Home

Please support mflenses.com if you need any graphic related work order it from us, click on above banner to order!
SearchSearch MemberlistMemberlist RegisterRegister ProfileProfile Log in to check your private messagesLog in to check your private messages Log inLog in

NEX-7: dpreview plays down noise issue
Post new topic   Reply to topic
Home   Official News and Announcements
View previous topic :: View next topic  


PostPosted: Fri Dec 16, 2011 10:39 pm    Post subject: NEX-7: dpreview plays down noise issue Reply with quote
http://www.dpreview.com/reviews/sonynex7/page16.asp Look at the RAW graph

I have expected a lot, but not THAT:

"high ISO noise performance can't match the competition"

"the E-P3's high ISO output now looks distinctly weak against the latest APS-C cameras"


PostPosted: Sat Dec 17, 2011 2:03 am    Post subject: Reply with quote
Of course noise is higher than previous NEX models, the 7 has a smaller pixel pitch and Sony are just trying to workaround the too small pixel pitch with aggressive noise reduction. The pixel pitch of the NEX-7 is 3.91, which compares very unfavourably to the EOS 5DMkII at 6.4 or even the NEX-3 at 4.77. Packing ever more pixels into the APS-C sensor just doesn't work with current technology as it negatively impacts noise and dynamic range too much.

http://www.dxomark.com/index.php/News/DxOMark-news/Sony-NEX-7-comparisons-and-review

Some interesting comparisons there.

Luminous Landscape does an interesting review of the NEX-7 too:

http://www.luminous-landscape.com/reviews/cameras/sony_nex_7_rolling_review.shtml#28


PostPosted: Sat Dec 17, 2011 7:16 am    Post subject: Reply with quote
Actually it isn't. If you compare the noise patterns of a 24 MPix cam with those of a 14 MPix cam, you need to downsample the higher-resolution image (or enlarge the other one). If you do that, the noise patterns are pretty similar. And the dynamic range of the NEX-7 is said to be excellent by many sources. How does that go together?

Yes, a higher pixel pitch theoretically means more noise, but we still are not at the problematic pixel pitch of regular compact cam sensors! So, Sony could cope with the 24 MPix on APS pretty well - much better than with the fullframe A900 by the way!
But newer and more sophisticated algorithms reduce visible noise - and I do not refer to noise reduction now but to D/A conversion.

Here is another review of the NEX-7 in which the cam excels: http://www.popphoto.com/gear/2011/12/camera-test-sony-nex-7-ilc

To be honest, I do not trust either of them. I comment on the noise the NEX-7 produces when I have used one!


PostPosted: Sat Dec 17, 2011 9:36 am    Post subject: Reply with quote
The RAW noise of the Nex 5 and 3 cameras seems the lowest, but after I read from Poilu that Sony applies noise reduction in RAW files too, I am perplexed.
Of course it is not fair to compare the RAW noise of a Nex-3 or Nex-5, which is processed, with that of a, say, Canon 7D, which is not processed.
This type of difference is a fundamental one, yet reviews barely mention that, when not at all. I feel it's a bit of cheating from Sony's side.


PostPosted: Sat Dec 17, 2011 10:38 am    Post subject: Reply with quote
What counts is what come out at the end. Wink

The NEX-5N deals with noise much better than my NEX-3 and it even does than the NEX-7 - as long as you keep the original file size.
If you resample the NEX-5N files to the size of the NEX-7 files, things look different.

Actually, trying to compare which cam offers less noise at ISO6400 is kind of pixel-peeping.
The images that the NEX-7 produces, as well as those of all the other NEXes, are perfectly usable at ISO1600 even at ISO3200.
What else do I want?

Combined with the extreme high resolution, great built and useful handling of the NEX-7 this offers everything I could wish for.

In case of a "low-light emergency" you can go up to what? 16000??? And you can shoot in RAW which offers you all possibilities.

With my Canon EOS cams I often shoot in RAW, almost always, with my NEX-3 I hardly do. The JPG files are visually extremely nicely processed. And I guess that will be true for the NEX-7 as well. I will report as soon as I have one. Wink


PostPosted: Sat Dec 17, 2011 11:34 am    Post subject: Reply with quote
LucisPictor wrote:

With my Canon EOS cams I often shoot in RAW, almost always, with my NEX-3 I hardly do. The JPG files are visually extremely nicely processed.


