Dxomark: A55, A33, Nex

[pakodominguez]

ACR might have improved (I remember the cries of defence about NR raw as well back then)

Nope, they still treat A700 RAW files really badly...

[bfitzgerald]

Ah but I never said CMOS was better for lower ISO levels. I'd not disagree with that say ISO 400 has less noise than comparable models using a CMOS sensor. Nowadays ISO 1600 is a walk in the park for most cameras. Try pushing a 350 to ISO 6400 and see what happens.
I have tried a Canon 7d and it looked pretty good to me even high ISO shots but it was only a quick play about I'll admit that. I'm not really into Canon though.

On the jpegs the trouble with the A2x-3xx series was excessive chroma noise (very little NR done in jpeg) and lots of luminance NR smearing away details. I'd not shoot the A200 over ISO 200 it was that noticeable, def a raw camera. Sony seem to have got more of a grip on colour noise reduction for jpegs recently but they are still too heavy for my tastes with luminance NR smearing away details too aggressively.

The A55 looks good at high ISO, but appears very much in line with other makers using the same/similar sensors.
Note to the above post..if you look around there are many raw samples available and plenty of controlled tests. IMO the 14mp CCD is not as good as the 10mp CCD high ISO (though neither are any great shakes) and the old 6mp CCD is better than both low light but beaten at very high ISO levels with the newer CMOS ones.

Any doubts about jpegs and raw noise lots of places to visit..though I've never really felt studio shots did much for low light testing..I guess it's consistent though.
http://www.dpreview.com/reviews/sonyslta55/page16.asp

[David Kilpatrick]

Just for comparison I also went back and looked at Pentax K20D 14 megapixel Samsung CMOS images taken at the same time as the A350. First of all - using ACR process 2003, they are terrible at ISO 800. But... exactly like the A350, simply switching to Process 2010 with 25/25/25 S/LNR/CNR settings produces results equal to the A350 and quite acceptable by 2010 camera standards.

I do believe it is not the silicon which has changed, it is the way the image data is processed.

I have managed to find some ISO 800 7D and 5D stuff and checked that too - at pixel level, it is not superior to the A350. It is much the same. Can't find any 3200 or anything extreme because we simply never used those settings.

Pako - they do not treat A700 badly! We can process Shirley's A700 files with beautiful results up to 1600 - suitable for Alamy or any professional use - grainless. A700 files have been transformed by using 2010. Mind you, if you are leaving old files with the exclamation mark appearing and 2003 process in place, which will happen with any previously processed file still complete with .XMP sidecar, you do need to change to 2010.

David

[pakodominguez]

Pako - they do not treat A700 badly! We can process Shirley's A700 files with beautiful results up to 1600 - suitable for Alamy or any professional use - grainless. A700 files have been transformed by using 2010. Mind you, if you are leaving old files with the exclamation mark appearing and 2003 process in place, which will happen with any previously processed file still complete with .XMP sidecar, you do need to change to 2010.

David

I did a photo-shoot last month (nothing really interesting, you know how this work: I have to pay the rent...) with both cameras A850 on a tripod and A700 for close up, same settings, same lighting, same everything. LR overexpose A700 files more than 1 stop. And it was really hard to match color and density. For the second part of the shooting, I only use the A850...

[pakodominguez]

Ah but I never said CMOS was better for lower ISO levels. I'd not disagree with that say ISO 400 has less noise than comparable models using a CMOS sensor. Nowadays ISO 1600 is a walk in the park for most cameras. Try pushing a 350 to ISO 6400 and see what happens.
I have tried a Canon 7d and it looked pretty good to me even high ISO shots but it was only a quick play about I'll admit that. I'm not really into Canon though.

Canon look is a trend, specially follow by non canon users... and I think it is a shame cuz I like, other than Sony, Nikon and Pentax colors better.

On the jpegs the trouble with the A2x-3xx series was excessive chroma noise (very little NR done in jpeg) and lots of luminance NR smearing away details. I'd not shoot the A200 over ISO 200 it was that noticeable, def a raw camera. Sony seem to have got more of a grip on colour noise reduction for jpegs recently but they are still too heavy for my tastes with luminance NR smearing away details too aggressively.

Early Sony Alpha JPGs are not that good. V4 for the A700 made things a little better, you can print decent 11*14 from them. But the best quality comes from RAW, that's for sure.

Any doubts about jpegs and raw noise lots of places to visit..though I've never really felt studio shots did much for low light testing..I guess it's consistent though.
http://www.dpreview.com/reviews/sonyslta55/page16.asp

As DK said, what really changed was the software treatment of RAW files. I have to reprocess the MRW i took with the 7Hi in West Africa in 2003 and see how can I improve those pics with LR3

[Dusty]

Huh? The dxomark scorers are calculated at *EQUAL SIZE* (8 MP). That's a fact. I am stating facts. You are not.

It's meaningless, you don't use a 24MP camera at 8MP, or downsize all your 24MP images too 8MP every time you take a shot. What about the 6MP camera? do you then up-res it to the arbitary 8MP figure (wherever they got that from, plucked it out of thin air probably...gee 8MP is a nice round figure, lets go with that).
Greg

There is nothing "meaningless" about it. The scores are calculated at equal print. And that's how it shoud be. Dxomark got it 100% right.

And I guess if you were testing racing engines you would de-tune the V-12's to V-8 horsepower levels to test them?

