R5II high ISO performance compared with R5

[Alastair Norcross]

I was surprised to read in another thread someone saying that the R5II has visibly more noise at high ISO than the R5. I have the R5II, but not the R5, so I can't do my own comparison shots of the things that I normally shoot. So I had to look at the DPR test shots, which I generally don't find that useful, because I don't usually shoot test charts, and extrapolation from test chart shots to the kind of subjects I shoot is very unreliable. However, I downloaded the RAWs for the R5 and R5II at 25,600. The R5II had RAWs in both EFCS and e-shutter, so that's a total of 3 RAWs. I first processed them in the way that I would never do, and I hope no-one else would either, which is using ACR without any noise reduction. But that seems to be the way that DPR displays things on its comparometer, so maybe that's what's giving people the idea that the R5II has higher visible noise at high ISO (it could also be that some people are going off measured results, without looking at images, even though that doesn't tell you about visible differences). Here are 100% crops of a large portion of the test chart for all three files:




14




13




15

These are all pretty noisy viewed at 100%, as you'd expect from ISO 25,600 shots with no noise reduction. I have a hard time having any preference between them, other than wanting to process them properly, which is what did (or tried to do) next. Here they are again using DXO Photolab 8 running Deep Prime XDs2:




3




10




8

Again, I have a hard time having a preference between them, but they are all clearly a lot better than the images with no NR, and have a lot more detail preserved (and no, it isn't false detail). Here are the ISO 100 images from both cameras:




6




1

There's clearly more detail in these images, as you'd expect for ISO 100 compared with 25,600. But that's mostly noticeable when viewing at 100% fairly close up, which is like peering at a massive poster print from a foot or so away.

Finally, I thought I'd see how the two cameras looked with a 3 and 4 stop underexposure, recovered in post. In over 20 years of digital processing, I don't think I've ever had to push a file more than 3 stops, and only very rarely as much as 2 stops, but it's interesting to see what happens if you do. First, the two files (one for each camera) at ISO 800, underexposed by 3 stops and recovered in post:




5





And lastly the two cameras at ISO 400, underexposed by 4 stops and recovered in post:




11




4

I haven't had a chance to see how these look on the DPR site (I'll do that when I post this), but I really don't have any preference between them, when I view them on my computer using Preview. This whole exercise has just confirmed to me my earlier opinion that any differences between the two sensors are actually imperceptible in terms of the final results. Modern noise reduction means we don't have to choose between noise and detail anymore for high ISO shots. It's the failure to use good noise reduction that is destructive to detail. Of course, you'll always get even better detail if you can shoot at lower ISO settings, but you won't get any advantage from choosing the R5 over the R5II in that respect (except for a much lower price, of course).

--
“When I die, I want to go peacefully in my sleep like my grandfather. Not screaming in terror, like the passengers in his car.” Jack Handey
Alastair

anorcross

Keen hobbyist since teaching myself film photography as a teenager in the 70s, using manual SLRs and black and white film. Love photographing people and animals.

anorcross.smugmug.com

Equipment in profile

 

[Ephemeris]

I was surprised to read in another thread someone saying that the R5II has visibly more noise at high ISO than the R5. I have the R5II, but not the R5, so I can't do my own comparison shots of the things that I normally shoot. So I had to look at the DPR test shots, which I generally don't find that useful, because I don't usually shoot test charts, and extrapolation from test chart shots to the kind of subjects I shoot is very unreliable. However, I downloaded the RAWs for the R5 and R5II at 25,600. The R5II had RAWs in both EFCS and e-shutter, so that's a total of 3 RAWs. I first processed them in the way that I would never do, and I hope no-one else would either, which is using ACR without any noise reduction. But that seems to be the way that DPR displays things on its comparometer, so maybe that's what's giving people the idea that the R5II has higher visible noise at high ISO (it could also be that some people are going off measured results, without looking at images, even though that doesn't tell you about visible differences). Here are 100% crops of a large portion of the test chart for all three files:


14


13


15

These are all pretty noisy viewed at 100%, as you'd expect from ISO 25,600 shots with no noise reduction. I have a hard time having any preference between them, other than wanting to process them properly, which is what did (or tried to do) next. Here they are again using DXO Photolab 8 running Deep Prime XDs2:


3


10


8

Again, I have a hard time having a preference between them, but they are all clearly a lot better than the images with no NR, and have a lot more detail preserved (and no, it isn't false detail). Here are the ISO 100 images from both cameras:


6


1

There's clearly more detail in these images, as you'd expect for ISO 100 compared with 25,600. But that's mostly noticeable when viewing at 100% fairly close up, which is like peering at a massive poster print from a foot or so away.

Finally, I thought I'd see how the two cameras looked with a 3 and 4 stop underexposure, recovered in post. In over 20 years of digital processing, I don't think I've ever had to push a file more than 3 stops, and only very rarely as much as 2 stops, but it's interesting to see what happens if you do. First, the two files (one for each camera) at ISO 800, underexposed by 3 stops and recovered in post:


5



And lastly the two cameras at ISO 400, underexposed by 4 stops and recovered in post:


11


4

I haven't had a chance to see how these look on the DPR site (I'll do that when I post this), but I really don't have any preference between them, when I view them on my computer using Preview. This whole exercise has just confirmed to me my earlier opinion that any differences between the two sensors are actually imperceptible in terms of the final results. Modern noise reduction means we don't have to choose between noise and detail anymore for high ISO shots. It's the failure to use good noise reduction that is destructive to detail. Of course, you'll always get even better detail if you can shoot at lower ISO settings, but you won't get any advantage from choosing the R5 over the R5II in that respect (except for a much lower price, of course).

--
“When I die, I want to go peacefully in my sleep like my grandfather. Not screaming in terror, like the passengers in his car.” Jack Handey
Alastair
http://anorcross.smugmug.com
Equipment in profile

There is an assumption I think in all the above im that both cameras received the same amount of light. This may not be the case and thus the shot noise component is harder to quantify.

Using NR would not be a way to compare the camera system. Similar to turning on lens corrections doesn't help us understand a lens performance.

It is possible to understand the NR tool performance but that's something separate.

The comment about detail, noise, modern tools doesn't make very much sense to me. Most of my real work requires detail and less worried about colour depth, accuracy or noise. I can't use NR for this work to be analysed but we do find a validated NR flow helps the humans look at images (although we loose detail as ISO increases and add NR to that and things continue to reduce).

