Rx100M6 magenta cast in shadows, test shots

[Digital Nigel]

As promised, here's a series of test shots with my RX100M6 and five of my other cameras. In each case, I'm showing an OOC JPEG and a processed RAW file; I'll let you do your own analyses (for David's benefit: every image is guaranteed to contain a bird, albeit a very boring one). All files are uploaded full size.

The camera settings were:

• f/4 (f/6.3 with the APS-c camera)
• 1/60 sec
• 35 equiv focal length
• 0 EV
• Auto ISO if available
• Neutral colours in both the OOC JPEGs and processed RAWs (not my usual, more colourful settings)
• RAW+JPEG shots
• All RAWs processed in DxO PhotoLab 2, with the same settings, including Centre-weighted average exposure

The cameras were:

• RX100m6
• A6500 with 24mm Zeiss f/1.8 lens
• RX100M1
• RX10M3
• Canon G7X
• Panasonic FZ1000

RX100M6:








RX100M6 OOC JPEG




RX100M6, processed from RAW



A6500, with 24mm lens:




A6500 OOC JPEG




A6500, processed from RAW

Original RX100, from 2012




RX100M1 OOC JPEG






RX100M1 processed from RAW



RX10M3:




RX10M3 OOC JPEG






RX10M3 processed from RAW

Canon G7X (same sensor as RX100M2/3):




G7X OOC JPEG






G7X processed from RAW



Panasonic FZ1000 (same sensor as G7X and RX100M2/3):




FZ1000 OOC JPEG






FZ1000 processed from RAW

 

[Horacecoker]

So you're leaving me to do all the hard work. Nigel? I don't mind.

Here are 100% crops from the RX100M6.

I've chosen two spots on each photo. I will do them all in time but first the RX100M6.

The crops are from the extreme left edge of the images so shouldn't really be used for image quality comparisons.

The tree branches against the white sky don't half show up magenta problems. The other crop is of dead leaves which in real life have no magenta tinge whatsoever.


RX100M6 jpeg crop 1




RX100M6 RAW Crop 1




RX100M6 jpeg crop 2




RX100M6 Raw crop2

I spotted the bird in question, a Woodpigeon. There are two in some photos.

David

 

[Bill Borne]

So you're leaving me to do all the hard work. Nigel? I don't mind.

Here are 100% crops from the RX100M6.

I've chosen two spots on each photo. I will do them all in time but first the RX100M6.

The crops are from the extreme left edge of the images so shouldn't really be used for image quality comparisons.

The tree branches against the white sky don't half show up magenta problems. The other crop is of dead leaves which in real life have no magenta tinge whatsoever.


RX100M6 jpeg crop 1


RX100M6 RAW Crop 1


RX100M6 jpeg crop 2


RX100M6 Raw crop2

I spotted the bird in question, a Woodpigeon. There are two in some photos.

David

Exactly what I thought viewing them

--
Bill
"Life's Too Short to Worry about the BS!"
So I Choose my Battles
Click for Wild Man's Photos
Using Rx10 IV at Present

 

[Jerry045]

So you're leaving me to do all the hard work. Nigel? I don't mind.

Here are 100% crops from the RX100M6.

I've chosen two spots on each photo. I will do them all in time but first the RX100M6.

The crops are from the extreme left edge of the images so shouldn't really be used for image quality comparisons.

The tree branches against the white sky don't half show up magenta problems. The other crop is of dead leaves which in real life have no magenta tinge whatsoever.


RX100M6 jpeg crop 1


RX100M6 RAW Crop 1

Could that magenta cast be uncorrected CA? But it is interesting that it is not on all the branches, just some, and not evenly distributed.


RX100M6 jpeg crop 2

The JPG crop looks sharper than the RAW crop. I am not seeing magenta in the leaves either.


RX100M6 Raw crop2

I spotted the bird in question, a Woodpigeon. There are two in some photos.

David

--
Jerry

Jerry Malin

Explore Jerry Malin’s 969 photos on Flickr!

www.flickr.com

 

[Digital Nigel]

So you're leaving me to do all the hard work. Nigel? I don't mind.

