussed and shown in this thread, even the smallest possible change to the applied black level can have a significant visible effect on deep shadows and associated read noise, especially when lightened in processing. One final point here about neutrality and consistency of the images presented: the screen grabs were made on my Mac Studio Monitor with display size set to 100% in RawDigger and on my Mac Studio. The screen grabs were copied to Photoshop, and output to sRGB JPEGs for upload here. No adjustments were applied to any of the screen grabs in Photoshop.

The RawDigger header portion of the screen grabs conveniently shows useful Exif info such as which camera, shutter speed, aperture, ISO, and lens was used. Note that only the camera body and shutter speed differs in any of these screen grabs. Also pay attention to the footer portion, where Black Level info is shown. The following exposures and corresponding Ev differences relative to the "base" G9 -7 Ev exposures (bolded) are shown below:

1. G9ii 1/1000 (auto black point) +2.67 Ev compared to the base G9 exposures
2. G9 1/1600 (auto black point) +2.00 Ev compared to the base G9 exposures
3. G9ii 1/4000 (auto black point) +0.67 Ev compared to the base G9 exposures
4. G9 1/6400 (auto black point) +0.00 Ev Base G9 exposure
5. G9 1/6400 (manual black point) +0.00 Ev Base G9 exposure
6. G9ii 1/8000 (auto black point) -0.33 E compared to the base G9 exposures

According to the Scientific DR metric used in IWE, which predicts about a +2.5 Ev advantage for the G9ii, #3 should appear to be less noisy than #2. Likewise, IWE predicts there should be a large and very visible advantage for #3 (over +2.00 Ev) compared to #4, and #6 should also have a large noise advantage over #4 and #5.

On the other hand, since PhotonsToPhotos' Photographic Dynamic Range (PDR) predicts just under +0.33 advantage for the G9ii, it predicts that #6 should show almost the same amount of noise as #4 and #5. Likewise, PDR predicts that there should be about a +1.00 Ev noise advantage for #3 compared to #4 and #5.

Considering the wide disparity in the predictions, the screen grabs should make things pretty clear as to which DR metric is more useful for "real life" shots like this test scene (and similarly low-lit shots that require a lot of lightening in conversion).

Here are the shots in the order listed above:


#1 g9ii 1/1000


#2 G9 1/1600 (Auto black point set to 142,143,142,143)


#3 g9ii 1/4000


#4 G9 1/6400 (Auto black point - 142, 142,142,142)


#5 G9 1/6400 (Black point manually set to 143,143,143,143)


#6 g9ii 1/8000

So, what are you seeing? A 2.5 Ev advantage for the G9ii or something much less impressive? Let's get some feedback and observations about the comps shown above and then we can continue with a discussion of why there's an obvious confounding factor based on which black level is applied to some lowlight G9 shots, and how that "problem" (and, yes, it's a problem) recurs when these lowlight G9 images are converted at default settings in the Adobe tools (and, likely, most other raw converters).

Click to expand...

The sigma has completely different meaning in presence of the image-signal like in all of your screenshot examples. I am not familiar with the RawDigger, but if the RawDigger has tools for the local analysis of the image regions, where the photon noise is absent or minimal then one can discuss the results for these regions. Up to now, there is nothing to discuss.

Click to expand...

A range of the jrsforums' -Ev test shots were included in order to provide some context for comparisons. However, as I noted the "base" shots for comparison purposes are the -7 Ev shots. These are extreme lowlight shots where read noise is certainly a significant contributor. And, yes, of course RawDigger allows you to target/select specific regions of an image for statistics and histogram display. See below showing 100x100 targeted boxes over the darkest patch in the ColorChecker. The statistics for the targeted box are in the header and outlined in red.


#4 G9 (1/6400) .


#6 G9ii (1/8000)

Click to expand...

