Ad
Ad
Ad
Pages: « 1 2 [3] 4 5 »   Bottom of Page
Print
Author Topic: "Diglloyd" is testing the Leica S2  (Read 19696 times)
ejmartin
Sr. Member
****
Offline Offline

Posts: 575


« Reply #40 on: July 16, 2010, 10:30:02 AM »
ReplyReply

Quote from: JeffKohn
Just because DSLR's have internal JPEG engine doesn't mean that the raw files are any less raw than from MFDB. I know of no evidence at all to suggest there is any processing happening on the RAW files for Nikon DSLR's. In fact there's plenty of evidence to the contrary. Nikon's CaptureNX RAW converter has things like dynamic lighting and automatic lens correction (for CA, vignetting, etc), and you don't get that functionality unless you use their converter. It's not baked into the RAW.

The only thing I'm aware of on Nikon is the use of NR on long exposures (greater than 1/4 sec on the D3x I believe), and if one uses dark frame subtraction that will of course be performed before the raw file is written.  Other than that, Nikon DSLR raw files are quite linear, and show no obvious signs of noise processing on shorter exposures.  The JPEG engine is used to generate thumbnails, and a full size jpeg is embedded in the raw for the purpose of in-camera review, etc.
« Last Edit: July 16, 2010, 10:32:39 AM by ejmartin » Logged

emil
ErikKaffehr
Sr. Member
****
Offline Offline

Posts: 8015


WWW
« Reply #41 on: July 16, 2010, 12:43:40 PM »
ReplyReply

Hi,

I spent some more time looking at the images I downloaded from the "Diglloyd" site. My initial peek was this morning after just having my first espresso. Initially I was doing some quite heavy sharpening on the Nikon images and very little on the Leica S2. This way I had similar edge sharpness but the Leica was resolving significantly more detail but also had a lot of false detail. The extensive sharpening induced a lot of noise in the Nikon image.

This evening, after having some more espresso, I looked at both images again. This time I sharpened the Nikon image less and applied noise reduction while the Leica image got more sharpening. On the second try the Leica image was much sharper. There is a lot of aliasing artifacts in the S2 image.

So what did I find?

1) Leica definitively resolves much higher (well, it has more pixels, so it should
2) Leica has lots of aliasing artifacts, very much visible on this subject
3) Nikon takes more sharpening and that sharpening may enhance noise
4) Smooth surfaces stay very smooth in the S2 image, on Nikon it takes some more work

Best regards
Erik


Quote from: ejmartin
In what aspects is the Leica result better?  Not being argumentative, just curious.  Can one put the difference down to the AA filter?  That will certainly rob the Nikon of a bit of MTF near Nyquist, while cutting down on aliasing artifacts.
Logged

JeffKohn
Sr. Member
****
Offline Offline

Posts: 1671



WWW
« Reply #42 on: July 16, 2010, 04:51:04 PM »
ReplyReply

Quote from: ejmartin
The only thing I'm aware of on Nikon is the use of NR on long exposures (greater than 1/4 sec on the D3x I believe), and if one uses dark frame subtraction that will of course be performed before the raw file is written.
I'm only aware of one type of long-exposure NR, the one that uses dark frames and is selectable in custom settings. On the D3x it kicks in at 8 seconds. What other NR are you referring to?
Logged

ejmartin
Sr. Member
****
Offline Offline

Posts: 575


« Reply #43 on: July 16, 2010, 05:16:08 PM »
ReplyReply

Quote from: JeffKohn
I'm only aware of one type of long-exposure NR, the one that uses dark frames and is selectable in custom settings. On the D3x it kicks in at 8 seconds. What other NR are you referring to?


http://forums.dpreview.com/forums/read.asp...essage=34309201

and some history:
http://forums.dpreview.com/forums/read.asp...essage=34318456

and AFAIK it cannot be turned off.
« Last Edit: July 16, 2010, 05:16:51 PM by ejmartin » Logged

emil
LKaven
Sr. Member
****
Online Online

Posts: 850


« Reply #44 on: July 16, 2010, 05:38:46 PM »
ReplyReply

Quote from: JeffKohn
I'm only aware of one type of long-exposure NR, the one that uses dark frames and is selectable in custom settings. On the D3x it kicks in at 8 seconds. What other NR are you referring to?
There is said to be an additional hot-pixel suppression algorithm as a part of LENR, and its effects were examined by Marianne Oelund and described here:
http://forums.dpreview.com/forums/read.asp...essage=34309201
Logged