Here are 100% crops from the RX100M6.

I've chosen two spots on each photo. I will do them all in time but first the RX100M6.

The crops are from the extreme left edge of the images so shouldn't really be used for image quality comparisons.

The tree branches against the white sky don't half show up magenta problems.

Isn't that purple fringing (chromatic aberration)? Having seen these images, I think I'll have to crank up the correction level for it with this camera.

The other crop is of dead leaves which in real life have no magenta tinge whatsoever.

I think I can happily live with that level.

RX100M6 jpeg crop 1


RX100M6 RAW Crop 1


RX100M6 jpeg crop 2


RX100M6 Raw crop2

I spotted the bird in question, a Woodpigeon. There are two in some photos.

Yes, I only spotted one of them when taking the pictures, but noticed the second when reviewing them later.

 

[sybersitizen]

As promised ...

There's a previous discussion about this?

here's a series of test shots with my RX100M6 and five of my other cameras. In each case, I'm showing an OOC JPEG and a processed RAW file; I'll let you do your own analyses ...

What are we looking for?

There's definitely heavy CA in the RAW (uncorrected?) from the RX100VI.

Beyond that, I see overall color differences among all the cameras, and overall differences between their in-camera JPEGs and their converted RAWs. Nothing surprising. All RAWs were processed in PhotoLab?

 

[Digital Nigel]

As promised ...

There's a previous discussion about this?

Yes, it came up in, and became a side discussion in this thread:
https://www.dpreview.com/forums/thread/4346983

here's a series of test shots with my RX100M6 and five of my other cameras. In each case, I'm showing an OOC JPEG and a processed RAW file; I'll let you do your own analyses ...

What are we looking for?

As the title says, magenta casts in shadows in RX100M6 processed RAW images (but not OOC JPEGs), but not other cameras.

There's definitely heavy CA in the uncorrected RAW from the RX100VI.

Yes, I deliberately used the same settings for all these images. In future, I'll make sure that RX100M6 images get enough CA correction.

Beyond that, I see overall color differences among all the cameras, and overall differences between their in-camera JPEGs and their converted RAWs. Nothing surprising.

The theory, to be proved or disproved, is that the RX100M6's processed RAWs have a unique magenta cast problem.

 

[sybersitizen]

There's a previous discussion about this?

Yes, it came up in, and became a side discussion in this thread:
https://www.dpreview.com/forums/thread/4346983

Okay, I focused on this: 'The magenta cast is only in the RAW's and is noticeable in shadows, especially if you lift them in post. There is no magenta cast in the SOOC JPEG's, as the inbody corrections fix it.'

But I didn't see if that was the case with more than one RAW converter.

The theory, to be proved or disproved, is that the RX100M6's processed RAWs have a unique magenta cast problem.

How about shooting a featureless white(ish) wall, underexposed by a few stops and pushed in PP? That type of test has identified some other issues in Sony cameras.

 

[Digital Nigel]

There's a previous discussion about this?

Yes, it came up in, and became a side discussion in this thread:
https://www.dpreview.com/forums/thread/4346983

Okay, I focused on this: 'The magenta cast is only in the RAW's and is noticeable in shadows, especially if you lift them in post. There is no magenta cast in the SOOC JPEG's, as the inbody corrections fix it.'

The theory, to be proved or disproved, is that the RX100M6's processed RAWs have a unique magenta cast problem.

How about shooting a featureless white(ish) wall, underexposed by a few stops and pushed in PP? That type of test has identified some other issues in Sony cameras.

Yes, that might work, although the theory was that they're visible in any image with shadows. I'd certainly never noticed them before, and I don't see any that would worry me in this test. I might look out for a white or grey (but no red/brown areas) building, with both bright and shadowed areas

The unexpected heavy CA is much more worrying, and I'm going to be looking more at that.

The other surprise (though it shouldn't be) is just how much better the A6500 (with high quality prime lens) image is. There's just so much more detail visible.

 

[adk38]

The other surprise (though it shouldn't be) is just how much better the A6500 (with high quality prime lens) image is. There's just so much more detail visible.

