Photons to Photos runs the R5

[ahsanford]

More pixels + more base ISO DR than the 1DX3? Yes, please.



That and more posted at photonstophotos.net.

Seems to be some weird mojo going on around ISOs 100-400 in the DR, noise, etc. Can anyone explain that to me?

- A

 

[privatebydesign]

This has already been debated here but no it doesn’t. The RAW file is cooked at sub 400iso for some reason, they are baking in noise reduction. The pixel density argument has been debunked for several generations of sensors just dive deeper into that site for proof.

Incidentally Bill Cliff himself posted on the DPReview thread when the figures went up and he estimates the low iso NR are creating an extra 2/3 stop of PDR.

 

[ahsanford]

This has already been debated here but no it doesn’t. The RAW file is cooked at sub 400iso for some reason, they are baking in noise reduction. The pixel density argument has been debunked for several generations of sensors just dive deeper into that site for proof.

Incidentally Bill Cliff himself posted on the DPReview thread when the figures went up and he estimates the low iso NR are creating an extra 2/3 stop of PDR.

Re: pixel density, I wasn't expecting bad news (the was more tongue in cheek).

But the NR was news to me. Catch me up, PBD -- is this an optional feature, or is the NR baked into every file? If it's optional, do we have a feeling what it would look like with the NR turned off? Is it just a straight 2/3 stop worse than what Bill is reporting?

- A

 

[privatebydesign]

Re: pixel density, I wasn't expecting bad news (the was more tongue in cheek).

But the NR was news to me. Catch me up, PBD -- is this an optional feature, or is the NR baked into every file? If it's optional, do we have a feeling what it would look like with the NR turned off? Is it just a straight 2/3 stop worse than what Bill is reporting?

- A

It isn’t optional it is baked in, to me this is far more serious than the special video mode limitations. In effect Bill Cliff estimated the files would have 2/3 less PDR if they did not have noise reduction already applied which would put them exactly the same as the 1DX III.

Of course there is the dual pixel RAW file and should we ever get usable access to those there is a spare stop of DR in them, so potentially there is technically a touch more DR possible.

 

[ahsanford]

It isn’t optional it is baked in, to me this is far more serious than the special video mode limitations. In effect Bill Cliff estimated the files would have 2/3 less PDR if they did not have noise reduction already applied which would put them exactly the same as the 1DX III.

Of course there is the dual pixel RAW file and should we ever get usable access to those there is a spare stop of DR in them, so potentially there is technically a touch more DR possible.

They took the F-150 of workhorse professional cams (i.e. the 5-series line in general) and saddled it with a mandatory NR feature? Really?

Have they ever done that before with a critical aspect of IQ at such a price point before?

(...besides an AA filter I mean)

- A

 

[Act444]

It isn’t optional it is baked in, to me this is far more serious than the special video mode limitations. In effect Bill Cliff estimated the files would have 2/3 less PDR if they did not have noise reduction already applied which would put them exactly the same as the 1DX III

Wow. That appears to be a first (for Canon anyway). I honestly detest low ISO NR because of how much detail it strips away. And a lot of times I don’t even need it so I turn it off during PP to get the most out of the sensor. The difference is very noticeable as default NR is often too aggressive for my liking (and this is across the ISO spectrum). If this is something that cannot be reversed after the fact in DPP, etc., I do not like this development.

 

[DrToast]

There is a lot of debate going on about dpreview about this. The short version seems to be is that it’s unclear if any noise reduction has been applied.

 

[Mt Spokane Photography]

I don't know all that much about NR, but I understood that all CMOS sensors have on chip NR, and one of the advantages of rear illuminated sensors is that more complex NR circuitry is possible for each pixel. Its there in raw and every image or CMOS would be terrible.

So, is Bill saying more NR is added in processing, or on chip?

 

[privatebydesign]

There is a lot of debate going on about dpreview about this. The short version seems to be is that it’s unclear if any noise reduction has been applied.

It seems Bill Cliff isn’t unsure, from his own personal figures he is sure the R5 applies NR at 100iso and 400iso, the case for 200iso and 800iso is less clear but the evidence of milder NR application is still there at the later two iso levels as well.

He also estimates the gain in PDR of that NR at 100 iso to be approximately +2/3 stop but won’t be drawn on the amount of detail ‘lost’ to that NR. Having said that the detail/resolution of the R5 compares very favorably to the higher mp Sony a7R IV.

 

[Sporgon]

A few of us here on CR some years ago and no doubt elsewhere prophesied this happening on Canon Raws thanks to underdeveloped testers reporting superiority in cooked raws when compared with Canon’s unadulterated file.

It reminded me of car magazines in the late ‘70s who, when testing a car, would put the driver’s seat fully back, irrespective of how much generous driver’s legroom the car may offer, and then criticised the car for poor leg room in the rear “when the driver’s seat is fully back”. Eventually car manufacturers responded by giving their cars limited leg room for the driver, meaning tall people couldn’t get comfortable, and let’s face it, cars are driven most of the time without adults in the back. After complaints from taller drivers, some manufacturers gave the drivers seat more travel backwards, but put a bolt in the frame that only dealers knew about which could be removed should a potential new owner complain of poor front leg room, but the underdeveloped testers would not be able to put the seat back too far and then report poor rear leg room.


Now thank goodness the car reviews have matured beyond this rubbish, and we are now back to cars that generally allow generous front legroom should the driver require it, even at the expense of rear leg room.

Hopefully in another decade camera reviewers will have matured to the same level, although I wouldn’t hold my breath.

 

[Sporgon]

A few of us here on CR some years ago and no doubt elsewhere prophesied predicted this happening on Canon Raws thanks to underdeveloped testers reporting superiority in cooked raws when compared with Canon’s unadulterated file.

Just amended my own post here because "prophesied" was the wrong word; there was much evidence that cooking raws was going to happen in the future,, so it was very much a prediction !

 

[DrToast]

It seems Bill Cliff isn’t unsure, from his own personal figures he is sure the R5 applies NR at 100iso and 400iso, the case for 200iso and 800iso is less clear but the evidence of milder NR application is still there at the later two iso levels as well.

He also estimates the gain in PDR of that NR at 100 iso to be approximately +2/3 stop but won’t be drawn on the amount of detail ‘lost’ to that NR. Having said that the detail/resolution of the R5 compares very favorably to the higher mp Sony a7R IV.

Yes, that remains his opinion; I didn't mean to suggest otherwise. But there are many other people in that thread that don't think it's clear. To me, the fact that it performs so well when compared to the R IV makes me suspicious of the claim that there is noise reduction applied.

 

[privatebydesign]

Yes, that remains his opinion; I didn't mean to suggest otherwise. But there are many other people in that thread that don't think it's clear. To me, the fact that it performs so well when compared to the R IV makes me suspicious of the claim that there is noise reduction applied.

That it beats the 1DX III by so much makes it obvious that it does have NR applied.

Photographic Dynamic Range versus ISO Setting

 

[DrToast]

That it beats the 1DX III by so much makes it obvious that it does have NR applied.

Photographic Dynamic Range versus ISO Setting

View attachment 192057

Then where's the lost detail from the NR? Also, what does the 1DXIII have to do with it? It's a different sensor?

 

[privatebydesign]

Then where's the lost detail from the NR? Also, what does the 1DXIII have to do with it? It's a different sensor?

As Bill Claff points out at such low ISO it is impossible to calculate the loss in resolution without a RAW file that doesn't have NR applied, and you currently can't get one of them. The larger point being at those low iso's it becomes an engineering difference not a visible one.

The relevance of the 1DX III is that it is the same generation and technology sensor but the RAW file is not cooked, it is the best indicator of where the PDR would be for an uncooked R5 RAW, sure the EDR will be different but that doesn't matter as we can still use the PDR as a normalized comparison.