 

[m100]

I was surprised to read in another thread someone saying that the R5II has visibly more noise at high ISO than the R5. I have the R5II, but not the R5, so I can't do my own comparison shots of the things that I normally shoot. So I had to look at the DPR test shots, which I generally don't find that useful, because I don't usually shoot test charts, and extrapolation from test chart shots to the kind of subjects I shoot is very unreliable. However, I downloaded the RAWs for the R5 and R5II at 25,600. The R5II had RAWs in both EFCS and e-shutter, so that's a total of 3 RAWs. I first processed them in the way that I would never do, and I hope no-one else would either, which is using ACR without any noise reduction. But that seems to be the way that DPR displays things on its comparometer, so maybe that's what's giving people the idea that the R5II has higher visible noise at high ISO (it could also be that some people are going off measured results, without looking at images, even though that doesn't tell you about visible differences). Here are 100% crops of a large portion of the test chart for all three files:


14


13


15

These are all pretty noisy viewed at 100%, as you'd expect from ISO 25,600 shots with no noise reduction. I have a hard time having any preference between them, other than wanting to process them properly, which is what did (or tried to do) next. Here they are again using DXO Photolab 8 running Deep Prime XDs2:


3


10


8

Again, I have a hard time having a preference between them, but they are all clearly a lot better than the images with no NR, and have a lot more detail preserved (and no, it isn't false detail). Here are the ISO 100 images from both cameras:


6


1

There's clearly more detail in these images, as you'd expect for ISO 100 compared with 25,600. But that's mostly noticeable when viewing at 100% fairly close up, which is like peering at a massive poster print from a foot or so away.

Finally, I thought I'd see how the two cameras looked with a 3 and 4 stop underexposure, recovered in post. In over 20 years of digital processing, I don't think I've ever had to push a file more than 3 stops, and only very rarely as much as 2 stops, but it's interesting to see what happens if you do. First, the two files (one for each camera) at ISO 800, underexposed by 3 stops and recovered in post:


5



And lastly the two cameras at ISO 400, underexposed by 4 stops and recovered in post:


11


4

I haven't had a chance to see how these look on the DPR site (I'll do that when I post this), but I really don't have any preference between them, when I view them on my computer using Preview. This whole exercise has just confirmed to me my earlier opinion that any differences between the two sensors are actually imperceptible in terms of the final results. Modern noise reduction means we don't have to choose between noise and detail anymore for high ISO shots. It's the failure to use good noise reduction that is destructive to detail. Of course, you'll always get even better detail if you can shoot at lower ISO settings, but you won't get any advantage from choosing the R5 over the R5II in that respect (except for a much lower price, of course).

There is an assumption I think in all the above im that both cameras received the same amount of light. This may not be the case and thus the shot noise component is harder to quantify.

Using NR would not be a way to compare the camera system. Similar to turning on lens corrections doesn't help us understand a lens performance.

It is possible to understand the NR tool performance but that's something separate.

The comment about detail, noise, modern tools doesn't make very much sense to me. Most of my real work requires detail and less worried about colour depth, accuracy or noise. I can't use NR for this work to be analysed but we do find a validated NR flow helps the humans look at images (although we loose detail as ISO increases and add NR to that and things continue to reduce).

This is what I know about modern NR or PL.

PL likes a lot of pixels.

PL cannot see a photo from 6 feet away.

All PL can see is 1s and 0s.

PL is very clever at moving pixels or 1s and 0s around.

If you want to play with the modern stuff it is PL8. Not just NR for sure.

--

 

[Ephemeris]

I was surprised to read in another thread someone saying that the R5II has visibly more noise at high ISO than the R5. I have the R5II, but not the R5, so I can't do my own comparison shots of the things that I normally shoot. So I had to look at the DPR test shots, which I generally don't find that useful, because I don't usually shoot test charts, and extrapolation from test chart shots to the kind of subjects I shoot is very unreliable. However, I downloaded the RAWs for the R5 and R5II at 25,600. The R5II had RAWs in both EFCS and e-shutter, so that's a total of 3 RAWs. I first processed them in the way that I would never do, and I hope no-one else would either, which is using ACR without any noise reduction. But that seems to be the way that DPR displays things on its comparometer, so maybe that's what's giving people the idea that the R5II has higher visible noise at high ISO (it could also be that some people are going off measured results, without looking at images, even though that doesn't tell you about visible differences). Here are 100% crops of a large portion of the test chart for all three files:


14


13


15

These are all pretty noisy viewed at 100%, as you'd expect from ISO 25,600 shots with no noise reduction. I have a hard time having any preference between them, other than wanting to process them properly, which is what did (or tried to do) next. Here they are again using DXO Photolab 8 running Deep Prime XDs2:


3


10


8

Again, I have a hard time having a preference between them, but they are all clearly a lot better than the images with no NR, and have a lot more detail preserved (and no, it isn't false detail). Here are the ISO 100 images from both cameras:


6


1

There's clearly more detail in these images, as you'd expect for ISO 100 compared with 25,600. But that's mostly noticeable when viewing at 100% fairly close up, which is like peering at a massive poster print from a foot or so away.

Finally, I thought I'd see how the two cameras looked with a 3 and 4 stop underexposure, recovered in post. In over 20 years of digital processing, I don't think I've ever had to push a file more than 3 stops, and only very rarely as much as 2 stops, but it's interesting to see what happens if you do. First, the two files (one for each camera) at ISO 800, underexposed by 3 stops and recovered in post:


5



And lastly the two cameras at ISO 400, underexposed by 4 stops and recovered in post:


11


4

I haven't had a chance to see how these look on the DPR site (I'll do that when I post this), but I really don't have any preference between them, when I view them on my computer using Preview. This whole exercise has just confirmed to me my earlier opinion that any differences between the two sensors are actually imperceptible in terms of the final results. Modern noise reduction means we don't have to choose between noise and detail anymore for high ISO shots. It's the failure to use good noise reduction that is destructive to detail. Of course, you'll always get even better detail if you can shoot at lower ISO settings, but you won't get any advantage from choosing the R5 over the R5II in that respect (except for a much lower price, of course).