;-)

 

[DenImage]

Thanks for posting your test images Nigel.

The magenta is definitely noticeable (if you go looking for it), but not prevalent in all images taken with the RX100vi.

I was wondering if the Xrite ColorChecker Passport would fix the magenta cast?

I presume it would, but has anyone got one to try?

I might go get one anyway.

Den

 

[Horacecoker]

So you're leaving me to do all the hard work. Nigel? I don't mind.

Here are 100% crops from the RX100M6.

I've chosen two spots on each photo. I will do them all in time but first the RX100M6.

The crops are from the extreme left edge of the images so shouldn't really be used for image quality comparisons.

The tree branches against the white sky don't half show up magenta problems.

Isn't that purple fringing (chromatic aberration)? Having seen these images, I think I'll have to crank up the correction level for it with this camera.

Yes, I do think it's largely to do with purple fringing, that's why I didn't use the word 'cast' in the above sentence. I do think it must be related to the magenta cast problem on the RAW files though.

The other crop is of dead leaves which in real life have no magenta tinge whatsoever.

I think I can happily live with that level.

Yes, it's not too obvious in this instance and I could also live with it. But Den has repeatedly informed us that the magenta cast is much more intrusive when lifting shadows. This has clearly been shown with your image samples in the other thread. It's telling that your posted images in this thread have practically no shadows to lift!! So they are far from being a good example for the magenta cast problem.

RX100M6 jpeg crop 1


RX100M6 RAW Crop 1


RX100M6 jpeg crop 2


RX100M6 Raw crop2

David

 

[Horacecoker]

The other surprise (though it shouldn't be) is just how much better the A6500 (with high quality prime lens) image is. There's just so much more detail visible.

I see the extra detail you speak of when viewing the images at 100% but you forgot to mention that the A6500 has a 24mp sensor as opposed to the 20mp sensor of the RX100 & RX10 cameras. I can tell you that the image detail on my A7Riii with its 42mp sensor and either of my prime lenses compared to my RX10iv is unbelievably massive. But when I downsize both images to my usual 8mp for viewing on a 4k TV or monitor I cannot for the life of me tell which image came from which camera when I deliberately do a test shot to make sure they both have the same DOF.

With that in mind here are two identical 100% crops of your posted images taken with RX10iii and A6500 having first downsized the A6500 image to 20mp.


RX10M3 100% crop


A6500 100% crop after downsizing to match the pixel count of the RX10M3

I still think the A6500 has the edge in quality but now don't see much in the way of the extra detail you speak of.

Another thing to take into consideration when talking about image quality and detail is that 400 ISO will never show a 1" sensor at its best.

David

 

[Digital Nigel]

The other surprise (though it shouldn't be) is just how much better the A6500 (with high quality prime lens) image is. There's just so much more detail visible.

I see the extra detail you speak of when viewing the images at 100% but you forgot to mention that the A6500 has a 24mp sensor as opposed to the 20mp sensor of the RX100 & RX10 cameras. I can tell you that the image detail on my A7Riii with its 42mp sensor and either of my prime lenses compared to my RX10iv is unbelievably massive. But when I downsize both images to my usual 8mp for viewing on a 4k TV or monitor I cannot for the life of me tell which image came from which camera when I deliberately do a test shot to make sure they both have the same DOF.

With that in mind here are two identical 100% crops of your posted images taken with RX10iii and A6500 having first downsized the A6500 image to 20mp.


RX10M3 100% crop


A6500 100% crop after downsizing to match the pixel count of the RX10M3

I still think the A6500 has the edge in quality but now don't see much in the way of the extra detail you speak of.

Another thing to take into consideration when talking about image quality and detail is that 400 ISO will never show a 1" sensor at its best.

I agree that they should be compared at the same res, so I generated JPEGs from RAW all with the same 5100x3400 dimensions, with CA correction increased. Here's a crop from each of the RX100M6, RX10M3 and A6500/24. It's from near the top edge, but not the corner:




RX100M6 crop






RX10M3 crop




A6500 with 24mm lens crop



I'd say the RX10 is very slightly better than the RX100 (as expected), but the A6500 is hugely better than both. Many smaller twigs and branches are visible in the latter, but in neither of the RX images.

I used the same generic colour rendering with all three, but while the two RX images have similar colour renderings, the A6500 image is different.

 

[Digital Nigel]

So you're leaving me to do all the hard work. Nigel? I don't mind.

Here are 100% crops from the RX100M6.

I've chosen two spots on each photo. I will do them all in time but first the RX100M6.

The crops are from the extreme left edge of the images so shouldn't really be used for image quality comparisons.

The tree branches against the white sky don't half show up magenta problems.

Isn't that purple fringing (chromatic aberration)? Having seen these images, I think I'll have to crank up the correction level for it with this camera.

Yes, I do think it's largely to do with purple fringing, that's why I didn't use the word 'cast' in the above sentence. I do think it must be related to the magenta cast problem on the RAW files though.

Yes, I agree, but with a twist: do you think the purple CA could be the cause of the magenta cast? In other words, the purple shows up more in dark shadows than brightly lit areas? Or, at least, the fact that it's caused by CA is hidden? It could be that the in-camera JPEGs have more effective CA correction than typical processed RAWs, hence the difference.

I've re-run the previous shot that you edited, with CA correction increased, and shadows slightly lifted:




Increased CA correction

Even if there is still a magenta cast in the shadows, I don't find it objectionable. I certainly wouldn't spot it if it wasn't pointed out.

 

[Horacecoker]

The other surprise (though it shouldn't be) is just how much better the A6500 (with high quality prime lens) image is. There's just so much more detail visible.

I see the extra detail you speak of when viewing the images at 100% but you forgot to mention that the A6500 has a 24mp sensor as opposed to the 20mp sensor of the RX100 & RX10 cameras. I can tell you that the image detail on my A7Riii with its 42mp sensor and either of my prime lenses compared to my RX10iv is unbelievably massive. But when I downsize both images to my usual 8mp for viewing on a 4k TV or monitor I cannot for the life of me tell which image came from which camera when I deliberately do a test shot to make sure they both have the same DOF.

With that in mind here are two identical 100% crops of your posted images taken with RX10iii and A6500 having first downsized the A6500 image to 20mp.


RX10M3 100% crop


A6500 100% crop after downsizing to match the pixel count of the RX10M3

I still think the A6500 has the edge in quality but now don't see much in the way of the extra detail you speak of.

Another thing to take into consideration when talking about image quality and detail is that 400 ISO will never show a 1" sensor at its best.

I agree that they should be compared at the same res, so I generated JPEGs from RAW all with the same 5100x3400 dimensions, with CA correction increased. Here's a crop from each of the RX100M6, RX10M3 and A6500/24. It's from near the top edge, but not the corner:


RX100M6 crop


RX10M3 crop


A6500 with 24mm lens crop

I'd say the RX10 is very slightly better than the RX100 (as expected), but the A6500 is hugely better than both. Many smaller twigs and branches are visible in the latter, but in neither of the RX images.

I think what you are showing here is the much better dynamic range available with the bigger sensor. The obliterated twigs on the RX100 & 10 are more likely due unrecoverable overexposed highlights than a lack of resolving power of the lenses or sensor size differences. Do a similar crop anywhere else on the main body of the picture away from the sky and I'm pretty sure there will be the exact same number of twigs, leaves etc, on all three cameras

I used the same generic colour rendering with all three, but while the two RX images have similar colour renderings, the A6500 image is different.

Yes, I cannot dispute (nor want to) the overall superiority of the colours and lack of CA in the A6500 shot. I think both the RX100M6 & RX10M3 crops are dreadful. But it's the fault of the RAW files, in both instances the jpegs don't have this excessive CA if any at all! There is something wrong somewhere!

David

 

[Digital Nigel]

The other surprise (though it shouldn't be) is just how much better the A6500 (with high quality prime lens) image is. There's just so much more detail visible.

