Ad
Ad
Ad
Pages: [1] 2 3 ... 6 »   Bottom of Page
Print
Author Topic: DXO now has MF cameras  (Read 29298 times)
lisa_r
Guest
« on: February 03, 2009, 08:38:21 AM »
ReplyReply

http://www.dxomark.com/index.php/eng/Image...d3)/Phase%20One

Here are some DR results for the a900, 5D2, P45+
http://www.dxomark.com/index.php/eng/Image...d3)/Phase%20One

« Last Edit: February 03, 2009, 08:44:40 AM by lisa_r » Logged
michele
Full Member
***
Offline Offline

Posts: 230


« Reply #1 on: February 03, 2009, 08:56:50 AM »
ReplyReply

I'm sure the numbers are right, but my experience and my daily work do not accept this... As Michael wrote, one thing is the test and one thing is the experience...
Oh they have to take my Phase back from my dead cold hands!
Logged

lisa_r
Guest
« Reply #2 on: February 03, 2009, 08:58:28 AM »
ReplyReply

Quote from: John Schweikert
Look at the DXOMark numbers for Aptus 75s, P45+ and Canon 5DII. It has the Canon with higher dynamic range (actual number of stops). That's the biggest bunch of bullshit I have ever seen.

I don't use these camera side by side, but when I shot a Phase p 30 (I think) and the 1Ds3 in a controlled test with strobes in the studio last fall, I did not see any appreciable difference in DR. This was shooting RAW and trying various converters. Even hunting around in shadows and using + exposure looking at the "shadow detail", there was very little difference in detail. The Canon did show more noise when boosting the shadows 2 or more stops looking for differences. Though I have never shot in image in real life where I have had to boost shadows by two or more stops.

The highlights seemed to clip right around the same point too, showing not much difference in DR.

Other than that, I have not shot MF and current 35 systems side by side in any controlled manner...
« Last Edit: February 03, 2009, 09:03:11 AM by lisa_r » Logged
footoograaf
Newbie
*
Offline Offline

Posts: 18


« Reply #3 on: February 03, 2009, 09:24:49 AM »
ReplyReply

Don't believe them, it's a conspiracy
« Last Edit: February 03, 2009, 09:25:45 AM by footoograaf » Logged
bcooter
Guest
« Reply #4 on: February 03, 2009, 11:08:04 AM »
ReplyReply

Quote from: lisa_r
I don't use these camera side by side, but when I shot a Phase p 30 (I think) and the 1Ds3 in a controlled test with strobes in the studio last fall, I did not see any appreciable difference in DR. This was shooting RAW and trying various converters. Even hunting around in shadows and using + exposure looking at the "shadow detail", there was very little difference in detail. The Canon did show more noise when boosting the shadows 2 or more stops looking for differences. Though I have never shot in image in real life where I have had to boost shadows by two or more stops.

The highlights seemed to clip right around the same point too, showing not much difference in DR.

Other than that, I have not shot MF and current 35 systems side by side in any controlled manner...


I have shot most 35mm cameras next to medium format in a lot of real world productions and I don't think anything I see on the dxo site is that far from my own viewing.

As Michael says trust your eyes, but I think you can take this one step further (at least if you work in commerce or editorial for clients) and say trust your client's eyes, or even trust the viewers eyes, because if anyone sees any difference between the current line of high end dslrs to medium format will be seen more in the ability to capture the moment or get an image in focus than any pixel peeping comparision or lab tests.

In the end it really doesn't matter to the viewer what it was shot on if the image is compelling and it doesn't have an obvious flaw like missed focus or wildly out of range exposure.

Recently I shot a campaign where a woman is running out of a subway platform with heavy cross side light.  I shot it with the Phase p30+ and a Canon 1dsIII and except for the different look in lens and AA filter I really could not see any appreciable difference in highlight to shadow.

In fact I shot some of this session with a Nikon D90 just because the focus sensor array covered more of the frame and when we put those d90 images in the computer on locaiton, I thought the d90 would really be awful compared to those other cameras.   To a person, we were shocked at what a small difference there was expecally from the d90 to the Canon.  Just nothing that noticeable and since more of the d90 images were in focus that ended up as the select.

Were talking about $900 cameras vs. $7,000 and $22,000 cameras and once again in the real world nobody noticed that much of a difference.

I'm not advocating everyone throw away their p45's or Leaf aptus and go buy a box ful of $900 cameras, but once again, in the real world this proved to me that medium grade consumer cameras have come a very long way and in fact have come much further than in a shorter amount of time than the high end specialty cameras.