LKaven
Sr. Member
****
Online Online

Posts: 850


« Reply #45 on: July 16, 2010, 05:39:52 PM »
ReplyReply

Ha, while I was looking up the link, Emil posted it, which I didn't see...sorry.
Logged

Guy Mancuso
Sr. Member
****
Offline Offline

Posts: 1123


WWW
« Reply #46 on: July 16, 2010, 06:15:46 PM »
ReplyReply

Quote from: JeffKohn
Just because DSLR's have internal JPEG engine doesn't mean that the raw files are any less raw than from MFDB. I know of no evidence at all to suggest there is any processing happening on the RAW files for Nikon DSLR's. In fact there's plenty of evidence to the contrary. Nikon's CaptureNX RAW converter has things like dynamic lighting and automatic lens correction (for CA, vignetting, etc), and you don't get that functionality unless you use their converter. It's not baked into the RAW.


Jeff maybe my bad choice of words saying processing. More like applying there algorithms at the camera/shooting level whereas the MF backs are doing it outside the cam but in the software.

Someone mentioned C1 is good for other cams besides there own backs and I 100 percent agree. C1 is one of the few that work hard on building profiles for other cams in there software. It is not just a I can see you type deal. They actually build profiles for each cam they support. I use C1 for every cam I have owned but with there own backs it seems they apply there secret sauce algorithms at the software level not in camera. Maybe that was better said. Sorry been a rough past couple days with family members here .
. Okay back to the s2
Logged

rainer_v
Sr. Member
****
Offline Offline

Posts: 1134


WWW
« Reply #47 on: July 16, 2010, 07:30:50 PM »
ReplyReply

dead  pixel and dead lines removement is done i the back via back specific individual reference files, blackframe subtraction removes hot pixels. tgis is either done from time to time for short exposures but  for every frame above a certain exposure length. this processings are independent of the conversion software. so the raw files which come out of the backs have applied these two fundamental processes, independent if 35mm of mf.

sometimes the softwares have some " hidden " functions, means even if you set sharpening at 0 it still may sharpen or even if you set noise reduction at 0 it still may apply some nr.
the files already come out ofvthe back with a maximum of cleanness and also with removed dead or hot pixels and dead lines,- which nearly all sensors have. if this would not be done in the back directly you would see these
in all other converters and no manufactor wants that the e.g.  user sees dead lines. it does not affect the image quality, but everybody would cry to get a "clean" back.

DR can and should be measured from the raw file directly, the software can smoothen the shadows to look better, but it cant increase the DR compared withbthe DR from the rwa directly. although it can make the DR significantly worse, if programmed bad, but thats another story. DXO is doing right to measure the DR right from the raws.
Logged

rainer viertlböck
architecture photographer
munich / germany

www.tangential.de
wildlightphoto
Sr. Member
****
Offline Offline

Posts: 671


« Reply #48 on: July 16, 2010, 08:17:22 PM »
ReplyReply

Quote from: rainer_v
DXO is doing right to measure the DR right from the raws.
I don't hang raw files on the wall.
Logged
ejmartin
Sr. Member
****
Offline Offline

Posts: 575


« Reply #49 on: July 16, 2010, 09:02:39 PM »
ReplyReply

Quote from: telyt
I don't hang raw files on the wall.

Why not?  I imagine it might have a certain postmodern cachet...

     
Logged

emil
Ray
Sr. Member
****
Offline Offline

Posts: 8943


« Reply #50 on: July 16, 2010, 09:11:48 PM »
ReplyReply

Quote from: telyt
I don't hang raw files on the wall.


Yes, you do, if you hang prints of images originating from a digital camera. All prints of digital images are processed RAW files. All digital cameras capture the image initially in some kind of RAW format. From there on, it's all processing of one kind or another and to one degree or another, whether the processing is done inside the camera or outside the camera, whether you are viewing a jpeg or tiff on your monitor, or a print on the wall. Untill the RAW file is processed, you can't even see the image. It's just a pile of numbers.

The bottom line here is, if the image information exists in the RAW file, it can be processed so that you will eventually see it on your wall. If it doesn't exist in the first instance, because of the limitations of the sensor design and capturing processes, then such information cannot be processed and you will not see it on your wall.
« Last Edit: July 16, 2010, 09:21:13 PM by Ray » Logged
Nick Rains
Sr. Member
****
Offline Offline

Posts: 704



WWW
« Reply #51 on: July 16, 2010, 10:37:51 PM »
ReplyReply

Quote from: ejmartin
In what aspects is the Leica result better?  Not being argumentative, just curious.  Can one put the difference down to the AA filter?  That will certainly rob the Nikon of a bit of MTF near Nyquist, while cutting down on aliasing artifacts.