There is an assumption I think in all the above im that both cameras received the same amount of light. This may not be the case and thus the shot noise component is harder to quantify.

Using NR would not be a way to compare the camera system. Similar to turning on lens corrections doesn't help us understand a lens performance.

It is possible to understand the NR tool performance but that's something separate.

The comment about detail, noise, modern tools doesn't make very much sense to me. Most of my real work requires detail and less worried about colour depth, accuracy or noise. I can't use NR for this work to be analysed but we do find a validated NR flow helps the humans look at images (although we loose detail as ISO increases and add NR to that and things continue to reduce).

This is what I know about modern NR or PL.

PL likes a lot of pixels.

PL cannot see a photo from 6 feet away.

All PL can see is 1s and 0s.

PL is very clever at moving pixels or 1s and 0s around.

If you want to play with the modern stuff it is PL8. Not just NR for sure.

How does that relate to understanding a cameras system performance?

--

 

[Quarkcharmed]

I was surprised to read in another thread someone saying that the R5II has visibly more noise at high ISO than the R5. I have the R5II, but not the R5, so I can't do my own comparison shots of the things that I normally shoot. So I had to look at the DPR test shots, which I generally don't find that useful, because I don't usually shoot test charts, and extrapolation from test chart shots to the kind of subjects I shoot is very unreliable. However, I downloaded the RAWs for the R5 and R5II at 25,600. The R5II had RAWs in both EFCS and e-shutter, so that's a total of 3 RAWs. I first processed them in the way that I would never do, and I hope no-one else would either, which is using ACR without any noise reduction. But that seems to be the way that DPR displays things on its comparometer, so maybe that's what's giving people the idea that the R5II has higher visible noise at high ISO (it could also be that some people are going off measured results, without looking at images, even though that doesn't tell you about visible differences). Here are 100% crops of a large portion of the test chart for all three files:

Out of these 3, #3 is the worst, #2 is the second. The #2 has the least noise. Apart from the noise, look at the text on the thread rolls and whether you can tell separate threads. apart. I think #3 is shot with ES, however the difference between ES and EFCS in the R5II is not as big as in the R5.

14


13


15

These are all pretty noisy viewed at 100%, as you'd expect from ISO 25,600 shots with no noise reduction. I have a hard time having any preference between them, other than wanting to process them properly, which is what did (or tried to do) next. Here they are again using DXO Photolab 8 running Deep Prime XDs2:


3


10


8

Again, I have a hard time having a preference between them, but they are all clearly a lot better than the images with no NR, and have a lot more detail preserved (and no, it isn't false detail).

Of course it is fake(false) detail. You still don't understand: it's physically impossible otherwise. After the NR, there's fake detail in either R5 and R5II images, but R5II has a little bit more of it.

Look at fine text (esp. vertical) and compare R5 to R5II.

you can shoot at lower ISO settings, but you won't get any advantage from choosing the R5 over the R5II in that respect (except for a much lower price, of course).

Yes you will - albeit the advantage is not big.

However there's another factor: IIRC DPR used different lenses for this test which may introduce a bias one way or another.

--

Login • Instagram

Welcome back to Instagram. Sign in to check out what your friends, family & interests have been capturing & sharing around the world.

www.instagram.com

500px

500px.com

 

[Alastair Norcross]

I was surprised to read in another thread someone saying that the R5II has visibly more noise at high ISO than the R5. I have the R5II, but not the R5, so I can't do my own comparison shots of the things that I normally shoot. So I had to look at the DPR test shots, which I generally don't find that useful, because I don't usually shoot test charts, and extrapolation from test chart shots to the kind of subjects I shoot is very unreliable. However, I downloaded the RAWs for the R5 and R5II at 25,600. The R5II had RAWs in both EFCS and e-shutter, so that's a total of 3 RAWs. I first processed them in the way that I would never do, and I hope no-one else would either, which is using ACR without any noise reduction. But that seems to be the way that DPR displays things on its comparometer, so maybe that's what's giving people the idea that the R5II has higher visible noise at high ISO (it could also be that some people are going off measured results, without looking at images, even though that doesn't tell you about visible differences). Here are 100% crops of a large portion of the test chart for all three files:


14


13


15

These are all pretty noisy viewed at 100%, as you'd expect from ISO 25,600 shots with no noise reduction. I have a hard time having any preference between them, other than wanting to process them properly, which is what did (or tried to do) next. Here they are again using DXO Photolab 8 running Deep Prime XDs2:


3


10


8

Again, I have a hard time having a preference between them, but they are all clearly a lot better than the images with no NR, and have a lot more detail preserved (and no, it isn't false detail). Here are the ISO 100 images from both cameras:


6


1

There's clearly more detail in these images, as you'd expect for ISO 100 compared with 25,600. But that's mostly noticeable when viewing at 100% fairly close up, which is like peering at a massive poster print from a foot or so away.

Finally, I thought I'd see how the two cameras looked with a 3 and 4 stop underexposure, recovered in post. In over 20 years of digital processing, I don't think I've ever had to push a file more than 3 stops, and only very rarely as much as 2 stops, but it's interesting to see what happens if you do. First, the two files (one for each camera) at ISO 800, underexposed by 3 stops and recovered in post:


5



And lastly the two cameras at ISO 400, underexposed by 4 stops and recovered in post:


11


4

I haven't had a chance to see how these look on the DPR site (I'll do that when I post this), but I really don't have any preference between them, when I view them on my computer using Preview. This whole exercise has just confirmed to me my earlier opinion that any differences between the two sensors are actually imperceptible in terms of the final results. Modern noise reduction means we don't have to choose between noise and detail anymore for high ISO shots. It's the failure to use good noise reduction that is destructive to detail. Of course, you'll always get even better detail if you can shoot at lower ISO settings, but you won't get any advantage from choosing the R5 over the R5II in that respect (except for a much lower price, of course).

There is an assumption I think in all the above im that both cameras received the same amount of light. This may not be the case and thus the shot noise component is harder to quantify.

I'm pretty sure the DPR staff are quite conscientious in their testing in this respect. You can't guarantee, of course, that the lighting is exactly the same, but it's certainly really close. Unless you think the DPR staff are incompetent, that is, but I'm pretty sure you're not saying that.

Using NR would not be a way to compare the camera system.