In fact if a Leaf, Phase of Hasselblad had the lcd screen, shutter response, higher iso capability and focusing of a D90 it would be cause for massive celebration.

I work with one in house pre press manager that I think is probably the best pre press person in the world. He has written his own cmyk conversion and moved this retailer to a complete digital workflow and his group processes over half a million images a year that run on building sized posters in times square all the way down to instore, catalog, print advertising and the web and he will be the very first to tell you that anything over 20mpx is just not that noticeable once it gets through retouching and into print.  In fact he would rather see a weeks shoot come to him from a dslr than from an medium format back because he sees less pattern moire usually more in focus images and finds the whole process of seeing a decent embedded preview for edititng to be a real benefit.

Then again if I was  staring down half a million images a year, I would probably feel the same.

DXO labs or not what all of us know is for medium format to prosper they are going to have to up their game and offer more than just more megpixels.   They really need to have a rethink on the whole process of how their cameras work in the real world and they are going to have to be positive that what they offer is more than 20% better image quality (whatever that means).

Still, and to be clear, I'm actually not advocating anything other than a compelling subject, beautiful lighting, a unique viewpoint is much more important than any chart comparing DR down to .003 of a difference or 5% more eyelash detail.    Don't just trust your eyes on what you see at 200% on a computer screen but also trust your mind in knowing that the camera you used allowed you to shoot what you wanted.

I'm setting on an invesment of tens of thousands of dollars in digital backs and cameras and though they still have a use, I have to keep in mind that just because I bought them doesn't mean I have to always use them.

I actually find all of this amazing that so many photographers will go into a catatonic state if anyone dare say a Canon or Nikon performs better than a more expensive system, but instead of sticking there noses on a computer and comparing each pixel, I suggest step back and look at the image and in fact look at it the way you want your viewer to sees it.  

In fact I would think that these are times to rejoice rather than condemn.    I'm personally happy that lower costs cameras perform so well, but then again I care a lot more about the photograph that I do camera.

« Last Edit: February 03, 2009, 11:13:39 AM by bcooter » Logged
Panopeeper
Sr. Member
****
Offline Offline

Posts: 1805


« Reply #5 on: February 03, 2009, 11:57:27 AM »
ReplyReply

Based on some of their charts with P45+ data I have the impression that they tested a P45 non-plus camera.

The P45 is probably an ISOless back, and this is visible from the dynamic range and from the ISO sensitivity charts; however, the P45+ does have analog ISO gain.

EDIT: perhaps the Phase raw images are not supported by DxO and they have been using ACR. That would explain some issues, as ACR is mistreating the P45+ raw images.
« Last Edit: February 03, 2009, 02:05:40 PM by Panopeeper » Logged

Gabor
KLaban
Sr. Member
****
Offline Offline

Posts: 1609



WWW
« Reply #6 on: February 03, 2009, 12:04:22 PM »
ReplyReply

OK MFD users, you're just having a nightmare, you'll wake up in an hour or two and all will be well...
Logged

Frank Kolwicz
Newbie
*
Offline Offline

Posts: 10


« Reply #7 on: February 03, 2009, 12:11:14 PM »
ReplyReply

In discussing the evaluation errors that DxO Mark introduces in measuring MF vs DSLR cameras, (Eyes vs. Numbers which to believe? ) Michael said "DSLRs also apply a lot of noise reduction algorithms before storing the data." That's the first time I've ever heard of that. I thought RAW was raw, not processed in any way. Where does that info come from?
Logged
Tomcat
Newbie
*
Offline Offline

Posts: 8


« Reply #8 on: February 03, 2009, 12:38:00 PM »
ReplyReply

I noticed that they compared the absolute newest DSLRs against the 2006 era MF backs.  I'm sure the MF sensors have gotten better with time too.  But then again, Dxo only sells software for DSLRs, so I guess they need to make sure that DSLRs win.

Logged
michael
Administrator
Sr. Member
*****
Offline Offline

Posts: 4731



« Reply #9 on: February 03, 2009, 01:06:54 PM »
ReplyReply

Quote from: Frank Kolwicz
In discussing the evaluation errors that DxO Mark introduces in measuring MF vs DSLR cameras, (Eyes vs. Numbers which to believe? ) Michael said "DSLRs also apply a lot of noise reduction algorithms before storing the data." That's the first time I've ever heard of that. I thought RAW was raw, not processed in any way. Where does that info come from?

Do some reading of the technical literature.

A sensor is an analogue device. It produces a voltage varying by how much light hits each photo site. That voltage then needs to be converted to a digital signal. Virtually all sensor chips have some image processing done on chip, sometimes at the analogue stage, sometimes at the digital stage.

