First downloadable RAW R5 and R6 files

yes.. Canon's word: sensor in R6 is a better sensor from a high ISO perspective than in R5.. mentioned multiple time in Canon product presentation videos, by Canon EOLs, product reviews. nothing unexpected. IDX3 sensor is excellent in low light. So this is one of the R6 features: low light performance.

R5 high ISO performance (3200 - 6400 and up) will NOT be better than R6 or 1Dx3...

let's focus on your original statement though: you expected R5 to perform at high ISO by at least.. at least 2 stops better...

myself and Privatebydesign addressed you on that number of time. An you keep coming back with more "information" to support your point.

let's agree / accept that your initial statement is false and move on from here.

P.S. so here is you post.. let see what you said back then:

Canon must have achieved current f/5.6-level IQ at f/11. So that's the amount of light/improvement we're talking about here – at minimum, IMO.


and then here again. much of the muchness:

I hope I'm right by saying 2 stops improvement (when sized down to R resolution for equivalency as you suggested), but even one stop would be nice.


Let’s see the samples compared to R on the DPR tool. I was already wrong about dual gain tech coming to the R5, so what’s one more time to be wrong. Oh, I also voted that the “one more feature for photographers” rumor might be pixel shift high res mode. Being wrong on a rumors site about rumors isn’t something I get too worked up about.

Still find it strange how several of you discount the future possibility of ISO gains. Seems like a rather narrow-minded vision of the future and possible advances in sensor design and AI-based imaging. Reminds me of people saying humans would never fly. The science behind the high ISO performance being limited as explained here sounded more like someone repeating what they read once rather than an deep understanding of the physics behind it. If that’s not the case, and one of you have a PhD in this field and wrote some papers on it, I‘m all ears. You guys talk about improving high ISO of the current sensors like it’s as unlikely as time travel.
 
Upvote 0
There is a physical limit for how much improvement in low light performance we can expect in the future though. Noise isn't just caused by the camera. It is also a characteristic of the light itself. Even with a perfect camera that introduces 0 noise due to heat and imperfections in the read out circuits, you'll still get noisy images in low light.

From what you've written I'm not sure if you are aware of that. If you are, and these aren't the claims you were referring to, than I can't judge if pseudoscience is an appropriate word to use. But otherwise, I disagree with its usage.

Especially since you yourself seem to prefer testing / comparing noise at 1:1 resolution, which goes against the scientific approach of only changing one variable at a time.

Is this topic something with which you have studied professionally, or if you’re going off an article you read once? We’re not asking to bend the fabric of the universe, just find a way to improve high ISO performance. Maybe the breakthrough at some point will be AI-based processing of some kind. Maybe the solution will be some sort of organic sensor that sees more like some animal species see in low light. I was hoping Canon made some sort of breakthrough like that having seen those f/11 lenses. So being wrong about that is a big fat “so what”.

It would be cool to see would be a chart that plotted the increase in ISO performance/sensitivity year over year from the invention of the first digital camera to today. It would be interesting to see if we’ve hit a plateau as you’re suggesting or if we’ve been on a rise closer to what we see with computing power and Moore’s Law.
 
Upvote 0

LSXPhotog

Automotive, Commercial, & Motorsports
CR Pro
Apr 2, 2015
787
980
Tampa, FL
www.diossiphotography.com
His video certainly taught me that the 800 f/11 may be a massively compromised ownership experience. These are sharp sample photos, but when you see how much he had to crank the ISO for the image of the cheetah laying in the shadows, that’s unacceptable for wildlife photos. Why? Because a look at the shutter speed of 1/320...that isn’t really going to work with a moving animal and you’re already pushing things at 800mm even with IS and IBIS. I am impressed by the performance of that aspect for sure.

So far, the noise performance of the camera is looking really nice.
 
