DxO has A77 test now

[bakubo]

For those who are interested in the DxO measurements here they are:

http://www.dxomark.com/index.php/News/D ... S-C-camera

[agorabasta]

DxO is useless, as usual.

They measure the SNR at 18% grey. Which means they measure the photon noise instead of evaluating the equipment noise. Hence the a77 looks as good as a55 at normalised res.
But the problem is that a77 has a lot of read noise at about 5% grey level. That noise is formed of random abnormally bright coloured pixels that simply cannot be reduced without either killing the detail or leaving some low-freq colour (and luma too) noise in the shadows. Then if you try and push the shadows a little bit, that ugly stuff promptly springs out of the shadows and kills the entire image.

I still hope it's all related to the way they drive the sensor analogue readout circuitry, like timing and voltages. So I still hope the Nex7 may turn out better in that regard. But my hopes are dwindling by the minute.

[bakubo]

DxO is useless, as usual.

They measure the SNR at 18% grey. Which means they measure the photon noise instead of evaluating the equipment noise. Hence the a77 looks as good as a55 at normalised res.
But the problem is that a77 has a lot of read noise at about 5% grey level. That noise is formed of random abnormally bright coloured pixels that simply cannot be reduced without either killing the detail or leaving some low-freq colour (and luma too) noise in the shadows. Then if you try and push the shadows a little bit, that ugly stuff promptly springs out of the shadows and kills the entire image.

I still hope it's all related to the way they drive the sensor analogue readout circuitry, like timing and voltages. So I still hope the Nex7 may turn out better in that regard. But my hopes are dwindling by the minute.

Interesting point.

There is a thread going on this other forum and if you want you might add your input to it:

http://forums.dpreview.com/forums/read. ... e=39421267

[agorabasta]

No Henry, I chose to stop posting there quite some time ago. Posting there would be well below my personal standards.

And then, anyway, the more intellectual posters from that forum do read this forum.

[bakubo]

No Henry, I chose to stop posting there quite some time ago. Posting there would be well below my personal standards.

Completely understand and you are right.

And then, anyway, the more intellectual posters from that forum do read this forum.

Good point.

[pakodominguez]

And then, anyway, the more intellectual posters from that forum do read this forum.

Good point. [/quote]
Flatering...

[bfitzgerald]

I'd had serious doubts about DxO so I can't really say I'm all that interested in their findings. Real world results are far more revealing.
When we see some good ones let me know

[OneGuyKs]

DxO is useless, as usual.

They measure the SNR at 18% grey. Which means they measure the photon noise instead of evaluating the equipment noise. Hence the a77 looks as good as a55 at normalised res.
But the problem is that a77 has a lot of read noise at about 5% grey level. That noise is formed of random abnormally bright coloured pixels that simply cannot be reduced without either killing the detail or leaving some low-freq colour (and luma too) noise in the shadows. Then if you try and push the shadows a little bit, that ugly stuff promptly springs out of the shadows and kills the entire image.

I still hope it's all related to the way they drive the sensor analogue readout circuitry, like timing and voltages. So I still hope the Nex7 may turn out better in that regard. But my hopes are dwindling by the minute.

Interesting point.

There is a thread going on this other forum and if you want you might add your input to it:

http://forums.dpreview.com/forums/read. ... e=39421267

Keep in mind agorabasta also made a big song & dance (that time on dpreview) about how A550 has the worst sensor, *ever*, worse than A850/A900 (that he claimed is one of the worst) and significantly worse than anything else on APSC (like 7D that was released at that time).

Pepole on dpreview obviously refuted his repeated assertion about that sensor (that sensor is pretty similar to Nex-5/3 sensor, rated highly).

Maybe that's why he no longer posts there?

In any case, I won't tale agorabasta claims as gospel (given that track record last time with A550).

[agorabasta]

My old posts are still there at DPR forums. Almost everything is well documented, except a few posts the DPR staff moderator-trolls removed being unable to respond factually.

Maybe that's why he no longer posts there?

That's your imagination.

[bfitzgerald]

Well they have some nice samples up at DPR jpeg and raw... this pretty much seals the case for not looking at the A77 for low light shooters. A77 is struggling at high ISO with lots of chroma noise, jpegs are fairly weak too. The light robbing SLT mirror only compounds the problem further

One for lower ISO shooters I think

This is why DxO isn't that interesting because real shots show far more than a graph ever can. Sony are going to have a hard time with the A77 for this type of user who not only demands top performance all round but superior IQ as well. I don't think many are going to be willing to compromise this for a fast FPS or an EVF

[David Kilpatrick]

The DPReview test is seriously flawed, as usual. If you examine the Canon 7D versus A77 say at ISO 100, and download the processed from raw JPEG via ACR prerelease, go to 100% and move over the colour checker's darkest shades it is immediately clear that the A77 has a stop or so shadow lift relative to the 7D - in other words, it's really being tested at ISO 200 versus 100. There is bound to be more noise if a curve which opens up the shadows and loses true maximum black, and also retains loads of colour information in the shadows, is used.

To make accurate comparisons, match the shadow density, don't just judge midtones.

David

[agorabasta]

While David is quite right that the 7D samples have the shadows effectively crushed, there's also some shadow brightening in the a77 samples due to noise. And that noise is there in the a77 samples even at ISO100.
And then that noise is simply making the shadows appear brighter.

And then we all remember how horrible are the deep shadows at low ISO with the 7D. So now the a77 is worse yet...

[pakodominguez]

And then we all remember how horrible are the deep shadows at low ISO with the 7D. So now the a77 is worse yet...

witch 7D? The Konika Minolta or the Canon?

[agorabasta]

witch 7D? The Konika Minolta or the Canon?

That's Canon 7D. Noise and banding in the shadows at base ISO.