A77 vs. Canon 60D vs. NEX 5N vs. Panasonic GH2 raw

[bakubo]

I don't know if anyone has seen the new A77 studio shot and comparison tool that dpreview posted today, but I decided to take a look since like most people the A77 has some attractive things about it (along with some unattractive things and some yet unknown things). I usually just take a cursory look at these sorts of pixel-peeping things, but what I was seeing surprised me a bit so I spent a few more minutes on it. Since I shoot an A700 and Canon 60D now those are the 2 most interesting bodies to compare the A77 to for me. Unfortunately, the A700 is not included (I think they only include current cameras). So, I compared the A77 24mp, 60D 18mp, NEX 5N 16mp, and Panasonic GH2 18mp (m4/3). I am looking at raw. Maybe a later raw processor will get better results.

Hmm, it isn't looking too good to me. At ISOs above, say, 400 or so the 60D and 5N look better to me and as the ISO goes up they get even better in comparison to the A77. If you downsize the A77 to 18mp then it might help the A77 a little bit (or downsize both of them to the 5N 16mp), but at the higher ISOs I don't think that it is likely at all that it would help enough. I am not a professional pixel-peeper, but I have looked at various parts of the image and the A77 just doesn't look so good even though the pixel density is similar to the 60D.

Not too long ago David had an interesting blog post about the new 24mp APS-C sensor:

http://www.photoclubalpha.com/2011/07/2 ... egapixels/

I have thought my 60D probably does a better job in general at higher ISO than my A700, but using the comparison tool and comparing it to the very highly regarded 5N I was surprised that the 60D looks pretty good in comparison to it too. The m4/3 18mp sensor looks pretty similar to the A77 24mp sensor at high ISO. Maybe the m4/3 is even a little better, but close enough not to be concerned about.

Here is the comparison:

http://www.dpreview.com/reviews/studioc ... 4527209579

I guess all of this is no surprise for agorabasta though. I think he is right that DxO graphs and ratings don't tell the whole story.

[agorabasta]

Henry,

What the DxO measures is just noise RMS value at 18% grey. There are two quite obvious problems with that. First is that 18% is too bright to reveal the read noise at lower-mid ISO values. The second is that RMS value tells nothing of the structure of the noise.

The more realistic gauge would be to measure the noise at about 2EV lower than 18%, i.e. at about 4%; and then to provide a spatial frequency plot of that noise. Obviously, the noise that's strong and tight (i.e. having more energy at higher spatial freqs) is much more tolerable than a weaker yet broader noise features. So it's also possible to create a perceptional frequency-weighted noise value gauge instead of a full spatial frequency plot.

The noise from the a77 sensor has a badly uneven structure due to a very high density of 'warm' stray pixel values. The density is so high that they cannot be filtered out effectively.
Then the a900 raw also initially suffered from those stray pixels, but due to their lower spatial density it finally became possible to filter them out though advanced algorithms of modern converters. But the a77 raw as it is by now is beyond such repair by converter means.

[Dr. Harout]

Then why would Sony do such a thing on a semi-professional camera? I am sure they know something that we're not aware of, yet.
I'm not a Sony fanboy, but I do believe that any brand definitely omits steps which might damage their career (well almost), and Sony is one of them.
How about the NEX-7?

[bfitzgerald]

Marketing is at work as well what better chance to get a "jump" over rivals than release the first 24mp APS-C camera?
I think in this case the high ISO performance of the new sensor leaves much to be desired though this could be the new A100 for Sony, nice but ultimately develop a reputation for poor high ISO performance.

The sensor alone could turn off many users.

[agorabasta]

My guess is that they made the flush and readout too fast. Slowing them down may really help, and the Nex version may be better since the speed is not as much of a priority with the slow CDAF.
But then there's that new 60/50p HD video thing that could be also pushing the flush/readout timings over the edge. But then again, they may drive the sensor very differently between stills and video capture.
So I still hope for much better results from the new Nex.