  • Like
Reactions: 1 user
Upvote 0

Joules

doom
CR Pro
Jul 16, 2017
1,801
2,247
Hamburg, Germany
Is this topic something with which you have studied professionally, or if you’re going off an article you read once? We’re not asking to bend the fabric of the universe, just find a way to improve high ISO performance. Maybe the breakthrough at some point will be AI-based processing of some kind. Maybe the solution will be some sort of organic sensor that sees more like some animal species see in low light.
That light behaves in a discrete way, like a particle, when coming in contact with surfaces is something that was briefly covered in school physics. If you want to call that professional study. I don't think that matters though, if you want to read up on the physics of light there is a ton of material out there. What we're talking about here is the noise inherent to the light itself, a property that is completly unaffacted by how good your sensor. Wikipedia has a small explanation for it: Shot noise.

As understanding and dealing with noise is a key aspect of astro photography, that's a good place to get information about it in the context of photography. Jon Rista has very detailed articles with example images and lots of math on his page (along some amazing deep sky images): Astrophotography Basics: Signal, Noise and Histograms & Astrophotography Basics: SNR. I believe he was also active on this forum under the account jrista, but haven't seen him post anything under the deep sky thread for a long while.

As for your second point, I though we were particularly concerned with the physical aspects of low light performance. That's why the focus was on RAW, not JPEG, or was it? Of course post processing techniques are continuing to improve. Slapping AI on something is no magical cure to any problem, but it can deliver impressive results for sure. Though, I personally think there's a pretty fine line between actually restoring details you would have gotten in an image by gathering more light, or just making the noise less apparent by smoothing it or replacing it with artificial detail.

As for animal eyes, not sure where you are going with this. As I said, the noise is in the light itself.

It would be cool to see would be a chart that plotted the increase in ISO performance/sensitivity year over year from the invention of the first digital camera to today. It would be interesting to see if we’ve hit a plateau as you’re suggesting or if we’ve been on a rise closer to what we see with computing power and Moore’s Law.
You are right, that would be cool. I'll see if I can put something together.

Just to clarify though, I am not saying that we have hit a plateau. I am saying that there is a physical limit for how noisy a low light image can look without further processing or techniques to improve it. That's what I consider to be meant when we talk about RAW file low light performance. And as progress is made, we approach that limit. As is generally the case when approaching these limits, it get's harder to make improvements as you get closer. I do think we are seeing that in recent releases, so hoping for really big steps forward appears to be futile to me. But for this last point, I don't have a proper source to point out at the moment. I'll get back to you if I can present the chart you mentioned, which I would also find very interesting myself.
 
  • Like
Reactions: 2 users
Upvote 0

SecureGSM

2 x 5D IV
Feb 26, 2017
2,360
1,231
Let’s see the samples compared to R on the DPR tool. I was already wrong about dual gain tech coming to the R5, so what’s one more time to be wrong. Oh, I also voted that the “one more feature for photographers” rumor might be pixel shift high res mode. Being wrong on a rumors site about rumors isn’t something I get too worked up about.

Still find it strange how several of you discount the future possibility of ISO gains. Seems like a rather narrow-minded vision of the future and possible advances in sensor design and AI-based imaging. Reminds me of people saying humans would never fly. The science behind the high ISO performance being limited as explained here sounded more like someone repeating what they read once rather than an deep understanding of the physics behind it. If that’s not the case, and one of you have a PhD in this field and wrote some papers on it, I‘m all ears. You guys talk about improving high ISO of the current sensors like it’s as unlikely as time travel.
We merely pointed out the fact that what you said in relation to possibility of R5 being at least two stops better in a high iso situation is not even remotely possible at this stage. What you now talking about is yet another hypothesis of “what if”
And in the outset. R6 RAW files in DPR sample images database look considerably better than R5. I would like to see what Photons to Photos will come up with. to my eye, R5 RAW files at iso 6400 look a bit mashie than I would like them to be. On an another hand, R6 RAW files do look impressive at ISO 6400 I am very much tempted at this stage with R6 high iso performance.... and the top dial. Just with my trusty 5D4s :)
 
  • Like
Reactions: 1 users
Upvote 0
That light behaves in a discrete way, like a particle, when coming in contact with surfaces is something that was briefly covered in school physics. If you want to call that professional study. I don't think that matters though, if you want to read up on the physics of light there is a ton of material out there. What we're talking about here is the noise inherent to the light itself, a property that is completly unaffacted by how good your sensor. Wikipedia has a small explanation for it: Shot noise.

As understanding and dealing with noise is a key aspect of astro photography, that's a good place to get information about it in the context of photography. Jon Rista has very detailed articles with example images and lots of math on his page (along some amazing deep sky images): Astrophotography Basics: Signal, Noise and Histograms & Astrophotography Basics: SNR. I believe he was also active on this forum under the account jrista, but haven't seen him post anything under the deep sky thread for a long while.

As for your second point, I though we were particularly concerned with the physical aspects of low light performance. That's why the focus was on RAW, not JPEG, or was it? Of course post processing techniques are continuing to improve. Slapping AI on something is no magical cure to any problem, but it can deliver impressive results for sure. Though, I personally think there's a pretty fine line between actually restoring details you would have gotten in an image by gathering more light, or just making the noise less apparent by smoothing it or replacing it with artificial detail.

As for animal eyes, not sure where you are going with this. As I said, the noise is in the light itself.


You are right, that would be cool. I'll see if I can put something together.

Just to clarify though, I am not saying that we have hit a plateau. I am saying that there is a physical limit for how noisy a low light image can look without further processing or techniques to improve it. That's what I consider to be meant when we talk about RAW file low light performance. And as progress is made, we approach that limit. As is generally the case when approaching these limits, it get's harder to make improvements as you get closer. I do think we are seeing that in recent releases, so hoping for really big steps forward appears to be futile to me. But for this last point, I don't have a proper source to point out at the moment. I'll get back to you if I can present the chart you mentioned, which I would also find very interesting myself.

That light behaves in a discrete way, like a particle, when coming in contact with surfaces is something that was briefly covered in school physics. If you want to call that professional study. I don't think that matters though, if you want to read up on the physics of light there is a ton of material out there. What we're talking about here is the noise inherent to the light itself, a property that is completly unaffacted by how good your sensor. Wikipedia has a small explanation for it: Shot noise.

As understanding and dealing with noise is a key aspect of astro photography, that's a good place to get information about it in the context of photography. Jon Rista has very detailed articles with example images and lots of math on his page (along some amazing deep sky images): Astrophotography Basics: Signal, Noise and Histograms & Astrophotography Basics: SNR. I believe he was also active on this forum under the account jrista, but haven't seen him post anything under the deep sky thread for a long while.

As for your second point, I though we were particularly concerned with the physical aspects of low light performance. That's why the focus was on RAW, not JPEG, or was it? Of course post processing techniques are continuing to improve. Slapping AI on something is no magical cure to any problem, but it can deliver impressive results for sure. Though, I personally think there's a pretty fine line between actually restoring details you would have gotten in an image by gathering more light, or just making the noise less apparent by smoothing it or replacing it with artificial detail.

As for animal eyes, not sure where you are going with this. As I said, the noise is in the light itself.


You are right, that would be cool. I'll see if I can put something together.

Just to clarify though, I am not saying that we have hit a plateau. I am saying that there is a physical limit for how noisy a low light image can look without further processing or techniques to improve it. That's what I consider to be meant when we talk about RAW file low light performance. And as progress is made, we approach that limit. As is generally the case when approaching these limits, it get's harder to make improvements as you get closer. I do think we are seeing that in recent releases, so hoping for really big steps forward appears to be futile to me. But for this last point, I don't have a proper source to point out at the moment. I'll get back to you if I can present the chart you mentioned, which I would also find very interesting myself.

Appreciate the thoughtful replies. Always nice when the harsh conversation turns more friendly.

If the R5 is one stop better than the R, can we not say I was half right and half wrong? :ROFLMAO:

One example of the kind of tech I was thinking Canon would bring to the R5 stills is illustrated by the 4K oversampled video that downsamples from 8K. From what I've read, it supposedly allows for full (I forget what the right word is) saturation(?) of each color channel, much like a still image from a Foveon sensor, plus producing higher dynamic range and sharpness. Not sure how I expected something like that to work for stills – perhaps processed the same way as the 4K oversampled video and delivered as S/MRAW.
 
Upvote 0