Surprising how much better the colors look with the g9m2 - for DR Serguei's iWE could be right but noise could be better estimated by Bill's PDR. The math is something else

Click to expand...

knickerhawk said:

Jens_ac said:

knickerhawk said:

Serguei Palto said:

knickerhawk said:

Serguei Palto said:

DaveyB said:

This thread is the equivalent of Jesuits arguing endlessly about how many angels can fit on the head of a pin. I’ve taken hundreds of thousands of wildlife photos with the G9 and a few thousand now with the G9 II. The new sensor is a definite step forward over the old 20 mp sensor. It produces more flexible files, you can raise shadows cleanly, crop a good bit more and there is also something undefinably more pleasing about the images when paired with the PL 100-400. The latter is something a photographer would recognise but a scientist might struggle to measure. On my social media, people who don’t know I’ve changed cameras have commented on the “wow” factor of some recent images and they aren’t tainted by some of the biases which inevitably taint all photography forums. Still, I guess this interminable measurebating keeps some people occupied and out of trouble instead of going out and taking photographs.

Click to expand...

Sure, with the G9II the Panasonic made a significant step ahead. I have said about this just after my first measurements of the G9II-DR many months ago.

The problem is in convincing "experts" like from "Photons to Photos" that there is the problem with their results showing very moderate improvement of G9II-DR compared to G9-DR. There is the problem with their methodology. The correct measurements definitely should correlate with real-life observations of photographers.

Click to expand...

OK. The "correct measurements" per IWE indicates about a +2.5 Ev advantage for the G9ii vs. the G9. Thanks to jrsforums, we have a set of well-controlled lowlight shots for the G9ii and G9. While these shots are of the test chart variety rather than real scenes that photographers typically shoot, they're more "real" than black frames and better controlled than you often get with scenes outside of a studio set-up. They definitely suffice for purposes of comparing the deep shadow behavior of the two cameras, which is where the purported +2.5 Ev advantage for the G9ii should be found. There's one challenge presented by these shots: the aperture setting and lens are constant, but the use of matrix metering resulted in non-matching shutter speeds (and, hence, non-matching exposures) between the two cameras. However, as you'll see from the range of displayed exposures, it's still possible to perform a good comparison and draw your own conclusions about the extent of the "real life" advantage of the G9ii.

In this post I'm using screen grabs from RawDigger. This is done for two reasons. First, RawDigger has been set to treat both sets of images in a very neutral and consistent way: As Shot white balance, no camera profile applied, and simple 2x2 pixel interpolation. RawDigger doesn't apply sharpening or noise reduction. Any "black box" processing differences that might arise later when we compare raws converted in ACR are minimized by my use of identical explicit presets in RawDigger. Second, RawDigger makes it very easy to see how black point settings impact the image display. As previously discussed and shown in this thread, even the smallest possible change to the applied black level can have a significant visible effect on deep shadows and associated read noise, especially when lightened in processing. One final point here about neutrality and consistency of the images presented: the screen grabs were made on my Mac Studio Monitor with display size set to 100% in RawDigger and on my Mac Studio. The screen grabs were copied to Photoshop, and output to sRGB JPEGs for upload here. No adjustments were applied to any of the screen grabs in Photoshop.

The RawDigger header portion of the screen grabs conveniently shows useful Exif info such as which camera, shutter speed, aperture, ISO, and lens was used. Note that only the camera body and shutter speed differs in any of these screen grabs. Also pay attention to the footer portion, where Black Level info is shown. The following exposures and corresponding Ev differences relative to the "base" G9 -7 Ev exposures (bolded) are shown below:

1. G9ii 1/1000 (auto black point) +2.67 Ev compared to the base G9 exposures
2. G9 1/1600 (auto black point) +2.00 Ev compared to the base G9 exposures
3. G9ii 1/4000 (auto black point) +0.67 Ev compared to the base G9 exposures
4. G9 1/6400 (auto black point) +0.00 Ev Base G9 exposure
5. G9 1/6400 (manual black point) +0.00 Ev Base G9 exposure
6. G9ii 1/8000 (auto black point) -0.33 E compared to the base G9 exposures

According to the Scientific DR metric used in IWE, which predicts about a +2.5 Ev advantage for the G9ii, #3 should appear to be less noisy than #2. Likewise, IWE predicts there should be a large and very visible advantage for #3 (over +2.00 Ev) compared to #4, and #6 should also have a large noise advantage over #4 and #5.