But the Sony cameras do process the RAW files also. This is the point. I do not like this, because I want to process my RAW files the way I like them. If I take an important photo, I am ready to work on it even one hour, to make it come out the way I want it. If the RAW file is clean, it will have more noise to start with, but I will be able to custom-denoise it just the way I want, just as much as I want and where I want.
If instead I receive a processed RAW file, I have to take what's been done by the camera, take it or leave it. Maybe I would not want to denoise some parts of the image, but the camera did it anyway. Maybe I want to use different amount of denoise in different parts, but the camera already denoised the image the same way all over it.
That's a kind of limited control that I do not wish to have when shooting RAW.


PostPosted: Sat Dec 17, 2011 12:50 pm    Post subject: Reply with quote
denoising algorithms get better every year
when I denoise old 800 iso negative film, I am surprised by the progress made in few years
they evolve from Gaussian blur who make everything soft to today's algorithms who differentiate dark parts from lighten or use fft on selected frequencies
keeping raws mean that when I will see them in 10 years, I will use the best algorithms not yet made today
cooked raw files is a problem as most of the details are already cleaned and will not be recovered


PostPosted: Sat Dec 17, 2011 1:25 pm    Post subject: Reply with quote
Orio wrote:
The RAW noise of the Nex 5 and 3 cameras seems the lowest, but after I read from Poilu that Sony applies noise reduction in RAW files too, I am perplexed

the reviewer notes that corrections are applied before Raw are written, so this is not Raw anymore; he say that it is better to let all the NR reduction to off but I read that NR is also made in dsp and cannot be turned off

the reviewer of Dpreview also note that photoshop apply noise reduction even with NR off and by a different amount by camera
in few years, it will be difficult to compare camera with all the added soft touch

Andy Westlake wrote:
Adobe does a degree of noise reduction even when the user-controlled NR is turned off.
The amount of NR applied 'under the hood' is not high, but it does vary by camera (Adobe is attempting to normalize output across different sensors), so we are still looking at a balance of noise and noise reduction, rather than pure noise levels


Andy Westlake wrote:
Any corrections that are enabled in-camera are also applied by Sony's Image Data Converter to its RAW conversions, but somewhat unexpectedly vignetting compensation also affects Adobe Camera Raw. This indicates that it's applied before the RAW file is written, so RAW shooters may prefer to turn it off


Last edited by poilu on Sat Dec 17, 2011 1:46 pm; edited 1 time in total


PostPosted: Sat Dec 17, 2011 1:45 pm    Post subject: Reply with quote
It depends on how much the RAW file is "preprocessed". I agree that a pure RAW file would be best.
BTW there is always some "softening" due to the BAYER algorithm and still the results are amazing and will be amazing in 10 years.

We shouldn't be too negative about it.


PostPosted: Sun Dec 18, 2011 6:25 pm    Post subject: Reply with quote
LucisPictor wrote:
We shouldn't be too negative about it.

No pun intended? Wink


PostPosted: Sun Dec 18, 2011 7:10 pm    Post subject: Reply with quote
LucisPictor wrote:
BTW there is always some "softening" due to the BAYER algorithm....

I'm not sure what you are trying to say here. What are you softening ? It's not the image because there isn't one without the Bayer filter - and subsequent de-Bayering of course.


PostPosted: Wed Dec 21, 2011 5:37 pm    Post subject: Reply with quote
LucisPictor wrote:
Actually it isn't. If you compare the noise patterns of a 24 MPix cam with those of a 14 MPix cam, you need to downsample the higher-resolution image (or enlarge the other one). If you do that, the noise patterns are pretty similar. And the dynamic range of the NEX-7 is said to be excellent by many sources. How does that go together?


Nobody would buy a 24MP camera just to downsample the images to 14MP or 16MP.
The sensor should be judged at the native resolution. Else, it is pointless.

And this is one reason I prefer the NEX 5N against the NEX 7: I think 24MP is too much for an APS-C sensor...

LucisPictor wrote:
It depends on how much the RAW file is "preprocessed". I agree that a pure RAW file would be best.
BTW there is always some "softening" due to the BAYER algorithm and still the results are amazing and will be amazing in 10 years.


Are you sure you can tell what will be regarded amazing in 10 years? I am not...

poilu wrote:

keeping raws mean that when I will see them in 10 years, I will use the best algorithms not yet made today
cooked raw files is a problem as most of the details are already cleaned and will not be recovered

Are you sure they do this? Any link/evidence?