CCD chips produce more heat and consume more power than CMOS and therefore their designers prefer to move processing to the computer rather that having to have power hungry and heat producing ASICs on board. Also, by moving the primary noise reduction to post processing they provide the user with more control over the process (at least in theory), since much of the fixed pattern NR is done to the image prior to delinerarization and matrix decoding.

Michael
Logged
etrexler
Newbie
*
Offline Offline

Posts: 23


« Reply #10 on: February 03, 2009, 01:35:07 PM »
ReplyReply

Could the low DxOMark Sensor scores be impacted by the generally low "Low-Light ISO" scores of these backs?  I'm wondering if they generate the DxOMark Sensor score based off an average of the ISO sensitivity, SNR,  Dynamic Range , tonal range, etc...?
Logged
NikosR
Sr. Member
****
Offline Offline

Posts: 622


WWW
« Reply #11 on: February 03, 2009, 01:43:47 PM »
ReplyReply

Quote from: michael
Do some reading of the technical literature.

A sensor is an analogue device. It produces a voltage varying by how much light hits each photo site. That voltage then needs to be converted to a digital signal. Virtually all sensor chips have some image processing done on chip, sometimes at the analogue stage, sometimes at the digital stage.

CCD chips produce more heat and consume more power than CMOS and therefore their designers prefer to move processing to the computer rather that having to have power hungry and heat producing ASICs on board. Also, by moving the primary noise reduction to post processing they provide the user with more control over the process (at least in theory), since much of the fixed pattern NR is done to the image prior to delinerarization and matrix decoding.

Michael

Huh?Huh

When you're taking measures to increase the S/N ratio in the analog domain this can hardly be called image processing (at least not anymore than an analogue sound amplifier using good quality capacitors vs bad quality ones can be said to do 'sound processing') not to mention noise reduction processing. Not doing that means you have to deal with a much dirtier signal in the digital domain. If you do noise processing you are really doing that after A/D conversion (which often takes place on the CMOS chip) and you're doing that in the digital domain.

Your last sentence sounds puzzling. What does user control have to do with NR being done prior to demosaicing?. There are theoretical benefits in doing noise reduction before demosaicing (you're getting cleaner data to demosaic thus you get more accurate demosaicing results) but what does it have to do with user control?

Also, are you suggesting that Nikon (for example) is doing noise reduction in the digital domain in camera prior to committing to raw now that they use CMOS while they were not doing that just 1 or 2 years ago when using CCD?

Could you point me to the technical literature that proves the dSLR manufacturers perform noise reduction (by definition in the digital domain) before committing to raw? Can they do that? Of course they can. Do they do that? I suspect that no one allowed to say knows, so I would be (pleasantly) surprised if this info exists in any technical literature in the public domain. *

You keep on maintaining both in your essay and in various threads that comparing the raw output of dSLRs vs the raw output of MFDBs puts the MFDBs at a disadvantage. You might have a point but I'm not sure what it is and I have yet to be convinced by your arguments. You either compare raw or compare the demosaiced results and there you have to cater for the intricacies of the raw converter. That's the same for all cameras. Noise results of the Nikons (for example) are better when using some NR in raw conversion and with Nikon's raw converter you are getting additional benefits like automatic CA control. Why is this any different from what MFDBs do with their proprietary raw conversion escapes my understanding.

I'm sorry to say this but it sounds a bit like you're parroting something some person working for an MFDB manufacturer told you without you really understanding what he was talking about. Either that or it was just marketing talk. I really hope I'm wrong.

I feel utterly confused....

*PS  Thanks to Gabor (panokeeper) who reminded me that Sony are doing this with their A700 / A900, which caused a big fuss in the internet, but thankfully nowdays this can be turned completely off.
« Last Edit: February 03, 2009, 03:05:15 PM by NikosR » Logged

Nikos
lisa_r
Guest
« Reply #12 on: February 03, 2009, 01:45:56 PM »
ReplyReply

I think the overall score is just about useless. Look at the various data parameters, use the cameras and draw your own conclusions.

Quote from: etrexler
Could the low DxOMark Sensor scores be impacted by the generally low "Low-Light ISO" scores of these backs?  I'm wondering if they generate the DxOMark Sensor score based off an average of the ISO sensitivity, SNR,  Dynamic Range , tonal range, etc...?


