DPReview Has 50D & D90 Sample Photos

[jpeyton]

50D

D90

Color me unimpressed by the noise levels and banding in the ISO 6400 shot in the 50D gallery. For a camera that was supposed to be a 1.5 stop improvement over the 40D, it doesn't deliver so far. Of course, a full review is still needed.

The ISO 6400 shot in the D90 gallery is quite pleasing; it does look about a 1/2 to 3/4 stop better noise performance than the D300. I wish they shot the same subjects side by side so we could make a fair comparison to the 50D.

 

[Deadtrees]

50D is supposed to have 4 levels of noise processing algorithms. To make any judgements, I need to know what level of noise algorithms were applied and how they work.
Anyway, in that ISO 6400 image of 50D, there're tons of color noise while maintaing details. I always favor that instead of smearing details to eliminate color noise. Getting rid of color noise only takes a few clicks but the details lost is just lost. (Sony A350 is the worst contender in that case.)

 

[soydios]

Originally posted by: Deadtrees
50D is supposed to have 4 levels of noise processing algorithms. To make any judgements, I need to know what level of noise algorithms were applied and how they work.
Anyway, in that ISO 6400 image of 50D, there're tons of color noise while maintaing details. I always favor that instead of smearing details to eliminate color noise. Getting rid of color noise only takes a few clicks but the details lost is just lost. (Sony A350 is the worst contender in that case.)

You're probably right in assuming that it's at Canon's default noise reduction setting: off. But it's the luminance noise that contains most of the detail; chroma noise reduction can be applied without affecting the image as much as luminance smoothing. Nikon and Pentax both take an aggressive approach to chroma noise reduction but are far lighter on the luminance smoothing, so detail is retained but the ugly color splotches are removed. It's a simple menu option to turn chroma noise reduction on in Canon cameras, but for some strange reason Canon leaves it off by default, which is how DPReview does all their noise and image quality tests.

Really, I would like RAW files at the same settings of the same subject from both cameras so that I could compare the undoctored output from the sensors myself. It won't affect my purchasing decision (D90 as soon as it's shipping), but I am curious if that 15MP 1.6x sensor lives up to the hype, or if it's following in the footsteps of the D2X.

The D2X was first to market by several years with a 12MP APS-C sensor, but the noise was just awful with that sensor. Nevertheless, it had the highest pixel density on the market in a CaNikon DSLR sensor from Fall 2004 until the D300 was announced in Fall 2007, and as such was the wildlife camera of choice for the longest time. The 10MP sensor of the D200/D80/D40x/D60/A100 wasn't much better. Things didn't improve until last year with the release of the 12MP D300. The 50D now has the highest pixel density of any DSLR sensor larger than Four-Thirds, and as the density leader might be too far ahead of its time, with the poor noise performance to match.

Remember that those crop factors drastically reduce pixel area.
1.0x = full frame (Canon 1Ds/5D, Nikon D3/D700)
1.3x = slightly larger than 1/2 (Canon 1D)
1.4x = 1/2
1.5x = slightly less than 1/2, aka APS-C (Nikon DX, Sony, Pentax)
1.6x = slightly larger than 1/3 (Canon EF-S)
1.7x = 1/3 (Sigma Foveon)
2.0x = 1/4 (Four-Thirds)

 

[Deadtrees]

Originally posted by: soydios

Originally posted by: Deadtrees
50D is supposed to have 4 levels of noise processing algorithms. To make any judgements, I need to know what level of noise algorithms were applied and how they work.
Anyway, in that ISO 6400 image of 50D, there're tons of color noise while maintaing details. I always favor that instead of smearing details to eliminate color noise. Getting rid of color noise only takes a few clicks but the details lost is just lost. (Sony A350 is the worst contender in that case.)

You're probably right in assuming that it's at Canon's default noise reduction setting: off. But it's the luminance noise that contains most of the detail; chroma noise reduction can be applied without affecting the image as much as luminance smoothing. Nikon and Pentax both take an aggressive approach to chroma noise reduction but are far lighter on the luminance smoothing, so detail is retained but the ugly color splotches are removed. It's a simple menu option to turn chroma noise reduction on in Canon cameras, but for some strange reason Canon leaves it off by default, which is how DPReview does all their noise and image quality tests.

Really, I would like RAW files at the same settings of the same subject from both cameras so that I could compare the undoctored output from the sensors myself. It won't affect my purchasing decision (D90 as soon as it's shipping), but I am curious if that 15MP 1.6x sensor lives up to the hype, or if it's following in the footsteps of the D2X.

The D2X was first to market by several years with a 12MP APS-C sensor, but the noise was just awful with that sensor. Nevertheless, it had the highest pixel density on the market in a CaNikon DSLR sensor from Fall 2004 until the D300 was announced in Fall 2007, and as such was the wildlife camera of choice for the longest time. The 10MP sensor of the D200/D80/D40x/D60/A100 wasn't much better. Things didn't improve until last year with the release of the 12MP D300. The 50D now has the highest pixel density of any DSLR sensor larger than Four-Thirds, and as the density leader might be too far ahead of its time, with the poor noise performance to match.

Remember that those crop factors drastically reduce pixel area.
1.0x = full frame (Canon 1Ds/5D, Nikon D3/D700)
1.3x = slightly larger than 1/2 (Canon 1D)
1.4x = 1/2
1.5x = slightly less than 1/2, aka APS-C (Nikon DX, Sony, Pentax)
1.6x = slightly larger than 1/3 (Canon EF-S)
1.7x = 1/3 (Sigma Foveon)
2.0x = 1/4 (Four-Thirds)

I agree with you but not entirly. I agree that Nikon/Pentax doesn't smear details much but even that is too much, I think. However, new Nikon cameras are better in that aspect: D80 was terrible.
Anyway, I'm interested in comparisons between 50D and K20D as they have same resolution(K20D is a bit off due to SR) that are based on the gap-less sensor technology.

 

[randomlinh]

nice on the D90. I'm really glad to see nikon jump this much. competition is gooood. I just hope canon seriously contends and doesn't just do a half ass job trying to maintain a small lead.

 

[ElFenix]

the canon has clearly not had any NR applied. dpreview always tests at default settings, and default on canon is NR off. trying to compare the noise of jpegs processed in camera with very different settings is ridiculous.

 

[ivan2]

FYI the 6400 shot of canon is the bananas and pears and for D90 is the news paper shot(last one)

hope this saved some of your time hunting through the typical pictures...

 

[andylawcc]

dpreview just updated their website saying that those pics are from a pre-production camera

 

[ElFenix]

We've just been informed by Canon that, contrary to our previous understanding, the camera we used for the initial samples gallery is still pre-production, not final. We've updated the news story for the samples gallery to reflect this.
Apologies to all for the confusion.

--
Andy Westlake
dpreview.com/lensreviews

assuming it's going into production shortly, i wonder how much tweaking still needs to be done?


edit: damnit! 1 minute!

 

[soydios]

Originally posted by: ElFenix
the canon has clearly not had any NR applied. dpreview always tests at default settings, and default on canon is NR off. trying to compare the noise of jpegs processed in camera with very different settings is ridiculous.

Aye, it beats me why Canon leaves their noise reduction off by default, particularly the chroma noise reduction. But it also beats me why DPReview doesn't do apples-to-apples comparisons.

 

[ElFenix]

Originally posted by: soydios
But it also beats me why DPReview doesn't do apples-to-apples comparisons.

i've never figured that out. they go through and test and provide crops for all the settings, but when it comes time to comparing head to head they go back and use the camera as it comes out of the box. that's why i just skip ahead to the raw samples near the end.

the one exception i've noticed is olympus SLRs. they did comparisons using both default and custom settings, and have used the custom settings for every review since.



edit: and if the camera somehow gets reset to factory defaults and i don't notice or forget to change settings, i'd rather have NR off. other than dark frame subtraction you can do NR in PP. you can't add detail.