PostPosted: Wed Dec 21, 2011 8:39 pm    Post subject: Reply with quote
nkanellopoulos wrote:
And this is one reason I prefer the NEX 5N against the NEX 7: I think 24MP is too much for an APS-C sensor...


I agree, 24mp is probably too much.


PostPosted: Wed Dec 21, 2011 8:52 pm    Post subject: Reply with quote
Ian wrote:
I agree, 24mp is probably too much

according to this discussion, Sony will give us 100 mega pixels soon
Sony wrote:
One hundred mega pixels not a strange thing soon


nkanellopoulos wrote:
Are you sure they do this? Any link/evidence?

just read the conclusion of the test

Andy Westlake wrote:
Any corrections that are enabled in-camera are also applied by Sony's Image Data Converter to its RAW conversions, but somewhat unexpectedly vignetting compensation also affects Adobe Camera Raw. This indicates that it's applied before the RAW file is written, so RAW shooters may prefer to turn it off


PostPosted: Wed Dec 21, 2011 9:06 pm    Post subject: Reply with quote
poilu wrote:
Ian wrote:
I agree, 24mp is probably too much

according to this discussion, Sony will give us 100 mega pixels soon
Sony wrote:
One hundred mega pixels not a strange thing soon


Oh no! I thought the megapixel war was over...


PostPosted: Wed Dec 21, 2011 9:50 pm    Post subject: Reply with quote
nkanellopoulos wrote:
poilu wrote:
Ian wrote:
I agree, 24mp is probably too much

according to this discussion, Sony will give us 100 mega pixels soon
Sony wrote:
One hundred mega pixels not a strange thing soon


Oh no! I thought the megapixel war was over...


100MP is useless if it means tiny pixels with tons of noise and low dynamic range.


PostPosted: Wed Dec 21, 2011 10:23 pm    Post subject: Reply with quote
Ian wrote:
100MP is useless if it means tiny pixels with tons of noise and low dynamic range

100MP may look big with today technologies but tomorrow it will look ordinary
next year, the Olympics will be recorded at 7.680×4.320, that means 33MP by frame
this the resolution that standard TV will have in 10 years
http://techcrunch.com/2011/11/23/japans-nhk-to-broadcast-london-olympics-in-ultra-hd-7680%C3%974320-pixels-video/

and if you want the 100MP technologies from today, just shot with the Ektar negative who resolve 154lp/mm, that correspond to 82MP


PostPosted: Thu Dec 22, 2011 1:36 am    Post subject: Reply with quote
nkanellopoulos wrote:

Nobody would buy a 24MP camera just to downsample the images to 14MP or 16MP.
The sensor should be judged at the native resolution. Else, it is pointless.


If your typical viewing of photos happens at 100% crop, then you are right. Otherwise, you are completely wrong: images should be viewed at their final output size (e.g. so that the frame fills the computer screen, or printed on paper at the same size).

As a film analogy, this is like saying “if you use film capable of capturing more details and you wish to compare its grain to other films, you must use a higher magnification loupe to view the higher resolution film and then compare it to lower resolution film viewed at lower magnification”. Of course the higher resolution film (or sensor) is then at a disadvantage; indeed, viewing at “native resolution” cancels out any resolution advantage exactly so for this kind of viewing it would be pointless to make (or buy) a camera with resolution higher than a computer screen (or whatever resolution is sufficient to make the size of print that a person is interested in).

(I think the film comparison makes the silliness of this kind of comparison obvious, but unfortunately digital has this “native resolution” that some people think are the intended viewing resolution. Personally I think DSLRs should be made with 100 mpix sensors without any antialising filter and then the camera should natively convert the GRGB bayer image to a 25MP RGB image and present that as the “raw”. The true 100MP raw would only be available by taking a test about understanding of how digital sensors work and signing an argeement not to make statements about the “native resolution”. =)


PostPosted: Thu Dec 22, 2011 8:09 am    Post subject: Reply with quote
Arkku wrote:

If your typical viewing of photos happens at 100% crop, then you are right. Otherwise, you are completely wrong: images should be viewed at their final output size (e.g. so that the frame fills the computer screen, or printed on paper at the same size).


Usually people don't take photos to use them at only one resolution. When I take a photo I want to be it the best possible image quality, so that if I want or need to print it large, I can. Then maybe I decide later to only see it on the monitor, or even to delete it. But this type of decision is normally taken after having taken the photograph, not before.