On the other hand, since PhotonsToPhotos' Photographic Dynamic Range (PDR) predicts just under +0.33 advantage for the G9ii, it predicts that #6 should show almost the same amount of noise as #4 and #5. Likewise, PDR predicts that there should be about a +1.00 Ev noise advantage for #3 compared to #4 and #5.

Considering the wide disparity in the predictions, the screen grabs should make things pretty clear as to which DR metric is more useful for "real life" shots like this test scene (and similarly low-lit shots that require a lot of lightening in conversion).

Here are the shots in the order listed above:


#1 g9ii 1/1000


#2 G9 1/1600 (Auto black point set to 142,143,142,143)


#3 g9ii 1/4000


#4 G9 1/6400 (Auto black point - 142, 142,142,142)


#5 G9 1/6400 (Black point manually set to 143,143,143,143)


#6 g9ii 1/8000

So, what are you seeing? A 2.5 Ev advantage for the G9ii or something much less impressive? Let's get some feedback and observations about the comps shown above and then we can continue with a discussion of why there's an obvious confounding factor based on which black level is applied to some lowlight G9 shots, and how that "problem" (and, yes, it's a problem) recurs when these lowlight G9 images are converted at default settings in the Adobe tools (and, likely, most other raw converters).

Click to expand...

The sigma has completely different meaning in presence of the image-signal like in all of your screenshot examples. I am not familiar with the RawDigger, but if the RawDigger has tools for the local analysis of the image regions, where the photon noise is absent or minimal then one can discuss the results for these regions. Up to now, there is nothing to discuss.

Click to expand...

A range of the jrsforums' -Ev test shots were included in order to provide some context for comparisons. However, as I noted the "base" shots for comparison purposes are the -7 Ev shots. These are extreme lowlight shots where read noise is certainly a significant contributor. And, yes, of course RawDigger allows you to target/select specific regions of an image for statistics and histogram display. See below showing 100x100 targeted boxes over the darkest patch in the ColorChecker. The statistics for the targeted box are in the header and outlined in red.


#4 G9 (1/6400) .


#6 G9ii (1/8000)

Click to expand...

Surprising how much better the colors look with the g9m2 - for DR Serguei's iWE could be right but noise could be better estimated by Bill's PDR. The math is something else

Click to expand...

With respect to color, what you should really be surprised about is how much better the -7 Ev G9 shot looks when the smallest possible adjustment to black level is applied to it (see #5 shot above). The haziness and lack of contrast is gone and the color is very similar to the G9ii rendering (and certainly nowhere near what I, at least, would expect to see from a purported 2.5 Ev difference in DR). The black level tweaked version of the G9 shot tells you that, for these extreme lowlight shots, you are likely to get better contrast and color straight from default settings for your G9ii shots compared to default G9 settings, but that you can rectify the effects of the suboptimal default black level setting by tweaking blacks and rgb tone curves in your raw converter.

Click to expand...

jrsforums said:

jrsforums said:

With respect to color, what you should really be surprised about is how much better the -7 Ev G9 shot looks when the smallest possible adjustment to black level is applied to it (see #5 shot above). The haziness and lack of contrast is gone and the color is very similar to the G9ii rendering (and certainly nowhere near what I, at least, would expect to see from a purported 2.5 Ev difference in DR). The black level tweaked version of the G9 shot tells you that, for these extreme lowlight shots, you are likely to get better contrast and color straight from default settings for your G9ii shots compared to default G9 settings, but that you can rectify the effects of the suboptimal default black level setting by tweaking blacks and rgb tone curves in your raw converter.

Click to expand...

Black level adjustment is interesting, but not really practical, as far as I know. How would I adjust it in LRc/ACR?

Plus, while it gets rid of cast, does it modify other colors?

Click to expand...

