G9 studio scene comparison

[JaKing]

I would have thought that it's pretty obvious that dynamic range is a fundamental characteristic of a sensor and that a tone curve is a secondary characteristic applied to the data captured by that sensor.

 

[whumber]

I would have thought that it's pretty obvious that dynamic range is a fundamental characteristic of a sensor and that a tone curve is a secondary characteristic applied to the data captured by that sensor.

 

[JaKing]

If you can't see that, then please don't hold your breath on my account ... ;-) .

 

[whumber]

If you can't see that, then please don't hold your breath on my account ... ;-) .

See, the problem is the fact that you don't see the point in comparing the underexposed and then pushed shots when comparing the dynamic range of the two cameras makes me seriously doubt that you understand any of the articles you linked. Answering the question would very easily clarify if you understand them correctly or not.

 

[JaKing]

Sorry, but I learned many years ago not to play these silly games here.

The fact that you seem to want to tells me something about you.

Bye.

 

[whumber]

Sorry, but I learned many years ago not to play these silly games here.

The fact that you seem to want to tells me something about you.

Bye.

This is not a trick question, I'm not sure I understand why you're acting like this is some kind of hostile interrogation.

You claim that there is a significant difference in dynamic range between the E-M1 and E-M1ii. I posted an example showing that was not the case. You can also check DxO's measurements as they show the same thing (E-M1 = 12.7 EV, E-M1ii = 12.8 EV).

In response you claim that you "didn't see the point" in an underexposed shot that is then pushed when characterizing dynamic range and that resampling images to a common output size "destroys" the comparison.

I do believe that we're talking about two separate things, I think you're talking about the default tone curve and I'm talking about sensor dynamic range. You could confirm this very easily by just answering the question I asked earlier. To repeat, would you say that in the images below, that the image on the right has better dynamic range than the image on the left? It's not a trick question, I don't understand your reluctance to answer the question.

 

[JaKing]

I don't believe in inquisitions, and I'm a bit old for being in a school classroom ... ;-) .

State whatever you want to, then support it with whatever evidence seems good to you.

 

[bobn2]

Yes but a high iso on the mk2 has more dynamic and tonality range than the others most of the noise can cleared at those iso,s which is same for all of them

I don't think it does despite the "different" sensor designations due to the specific customisations required by the brand, be it the CDAF layer on the E-M1 II or the video centric features of the GH5.

Reference for your assertion about it being a layer? It is my understanding that the focusing pixels are part of the imaging surface of the sensor.

It's an adaptation of the microlens layer, masking or profiling the microlenses to receive light from one side or the other of the exit pupil. It is easily added to standard sensor silicon with a new microlens layer. I would hazard a guess that this is the root of Sony's different model numbers for very similar sensors. They have essentially the same silicon but different microlens layers.

My apologies in advance in this post appears to be professing to be an expert in everything, I just thought that it might be useful to apply what knowledge I do have to clarify this particular point.

 

[bobn2]

Why does the G9 produce much better detail than the Oly EM 1.2?

They tested the G9 with the Nocticron 42.5mm f/1.2, and the Oly EM 1.2 was tested with the Oly 45mm 1.8.

A sharper lens produces more detail.

The E-M1ii is also slightly misfocused as it was done before DPReview implemented their new focusing procedure. I spoke with Rishi about it when they wrote the article related to the early soft A9 results but I guess they've been too busy to go back and redo the E-M1ii.

Cameras are so good these days that the test procedures have to be very, very precise if tolerance in them isn't going to skew the results more than some performance differences in the cameras on test. They days are past when you could do the required testing in your spare bedroom with normal photographic equipment.

People also tend to concentrate on the wrong aspects. Focussing properly is vitally important but not easy.

 

[Adrian Harris]

Why does the G9 produce much better detail than the Oly EM 1.2?

They tested the G9 with the Nocticron 42.5mm f/1.2, and the Oly EM 1.2 was tested with the Oly 45mm 1.8.

A sharper lens produces more detail.

The E-M1ii is also slightly misfocused as it was done before DPReview implemented their new focusing procedure. I spoke with Rishi about it when they wrote the article related to the early soft A9 results but I guess they've been too busy to go back and redo the E-M1ii.

This raises an interesting point in that it does not matter how good a camera is in theory, or on a test rig, if in the real world the focusing is likely to be regularly slightly off (or some other factor such as stabilisation causing frequent soft images).

Unfortunately it now looks as if modern hi-res lenses and high megapixel cameras are starting to show up these other inconsistencies.

My Panasonic gx8 would give me the occasional problem like this and my em1-mk2 is all over the place regarding trying to get consistently sharp images (as other truthful owners have also noted).

So unless we all use a tripod with a focus rail and focus manually for every shot, then a lot of the talk regarding testing is basically 'bench racing'.

Consistent focusing and effective stabilisation (that doesn't detract from sharpness) is I think now far more important than any miniscule non practical resolution increase.

 

[bclaff]

...

I have read it explained that Bill's DR testing method reduces the impact of read noise ...

At best that statement is misleading.

DxOMark uses only read noise in determining their Dynamic Range (Landscape score).
PhotonsToPhotos (my site) uses all noise sources in determining Photographic Dynamic Range (PDR)

When you take a picture the noise in that image is not limited to only read noise.
So the PhotonsToPhotos approach is clearly more realistic than that of DxOMark.

