The Noise thing has me bothered.

[wayne_s]

"People" ... thinly veiled reference to me I'm sure :ROTFL: Evidently no critical observations are allowed here.

Why is it that all this has to be "for or against?" :wtf: Or did Jack and Guy decide that this Sony forum is only for "mutual admiration"?

Objective criticism concerning performance of this relatively new camera is what leads to possible solutions isn't it?

Now some folks are experimenting with post techniques to lessen that effect which I am pretty certain came from reacting to observations concerning the noise issue. Without objective critical comments nothing gets done, or gets done slowly. It's the same process many Zeiss lovers went through with the Contax ND which was a LOT more plagued with issues ... many of which got solved, albeit too late (I say that because I see this camera as the natural successor to the ND which also featured AF Zeiss lenses).

When I said that IMO the true test is prints not 100% screen shots, I meant exactly that. IMO, we should evaluate it that way because it's the end intent that counts. At that time I hadn't printed all that much from this camera.

That was then. This is now.

I've now printed over 400 A900 files, some at 17X22, most at 8X10 ... but some of those are crops, with a few severe crops.

Noise shows up in prints also.

Most certainly not as bad as the 100% screen grabs show ... but it's there to contend with. And depending on the ambient conditions, it isn't just in high ISO files either. It manifests itself differently at different luminance levels. Just to mention one thing I've noticed, it can adversely affect the subtile OOF transitions of the bokeh that the Zeiss lenses deliver because of the more abrupt noise transitions.

To be clear, I am not looking for NO noise, and in some conditions the noise is perfectly acceptable even more film like than other cameras deliver. It's the other real world situations where blotchiness, uneven noise levels, less ability to lift shadow areas and so on ... is in need of attention.

Some of that can be addressed and is being addressed ... and some of it we'll learn to live with. That will help all of us define the boundaries of this camera to maximize its performance and avoid it's shortcomings.

However, implying there are no shortcomings is useless internet chatter of zero value to anyone IMHO.

I also don't think we make this camera better by trashing other really good ones ... it just makes it all look very defensive and lacks credibility.

Marc,
This last sentence makes me laugh. You trashed the 1ds3 and 5d2 canon cameras as producing plasticy looking files without any example comparison shots. If I look at the unbiased noise test the 5d2 competes very well with the D3x and the 1ds3 competes well up to 1600. I think there needs to be alot more good fair example shots and lot less nebulous bs about this camera's files are crap and this one's files are superior without showing any good examples. Lloyd Chambers of Digilloyd is saying crap like D3x is another league compared to the 5d2/1ds3 and when I look at his few comparison shots on the blog they are bad/faulty comparison shots.Unbiased view points where the reviewer doesn't have a gain from his connection with the camera gear maker is very rare these days. That is why I go to forums like this to get my information. All three cameras are very good and have strengths and weaknesses and we all have different requirements. But I don't mind learning about the weaknesses of my camera relative to another if I am presented with fair comparison shots which show it. Thanks.

 

[douglasf13]

That's why I've become so Iliah Borg-centric. He's a master at details and testing, and he owns all of these cameras, so there is little emotion there. At the end of the day, though, all of these cameras are outstanding.

 

[Eoin]

I don't know what all the fuss is about, "digital noise" (a catch all term) is a fact of life in digital photography as it was with film. Give it another couple of years and we'll have noise-less ISO 12800 files with 10 stops of dynamic range at that ISO.

If you're shooting for a living, then by all means and quite reasonably you may have different needs or wants. Odds are you'll have a couple of systems to choose from to meet your or your customers needs.

In the mean time, for me at least, I'm reminded every time I open a file from the a900 in Aperture as to how good this £1600 full frame 24mp camera actually is. Printing output is also excellent. I may only have an amateurs perspective on all of this, but the a900 with the Zeiss lenses will see me through for quite a few years. It's a simple, capable, cost effective system.

Nirvana in digital photography does not exist, yet!.

 

[fotografz]

Marc,
This last sentence makes me laugh. You trashed the 1ds3 and 5d2 canon cameras as producing plasticy looking files without any example comparison shots. If I look at the unbiased noise test the 5d2 competes very well with the D3x and the 1ds3 competes well up to 1600. I think there needs to be alot more good fair example shots and lot less nebulous bs about this camera's files are crap and this one's files are superior without showing any good examples. Lloyd Chambers of Digilloyd is saying crap like D3x is another league compared to the 5d2/1ds3 and when I look at his few comparison shots on the blog they are bad/faulty comparison shots.Unbiased view points where the reviewer doesn't have a gain from his connection with the camera gear maker is very rare these days. That is why I go to forums like this to get my information. All three cameras are very good and have strengths and weaknesses and we all have different requirements. But I don't mind learning about the weaknesses of my camera relative to another if I am presented with fair comparison shots which show it. Thanks.

I did post comparison shots ... even though I actually don't think they prove anything. Their only value is in stimulating conversation ... which in this case it did, and a thread about different techniques to deal with A900 files is underway ... wahoo. less pissing contests and more useful info.

Guilty as charged concerning the Canons BTW ...

I'm not Digilloyd ... but having owned the Canons and now the D3X, I'd have to agree with him no matter what you think of his comparison shots. I don't necessarily trust comparison shots that are sub one meg jpg compressed sRGB files viewed on a rear lit computer screen anyway ... even though I sometimes post them myself :ROTFL: I trust what comes off my printer ... :thumbup:

 

[fotografz]

Well, Marc, it wasn't meant in a malicious way I respect your work and opinion and any commenst are given in a friendly way like every one else on this forum.

But my point is, it's the very reason that gives us these wonderful low iso files that cause the blotchy chroma noise at high iso. We all know it's the dense CFA and the lack of noise reduction at the hardware level (and it seems both Canon and Nikon use different CFA densities, and do apply NR at the hardware level). It's a fact and a trade off we have to live with. I understand that this chroma noise is no good for wedding work. But for me, as Quentin and Peter said, I prefer the non-processed look of the A900 high iso files, and find it more appealing to remove most if all the chroma noise in post without getting that processed look that is so evident in the 5D2 files provided by Georg.

Perhaps Sony can do something about it in a firmware update, but I have the impression that anything done on the CFA density (in future models) or in camera processing will inevitably have an unwanted effect on the lovely low-iso photos.

Totally agree. I still suspect we aren't getting all this camera is capable of and if we keep flogging it it'll get better :thumbs: I'd hazard a quess that what we're doing with it now is better than a month or two ago ... at least my stuff is looking better than the first shots.

 

[Greg Seitz]

For those interested DXO has a case study of the Nikon D5000 and Canon 500D. While not directly relevant to this discussion it provides lots of insight on how to read and interpret their detailed info for each camera.

http://www.dxomark.com/index.php/eng/Insights/Canon-500D-T1i-vs.-Nikon-D5000

There is plenty of info when you dig down into each cameras specifics (rather than the high level overviews). You can take this knowledge and apply it to any camera in their database and learn quite a bit about how each sensor behaves.

I have found that what I see in their charts mimics very closely what I've found in real life as I've compared various cameras.

A couple of tidbits I've gleaned for instance by looking at the details for the A900, D3X, and 5D2:

The color response (i.e the Color Filter Array behavior) of the D3X and A900 is almost identical and very good at separating the 3 color channels. The 5D2 does not have the clean separation that the other two have in particular in the red channel. Unfortunately the extra noise in the Sony throws away that advantage and the Canon performs slightly better overall. The Nikon retains it's advantage due to its low noise and bests the Canon and the Sony.

See the color sensitivity tab on this page:
http://www.dxomark.com/index.php/en.../0/(brand)/Sony/(brand2)/Nikon/(brand3)/Canon

Read noise: the D3X is far better than both the Canon or the Sony. This is one of the reasons it has significantly greater dynamic range at low ISOs than the other two. In the higher ISO range the Sony can't keep up (by about 1 stop) vs the other two and the Canon shows more of its potential since read noise becomes less of an issue relative to the photon noise and the Canon pulls very slightly ahead of the D3x from ISO 1600 on.

Again, what's interesting is that I have found the results they publish do indeed correspond to my own findings for noise and dynamic range for cameras I've owned so I tend to give credence to them for cameras that I haven't used. It's not the be all end all of measurements but for the areas they do cover they seem to be realistic on a relative basis.

 

[douglasf13]

Color sensitivity graphs that DxO Mark use are not the same as the spectral properties that I've been mentioning. I'm not talking about how many shades of green that we see in Sony vs. Nikon, but rather how much resolution is in the greens, due to the narrow CFA spectrum. The trade off is that these denser CFAs require more amplification, and that's a big reason why our Sony files are noisier than others. D2x had similar "issues," and there's no free lunch. The difference in color range between these three cameras, like many of the DxO mark results, are withing the margin of error, and basically negligible. The difference in spectral properties is a different story, and both Canon and Nikon are trending towards high ISO performance over dense CFAs, it appears. Not better or worse, just different. Marc is in great shape, because he is able to use all kinds of tools in different situations, as each has its strengths/weaknesses.

 

[wayne_s]

Thanks Douglas for clearing that up, because I saw both the tonal and color sensitivity graphs are pretty similar between the three. The big difference is like Greg mentions is the Dynamic Range graph at low iso where the D3x has a clear advantage. But since the S/N graphs looks similar between the three that this extra Dynamic Range is a benefit of the D3x reading the sensor twice in order to lower the Read noise. Which comes at the sacrifice of FPS. It would be nice to see the difference made with this mode disabled in the D3x.
It also interesting that the final Dynamic Range metric # for a camera is highly weighted to its sub 200 iso performance. This really makes the difference in the final DXO mark being higher than the 5d2 even thought the Dynamic Range of the 5d2 beats the Dx3 at higher iso.
Also since the sensor S/N ration is very similar between the three, the poorer high iso performance of the A900 must be a result of only having 12-bit A/D's instead of 14 and maybe less noise reducing going on in camera? Just wanted to know what people think about that. Maybe it is only the dense CFA thing that you mention.