• Start with a linear profile if you can. Otherwise, I usually prefer Adobe Standard.
• Make the basic big adjustments in the Lightness tab, starting of course with the necessary positive Exposure slider push.
• I usually just leave WB at As Shot, but some experimentation might be appropriate. Most of the color fixes come later with curves adjustments to the individual color channels, so don't try to fix any big color cast present in the image with WB tweaking. WB "fixing" is a sledge hammer for these lowlight color cast-infected shots.
• If it's obvious that the red and green channels are consistently imbalanced in the same direction from the left end to the right end of the histogram, then it often helps to attack that by a corresponding adjustment of the Shadows slider in the Calibration tab. It doesn't have to be perfect since final tweaking will happen in the Curve tool. The shadow adjustment might make things easier in the next step, but don't over-do it. Be conservative with any shadow adjustment you apply and finish the color cast correction in the curve tool.
• The curve tool is where the targeted fixing takes place. Start with the Gray (Point Curve) and perform any all-channel tweaks that are desired, paying attention to any clipping at both ends that need to be addressed. (Alternatively go back to the Light tab and fix clipping and global lightness/contrast there.)
• Using the histogram and maybe sampled spots in the image you're targeting to be neutral (uncolored), adjust the individual color channels. If there's a magenta color cast, this usually means working mostly with the red and blue curves to bring them down to match where the greens are set in the RGB/colors histogram. However, you might also need to tweak the green channel here and there as well. Use your eyes to see what happens when you tweak the curves, but also watch what's going on in the histogram and spot samples to see how the channels come together (good) or separate (bad) as you tweak the individual color channel curves.

knickerhawk said:

jrsforums said:

knickerhawk said:

With respect to color, what you should really be surprised about is how much better the -7 Ev G9 shot looks when the smallest possible adjustment to black level is applied to it (see #5 shot above). The haziness and lack of contrast is gone and the color is very similar to the G9ii rendering (and certainly nowhere near what I, at least, would expect to see from a purported 2.5 Ev difference in DR). The black level tweaked version of the G9 shot tells you that, for these extreme lowlight shots, you are likely to get better contrast and color straight from default settings for your G9ii shots compared to default G9 settings, but that you can rectify the effects of the suboptimal default black level setting by tweaking blacks and rgb tone curves in your raw converter.

Click to expand...

Black level adjustment is interesting, but not really practical, as far as I know. How would I adjust it in LRc/ACR?

Plus, while it gets rid of cast, does it modify other colors?

Click to expand...

RawDigger is not intended for anything other than basic/neutral renderings for comparative analysis. One of its few rendering options is to adjust black level, and it's really only useful for getting a general idea of how much low level tonal adjustment is available in the raw before all hell breaks loose - such as color casts, haziness, splotchy/blocky noise, etc. For your -7 Ev G9 shot, the 1 DN adjustment I applied resulted in an overall improvement in contrast, reduced color cast and no visible introduction of other problems. In other words, it showed that the raw file has the potential to produce an excellent rendering as long as appropriate settings are used during conversion. Those appropriate settings may very well NOT be the default ones in your raw converter of choice.

In LR/ACR, It's not necessary (or advisable) to limit yourself to the Blacks slider, which is the single most comparable tonal adjustment to RawDigger's black level adjustment. One shortcoming is that the Blacks slider affects all color channels equally. The same goes for the other "Lightness" controls. Thus, to attack color casts (which are likely to be rather obvious in the histogram), you need channel-specific controls such as the Curve tool and the Shadows adjustment in the Calibration tab. My general strategy and workflow for conversions involving major shadow pushes:

• Start with a linear profile if you can. Otherwise, I usually prefer Adobe Standard.
• Make the basic big adjustments in the Lightness tab, starting of course with the necessary positive Exposure slider push.
• I usually just leave WB at As Shot, but some experimentation might be appropriate. Most of the color fixes come later with curves adjustments to the individual color channels, so don't try to fix any big color cast present in the image with WB tweaking. WB "fixing" is a sledge hammer for these lowlight color cast-infected shots.
• If it's obvious that the red and green channels are consistently imbalanced in the same direction from the left end to the right end of the histogram, then it often helps to attack that by a corresponding adjustment of the Shadows slider in the Calibration tab. It doesn't have to be perfect since final tweaking will happen in the Curve tool. The shadow adjustment might make things easier in the next step, but don't over-do it. Be conservative with any shadow adjustment you apply and finish the color cast correction in the curve tool.
• The curve tool is where the targeted fixing takes place. Start with the Gray (Point Curve) and perform any all-channel tweaks that are desired, paying attention to any clipping at both ends that need to be addressed. (Alternatively go back to the Light tab and fix clipping and global lightness/contrast there.)
• Using the histogram and maybe sampled spots in the image you're targeting to be neutral (uncolored), adjust the individual color channels. If there's a magenta color cast, this usually means working mostly with the red and blue curves to bring them down to match where the greens are set in the RGB/colors histogram. However, you might also need to tweak the green channel here and there as well. Use your eyes to see what happens when you tweak the curves, but also watch what's going on in the histogram and spot samples to see how the channels come together (good) or separate (bad) as you tweak the individual color channel curves.

If all this sounds like a lot to do, well...it IS a lot to do! Just bear in mind that it gets easier over time as you get familiar with performing complex curve adjustments, And bear in mind that, unless you regularly do lots of very lowlight photography, so much channel-specific fine tuning is rarely needed. If you do a lot of lowlight photography, you will probably want to create DNG profiles and presets that give you a head start on fixing color casts that consistently show up at default ACR/LR settings in your shots.

Below is the -7 Ev G9ii shot processed in ACR to default settings (Adobe Standard) with no sharpening and no denoising (Luminance=0 and Color=0) and +5 Exposure slider adjustment. This is followed by the -7 Ev G9 shot converted in ACR using the workflow/strategy described above (also with no denoising applied and +5 Exposure adjustment. There's no doubt that the G9ii shot has the advantage of requiring no special effort to get a reasonable conversion, free of color cast. On the other hand, the G9 shot is fully capable of being rendered without color cast, haziness and with no more visible noise (actually less noise compared to the default G9ii rendering). Where's the purported 2.5 Ev difference???


-7 Ev G9ii shot converted to default settings, +5 Exposure push and no denoising in ACR


-7 Ev G9 shot with described workflow/strategy to closely match G9ii look, +5 Exposure push and no denoising in ACR

Click to expand...

JosephScha said:

Who in real life exposes 7 stops underexposed, and raises 5 stops in post?

Click to expand...

knickerhawk said:

jrsforums said:

knickerhawk said:

With respect to color, what you should really be surprised about is how much better the -7 Ev G9 shot looks when the smallest possible adjustment to black level is applied to it (see #5 shot above). The haziness and lack of contrast is gone and the color is very similar to the G9ii rendering (and certainly nowhere near what I, at least, would expect to see from a purported 2.5 Ev difference in DR). The black level tweaked version of the G9 shot tells you that, for these extreme lowlight shots, you are likely to get better contrast and color straight from default settings for your G9ii shots compared to default G9 settings, but that you can rectify the effects of the suboptimal default black level setting by tweaking blacks and rgb tone curves in your raw converter.

Click to expand...

Black level adjustment is interesting, but not really practical, as far as I know. How would I adjust it in LRc/ACR?

Plus, while it gets rid of cast, does it modify other colors?

Click to expand...

RawDigger is not intended for anything other than basic/neutral renderings for comparative analysis. One of its few rendering options is to adjust black level, and it's really only useful for getting a general idea of how much low level tonal adjustment is available in the raw before all hell breaks loose - such as color casts, haziness, splotchy/blocky noise, etc. For your -7 Ev G9 shot, the 1 DN adjustment I applied resulted in an overall improvement in contrast, reduced color cast and no visible introduction of other problems. In other words, it showed that the raw file has the potential to produce an excellent rendering as long as appropriate settings are used during conversion. Those appropriate settings may very well NOT be the default ones in your raw converter of choice.

In LR/ACR, It's not necessary (or advisable) to limit yourself to the Blacks slider, which is the single most comparable tonal adjustment to RawDigger's black level adjustment. One shortcoming is that the Blacks slider affects all color channels equally. The same goes for the other "Lightness" controls. Thus, to attack color casts (which are likely to be rather obvious in the histogram), you need channel-specific controls such as the Curve tool and the Shadows adjustment in the Calibration tab. My general strategy and workflow for conversions involving major shadow pushes:

• Start with a linear profile if you can. Otherwise, I usually prefer Adobe Standard.
• Make the basic big adjustments in the Lightness tab, starting of course with the necessary positive Exposure slider push.
• I usually just leave WB at As Shot, but some experimentation might be appropriate. Most of the color fixes come later with curves adjustments to the individual color channels, so don't try to fix any big color cast present in the image with WB tweaking. WB "fixing" is a sledge hammer for these lowlight color cast-infected shots.
• If it's obvious that the red and green channels are consistently imbalanced in the same direction from the left end to the right end of the histogram, then it often helps to attack that by a corresponding adjustment of the Shadows slider in the Calibration tab. It doesn't have to be perfect since final tweaking will happen in the Curve tool. The shadow adjustment might make things easier in the next step, but don't over-do it. Be conservative with any shadow adjustment you apply and finish the color cast correction in the curve tool.
• The curve tool is where the targeted fixing takes place. Start with the Gray (Point Curve) and perform any all-channel tweaks that are desired, paying attention to any clipping at both ends that need to be addressed. (Alternatively go back to the Light tab and fix clipping and global lightness/contrast there.)
• Using the histogram and maybe sampled spots in the image you're targeting to be neutral (uncolored), adjust the individual color channels. If there's a magenta color cast, this usually means working mostly with the red and blue curves to bring them down to match where the greens are set in the RGB/colors histogram. However, you might also need to tweak the green channel here and there as well. Use your eyes to see what happens when you tweak the curves, but also watch what's going on in the histogram and spot samples to see how the channels come together (good) or separate (bad) as you tweak the individual color channel curves.

If all this sounds like a lot to do, well...it IS a lot to do! Just bear in mind that it gets easier over time as you get familiar with performing complex curve adjustments, And bear in mind that, unless you regularly do lots of very lowlight photography, so much channel-specific fine tuning is rarely needed. If you do a lot of lowlight photography, you will probably want to create DNG profiles and presets that give you a head start on fixing color casts that consistently show up at default ACR/LR settings in your shots.

Below is the -7 Ev G9ii shot processed in ACR to default settings (Adobe Standard) with no sharpening and no denoising (Luminance=0 and Color=0) and +5 Exposure slider adjustment. This is followed by the -7 Ev G9 shot converted in ACR using the workflow/strategy described above (also with no denoising applied and +5 Exposure adjustment. There's no doubt that the G9ii shot has the advantage of requiring no special effort to get a reasonable conversion, free of color cast. On the other hand, the G9 shot is fully capable of being rendered without color cast, haziness and with no more visible noise (actually less noise compared to the default G9ii rendering). Where's the purported 2.5 Ev difference???


-7 Ev G9ii shot converted to default settings, +5 Exposure push and no denoising in ACR


-7 Ev G9 shot with described workflow/strategy to closely match G9ii look, +5 Exposure push and no denoising in ACR

Click to expand...

chriss43 said:

Can you please double-check the images you posted? EXIF information says both images are from the G9 M2. So it seems that it is not possible to draw conclusions for G9 vs G9 M2. In addition the images are at ISO = 200. For a fair comparison of shadow recovery capabilities, both cameras should operate at base ISO, which would be ISO = 100 for the G9 M2. Thank you!

Click to expand...

Interceptor121 said:

No bashing if I recall the discussion was about the electronic shutter to which there were no good answers for a while

In terms of dynamic range if you look at this chart the situation has stayed unchanged as this is only mechanical shutter

Some people here have gone on a tanget to say the G9M2 was amazingly better however the best case data point is opyczne saying at SNR=1 the DR would be 13.3 Ev mostly due to increased bit depth which is 1.3 Ev more than any other MFT camera

And there is no agenda this forum is just full of people with some inferiority complex or endlessly defending turf

The G9M2 at base ISO is marginally better than other MFT camera and mostly due to bit depth. At higher ISO is worse that is what the chart says and same as before

Click to expand...

