Micro 4/3 Panasonic GX80/GX85: the real base ISO is 100?

[mike3996]

Some time ago I noticed on PhotonsToPhotos.net that GX80 might have its best DR at its "pull ISO" setting of 100. Usually DR suffers greatly if you let the camera overexpose by a stop to achieve this pull setting.

But the graph at PhotonsToPhotos begged to differ. I made sample shots just out of curiosity.

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

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

Both are f/7.1, one is 1/1000 sec ISO 200, the other is 1/500 sec at ISO 100. Opened in darktable without any manual manipulations, both exposures very neatly touch the right edge of the histogram (Panasonics have always metered the most accurate exposures). Neither has any clipped pixels outside margins of usual variance and error.

What do you think about this? Is the real usable base ISO of this camera actually the pull setting?

 

[agentlossing]

I'm struggling to see any difference at all, personally.

 

[DeeJayK]

I also don't see any significant difference in these JPGs. Maybe 100% crops would show differences better.

I typically completely discount any "extended" ISOs as just something to puff up a spec sheet. With most M43 cameras I've seen under test bears out that the "extended" ISOs measure identically to the lowest "real" ISO.

For example, here's the chart for the E-M1 Mark II from DXOMark's testing showing that the base ISO 200 measures identically to the extended ISO 64:

But then if we look at the similar chart for the GX80, the extended ISO measures a bit lower (although not as low as claimed):

So maybe there is a little something to the assertion that the extended ISO of the GX80 is a good choice.

- K

 

[John King]

I'm struggling to see any difference at all, personally.

The range of whites is slightly wider in the ISO 200 shot on my tablet display. Would really need to look at them both on my PC ...

 

[Brownie]

Shooting @100 ISO on all my M-4/3 has resulted in a lot of magentas in the whites.

 

[mike3996]

I'm struggling to see any difference at all, personally.

That's my exact point. There's no difference in files despite one using a pull ISO. Usually using a pull ISO means one stop overexposure and in this scene here, it would mean much clipping.

Shooting @100 ISO on all my M-4/3 has resulted in a lot of magentas in the whites.

Is this with raw files? Some raw developers do develop magenta artifacts when trying to recover clipped highlight data.

 

[Brownie]

That's my exact point. There's no difference in files despite one using a pull ISO. Usually using a pull ISO means one stop overexposure and in this scene here, it would mean much clipping.




Is this with raw files? Some raw developers do develop magenta artifacts when trying to recover clipped highlight data.

Yes.

 

[mike3996]

Yes.

By the way, the artifacts are magenta because the highlight recovery algorithms work with the raw Bayer pixels, which are 50 % green (complementary of which is magenta).

 

[Brownie]

Example

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

P1026103 by Shotglass Photo, on Flickr

 

[Joris]

Some time ago I noticed on PhotonsToPhotos.net that GX80 might have its best DR at its "pull ISO" setting of 100. Usually DR suffers greatly if you let the camera overexpose by a stop to achieve this pull setting.

But the graph at PhotonsToPhotos begged to differ. I made sample shots just out of curiosity.

Both are f/7.1, one is 1/1000 sec ISO 200, the other is 1/500 sec at ISO 100. Opened in darktable without any manual manipulations, both exposures very neatly touch the right edge of the histogram (Panasonics have always metered the most accurate exposures). Neither has any clipped pixels outside margins of usual variance and error.

What do you think about this? Is the real usable base ISO of this camera actually the pull setting?

I once put that exact same question to Bill Claff after seeing Photons to Photos' graphs for the GX80. His answer was : "Native ISO is ISO 100 but below ISO 160 is affected by 12-bit ADC instead of 14-bit ADC so they list ISO 200 as native."
I do not really know how 'affected by 12-bit ADC instead of 14-bit ADC' translates... Always shoot RAW at 100 ISO with excellent results noise-wise. However, I sometimes have magenta highlights, even when not yet clipped according to the histogram.

 

[mike3996]

Wow. Thank you for sharing this information. Oh, what tricks do the engineers have in their sleeves.

I am going to give the PULL ISO a more frequent spin from now on, and see if I can get differences visible in darktable.

DR is always welcome, I shoot difficult light often. I will have to watch for those magenta highlights and cease if I start seeing them.

 

[Brownie]

I once put that exact same question to Bill Claff after seeing Photons to Photos' graphs for the GX80. His answer was : "Native ISO is ISO 100 but below ISO 160 is affected by 12-bit ADC instead of 14-bit ADC so they list ISO 200 as native."
I do not really know how 'affected by 12-bit ADC instead of 14-bit ADC' translates... Always shoot RAW at 100 ISO with excellent results noise-wise. However, I sometimes have magenta highlights, even when not yet clipped according to the histogram.

Hey Joris!

Agree, highlights don't have to be clipped for the magenta to show up. Interesting to note too that the example I posted above was corrected as much as I could get it at the time. It was really pink SOOC.

 

[Joris]

Interesting to note too that the example I posted above was corrected as much as I could get it at the time

Hi Tim, in my shots it usually shows in clouds in the sky. I am slightly paranoid about this since, colorblind, I risk not seeing what migh be overly obvious to someone else. I mostly exposure bracket for this reason. But if needs be, I push magenta (and purple) at 0 saturation in HSL and at that add a brush at 0 saturation.