Of course it would, given that the system itself included processing. And even if you want to exclude that, since you have to use some processing or other, the only relevant way to compare cameras for your own usage is to use the processing that you use. That's why I used this processing for this, because it's the processing I use. As I've said in other threads, there's simply no such thing as a camera's (in this case meaning a sensor's) performance in isolation from processing. Cameras produce images in concert with processing. We use cameras to produce images, so we have to use some processing or other to compare them. Given that, why would we choose to compare using processing that we'll never actually use to produce our own images? That makes no sense whatsoever.

Similar to turning on lens corrections doesn't help us understand a lens performance.

Of course it does. Many modern lenses are designed to work with lens corrections. You can't understand such a lens's performance without using lens corrections. It's the whole system that produces the image. It looks like your thinking is stuck in the film era, which is understandable. But even then, processing played a major part in many photographers' workflows.

It is possible to understand the NR tool performance but that's something separate.

No.

The comment about detail, noise, modern tools doesn't make very much sense to me. Most of my real work requires detail and less worried about colour depth, accuracy or noise. I can't use NR for this work to be analysed but we do find a validated NR flow helps the humans look at images (although we loose detail as ISO increases and add NR to that and things continue to reduce).

It could be that your work is so far from most ordinary photography that what you need is simply different from what others need. I, and I think most other photographers, produce images that are intended to be looked at by humans. It is how those images look to the humans looking at them that is most important. You might just have different priorities, and that's OK. This whole discussion arose from a comment made on another thread about a putative difference between the R5 and R5II for landscape use. I took such use to be taking landscape pictures to be looked at by humans. I said that the measured technical differences between the two sensors would make no difference to such pictures. My further testing has only strengthened this view of mine. It's possible that whatever (slight) differences there may be between the sensors would make some difference for at least some of your uses. But what I've seen is that those sensor differences would make no difference whatsoever to how a landscape image produced by either camera and a competent photographer would look to a viewer.

--
“When I die, I want to go peacefully in my sleep like my grandfather. Not screaming in terror, like the passengers in his car.” Jack Handey
Alastair

anorcross

Keen hobbyist since teaching myself film photography as a teenager in the 70s, using manual SLRs and black and white film. Love photographing people and animals.

anorcross.smugmug.com

Equipment in profile

 

[SrMi]

I was surprised to read in another thread someone saying that the R5II has visibly more noise at high ISO than the R5. I have the R5II, but not the R5, so I can't do my own comparison shots of the things that I normally shoot. So I had to look at the DPR test shots, which I generally don't find that useful, because I don't usually shoot test charts, and extrapolation from test chart shots to the kind of subjects I shoot is very unreliable. However, I downloaded the RAWs for the R5 and R5II at 25,600. The R5II had RAWs in both EFCS and e-shutter, so that's a total of 3 RAWs. I first processed them in the way that I would never do, and I hope no-one else would either, which is using ACR without any noise reduction. But that seems to be the way that DPR displays things on its comparometer, so maybe that's what's giving people the idea that the R5II has higher visible noise at high ISO (it could also be that some people are going off measured results, without looking at images, even though that doesn't tell you about visible differences). Here are 100% crops of a large portion of the test chart for all three files:


14


13


15

These are all pretty noisy viewed at 100%, as you'd expect from ISO 25,600 shots with no noise reduction. I have a hard time having any preference between them, other than wanting to process them properly, which is what did (or tried to do) next. Here they are again using DXO Photolab 8 running Deep Prime XDs2:


3


10


8

Again, I have a hard time having a preference between them, but they are all clearly a lot better than the images with no NR, and have a lot more detail preserved (and no, it isn't false detail). Here are the ISO 100 images from both cameras:


6


1

There's clearly more detail in these images, as you'd expect for ISO 100 compared with 25,600. But that's mostly noticeable when viewing at 100% fairly close up, which is like peering at a massive poster print from a foot or so away.

Finally, I thought I'd see how the two cameras looked with a 3 and 4 stop underexposure, recovered in post. In over 20 years of digital processing, I don't think I've ever had to push a file more than 3 stops, and only very rarely as much as 2 stops, but it's interesting to see what happens if you do. First, the two files (one for each camera) at ISO 800, underexposed by 3 stops and recovered in post:


5



And lastly the two cameras at ISO 400, underexposed by 4 stops and recovered in post:


11


4

I haven't had a chance to see how these look on the DPR site (I'll do that when I post this), but I really don't have any preference between them, when I view them on my computer using Preview. This whole exercise has just confirmed to me my earlier opinion that any differences between the two sensors are actually imperceptible in terms of the final results. Modern noise reduction means we don't have to choose between noise and detail anymore for high ISO shots. It's the failure to use good noise reduction that is destructive to detail. Of course, you'll always get even better detail if you can shoot at lower ISO settings, but you won't get any advantage from choosing the R5 over the R5II in that respect (except for a much lower price, of course).

There is an assumption I think in all the above im that both cameras received the same amount of light. This may not be the case and thus the shot noise component is harder to quantify.

I'm pretty sure the DPR staff are quite conscientious in their testing in this respect.

Not really . I have regularly found cases where the scene light differs, i.e., the exposure differs. Also, DPR had to move the studio, which introduced some changes.

One can estimate the difference in exposure by looking at the raw histogram. Regarding R5 vs R5II, R5II seems to have received about 1/3 stop more exposure.

You can't guarantee, of course, that the lighting is exactly the same, but it's certainly really close. Unless you think the DPR staff are incompetent, that is, but I'm pretty sure you're not saying that.

Using NR would not be a way to compare the camera system.

Of course it would, given that the system itself included processing. And even if you want to exclude that, since you have to use some processing or other, the only relevant way to compare cameras for your own usage is to use the processing that you use. That's why I used this processing for this, because it's the processing I use. As I've said in other threads, there's simply no such thing as a camera's (in this case meaning a sensor's) performance in isolation from processing. Cameras produce images in concert with processing. We use cameras to produce images, so we have to use some processing or other to compare them. Given that, why would we choose to compare using processing that we'll never actually use to produce our own images? That makes no sense whatsoever.

Similar to turning on lens corrections doesn't help us understand a lens performance.

Of course it does. Many modern lenses are designed to work with lens corrections. You can't understand such a lens's performance without using lens corrections. It's the whole system that produces the image. It looks like your thinking is stuck in the film era, which is understandable. But even then, processing played a major part in many photographers' workflows.