PostPosted: Thu Dec 22, 2011 11:30 am    Post subject: Reply with quote
Orio wrote:
Arkku wrote:

If your typical viewing of photos happens at 100% crop, then you are right. Otherwise, you are completely wrong: images should be viewed at their final output size (e.g. so that the frame fills the computer screen, or printed on paper at the same size).


Usually people don't take photos to use them at only one resolution. When I take a photo I want to be it the best possible image quality, so that if I want or need to print it large, I can. Then maybe I decide later to only see it on the monitor, or even to delete it. But this type of decision is normally taken after having taken the photograph, not before.


+10

Yesterday I ordered 50X75 cm prints of photos I took more than a year ago.


PostPosted: Thu Dec 22, 2011 2:18 pm    Post subject: Reply with quote
Orio wrote:

Usually people don't take photos to use them at only one resolution. When I take a photo I want to be it the best possible image quality, so that if I want or need to print it large, I can. Then maybe I decide later to only see it on the monitor, or even to delete it. But this type of decision is normally taken after having taken the photograph, not before.


Of course; this is why higher resolution sensors are better: they add more possibility for cropping and more flexibility for choosing the print size (for digital viewing we are currently at a point where the megapixel count does not matter unless cropped, but when we start getting 300dpi displays on our computers then it will again be relevant).

My point was that since what you said is true, it is also why photos from different sensor resolutions should not be judged at “native resolution”; doing so is like saying that you do not want high resolution but rather that you only want to shoot at the one resolution (native of the lower resolution sensor) because this rigs the comparison to always prefer that resolution.

(As another film analogy, if you were comparing prints, comparing at “native resolution” would be exactly like printing an image taken with a higher resolution film at a larger size and an image with lower resolution film at a smaller size and then judging the amount of visible grain by viewing both prints at the same distance—if someone did this comparison and posted it here, I don't think many would agree that it is a valid assessment of film quality. So why do people think this is a good way to compare digital images?)

If you want the best overall quality at the higher resolution, then you must also resize the lower resolution image up to that higher resolution—why would your needs for quality be different with a lower res camera? In practice, though, most uses and users do not require quality at the native resolution of current high res sensors, so the more common and practical suggestion is to downsample the higher res (or both) image(s). Ideally both images would be sampled up or down to your actual quality requirement; if you require the best possible resolution then the size should be the resolution of the highest res camera being compared to.)

Last edited by Arkku on Thu Dec 22, 2011 6:27 pm; edited 3 times in total


PostPosted: Thu Dec 22, 2011 2:24 pm    Post subject: Reply with quote
nkanellopoulos wrote:

Yesterday I ordered 50X75 cm prints of photos I took more than a year ago.


And if your camera had had 10 more megapixels, would you have ordered a 60×90cm print just because you had more megapixels? If yes, then you should compare your current camera to a higher res camera by upsampling to the higher resolution. If no, then you should compare your current camera to a higher res camera by downsampling to the lower res, or by downsampling both to a common lower resolution, or by printing both at 50×75cm which is the same thing.

Comparing both at native resolution is like you printed one at 50×75cm and the other at 60×90cm and then complained about the larger image having more visible noise at the same viewing distance. (Unless the higher res camera was sufficiently more advanced, or both cameras good enough for 60×90cm at your viewing distance.)

Last edited by Arkku on Thu Dec 22, 2011 8:04 pm; edited 1 time in total


PostPosted: Thu Dec 22, 2011 5:13 pm    Post subject: Reply with quote
Agreed with Arkku.

The NEX-7 looks pretty good to me here: http://www.fredmiranda.com/forum/topic/1070774/
Very usable at least.


PostPosted: Sat Dec 24, 2011 12:25 pm    Post subject: Reply with quote
More 100% crops: http://www.fredmiranda.com/forum/topic/1071010 (click big image to zoom)
Now complain that the pixel level detail isn't good enough... Wink


PostPosted: Thu Dec 29, 2011 11:14 am    Post subject: Reply with quote
I found this topic very interesting :
http://petersills.wordpress.com/2011/12/08/sony-nex-7-a-week-in-havana/

Viewing this and luminous-lanscape review, undoubtly this Nex-7 is a great improvement in CSC market.
Post new topic   Reply to topic
Home   Official News and Announcements
Page 1 of 1