I think DXO does some good work - I really like their software - but on this point I have to scratch my head and go - HUH? I think that if I produce a medium-quality result at 50 MPs and it prints better than your fantastic 8 MP shot, mine's better. If you jerk my shot around to make it on "equal" footing as your 8 MP shot, and then it doesn't hold up, you've cheated.

Any way you cut it, you need 2 scores - one at full resolution and another at equal resolution to get a good comparison.

Dusty

[David Kilpatrick]

Pako - they do not treat A700 badly! We can process Shirley's A700 files with beautiful results up to 1600 - suitable for Alamy or any professional use - grainless. A700 files have been transformed by using 2010. Mind you, if you are leaving old files with the exclamation mark appearing and 2003 process in place, which will happen with any previously processed file still complete with .XMP sidecar, you do need to change to 2010.

David

I did a photo-shoot last month (nothing really interesting, you know how this work: I have to pay the rent...) with both cameras A850 on a tripod and A700 for close up, same settings, same lighting, same everything. LR overexpose A700 files more than 1 stop. And it was really hard to match color and density. For the second part of the shooting, I only use the A850...

Are you sure you had the same camera raw defaults in Lightroom? Or that you should have had? For example, with ACR I use Brightness 25 as a default for the A700 and Brightness 50 as a defaults for the A900. Each camera needs slightly different settings. What I is to do open a normal (non-DRO) Standard JPEG next to an LR/ACR preview window, and then adjust the settings until my conversion most closely matches the in-camera JPEG for tone and colour.

Every camera is different. I do not touch the Exposure control or Fill Light or Recovery of course, but I may use Brightness, Contrast, Black point, Tone Curve, Saturation, and some of the colour controls if necessary (not usually though).

Some cameras are way off a match by default. Others are very close with the standard default settings.

David

[Neonsquare]

And I guess if you were testing racing engines you would de-tune the V-12's to V-8 horsepower levels to test them?

Comparing horsepower to MP eh? You're funny. The whole point about normalizing on a particular print size and resolution is to get a more practical measure of comparison for high iso image quality. Do you just use/print sizes based on what the camera spits out? Most of the time I get 30x20 cm prints of my photos and I want them to be as good as possible on that size. I do not do "100% prints".

If I'm able to get big poster sized prints with an A900 in superb quality on low ISO material - that is a plus against not getting big poster sized prints on low ISO using a "High-ISO"-Monster cam with 12MP. So with the A900 you can choose: Good High-ISO on smaller prints or good quality on big prints when using low iso. With a cam that just gets you the good High-ISO quality of 12MP reduced A900 files and not bigger resolution on lower ISO you would get less.

Any way you cut it, you need 2 scores - one at full resolution and another at equal resolution to get a good comparison.

Which is what they actually do - you can switch between "screen" and "print" in the diagrams. I think you would have noticed that if you really looked at their results.

[Greg Beetham]

That's not telling us anything we didn't already know before from using common sense, without any hypothetical 8MP DxO 'print' i.e. the more MP's you have the bigger the print you can do with good results; and high ISO ability is not as good as a camera with less MP's. My issue is the reduction of every camera too the same size at 8MP, too my mind that invalidates the usefulness of the 'print' test.
Greg

[Neonsquare]

@Greg
The whole point of this idea is, that there are more reasons to have more MP than just doing bigger prints. It is all about recording more information on the same area. If the signal/noise ratio of a single pixel is constant and the sensor area is constant, but the pixel density is bigger, then you actually have more information in the resulting RAW image. This also enables better noise reduction. Signal/noise-Ratio is not only dependent on the pixel density, but also on the whole progress in signal processing. The progress in sensors of the last years allowed a bigger pixel density while not getting worse signal/noise-ratio. This may enable bigger prints on one side but it particularily enables better image quality in smaller prints by showing more details on high iso images or by providing a bigger tonal range and dynamic range. You're essentially increasing the DPI of the image instead of the print area of a 300dpi print. You can see this increase in quality by using tests like the "print" test done by DxO. And yes; this actually IS common sense; but common nonsense is telling people that image quality of sensors gets worse the more MP a sensor has.

[Greg Beetham]

Ok, I have too agree there has been obvious improvement in sensors and sensor management but I don't think there is a printer out there that can actually show the improvement, I've yet to notice any difference to the eye between my 6MP 10MP and 12MP cameras using a dedicated photo printer at A4 size. (I don't print any bigger, I'm not sure what one is supposed to do with giant prints anyway)
If the current cameras can pre-render an image and control noise using the higher MP's for the printer better with more DR and finer tonal graduations than my older cameras can and the printer in real life can actually show it then I agree that that is an advancment.
Greg

[Neonsquare]

@Greg
Look at it at this: On levels of noise which printed perfectly beautiful with a 10MP camera - you will most certainly not see any difference. Tough there may be images of this 10MP which didn't look good anymore; this are cases were the noise actually reduces the effective resolution below the wanted print size (be it 8MP @300dpi or whatever). In those cases a 16 or even 24MP image with the same pixelwise SNR has more information in it and will show less noise/more details in the final results.

[bakubo]

Can't you just press the Screen button on dxomark and not get the 8mp print stuff if you don't think it is relevant?

[bfitzgerald]

Hey DxO have the scores up for the K-5 16mp CMOS sensor.
And erm it blows the A55 away


Looks like that translucent mirror is not doing the IQ and favours

[Dr. Harout]

Barry, I've read that and am really glad for the K5. But isn't that camera on a professional level? Come on, at least DxO is comparing a55 with a professional level camera, boy am I glad for that too.