 

[wayne_s]

Sorry, the other thing I wanted to point out is that the Dynamic Range graph doesn't show the A900 having more dynamic range than the 5d2 except a tad more at ISO 200.
This doesn't seem to jive with some of the reviews and comments on forums that the A900 has a stop more dynamic range than the 5d2.

 

[Greg Seitz]

Color sensitivity graphs that DxO Mark use are not the same as the spectral properties that I've been mentioning. I'm not talking about how many shades of green that we see in Sony vs. Nikon, but rather how much resolution is in the greens, due to the narrow CFA spectrum. The trade off is that these denser CFAs require more amplification, and that's a big reason why our Sony files are noisier than others. D2x had similar "issues," and there's no free lunch. The difference in color range between these three cameras, like many of the DxO mark results, are withing the margin of error, and basically negligible. The difference in spectral properties is a different story, and both Canon and Nikon are trending towards high ISO performance over dense CFAs, it appears. Not better or worse, just different. Marc is in great shape, because he is able to use all kinds of tools in different situations, as each has its strengths/weaknesses.

Not talking about the graph. They have a page devoted to spectral response for the individual cameras. Pull up the details for a single camera such as:

http://www.dxomark.com/index.php/eng/Image-Quality-Database/Sony/Alpha-900

and then along the tabs at the top there is a right arrow, hold your mouse over that until the "Color Response" tab shows and click on it which takes you to graphs for the color response for each channel.

 

[Greg Seitz]

Sorry, the other thing I wanted to point out is that the Dynamic Range graph doesn't show the A900 having more dynamic range than the 5d2 except a tad more at ISO 200.
This doesn't seem to jive with some of the reviews and comments on forums that the A900 has a stop more dynamic range than the 5d2.

Actually I'm finding not much between them myself. I think the perception of more DR on the Sony is that it tends to meter a bit darker for some scenes and that converters such as Adobe push up the exposure transparently behind the scenes by differing amounts for different cameras. I have found that my Nikon D5000 does indeed have more range than either which does jive with what DXO shows as well.

Oh, and regarding noise reduction in raw, according the DXO some of the Pentax cameras do it and the Sony does at high ISO (they may have been using it as an example with NR on but it's not clear). No specific mention of the 5D2 or D3X is made and they generally call that out if it appears to be taking place like they did with the Pentax models. Here's some info on how they determine if a camera is doing noise reduction in the raw files:

http://www.dxomark.com/index.php/eng/Insights/Half-cooked-RAW

 

[douglasf13]

Thanks, Greg. I'll have to look at those when I get home. We are getting to the limit of my understanding, and we really need Iliah on this forum to help explain things. I do know that Iliah doesn't have a high opinion of DxO Mark, and the spectral response that he tested is similar, but not identical, in all Sony cameras. Maybe Andrey (hardloaf) can further the conversation from this point?

 

[nostatic]

not to delve too deeply into semantics, but the reality is there is no such thing as an "unbiased" test. Every test has some inherent bias, and there will be artifacts.

The significance of the bias (or better still, recognition of it) is what is important to consider. But just as there really is no such thing as a purely unbiased documentary (even a surveillance camera has bias), no assay is free from a POV either. And perhaps even more dangerous is the application of machine tests for tools used in the creation of art. The extrapolations made from these numbers are a never ending source of amusement.

 

[wayne_s]

Actually I'm finding not much between them myself. I think the perception of more DR on the Sony is that it tends to meter a bit darker for some scenes and that converters such as Adobe push up the exposure transparently behind the scenes by differing amounts for different cameras. I have found that my Nikon D5000 does indeed have more range than either which does jive with what DXO shows as well.

Oh, and regarding noise reduction in raw, according the DXO some of the Pentax cameras do it and the Sony does at high ISO (they may have been using it as an example with NR on but it's not clear). No specific mention of the 5D2 or D3X is made and they generally call that out if it appears to be taking place like they did with the Pentax models. Here's some info on how they determine if a camera is doing noise reduction in the raw files:

http://www.dxomark.com/index.php/eng/Insights/Half-cooked-RAW

Interesting, Nikon's "trick" to raise the Dynamic Range in the low iso area is not there in the D3 but is in the D3x and now the D5000 like Greg mentions.
I would like to know more how Nikon is doing this. Is it just multiple sensor reads?

I didn't understand why DXO is doing all their measurements only on the green channel. I thought all Bayer pattern sensors would have more noise on the blue channel, right?When camera's sharing the same sensor like the D3x/A900 and 5D2/1DS3 have quite different high iso capabilities, I think alot of that is due to some noise removal going on. I don't think it is just the difference in CFA densities.

I agree that these DXO tests are probably not perfect either but it is just another piece of information which can hopefully corroborate with what people see.