 

[bclaff]

Looking at the near identical RAW results across the board with the various 20mp m43 model , it is reasonable to suppose that they all { not surprisingly } share the same 20mp Sony base sensor. Hence why other than processing tweaks { NR levels etc } and AA filter strength/weakness the output is so close,

Bill Claff's measurements show otherwise, very distinctly.

Not really, Bill plots PDR vs specified ISO setting rather than some sort of measured ISO value. Olympus decided to pull a Fuji with the ISO definitions on the E-M1ii so that plot looks much better than the camera performs in reality.

The dynamic range values are correct but where middle gray is located in the dynamic range may be shifted depending on "measured ISO".
If you have plenty of available light then whether "measured ISO" is relevant is debatable.
In any case, feel free to visually shift PhotonsToPhotos curves left or right as you like to make comparisons; if you think it important.

 

[bclaff]

I would have thought that it's pretty obvious that dynamic range is a fundamental characteristic of a sensor and that a tone curve is a secondary characteristic applied to the data captured by that sensor.

You may or may not realize that this touches on one reason that "measured ISO" is not important. Middle gray in the raw data and middle gray in the final image are affected by a tone curve. So long as middle gray winds up in the right place in the final image you are "apples to apples" using ISO setting rather than "measured ISO".

 

[whumber]

Looking at the near identical RAW results across the board with the various 20mp m43 model , it is reasonable to suppose that they all { not surprisingly } share the same 20mp Sony base sensor. Hence why other than processing tweaks { NR levels etc } and AA filter strength/weakness the output is so close,

Bill Claff's measurements show otherwise, very distinctly.

Not really, Bill plots PDR vs specified ISO setting rather than some sort of measured ISO value. Olympus decided to pull a Fuji with the ISO definitions on the E-M1ii so that plot looks much better than the camera performs in reality.

The dynamic range values are correct but where middle gray is located in the dynamic range may be shifted depending on "measured ISO".
If you have plenty of available light then whether "measured ISO" is relevant is debatable.
In any case, feel free to visually shift PhotonsToPhotos curves left or right as you like to make comparisons; if you think it important.

 

[bobn2]

...

I have read it explained that Bill's DR testing method reduces the impact of read noise ...

At best that statement is misleading.
DxOMark uses only read noise in determining their Dynamic Range (Landscape score).
PhotonsToPhotos (my site) uses all noise sources in determining Photographic Dynamic Range (PDR)

To be fair around,Bill, that is also slightly misleading. Yes, your metric includes all sources of noise, but it is based on an arbitrarily selected threshold value, and the choice of that threshold determines the mix of the different noises.

When you take a picture the noise in that image is not limited to only read noise.
So the PhotonsToPhotos approach is clearly more realistic than that of DxOMark.

I think you're over-egging the pudding when you say 'clearly'. Your PDR metric is not based on any perceptual research, simply your idea of what is reasonable. If you read some of John Sheehy's thoughts on read noise, and compare with your PDR, it seems that you're overestimating the SNR required in the shadow areas for acceptable results. It seems possible that we don't expect high levels of detail in the shadows, but some is still better than completely plugged.

 

[bobn2]

Looking at the near identical RAW results across the board with the various 20mp m43 model , it is reasonable to suppose that they all { not surprisingly } share the same 20mp Sony base sensor. Hence why other than processing tweaks { NR levels etc } and AA filter strength/weakness the output is so close,

Bill Claff's measurements show otherwise, very distinctly.

Not really, Bill plots PDR vs specified ISO setting rather than some sort of measured ISO value. Olympus decided to pull a Fuji with the ISO definitions on the E-M1ii so that plot looks much better than the camera performs in reality.

The dynamic range values are correct but where middle gray is located in the dynamic range may be shifted depending on "measured ISO".
If you have plenty of available light then whether "measured ISO" is relevant is debatable.
In any case, feel free to visually shift PhotonsToPhotos curves left or right as you like to make comparisons; if you think it important.

 

[bclaff]

...

When you take a picture the noise in that image is not limited to only read noise.
So the PhotonsToPhotos approach is clearly more realistic than that of DxOMark.

I think you're over-egging the pudding when you say 'clearly'. Your PDR metric is not based on any perceptual research, simply your idea of what is reasonable. ...

I think it is rooted in perceptual research such as known characteristics of visual acuity.
It is true that within that framework I chose certain thresholds; also based on ISO definitions (like acceptable SNR) and and accepted standards (like CoC).

 

[whumber]

The dynamic range values are correct but where middle gray is located in the dynamic range may be shifted depending on "measured ISO".
If you have plenty of available light then whether "measured ISO" is relevant is debatable.
In any case, feel free to visually shift PhotonsToPhotos curves left or right as you like to make comparisons; if you think it important.

 

[bobn2]

The dynamic range values are correct but where middle gray is located in the dynamic range may be shifted depending on "measured ISO".
If you have plenty of available light then whether "measured ISO" is relevant is debatable.
In any case, feel free to visually shift PhotonsToPhotos curves left or right as you like to make comparisons; if you think it important.

 

[bclaff]

As a resource for DxOMark Measured ISO values one can consult the sort-able table below the DxOMark Derived Measured ISO chart at PhotonsToPhotos.

Here are the cameras that shift more than 2/3 stop toward highlight protection:



Lots of Olympus cameras there as well as some other surprise contestants.

--
Bill ( Your trusted source for independent sensor data at PhotonsToPhotos )