By using the same focal lengths I was expecting to see very little difference between the images as each camera was applying a similar number of pixels to the same field of view ( the pixel pitch is comparable between the two, 5.94 - 6)  It appeared to me that the S2 resolved better detail as well as a subtlety of tone that was lacking from the Nikon. The D3X looked slightly plastic whilst the S2 looked slightly filmic if that makes sense.

[attachment=23187:NIRA0010_2.jpg]

[attachment=23188:_DSC0015.jpg]

I'm a bit reluctant to post 100% crops for all the usual reasons but anyway...shot in a church in Brisbane yesterday. Nikon 24-70 @ 70mm / Leica 70mm. Both at f5.6, MU, tripod etc. Processed in ACR but the relative differences are there in C1 too.

Obviously the S2 will resolve more details if I matched the fields of view, but this is a comparison of the same sensor areas.

I prefer to judge from prints, which I will do in a couple of weeks when I get a 7900. I plan to do some 20x30s and see how the two cameras stack up. I know the S2 will be 'better', no doubt in my mind, it's just  a question of how much better.

What's exciting is just how amazingly good both these cameras are!


ps the top image is from the Leica. Check out the wooden leaf designs at the ends of Christ's arms. To me the Nikon has some jaggies in the curves and the S2 none.
« Last Edit: July 16, 2010, 10:43:44 PM by Nick Rains » Logged

Nick Rains
Australian Photographer
Leica Akademie Instructor
www.nickrains.com
ejmartin
Sr. Member
****
Offline Offline

Posts: 575


« Reply #52 on: July 16, 2010, 11:40:42 PM »
ReplyReply

Quote from: Nick Rains
Check out the wooden leaf designs at the ends of Christ's arms. To me the Nikon has some jaggies in the curves and the S2 none.

The jaggies are demosaic errors.  Any chance I could play with the files?  PM me if interested.
Logged

emil
ErikKaffehr
Sr. Member
****
Offline Offline

Posts: 8015


WWW
« Reply #53 on: July 17, 2010, 12:28:29 AM »
ReplyReply

Hi,

Some cameras do some manipulation on RAW. The Sonya Alpha 900 and the Leica M9 does it for sure. This can be seen by doing autocorrelation on the raw images. DxO does that and have clearly pointed out this on both the Leica M9 and the Sony Alpha 900.

It's quite possible that C1 is better in phase with Phase One backs than others.

Best regards
Erik

Quote from: Guy Mancuso
Jeff maybe my bad choice of words saying processing. More like applying there algorithms at the camera/shooting level whereas the MF backs are doing it outside the cam but in the software.

Someone mentioned C1 is good for other cams besides there own backs and I 100 percent agree. C1 is one of the few that work hard on building profiles for other cams in there software. It is not just a I can see you type deal. They actually build profiles for each cam they support. I use C1 for every cam I have owned but with there own backs it seems they apply there secret sauce algorithms at the software level not in camera. Maybe that was better said. Sorry been a rough past couple days with family members here .
. Okay back to the s2
Logged

wildlightphoto
Sr. Member
****
Offline Offline

Posts: 671


« Reply #54 on: July 17, 2010, 07:10:59 AM »
ReplyReply

Quote from: Ray
Yes, you do, if you hang prints of images originating from a digital camera.

No I don't.  They're processed, not straight from the camera.
Logged
madmanchan
Sr. Member
****
Offline Offline

Posts: 2110


« Reply #55 on: July 17, 2010, 10:25:08 AM »
ReplyReply

There is the data, and then there are the algorithms used to process that data. To the extent that developers measure, characterize, and establish default processing values (i.e., baseline values) for a given sensor, I contend that the end results can be excellent regardless of who makes the raw conversion software. I will frankly admit that for the Phase backs (and in general the medium format backs), Adobe has not historically done all the appropriate steps to get the most out of the files, but we are aiming to change that. So I understand where Guy is coming from with regards to tuned software from the camera maker, but I don't believe this automatically means that results from third-party software necessarily have to be inferior.

On a related topic: Cameras will do varying degrees of processing on the data before writing out the raw file to the card. In practice there is no such thing as an "unprocessed" raw file; some raw files undergo fairly little internal processing by the camera prior to being written out, whereas others undergo quite a bit more. Most cameras for example will at least do in-camera black subtraction and some flat field calibration (e.g., for the two green pixel types). Some cameras will even apply optical configuration-specific processing to the image data before writing out the raw.
Logged

Ray
Sr. Member
****
Offline Offline

Posts: 8943


« Reply #56 on: July 17, 2010, 09:45:49 PM »
ReplyReply

Quote from: telyt
No I don't.  They're processed, not straight from the camera.