[bakubo]

The noise from the a77 sensor has a badly uneven structure due to a very high density of 'warm' stray pixel values. The density is so high that they cannot be filtered out effectively.
Then the a900 raw also initially suffered from those stray pixels, but due to their lower spatial density it finally became possible to filter them out though advanced algorithms of modern converters. But the a77 raw as it is by now is beyond such repair by converter means.

If a new raw converter can't fix this then that is really unfortunate. I will keep watching to see how things turn out, but this is rather discouraging. An A75 and NEX 6 with the 16mp sensor would be nice.

[agorabasta]

An A75 and NEX 6 with the 16mp sensor would be nice.

Well, Henry, that's a 'Yes&No'. The problem is not the pixel count, and, most likely, it's not the sensor tech.
The real problem is that they had to increase the speed of the silicon while, at the same time, trying to lower the power consumption.
That's the real problem. Without the drastic changes to basic sensor design, lowering power consumption while boosting speed means higher noise. The current sensor tech is at the its basic physical limits. Basta!

But it's not impossible that (alternatively, to the user choice) they could provide some more power-hungry firmware variation without those 60/50p HD video capabilities and sans that electronic first curtain shutter neither sans that stupid 12fps mode. It could deliver far better quality stills, if only the hardware permits such readjustments.

P.S. And btw, trying to lower power consumption and alleviate overheating issues they also added quite a lot of noise into the a35 raw files. And the joke DxO has promptly never noticed that. Deserves considering, uh?

[bakubo]

Today dpreview reprocessed their A77 photo using ACR 6.5. Here is the comparison with the same cameras again:

http://www.dpreview.com/reviews/studioc ... 6658232477

What do you think?

[bakubo]

I have 4 screenshots but they are 1600x900 so too big to upload here and too big to upload to my photobucket.com account. There are 2 before and after screenshots of one part of the dpreview image and then 2 more of another part of the image. If anyone wants to see them let me know or if there is some way to upload them here please tell me. I wish I had thought to crop just the important part before saving. Unfortunately, the previous dpreview image is gone from their site. I saved the screenshots as level 10 jpegs. I could go back and crop them now and resave, but there would be a very negligible, probably totally unseeable, loss when the jpeg is resaved. If you want me to do it I can though.

In one of them the new version has less color saturation, but the noise doesn't look much different. The other one is also not much different.

It seems to me that the new ACR in this particular case changed things very little.

[Greg Beetham]

Henry I took the liberty of replacing the Panasonic 4/3rd camera (hopeless) with a proper camera (Nikon D7000) and selected a different area with dark tones, overall the NEX 5n is very, very good from ISO100-3200, no wonder it gets such a good rep. The Nikon is about as good I think. There was an anomaly with the 60D, it re-arranged the location of the green ball and couldn’t resolve the round pipe at any ISO (despite having the largest RAW files ?), but the 60D seems much less noisy than the A77 just the same, but based on that comparison the A77 is very poor at high ISO, and average at low ISO.
Greg
http://www.dpreview.com/reviews/studioc ... 5373471927

[stevecim]

Hi All

I was looking at these images and noticed something, don't know if it has any effect, I'm sure someone here can tell me
The Sony iso6400 shot was taken at 1/1250 f9 the canon was taken at 1/1250 f8
Same at iso100 Sony 1/20 f9 Canon 1/20 f8 Does 1/3 of stop less light make any real difference?

Can some explain why the Canon crops are zoomed in less the 5N, is the Canon sensor that much smaller than the sony?

[Greg Beetham]

I went back and had a look, I didn't see any shutter speeds? just ISO and f-stop, (maybe you get more info if you download ?) and I wonder why they used different f-stops between cameras anyway?
No stevecim I don't know what's going on there, we know Sony takes up the light loss in the SLT's by bumping the ISO but why that would result in having to stop down an extra stop over the exposure
in the Canon is a mystery too me, one would think that if Sony did the ISO boosting properly the exposures would be the same.
Greg