I see the extra detail you speak of when viewing the images at 100% but you forgot to mention that the A6500 has a 24mp sensor as opposed to the 20mp sensor of the RX100 & RX10 cameras. I can tell you that the image detail on my A7Riii with its 42mp sensor and either of my prime lenses compared to my RX10iv is unbelievably massive. But when I downsize both images to my usual 8mp for viewing on a 4k TV or monitor I cannot for the life of me tell which image came from which camera when I deliberately do a test shot to make sure they both have the same DOF.

With that in mind here are two identical 100% crops of your posted images taken with RX10iii and A6500 having first downsized the A6500 image to 20mp.


RX10M3 100% crop


A6500 100% crop after downsizing to match the pixel count of the RX10M3

I still think the A6500 has the edge in quality but now don't see much in the way of the extra detail you speak of.

Another thing to take into consideration when talking about image quality and detail is that 400 ISO will never show a 1" sensor at its best.

I agree that they should be compared at the same res, so I generated JPEGs from RAW all with the same 5100x3400 dimensions, with CA correction increased. Here's a crop from each of the RX100M6, RX10M3 and A6500/24. It's from near the top edge, but not the corner:


RX100M6 crop


RX10M3 crop


A6500 with 24mm lens crop

I'd say the RX10 is very slightly better than the RX100 (as expected), but the A6500 is hugely better than both. Many smaller twigs and branches are visible in the latter, but in neither of the RX images.

I think what you are showing here is the much better dynamic range available with the bigger sensor. The obliterated twigs on the RX100 & 10 are more likely due unrecoverable overexposed highlights than a lack of resolving power of the lenses or sensor size differences.

It's one reason why I usually shoot -0.7EV with my 1" cameras, to preserve highlights. It does make the images noisier, but that's easier to fix in post-processing than blown highlights. It's not needed with the ARS-C cameras, though I sometimes shoot -0.3EV with them.

Do a similar crop anywhere else on the main body of the picture away from the sky and I'm pretty sure there will be the exact same number of twigs, leaves etc, on all three cameras

Yes, though that may also be partly that the image quality is better towards the centre.

I used the same generic colour rendering with all three, but while the two RX images have similar colour renderings, the A6500 image is different.

Yes, I cannot dispute (nor want to) the overall superiority of the colours and lack of CA in the A6500 shot. I think both the RX100M6 & RX10M3 crops are dreadful. But it's the fault of the RAW files, in both instances the jpegs don't have this excessive CA if any at all! There is something wrong somewhere!

Yes, Sony seems to be correcting the CA very aggressively in-camera, more than RAW processors are geared-up for.

 

[Horacecoker]

So you're leaving me to do all the hard work. Nigel? I don't mind.

Here are 100% crops from the RX100M6.

I've chosen two spots on each photo. I will do them all in time but first the RX100M6.

The crops are from the extreme left edge of the images so shouldn't really be used for image quality comparisons.

The tree branches against the white sky don't half show up magenta problems.

Isn't that purple fringing (chromatic aberration)? Having seen these images, I think I'll have to crank up the correction level for it with this camera.

Yes, I do think it's largely to do with purple fringing, that's why I didn't use the word 'cast' in the above sentence. I do think it must be related to the magenta cast problem on the RAW files though.

Yes, I agree, but with a twist: do you think the purple CA could be the cause of the magenta cast? In other words, the purple shows up more in dark shadows than brightly lit areas? Or, at least, the fact that it's caused by CA is hidden? It could be that the in-camera JPEGs have more effective CA correction than typical processed RAWs, hence the difference.

I now think the magenta cast and excessive purple fringing in the processed RAW images are not linked in any way. See below.

I've re-run the previous shot that you edited, with CA correction increased, and shadows slightly lifted:


Increased CA correction

Even if there is still a magenta cast in the shadows, I don't find it objectionable. I certainly wouldn't spot it if it wasn't pointed out.

I know the magenta cast does not worry you and you don't even notice it until it's pointed out. But it does exist and it obviously bothers Den to the point where he was thinking of returning the camera.

There wasn't much point posting the image above in isolation with CA correction increased to see if it decreases or eliminates the magenta cast issue. There is nothing to compare it with!