It is possible to understand the NR tool performance but that's something separate.

No.

The comment about detail, noise, modern tools doesn't make very much sense to me. Most of my real work requires detail and less worried about colour depth, accuracy or noise. I can't use NR for this work to be analysed but we do find a validated NR flow helps the humans look at images (although we loose detail as ISO increases and add NR to that and things continue to reduce).

It could be that your work is so far from most ordinary photography that what you need is simply different from what others need. I, and I think most other photographers, produce images that are intended to be looked at by humans. It is how those images look to the humans looking at them that is most important. You might just have different priorities, and that's OK. This whole discussion arose from a comment made on another thread about a putative difference between the R5 and R5II for landscape use. I took such use to be taking landscape pictures to be looked at by humans. I said that the measured technical differences between the two sensors would make no difference to such pictures. My further testing has only strengthened this view of mine. It's possible that whatever (slight) differences there may be between the sensors would make some difference for at least some of your uses. But what I've seen is that those sensor differences would make no difference whatsoever to how a landscape image produced by either camera and a competent photographer would look to a viewer.

 

[Ephemeris]

I was surprised to read in another thread someone saying that the R5II has visibly more noise at high ISO than the R5. I have the R5II, but not the R5, so I can't do my own comparison shots of the things that I normally shoot. So I had to look at the DPR test shots, which I generally don't find that useful, because I don't usually shoot test charts, and extrapolation from test chart shots to the kind of subjects I shoot is very unreliable. However, I downloaded the RAWs for the R5 and R5II at 25,600. The R5II had RAWs in both EFCS and e-shutter, so that's a total of 3 RAWs. I first processed them in the way that I would never do, and I hope no-one else would either, which is using ACR without any noise reduction. But that seems to be the way that DPR displays things on its comparometer, so maybe that's what's giving people the idea that the R5II has higher visible noise at high ISO (it could also be that some people are going off measured results, without looking at images, even though that doesn't tell you about visible differences). Here are 100% crops of a large portion of the test chart for all three files:


14


13


15

These are all pretty noisy viewed at 100%, as you'd expect from ISO 25,600 shots with no noise reduction. I have a hard time having any preference between them, other than wanting to process them properly, which is what did (or tried to do) next. Here they are again using DXO Photolab 8 running Deep Prime XDs2:


3


10


8

Again, I have a hard time having a preference between them, but they are all clearly a lot better than the images with no NR, and have a lot more detail preserved (and no, it isn't false detail). Here are the ISO 100 images from both cameras:


6


1

There's clearly more detail in these images, as you'd expect for ISO 100 compared with 25,600. But that's mostly noticeable when viewing at 100% fairly close up, which is like peering at a massive poster print from a foot or so away.

Finally, I thought I'd see how the two cameras looked with a 3 and 4 stop underexposure, recovered in post. In over 20 years of digital processing, I don't think I've ever had to push a file more than 3 stops, and only very rarely as much as 2 stops, but it's interesting to see what happens if you do. First, the two files (one for each camera) at ISO 800, underexposed by 3 stops and recovered in post:


5



And lastly the two cameras at ISO 400, underexposed by 4 stops and recovered in post:


11


4

I haven't had a chance to see how these look on the DPR site (I'll do that when I post this), but I really don't have any preference between them, when I view them on my computer using Preview. This whole exercise has just confirmed to me my earlier opinion that any differences between the two sensors are actually imperceptible in terms of the final results. Modern noise reduction means we don't have to choose between noise and detail anymore for high ISO shots. It's the failure to use good noise reduction that is destructive to detail. Of course, you'll always get even better detail if you can shoot at lower ISO settings, but you won't get any advantage from choosing the R5 over the R5II in that respect (except for a much lower price, of course).

There is an assumption I think in all the above im that both cameras received the same amount of light. This may not be the case and thus the shot noise component is harder to quantify.

I'm pretty sure the DPR staff are quite conscientious in their testing in this respect. You can't guarantee, of course, that the lighting is exactly the same, but it's certainly really close. Unless you think the DPR staff are incompetent, that is, but I'm pretty sure you're not saying that.

Hangon, that's a pretty unpleasant thing to throw out in to the universe from any perspective.

I have never, ever suggested they are incompetent. Moreover, I have never ever discussed my views upon the competence period.

That does not preclude there being differences between the two scenes. Until this is understood well the discussion utilising the scenes may not be helpful.

Over time different lenses have been used, the whole scene was moved from one building to another and revalidated.

If all things are equal then great, but this needs to be demonstrated.

Using NR would not be a way to compare the camera system.

Of course it would, given that the system itself included processing.

No, the system does not have to use processing that adds NR. This can be done using 3rd party tools, or DPP (turn those features off).

The camera is a system, that's what we are examining.

We are not examining post processing tools, monitor calibrations etc.

And even if you want to exclude that, since you have to use some processing or other, the only relevant way to compare cameras for your own usage is to use the processing that you use.

No that's not correct or scientific. If I wish to compare equivocal systems then I placed a boundary around it and be very careful what I am testing (here is where the studio scene affects matters).

If I wish to test post processing systems that's a different task.

If I want to test camera plus post processing, that is yet another task, and I would only ever do this after the first two parallel tasks.

That's why I used this processing for this, because it's the processing I use.

That's great but it's not a camera system test and may not be relevant to others.

As I've said in other threads, there's simply no such thing as a camera's (in this case meaning a sensor's) performance in isolation from processing.

There is. It has a clear digital output pathway (pathways). We can analyse that data without any manipulation.

Cameras produce images in concert with processing.

Cameras produce data. It is that we are interested in . That's the system .

We use cameras to produce images, so we have to use some processing or other to compare them.

That's a different task.

Given that, why would we choose to compare using processing that we'll never actually use to produce our own images?

One doesn't need to process (alter) the data to analyse it. One would do this to get a characterisation of the camera system. If one distorts matters by taking that output data, manipulating in an unknown way then how can one extrapolate the original system performance.

This is the scientific way.

That makes no sense whatsoever.

It's how scientists and engineers have operated for a very long time so it's likely to make some sense.

Similar to turning on lens corrections doesn't help us understand a lens performance.

Of course it does. Many modern lenses are designed to work with lens corrections. You can't understand such a lens's performance without using lens corrections.

I can't understand the lens with it turned on. See recent posts related to RF wide angle zooms.

It's the whole system that produces the image.

Without any corrections I still have an image. That image is very important to me.

It looks like your thinking is stuck in the film era,

I'm stuck in the long standing scientific and engineering process era. Given we design systems for cars I don't think we have decided to use film.

which is understandable. But even then, processing played a major part in many photographers' workflows.

I think perhaps you are mistaken.

It is possible to understand the NR tool performance but that's something separate.

No.

It isn't possible to understand the tool? Why do you think that is the case? Would a manufacturer agree?

The comment about detail, noise, modern tools doesn't make very much sense to me. Most of my real work requires detail and less worried about colour depth, accuracy or noise. I can't use NR for this work to be analysed but we do find a validated NR flow helps the humans look at images (although we loose detail as ISO increases and add NR to that and things continue to reduce).

It could be that your work is so far from most ordinary photography that what you need is simply different from what others need.

That is highly likely, however we do cover a lot of sports invents and dog events. This has no bearing on staying solid and faithful to ways in which science, engineering and systems engineering would describe we should act.

I, and I think most other photographers, produce images that are intended to be looked at by humans.

Possibly, but the human is not what is affecting the system capability and performance.

It is how those images look to the humans looking at them that is most important. You might just have different priorities, and that's OK.

And for those images to look better than last year's models we need to follow a clear understanding and method to validate and characterise the camera system.

Processing is another system

The systems can be combined, but taking this combined approach makes understanding the two input systems very tricky. We are interested in the camera system performance (high iso compared to the R5)

This whole discussion arose from a comment made on another thread about a putative difference between the R5 and R5II for landscape use. I took such use to be taking landscape pictures to be looked at by humans.

The human, at the end of these combined systems will find it hard to evidence the system performance of the camera. Which is what the title is describing.

I said that the measured technical differences between the two sensors would make no difference to such pictures.

Why the sensors? Why not cameras? Sensors are but a sub-system.

My further testing has only strengthened this view of mine. It's possible that whatever (slight) differences there may be between the sensors would make some difference for at least some of your uses.

I'm interested in camera system performance. If that is driven by sensor technology great, but that's not what I'm interacting with.

But what I've seen is that those sensor differences would make no difference whatsoever to how a landscape image produced by either camera and a competent photographer would look to a viewer.

I haven't seen evidence that would say one way or another.

Sadly I'm not at home or I could help the conversation because we own both systems.

Your describing not a sensor, not even a camera, but a viewer at the far right of this chain, looking at an image on some screen. I don't know how one could quantify the system difference between the two bodies this way.

There are differences so that would suggest this isn't a suitable way of testing.

Which reminds me, I really aught to build the rig so the rolling shutter can be measured in all modes.

--
“When I die, I want to go peacefully in my sleep like my grandfather. Not screaming in terror, like the passengers in his car.” Jack Handey
Alastair
http://anorcross.smugmug.com
Equipment in profile

 

[J A C S]

I noticed a similar difference in the RAWs, too, as I said in the other thread. One might argue that RAWs from different sensors should not be compared but the default rendering with DPP look darker as well with the R5II. I would expect Canon to calibrate cameras of the recent generation in the same way.

Also, different lenses.

 

[ThrillaMozilla]

I was surprised to read in another thread someone saying that the R5II has visibly more noise at high ISO than the R5. I have the R5II, but not the R5, so I can't do my own comparison shots of the things that I normally shoot. So I had to look at the DPR test shots, which I generally don't find that useful, because I don't usually shoot test charts, and extrapolation from test chart shots to the kind of subjects I shoot is very unreliable.

It shouldn't be. Subject detail hides noise, and noise hides detail. Test shots are just the thing.

However, I downloaded the RAWs for the R5 and R5II at 25,600. The R5II had RAWs in both EFCS and e-shutter, so that's a total of 3 RAWs. I first processed them in the way that I would never do, and I hope no-one else would either, which is using ACR without any noise reduction. But that seems to be the way that DPR displays things on its comparometer, so maybe that's what's giving people the idea that the R5II has higher visible noise at high ISO (it could also be that some people are going off measured results, without looking at images, even though that doesn't tell you about visible differences).

...or maybe they are looking at the DPR test shots, which definitely show visible differences. Whether the measurements and the test shots are reliable is the question.

Here are 100% crops of a large portion of the test chart for all three files:

These have very little chromaticity noise, and they bear little resemblance to the DPR tests. Your pictures appear to show considerable noise reduction. I doubt that they represent the raw data accurately.

Because of the processing and exposure difference, the pictures don't tell us much in detail about the camera performance except that they are fairly similar in image quality. You do have a good point, and I'm not sure if it's very productive to try to look more carefully for differences.

On the bright side, your reprocessed pictures show very good noise reduction indeed. In that limited test I don't see any spurious detail, and that's good news. Thanks for presenting this.

 

[aftab]

Does R5II produce more noise than R5 at high ISO?

Most people agree that up to ISO 6400, there is virtually no difference between the two cameras. I downloaded ISO 12800 files for both cameras from DPR and opened them in Camera RAW with default parameters. At 100%, I couldn’t tell one from the other. At 200%, slightly more noise was visible in the R5II image in a few parts. For ISO 25600, slightly more noise was visible at 100% in some parts.

When is this excess noise visible?

For ISO 12800, excess noise can be seen at 200%. For ISO 25600, excess noise can be seen at 100%. At smaller sizes (6000x4000, for example), I couldn’t see any difference. In other words, above ISO 12800 pixel-peeping would show slightly increased noise. This excess noise is about the same as the difference between electronic and mechanical shutter noise produced by R5II at ISO 12800.

Now, what are we going to do with these images? Unless you are DPR and have an image comparison page, you will apply NR. So, your assertion about considering editing software in the equation is not without merit. With modern FF cameras noise comparison without considering editing software is mostly academic.

How about shadow lifting?

I have done a few tests with the mechanical shutter. As long as there is some light in an area of the image before it was darkened (to preserve highlight for example), shadow lifting can be done up to 5 stops without significant image penalty (at base ISO). Up to 3 stops is easy. I don’t know how this compares with R5 MKI, DPR didn’t provide any mechanical shutter image for comparison.

My feeling so far is that for all practical purposes, R5 II IQ is similar to R5I.

 

[Ephemeris]

Does R5II produce more noise than R5 at high ISO?

Most people agree that up to ISO 6400, there is virtually no difference between the two cameras. I downloaded ISO 12800 files for both cameras from DPR and opened them in Camera RAW with default parameters. At 100%, I couldn’t tell one from the other. At 200%, slightly more noise was visible in the R5II image in a few parts. For ISO 25600, slightly more noise was visible at 100% in some parts.

When is this excess noise visible?

For ISO 12800, excess noise can be seen at 200%. For ISO 25600, excess noise can be seen at 100%. At smaller sizes (6000x4000, for example), I couldn’t see any difference. In other words, above ISO 12800 pixel-peeping would show slightly increased noise. This excess noise is about the same as the difference between electronic and mechanical shutter noise produced by R5II at ISO 12800.

Now, what are we going to do with these images? Unless you are DPR and have an image comparison page, you will apply NR. So, your assertion about considering editing software in the equation is not without merit. With modern FF cameras noise comparison without considering editing software is mostly academic.

How about shadow lifting?

I have done a few tests with the mechanical shutter. As long as there is some light in an area of the image before it was darkened (to preserve highlight for example), shadow lifting can be done up to 5 stops without significant image penalty (at base ISO). Up to 3 stops is easy. I don’t know how this compares with R5 MKI, DPR didn’t provide any mechanical shutter image for comparison.

My feeling so far is that for all practical purposes, R5 II IQ is similar to R5I.

--
https://www.flickr.com/photos/catch45/
https://www.flickr.com/photos/aftab/

We have an open topic where the two test scene setups may not be equal.

Still to resolve it the same amount of light is present to both camera sensors.

Test scene was moved building and recommissioned between these two cameras test data production and release.

So, utilising these scenes to compare such camera characteristics may (may) not be ideal.

 

[Quarkcharmed]

Does R5II produce more noise than R5 at high ISO?

Most people agree that up to ISO 6400, there is virtually no difference between the two cameras. I downloaded ISO 12800 files for both cameras from DPR and opened them in Camera RAW with default parameters. At 100%, I couldn’t tell one from the other. At 200%, slightly more noise was visible in the R5II image in a few parts. For ISO 25600, slightly more noise was visible at 100% in some parts.

Do you mind posting the images you got and used for comparison?

How about shadow lifting?

I have done a few tests with the mechanical shutter. As long as there is some light in an area of the image before it was darkened (to preserve highlight for example), shadow lifting can be done up to 5 stops without significant image penalty (at base ISO). Up to 3 stops is easy. I don’t know how this compares with R5 MKI,

And in this case as well - do you have the images?

DPR didn’t provide any mechanical shutter image for comparison.

Mechanical and EFC shutters will be exactly the same for the purposes of this test, only you get a shutter stock in mechanical in the certain shutter speed range.

My feeling so far is that for all practical purposes, R5 II IQ is similar to R5I.

Feelings are deceitful

 

[Higgins2002]

Looking at the evolution in sensors.

I wonder why R5II is slightly worse than R5 when R3 is slightly better in all cases than R6II.

 

[ThrillaMozilla]

Looking at the evolution in sensors.

I wonder why R5II is slightly worse than R5 when R3 is slightly better in all cases than R6II.

For a start, you're looking at a more highly magnified view of the R5 and R5 II sensors. You're also looking at different images displayed with different contrast and brightness settings. And then there's the structure of the noise to consider. If you are looking for small differences, it's harder than you think to make the comparison.

Try this.

 

[Higgins2002]

Looking at the evolution in sensors.

I wonder why R5II is slightly worse than R5 when R3 is slightly better in all cases than R6II.

For a start, you're looking at a more highly magnified view of the R5 and R5 II sensors. You're also looking at different images displayed with different contrast and brightness settings. And then there's the structure of the noise to consider. If you are looking for small differences, it's harder than you think to make the comparison.

Try this.

No I was not comparing R5's against R3 and R6II

I was comparing R5II with R5 and R3 with R6II

"I wonder why R5II is slightly worse than R5 when R3 is slightly better in all cases than R6II."

 

[ThrillaMozilla]

No I was not comparing R5's against R3 and R6II

I was comparing R5II with R5 and R3 with R6II

"I wonder why R5II is slightly worse than R5 when R3 is slightly better in all cases than R6II."

Yes, I misread it.

 

[RLight]

DPR never re-did the studio scene for the R5 with the 85mm f/1.2L, makes a difference as you can tell with just the improved contrast in the center of the scene with the 85mm f/1.2L on the R5 II, and, there's perceivable CA on the corners of the studio scene with the original R5 vs on the R5 II, it's virtually nil with the 85mm f/1.2L. The RF 50mm f/1.2L and 85mm f/1.2L are neither slouches at f/5.6 in vignetting, however, there's a 0.20EV difference between the two, in favor of the 85mm f/1.2L according to optical limits in case you're curious.

https://opticallimits.com/canon/canon-rf-85mm-f-1-2-usm-l-review-30mp/

https://opticallimits.com/canon/canon-rf-50mm-f-1-2-usm-l-review/

.

In my observation, the R5 II is "better" than the R5 in terms of SOOC jpegs in low light. This is largely in my opinion, due to smarter metering and AWB, combined with slightly more intelligent sharpening and noise reduction even without Neural Network Noise reduction. Now in RAWs? Very similar results. Now in terms of read noise? They're on par with each other, which makes sense as they're both the same pixel pitch, however the original EOS R5 pulls ahead just slightly. But, if you compare them with a less dense EOS R3? They both get left in the dust in noise...

https://www.photonstophotos.net/Cha...R3_14,Canon EOS R5_14,Canon EOS R5 Mark II_14

.

My 2 cents? Don't buy an R5 II expecting better noise handling than the R5, or keep an original R5 because it has slightly to almost no advantage in low light RAW only vs the R5 II. Now, if you're concerned with noise in low light? Might I draw your attention to that R3, R6 II or R8, or upcoming R1, which all are less dense pixel pitches and full frame. Those have notably lower read noise than either than either the R5 or R5 II, but aren't 45MP, so it's a double edged sword of trading resolution for greater color fidelity in low light.

 

[m100]

I was surprised to read in another thread someone saying that the R5II has visibly more noise at high ISO than the R5. I have the R5II, but not the R5, so I can't do my own comparison shots of the things that I normally shoot. So I had to look at the DPR test shots, which I generally don't find that useful, because I don't usually shoot test charts, and extrapolation from test chart shots to the kind of subjects I shoot is very unreliable. However, I downloaded the RAWs for the R5 and R5II at 25,600. The R5II had RAWs in both EFCS and e-shutter, so that's a total of 3 RAWs. I first processed them in the way that I would never do, and I hope no-one else would either, which is using ACR without any noise reduction. But that seems to be the way that DPR displays things on its comparometer, so maybe that's what's giving people the idea that the R5II has higher visible noise at high ISO (it could also be that some people are going off measured results, without looking at images, even though that doesn't tell you about visible differences). Here are 100% crops of a large portion of the test chart for all three files:


14


13


15

These are all pretty noisy viewed at 100%, as you'd expect from ISO 25,600 shots with no noise reduction. I have a hard time having any preference between them, other than wanting to process them properly, which is what did (or tried to do) next. Here they are again using DXO Photolab 8 running Deep Prime XDs2:


3


10


8

Again, I have a hard time having a preference between them, but they are all clearly a lot better than the images with no NR, and have a lot more detail preserved (and no, it isn't false detail). Here are the ISO 100 images from both cameras:


6


1

There's clearly more detail in these images, as you'd expect for ISO 100 compared with 25,600. But that's mostly noticeable when viewing at 100% fairly close up, which is like peering at a massive poster print from a foot or so away.

Finally, I thought I'd see how the two cameras looked with a 3 and 4 stop underexposure, recovered in post. In over 20 years of digital processing, I don't think I've ever had to push a file more than 3 stops, and only very rarely as much as 2 stops, but it's interesting to see what happens if you do. First, the two files (one for each camera) at ISO 800, underexposed by 3 stops and recovered in post:


5



And lastly the two cameras at ISO 400, underexposed by 4 stops and recovered in post:


11


4

I haven't had a chance to see how these look on the DPR site (I'll do that when I post this), but I really don't have any preference between them, when I view them on my computer using Preview. This whole exercise has just confirmed to me my earlier opinion that any differences between the two sensors are actually imperceptible in terms of the final results. Modern noise reduction means we don't have to choose between noise and detail anymore for high ISO shots. It's the failure to use good noise reduction that is destructive to detail. Of course, you'll always get even better detail if you can shoot at lower ISO settings, but you won't get any advantage from choosing the R5 over the R5II in that respect (except for a much lower price, of course).

There is an assumption I think in all the above im that both cameras received the same amount of light. This may not be the case and thus the shot noise component is harder to quantify.

Using NR would not be a way to compare the camera system. Similar to turning on lens corrections doesn't help us understand a lens performance.

It is possible to understand the NR tool performance but that's something separate.

The comment about detail, noise, modern tools doesn't make very much sense to me. Most of my real work requires detail and less worried about colour depth, accuracy or noise. I can't use NR for this work to be analysed but we do find a validated NR flow helps the humans look at images (although we loose detail as ISO increases and add NR to that and things continue to reduce).

This is what I know about modern NR or PL.

PL likes a lot of pixels.

PL cannot see a photo from 6 feet away.

All PL can see is 1s and 0s.

PL is very clever at moving pixels or 1s and 0s around.

If you want to play with the modern stuff it is PL8. Not just NR for sure.

How does that relate to understanding a cameras system performance?

I read the user manual and check out how the raw files process in PL when a camera interests me.

How a camera's system performs with PL is very important to me.

Love that 45mp grain.

--

 

[Higgins2002]

DPR never re-did the studio scene for the R5 with the 85mm f/1.2L, makes a difference as you can tell with just the improved contrast in the center of the scene with the 85mm f/1.2L on the R5 II, and, there's perceivable CA on the corners of the studio scene with the original R5 vs on the R5 II, it's virtually nil with the 85mm f/1.2L. The RF 50mm f/1.2L and 85mm f/1.2L are neither slouches at f/5.6 in vignetting, however, there's a 0.20EV difference between the two, in favor of the 85mm f/1.2L according to optical limits in case you're curious.

https://opticallimits.com/canon/canon-rf-85mm-f-1-2-usm-l-review-30mp/

https://opticallimits.com/canon/canon-rf-50mm-f-1-2-usm-l-review/

.

In my observation, the R5 II is "better" than the R5 in terms of SOOC jpegs in low light. This is largely in my opinion, due to smarter metering and AWB, combined with slightly more intelligent sharpening and noise reduction even without Neural Network Noise reduction. Now in RAWs? Very similar results. Now in terms of read noise? They're on par with each other, which makes sense as they're both the same pixel pitch, however the original EOS R5 pulls ahead just slightly. But, if you compare them with a less dense EOS R3? They both get left in the dust in noise...

https://www.photonstophotos.net/Cha...R3_14,Canon EOS R5_14,Canon EOS R5 Mark II_14

.

My 2 cents? Don't buy an R5 II expecting better noise handling than the R5, or keep an original R5 because it has slightly to almost no advantage in low light RAW only vs the R5 II. Now, if you're concerned with noise in low light? Might I draw your attention to that R3, R6 II or R8, or upcoming R1, which all are less dense pixel pitches and full frame. Those have notably lower read noise than either than either the R5 or R5 II, but aren't 45MP, so it's a double edged sword of trading resolution for greater color fidelity in low light.

Yes I agree with your statement and I've been waiting for my R5II for 3.5 months now.

I had the R5 for 3 years and took 250 000 images with it so I'm aware of the R5's capabilities and flaws.

I'm in for the faster readoutspeed of R5II, but I'm a bit concerned by the result with raw files available on the web. Comparing IQ on R3 and R6II the R3 is better in every way but size and price, R5II is not better in the same way, better in some but worse in some.....

I'm concerned about higher iso because I'm going to use the R5II with f7.1-f10 maximum light gathering.

This f7.1-10 was somewhat a struggle with R5 (no problem in lighter days of the year late april to september where I live).

So I'm a bit dissapionted that the upgrade from an FSI sensor to a BSI Stacked sensor is not the stepup that was in R3 vs R6II