Of course they're processed. What isn't? You can't even see a RAW file until it's processed, whether it comes straight from the camera or not. Nor can you see a TIFF or JPEG until the information contained in such file formats has been processed in such a way that a group of phosphorescent dots on your monitor gives the impression of a picture.

By putting that information in the TIFF file through further processing, the luminescent dots on your monitor can be transformed into fine dots of ink on paper.

There's also a type of processing that takes place prior to the RAW image being written to the camera's memory card. It's the quality of the results from that processing, which is largely due to a camera's design and sophistication of componentry, that DXO are testing.

It's standard scientific procedure when testing for a specific quality to exclude all variables that may affect the accuracy of the results, if possible, or at least keep such variables constant.

The amount of processing that takes place between the stage when a RAW file is written to memory, and the stage when such information contained within that RAW file is transformed into dots of ink on paper, is huge and varied. It involves not only the sophistication of RAW converters, interpolation and sharpening algorithms, noise reduction software, monitors and calibration devices, editing programs such as photoshop, but also another major piece of hardware called a printer, with its additional variables of ink, paper and profile qualities.

Add to that the even more variable factors, such as human skill and subjective taste which can also be involved at each stage of such processing to some degree, then it makes complete sense for DXOmark to directly test the quality of the RAW file. With such results in hand, one is then in a better position to compare other factors in the post-RAW processing chain, such as RAW converters, for example.
Logged
wildlightphoto
Sr. Member
****
Offline Offline

Posts: 671


« Reply #57 on: July 17, 2010, 11:09:54 PM »
ReplyReply

Quote from: Ray
Of course they're processed. What isn't? You can't even see a RAW file until it's processed, whether it comes straight from the camera or not. Nor can you see a TIFF or JPEG until the information contained in such file formats has been processed in such a way that a group of phosphorescent dots on your monitor gives the impression of a picture.

By putting that information in the TIFF file through further processing, the luminescent dots on your monitor can be transformed into fine dots of ink on paper.

There's also a type of processing that takes place prior to the RAW image being written to the camera's memory card. It's the quality of the results from that processing, which is largely due to a camera's design and sophistication of componentry, that DXO are testing.

It's standard scientific procedure when testing for a specific quality to exclude all variables that may affect the accuracy of the results, if possible, or at least keep such variables constant.

The amount of processing that takes place between the stage when a RAW file is written to memory, and the stage when such information contained within that RAW file is transformed into dots of ink on paper, is huge and varied. It involves not only the sophistication of RAW converters, interpolation and sharpening algorithms, noise reduction software, monitors and calibration devices, editing programs such as photoshop, but also another major piece of hardware called a printer, with its additional variables of ink, paper and profile qualities.

Add to that the even more variable factors, such as human skill and subjective taste which can also be involved at each stage of such processing to some degree, then it makes complete sense for DXOmark to directly test the quality of the RAW file. With such results in hand, one is then in a better position to compare other factors in the post-RAW processing chain, such as RAW converters, for example.

You're preaching to the choir.  I write engineering software for a living.  However, when I want to see what camera will produce the best prints I look at prints, not raw files.  Raw files are only an intermediate step.
Logged
Rob C
Sr. Member
****
Offline Offline

Posts: 12213


« Reply #58 on: July 18, 2010, 03:57:22 AM »
ReplyReply

Quote from: telyt
You're preaching to the choir.



And that's the danger with these sorts of topics: only the committed have the energy to stick around for long, and I doubt that includes many people who actually go out on a regular basis and make photographs happen, though I accept that there will be exceptions.

So why am I still here? Hope beats long in the human breast, even as the waves close over one's head. Not, of course, that anyone having just drowned has come back momentarily (in my admittedly limited experience) to prove this belief. But on balance, it's an idea that seems to live happily enough with what might be the reality. Much as with the multifarious, disparate photographic beliefs, you might say; a broad church indeed.

Rob C

Logged

tho_mas
Sr. Member
****
Offline Offline

Posts: 1697


« Reply #59 on: July 18, 2010, 06:36:42 AM »
ReplyReply

Quote from: madmanchan
(...) but I don't believe this automatically means that results from third-party software necessarily have to be inferior.
true. For instance RAW Developer basically does a nice job even with Phase One files. And also with S2 files.
And vice versa Capture One does a nice job with files from DSLRs ...
Logged
Pages: « 1 2 [3] 4 5 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad