fredmiranda.com
Login

Moderated by: Fred Miranda
  New fredmiranda.com Mobile Site
  New Feature: SMS Notification alert
  New Feature: Buy & Sell Watchlist
  

FM Forums | Canon Forum | Join Upload & Sell

       2       3       end
  

Archive 2012 · visualized read noise for TONS of different cameras

  
 
skibum5
Offline
• • • • • •
Upload & Sell: Off
p.1 #1 · visualized read noise for TONS of different cameras


Wow, someone put together quite the comprehensive visual comparison, sooo many different bodies compared:
http://a2bart.com/tech/allcamdknz.htm

One thing to keep in mind is that he didn't account for differences in gain delivered by stated ISO number (for instance the 5D2 and 50D, in particular, give less gain for a given ISO than many of the others, so in reality, would look worse than shown here).

Another thing to keep in mind is that he didn't account for MP differences, these are all 100% crop comparisons, so the lower MP cameras here should actually look a bit worse in comparison than they do compared to high MP ones like 5D2/5D3/D800/Medium Format, in particular.

Also unless he used a true RAW examination program the results might be muddled since who knows what ACR does for each camera behind the scenes, it may do different baked in NR for each and set base levels differently, etc.

So it's not really quite fair test, but quite interesting all the same.

One clear thing is that Canon DSLR are simply poor at low ISO shadow noise, worse than basically any type of camera there is other than some (but not all!) point and shoots. The 5D3 did nothing at all to improve this. Even the D4 which has no fancy Exmor patent tech did way better.

ISO6400+ the 5D3 shadow quality should be better than D800 though, but then again most shots probably are taken ISO3200 and under, but for scenes with tons of near black areas at super high ISO the 5D3 definitely should be better than the D800.

It does fit expectations, such as that (going by 100% view with nothing normalized for MP count of ISO gain) the 40D/1D3 era of Canon was much more pushable than the 5D2/50D era.

That exmor sensor push great at low ISO.

The D4 looks better than expected, especially at low ISO. This also shows that Canon COULD do MUCH better for low ISO shadows even without Exmor patents since the D4 doesn't use any Exmor patented tech. It is just a regular sensor design. Could it be as a couple of semi-insiders hinted at that Canon saved the best tech for the 1DX. I mean even if you normalized for 22MP vs 16MP that D4 would still have heck of better low ISO shadows than the 5D3. And once again it says nobody is asking for magic or fairly land, just look at how all those Nikons, even ones without Exmor, do at low ISO. This is reality.

The 5D3 finally pulls ahead of D800 shadow quality at ISO6400 (it looks like it at ISO3200 but I think after MP and ISO gian normalization it probably won't happen until ISO6400, although it is hard to be 100% sure, maybe ISO4000 or 5000?).

The 5D3 is a bit of an outdated let down at low ISO, at 100% view it's not even quite to the 60D level even and the NON-Exmor D4 just blow it to pieces, so forget crying about not having patent access....

You can see the unique form of vertical banding the 7D has (it varies a fair amount copy to copy though some will have a lot of it and some much less).

You can see they did fix up the shadow noise quality ISO1600 and up compared to the 5D2 quite a bit (even more than apparent since the 5D3 appears to give more gain for a given ISO label at higher ISOs).

The below ISO800 shadow quality from Canon DSLRs appears to be second rate though. Nikon, even ones not even using fancy Exmor patents, are just simply better. Even Canon P&S are better shadow noise at ISO100.

But on the positive side, the deepest high iso shadows should look better than with the 5D2, by well over a stop,if this holdsup.



Edited on Apr 15, 2012 at 04:28 PM · View previous versions



Apr 15, 2012 at 03:18 PM
skibum5
Offline
• • • • • •
Upload & Sell: Off
p.1 #2 · visualized read noise for TONS of different cameras


You can see Nikon uses non-exmor for their action/speed/lowlight camera since it gets much better shadows at high iso than the exmor stuff.

But you can see that with the new generation they really did go to a new generation since they also fixed up non-exmor shadows this time while Canon is clearly STILL using old generation sensor fab and didn't put in to update the way Nikon did since they didnt improve at all at low iso and not as much at high iso.

So Nikon's own sensors are quite an all around balance, the best at high iso and yet pretty good at low iso (if not quite a match for Sony Exmor).


Talking strictly Canon vs Canon, they did fix up the high iso read noise on the 5D3 vs 5D2 it seems, the super deep tone read noise does look much better than for the 5D2, so the very depths of the darkest parts of high iso images should be a lot nicer on 5D3 than 5D2, what the mid-tone SNR difference will be is not known though).




Edited on Apr 15, 2012 at 04:28 PM · View previous versions



Apr 15, 2012 at 03:26 PM
skibum5
Offline
• • • • • •
Upload & Sell: Off
p.1 #3 · visualized read noise for TONS of different cameras


One caveat is that if he used lightroom or something instead of a true raw examination program the behind the scenes fiddling of ACR and where it set levels and how much under the scenes NR it did could have made the results potentially unreliable. I have to find out what he did.



Apr 15, 2012 at 03:27 PM
skibum5
Offline
• • • • • •
Upload & Sell: Off
p.1 #4 · visualized read noise for TONS of different cameras


But yeah ISO1600+ the 5D3 deep shadows do seem to be improved quite a bit from the 5D2 (again especially since you probably need to give the 5D3 another 1/3 stop+ boost compared to what it shown since the 5D2 doesn't seem to apply as much gain per ISO setting as the 5D3 does at high ISO).


Apr 15, 2012 at 03:30 PM
skibum5
Offline
• • • • • •
Upload & Sell: Off
p.1 #5 · visualized read noise for TONS of different cameras


Making some guesses, ISO3200+ the 5D3 deepest black tones might be almost two stops better than the 5D2 although the brighter dark tones and up SNR will probably only be be between 1/3 to 2/3 of a stop better (the latter stuff is not demonstrated here at all and depends more upon light collection, which no way has been improved so much, it's not even possible for it to have been). Even at ISO1600 it might be almost 1.5 stops better deepest black tones (again we are talking only for the deepest, darkest, near pitch black parts of an image, some shots will have none of the shot even in this realm, some with tons of dark shadows all around the subject may look much better though).

(oddly though when I measured read noise in a true raw program, without any processing having been done, it seemed to be only 1/4 stop better, although the gain differential might bring it to 2/3 of a stop, so that does make one wonder, 2/3 of a stop if hardly 1.5-2 stops, so hopefully what he shows is real and it's not just some ACR artifact....)



Apr 15, 2012 at 03:40 PM
skibum5
Offline
• • • • • •
Upload & Sell: Off
p.1 #6 · visualized read noise for TONS of different cameras


anyone enough about this stuff, out to shoot

although something about his test does seem a bit odd, since it would imply DR increases at high iso for D4 not seen by DxO and a better improvent for 5D3 high iso DR than seems possible as well



Apr 15, 2012 at 03:40 PM
willis
Offline
• • •
Upload & Sell: Off
p.1 #7 · visualized read noise for TONS of different cameras


Why is the G12 so good at 100 ISO? Better than any of Canons DSLRs and the G1X. Presumably there must be some noise reduction going on there.


Apr 15, 2012 at 03:59 PM
mttran
Offline
• • • • •
Upload & Sell: Off
p.1 #8 · visualized read noise for TONS of different cameras


Black frame is your cook book, best test to explore your camera reading noise. Pushing the black frame simulates how cam I/O performance against wider DR scenes. It's is not a 100% perfect test but pretty much giving you the DR margin. Based on these samples...you can say Canon low iso range is not that hot for any isoless application


Apr 15, 2012 at 05:23 PM
Pixel Perfect
Offline
• • • • • • •
Upload & Sell: Off
p.1 #9 · visualized read noise for TONS of different cameras


Love the banding on the Canon G3, G5 and G6. But Pentax Q must take the cake, it looks bad from ISO 200 but no banding.


Apr 15, 2012 at 05:43 PM
surf monkey
Offline
• • • •
Upload & Sell: Off
p.1 #10 · visualized read noise for TONS of different cameras


In this test it looks like the 7D is about as good as the 5D2, but I find the 7D to have much more noise in my images. What's the explanation for this?

Edited on Apr 15, 2012 at 08:24 PM · View previous versions



Apr 15, 2012 at 08:19 PM
surf monkey
Offline
• • • •
Upload & Sell: Off
p.1 #11 · visualized read noise for TONS of different cameras


If the D4 doesn't use exmor then what is Nikon doing that is so much better than Canon?


Apr 15, 2012 at 08:20 PM
morganb4
Offline
• • • • •
Upload & Sell: Off
p.1 #12 · visualized read noise for TONS of different cameras


surf monkey wrote:
If the D4 doesn't use exmor then what is Nikon doing that is so much better than Canon?


Competing - whilst Canon sat on laurels and invested in cinematography instead.



Apr 15, 2012 at 08:44 PM
thedutt
Offline
• • • •
Upload & Sell: Off
p.1 #13 · visualized read noise for TONS of different cameras


surf monkey wrote:
If the D4 doesn't use exmor then what is Nikon doing that is so much better than Canon?

I am sure that these photos have a place, but if you use the latest canon camera (5DMKIII) @ ISO 12800 & 25600, the results are very useable under difficult circumstances. (And the camera can focus in these situations). I have never used a Nikon camera so do not have anything to compare to, but I really doubt that 5DMKIIIs output is in any way limiting my photography in low light.

If you shoot Canon and are concerned if you should switch due to better sensor in D800 (a very fine camera), do yourself a favor. Rend a 5DIII for a weeks before to spend all that energy in switching.




Apr 15, 2012 at 08:46 PM
morganb4
Offline
• • • • •
Upload & Sell: Off
p.1 #14 · visualized read noise for TONS of different cameras


^This has more to do with good light.


Apr 15, 2012 at 08:50 PM
thw2
Offline
• • • •
Upload & Sell: Off
p.1 #15 · visualized read noise for TONS of different cameras


surf monkey wrote:
In this test it looks like the 7D is about as good as the 5D2, but I find the 7D to have much more noise in my images. What's the explanation for this?


Because you EXPECT the 5D2 to be better when in reality, they are really close in use. It's been verified by others:
"When the 7D has been announced, I was doubtful about its image quality at high ISO...18 megapixels on APS-C are really a lot! Surprisingly, the 7D is relatively good in this aspect. It is not as good as the 1DsIII, but the difference is small - I'd say about 0.5 stop..."

Edited on Apr 15, 2012 at 08:58 PM · View previous versions



Apr 15, 2012 at 08:55 PM
thw2
Offline
• • • •
Upload & Sell: Off
p.1 #16 · visualized read noise for TONS of different cameras


morganb4 wrote:
Competing - whilst Canon sat on laurels and invested in cinematography instead.


Ya, Canon is going where they think the profits lie.

Pixel Perfect wrote:
Love the banding on the Canon G3, G5 and G6.


And these use Sony CCDs.

Edited on Apr 15, 2012 at 09:14 PM · View previous versions



Apr 15, 2012 at 08:58 PM
wfektar
Offline
• •
Upload & Sell: Off
p.1 #17 · visualized read noise for TONS of different cameras


surf monkey wrote:
In this test it looks like the 7D is about as good as the 5D2, but I find the 7D to have much more noise in my images. What's the explanation for this?


The 5D2 is using more than twice as much information in describing a scene as the 7D.



Apr 15, 2012 at 09:13 PM
skibum5
Offline
• • • • • •
Upload & Sell: Off
p.1 #18 · visualized read noise for TONS of different cameras


surf monkey wrote:
In this test it looks like the 7D is about as good as the 5D2, but I find the 7D to have much more noise in my images. What's the explanation for this?


This only looks at the deepest shadows, the very bottom end stops. Once you get a bit brighter photon collection efficiency starts mattering more and more so in the light shadows and brighter it's more about the photo conversion efficiency and since the 5D2 has 2.56x larger light collecting area it is not a surprise that it has better SNR for those regions.

So if you shoot a gray wall surrounded by dark shadows the dark shadows might be pulled to similar to detail but the gray wall might have a lot more noise on the 7D.

These are not middle gray patches compared for noise to see typical SNR these are tests of what detail you might see in the darker stops and thus it's more of a dynamic range test than a SNR test.



Apr 15, 2012 at 09:26 PM
skibum5
Offline
• • • • • •
Upload & Sell: Off
p.1 #19 · visualized read noise for TONS of different cameras


thedutt wrote:
I am sure that these photos have a place, but if you use the latest canon camera (5DMKIII) @ ISO 12800 & 25600, the results are very useable under difficult circumstances. (And the camera can focus in these situations). I have never used a Nikon camera so do not have anything to compare to, but I really doubt that 5DMKIIIs output is in any way limiting my photography in low light.

If you shoot Canon and are concerned if you should switch due to better sensor in D800 (a very fine camera), do yourself a favor. Rend a 5DIII for a
...Show more

It doesn't limit in low light, you can even see here that in the upper ISO it might have better deep shadows than the D800 actually, it's in good light where it differs in not having, by today's standards, very good dynamic range, same as the old 1Ds3 half a decade ago. In the high ISO you see it does look better than the 5D2 though and even D800 if you go really high, for the deep shadows up there.

If these can be trusted, and some of the differences up top seem a bit larger than expected, then it, the 5D3, might have better dynamic range than the 5D2 at any ISO1600+ by a noticeable amount and better than the D800 ISO6400ish and up. Maybe that is why some say it look 1.5+ stops better at high ISO and some say only 1/3-2/3 better. Maybe some are shooting stuff where DR matter more than SNR?

It will be interesting to see what DxO charts look like. It's pretty strange they don't have 5D3 tests up yet.


Edited on Apr 15, 2012 at 09:37 PM · View previous versions



Apr 15, 2012 at 09:29 PM
skibum5
Offline
• • • • • •
Upload & Sell: Off
p.1 #20 · visualized read noise for TONS of different cameras


surf monkey wrote:
If the D4 doesn't use exmor then what is Nikon doing that is so much better than Canon?


dunno, maybe they were willing to spend to upgrade their fab and canon wanted to re-use old stuff for higher profit per body. It will be interesting to see what the 1DX does.

I think EOSfun said something about they reserved the new fab for the 1DX and wanted max profit margin for 5 series Maybe Canon jsut cheaped out

OTOH, someone else said that to get good low iso dynamic range without exmor you might need to use expensive fab, maybe a $7000 D4 can cover that expense and a $3500 and under model can't Maybe it costs a ton to do much better without Exmor



Apr 15, 2012 at 09:34 PM
       2       3       end




FM Forums | Canon Forum | Join Upload & Sell

       2       3       end
    
 

Welcome back
Log in to your account