I've done an identical crop on the above image as the ones I did in the other thread. Here are all three.


OOC jpeg with shadows lifted slightly


processed RAW with shadows lifted slightly


processed RAW with CA correction increased and shadow lightening increased

Now we can compare your increased CA correction image we can clearly see it hasn't made the slightest bit of difference to the magenta cast. So no, I now don't think there is any link with the excessive CA and magenta cast in RAW files. Instead I think there are two separate issues with the RAW files, not just one.

The original images you posted in this thread highlighted the CA problem extremely well but failed completely to demonstrate the magenta cast problem in the shadows, simply because there are no shadows in these images. That is another reason I've posted these crops in this thread - so folk can actually see what we are talking about!

David

 

[Digital Nigel]

The other surprise (though it shouldn't be) is just how much better the A6500 (with high quality prime lens) image is. There's just so much more detail visible.

I see the extra detail you speak of when viewing the images at 100% but you forgot to mention that the A6500 has a 24mp sensor as opposed to the 20mp sensor of the RX100 & RX10 cameras. I can tell you that the image detail on my A7Riii with its 42mp sensor and either of my prime lenses compared to my RX10iv is unbelievably massive. But when I downsize both images to my usual 8mp for viewing on a 4k TV or monitor I cannot for the life of me tell which image came from which camera when I deliberately do a test shot to make sure they both have the same DOF.

With that in mind here are two identical 100% crops of your posted images taken with RX10iii and A6500 having first downsized the A6500 image to 20mp.


RX10M3 100% crop


A6500 100% crop after downsizing to match the pixel count of the RX10M3

I still think the A6500 has the edge in quality but now don't see much in the way of the extra detail you speak of.

Another thing to take into consideration when talking about image quality and detail is that 400 ISO will never show a 1" sensor at its best.

I agree that they should be compared at the same res, so I generated JPEGs from RAW all with the same 5100x3400 dimensions, with CA correction increased. Here's a crop from each of the RX100M6, RX10M3 and A6500/24. It's from near the top edge, but not the corner:


RX100M6 crop


RX10M3 crop


A6500 with 24mm lens crop

I'd say the RX10 is very slightly better than the RX100 (as expected), but the A6500 is hugely better than both. Many smaller twigs and branches are visible in the latter, but in neither of the RX images.

I think what you are showing here is the much better dynamic range available with the bigger sensor. The obliterated twigs on the RX100 & 10 are more likely due unrecoverable overexposed highlights than a lack of resolving power of the lenses or sensor size differences.

It's one reason why I usually shoot -0.7EV with my 1" cameras, to preserve highlights. It does make the images noisier, but that's easier to fix in post-processing than blown highlights. It's not needed with the ARS-C cameras, though I sometimes shoot -0.3EV with them.

Do a similar crop anywhere else on the main body of the picture away from the sky and I'm pretty sure there will be the exact same number of twigs, leaves etc, on all three cameras

Yes, though that may also be partly that the image quality is better towards the centre.

I used the same generic colour rendering with all three, but while the two RX images have similar colour renderings, the A6500 image is different.

Yes, I cannot dispute (nor want to) the overall superiority of the colours and lack of CA in the A6500 shot. I think both the RX100M6 & RX10M3 crops are dreadful. But it's the fault of the RAW files, in both instances the jpegs don't have this excessive CA if any at all! There is something wrong somewhere!

Yes, Sony seems to be correcting the CA very aggressively in-camera, more than RAW processors are geared-up for.

I've done some more tests, and now disagree with myself above! With some more CA correction adjustments in PhotoLab (so these are more processed than the previous examples), here's a couple of examples from RAW+JPEG shots:

First a pair at 0EV:


Cropped OOC JPEG






Cropped processed RAW

Now a pair at -0.7EV:




Cropped OOC JPEG




Cropped processed RAW

So there's less detail, worse blown highlights and more CA in the OOC JPEGs than the processed RAWs.

 

[DenImage]

Just to follow up.

I have tested X-rite Color Passport (software only), and it doesn't correct the magenta Color cast.

Den