 

[agentlossing]

Wow. Thank you for sharing this information. Oh, what tricks do the engineers have in their sleeves.

I am going to give the PULL ISO a more frequent spin from now on, and see if I can get differences visible in darktable.

DR is always welcome, I shoot difficult light often. I will have to watch for those magenta highlights and cease if I start seeing them.

I guess I wasn't thinking about a normal exposure using ISO 100 as being a "pull" ISO as you termed it. When I think about digital RAW files, I think of pushing and pulling as post-work in a RAW converter. Wouldn't highlight recovery in something like LR be a better indicator of highlight clipping/lower DR than straight converted images?

 

[mike3996]

I guess I wasn't thinking about a normal exposure using ISO 100 as being a "pull" ISO as you termed it. When I think about digital RAW files, I think of pushing and pulling as post-work in a RAW converter. Wouldn't highlight recovery in something like LR be a better indicator of highlight clipping/lower DR than straight converted images?

You're correct in that I did provide misleading sample photos while talking about DR in the same post.

But "pull ISO" is called so because without exception it's achieved by overexposing the shot at camera's native ISO and then pulled down to achieve the simulated sensitivity. My shots gave "proof" that GX80 does no such pulling at ISO 100. If it did, my ETTR settings would have clipped (unrecoverably) a lot of snow.

Yet, I admit that my "proof" by providing processed jpeg files was half-hearted at best. I did try to attach original RW2 files but they weren't supported file types by the forum...

 

[agentlossing]

Have you experimented with how much highlight latitude the ISO 100 shots give you? I'd be interested to know if they behave similar to normal ISO range, though I don't have the GX85 anymore. I still have a lot of files, though (not sure if many are ISO 100).

 

[mike3996]

I will be doing all my daylight shooting at 100 from now on, and monitor for ill effects. If the Panasonic engineers say it's using a 12-bit ADC and thus say 200 is safer/overall better, I believe them. But I want to see what sort of stuff I am getting.

 

[mike3996]

It didn't take long for me to find where and how the adverse effects of the pull ISO present themselves reliably. It's whenever I let highlights clip, there'll be very difficult magenta cast present, more difficult to deal with than some of the normal magenta cast sometimes present after highlight recovery.

Panasonic cameras feature undistracting zebra patterns and histograms, so clipping highlights is usually not a mistake but rather a technical decision based on subject matter.


I played with the clipped files more, and at least on darktable there's a way to control the threshold where highlight recovery starts to work. Normally the default threshold of "1.00" is fine but with the pull ISO files you have to lower it down to 0.66 - 0.7 range. Maybe this goes hand in hand with the 12/14 bit ADC processing, I don't know.

I should make more pairs of test shots (taken at ISO 100/200) to see if the files are correctable to the same degree. I don't think they are, otherwise they wouldn't call it a pull ISO.

It's highly scene dependent.

 

[agentlossing]

It didn't take long for me to find where and how the adverse effects of the pull ISO present themselves reliably. It's whenever I let highlights clip, there'll be very difficult magenta cast present, more difficult to deal with than some of the normal magenta cast sometimes present after highlight recovery.

Panasonic cameras feature undistracting zebra patterns and histograms, so clipping highlights is usually not a mistake but rather a technical decision based on subject matter.


I played with the clipped files more, and at least on darktable there's a way to control the threshold where highlight recovery starts to work. Normally the default threshold of "1.00" is fine but with the pull ISO files you have to lower it down to 0.66 - 0.7 range. Maybe this goes hand in hand with the 12/14 bit ADC processing, I don't know.

I should make more pairs of test shots (taken at ISO 100/200) to see if the files are correctable to the same degree. I don't think they are, otherwise they wouldn't call it a pull ISO.

It's highly scene dependent.

The highlight clipping is the one and only thing I rail against with M4/3 sensors, it's what frustrates me more than anything else. I want nothing more than to be able to have a gentle rolloff into white, like you get with color film, and like what's easier to get an approximation of even with the APS-C sensor in the GR III/x. 14-bit RAWs help. But, with the 16mp sensors especially, you have to watch the highlights like a hawk if you don't want to have to fight with ugliness. I have mostly adopted a style of letting the shadows go down into unrecoverable lack of detail where necessary in order to have clean highlights. Or, only raising the shadows a little bit in order to give more of a suggestion of the detail and not unduly raise noise or lay bare the parts with no detail.

 

[mike3996]

The highlight clipping is the one and only thing I rail against with M4/3 sensors, it's what frustrates me more than anything else. I want nothing more than to be able to have a gentle rolloff into white, like you get with color film, and like what's easier to get an approximation of even with the APS-C sensor in the GR III/x. 14-bit RAWs help. But, with the 16mp sensors especially, you have to watch the highlights like a hawk if you don't want to have to fight with ugliness.

Compared to the Leica M, I don't find the highlights behaving much differently with my Panasonics. The Ricoh processing is just particularly good, probably.