ChristianC1975 said:

ChristianC1975 said:

No bashing if I recall the discussion was about the electronic shutter to which there were no good answers for a while

In terms of dynamic range if you look at this chart the situation has stayed unchanged as this is only mechanical shutter

Some people here have gone on a tanget to say the G9M2 was amazingly better however the best case data point is opyczne saying at SNR=1 the DR would be 13.3 Ev mostly due to increased bit depth which is 1.3 Ev more than any other MFT camera

And there is no agenda this forum is just full of people with some inferiority complex or endlessly defending turf

The G9M2 at base ISO is marginally better than other MFT camera and mostly due to bit depth. At higher ISO is worse that is what the chart says and same as before

Click to expand...

Ciao Massimo (nicer name than Interceptor!),

I fail to understand how, in light of measurements and examples you keep up in the castle saying that, at base ISO, the G9ii is "marginally better" than other MfT sensors, being here "better" = higher DR.

Maybe a picture worth more than 1000 words:



Remember that it has been shown repeatedly that the G9ii has about 2 stops of highlight latitude, similar (a tiny bit better) when compared to the G9 og. That means that it is the shadow noise the factor that defines the DR.

Cameras are just tools. I loved my G9 og, but the mark ii gives a much improved picture at low ISO.

At high ISO I see no difference, no advantage. Nobody is fighting that.

Click to expand...

ChristianC1975 said:

ChristianC1975 said:

No bashing if I recall the discussion was about the electronic shutter to which there were no good answers for a while

In terms of dynamic range if you look at this chart the situation has stayed unchanged as this is only mechanical shutter

Some people here have gone on a tanget to say the G9M2 was amazingly better however the best case data point is opyczne saying at SNR=1 the DR would be 13.3 Ev mostly due to increased bit depth which is 1.3 Ev more than any other MFT camera

And there is no agenda this forum is just full of people with some inferiority complex or endlessly defending turf

The G9M2 at base ISO is marginally better than other MFT camera and mostly due to bit depth. At higher ISO is worse that is what the chart says and same as before

Click to expand...

Ciao Massimo (nicer name than Interceptor!),

I fail to understand how, in light of measurements and examples you keep up in the castle saying that, at base ISO, the G9ii is "marginally better" than other MfT sensors, being here "better" = higher DR.

Maybe a picture worth more than 1000 words:



Remember that it has been shown repeatedly that the G9ii has about 2 stops of highlight latitude, similar (a tiny bit better) when compared to the G9 og. That means that it is the shadow noise the factor that defines the DR.

Cameras are just tools. I loved my G9 og, but the mark ii gives a much improved picture at low ISO.

At high ISO I see no difference, no advantage. Nobody is fighting that.

Click to expand...

chriss43 said:

Can you please double-check the images you posted? EXIF information says both images are from the G9 M2. So it seems that it is not possible to draw conclusions for G9 vs G9 M2. In addition the images are at ISO = 200. For a fair comparison of shadow recovery capabilities, both cameras should operate at base ISO, which would be ISO = 100 for the G9 M2. Thank you!

Click to expand...

Jens_ac said:

Thanks for you summary about shadow recovery. You did all that typical work for G9 but not for the G9m2, thus there could be a stop more (or not). Another stop may or may not be possible to recover in the highlights. Your test is not complete yet - I suggest to follow CineD with their lattitude test: Usuable DR can be found both in shadow recovery and in highlight recovery and since cameras may have different "zero exposure change", that might be non symmetric.

On the other hand: with iWE those missing stopps may be found in the normalization, since here only ADC exists and pixel size does not matter. This should give the G9m2 with the smaller pixel area an "math-advantage" somehow.

But I doubt that you can find that "math-advantage" in the images...at least I could not but I do not compare cameras like that. Most comparisons here are flawed in similar way. PDR also is a methode easy to understand and simple to apply, but difficult to interpret in real live - especially the DRboost confuses the tests often but is very helpful to use.

Click to expand...