As to the above statement, I think "generally Low-Light ISO scores" are indeed very important, and should drag a score down if you care about scores. In my experience, just about any modern camera can make a pretty picture with enough controlled light/strobes. It's when the light gets low that you really need a camera with great tricks - good mannered in the dark in terms of IQ, and with good low-light AF.
Logged
markowich
Full Member
***
Offline Offline

Posts: 119


WWW
« Reply #13 on: February 03, 2009, 02:55:17 PM »
ReplyReply

Quote from: michael
Do some reading of the technical literature.

A sensor is an analogue device. It produces a voltage varying by how much light hits each photo site. That voltage then needs to be converted to a digital signal. Virtually all sensor chips have some image processing done on chip, sometimes at the analogue stage, sometimes at the digital stage.

CCD chips produce more heat and consume more power than CMOS and therefore their designers prefer to move processing to the computer rather that having to have power hungry and heat producing ASICs on board. Also, by moving the primary noise reduction to post processing they provide the user with more control over the process (at least in theory), since much of the fixed pattern NR is done to the image prior to delinerarization and matrix decoding.

Michael

what you say is true of the previous (three year old) CCD sensor generation. today's CCDs (H3DII 50, P65, Aptus A10..) consume much less power and produce much less heat then the old ones,  not even one order of magnitude more than CMOS.
also, most of the on-board noise reduction is better than what 99.9% of the users can do 'manually' by running noise ninja etc. one has to be very sophisticated if one wants to beat it, certainly tuning a few parameters in a standard commercial software does not do it. MF back producer also use your argument of maximizing control for the user's benefit, but most users (including 99.9999999999999999% of professional photographers) would be better of without it.
on a sideline, those DXO people are super sophisticated (i know what i am talking about here) and i do believe their measurements without restrictions. as michael says in his article, measured numbers have to be taken relatively to human perceptional
constraints but this is a triviality anyway.
peter
Logged
Anders_HK
Sr. Member
****
Offline Offline

Posts: 1001



WWW
« Reply #14 on: February 03, 2009, 03:03:41 PM »
ReplyReply

Quote from: John Schweikert
Look at the DXOMark numbers for Aptus 75s, P45+ and Canon 5DII. It has the Canon with higher dynamic range (actual number of stops). That's the biggest bunch of bullshit I have ever seen.

I agree John.

Plugging in some cameras I have owned reveals obvious deficiencies in DxO;s numbers;

D200 + ZD + Aptus 65 (theirs 75S, same identical performance)

= shows D200 higher DR than both ZD and Aptus    Gee wiz, why did I not like the D200 or ZD but find Aptus to yield MUCH more improved image quality on all counts?? I would even say better ISO than D200... D200 files simply falls apart in comparison. D200 files also falls apart in DR compared to ZD...

or...

plug in G9 ... whee... the little G9 near same DR...

It is very obvious the numbers are flawed or they have no idea how to do proper scientific testing. I would have thought DxO was reputable... guess not.

Regards
Anders


Logged
erick.boileau
Sr. Member
****
Offline Offline

Posts: 469


WWW
« Reply #15 on: February 03, 2009, 03:22:26 PM »
ReplyReply

DXO has no credibility at all ... without any interest for me
Logged
NikosR
Sr. Member
****
Offline Offline

Posts: 622


WWW
« Reply #16 on: February 03, 2009, 03:51:34 PM »
ReplyReply

Galileo Galilei was talking bull because anybody could see the sun is circling the earth.
Logged

Nikos
bjanes
Sr. Member
****
Offline Offline

Posts: 2714



« Reply #17 on: February 03, 2009, 04:27:36 PM »
ReplyReply

Quote from: NikosR
Huh?Huh

Could you point me to the technical literature that proves the dSLR manufacturers perform noise reduction (by definition in the digital domain) before committing to raw? Can they do that? Of course they can. Do they do that? I suspect that no one allowed to say knows, so I would be (pleasantly) surprised if this info exists in any technical literature in the public domain.

The Canon CMOS White Paper describes Canon's on chip noise reduction (see Noise Issues,  page 17 of the PDF). One should remember that the output of a CMOS chip is bits (digital), whereas the output of a CCD chip is voltage (Dalsa Paper). The output of a pixel in CMOS is voltage while the output of a pixel in CCD is an electron packet. Both of these outputs could be manipulated in the analog domain on chip.

Since CMOS chips have a digital output, any processing of data outside of the chip would have to be in the digital domain. Some NR is done on the chip, but further NR could be done to the raw data before it is written to the memory chip. It is difficult get this information out of the camera makers.

Bill
Logged
bjanes
Sr. Member
****
Offline Offline

Posts: 2714



« Reply #18 on: February 03, 2009, 04:46:42 PM »
ReplyReply

Quote from: erick.boileau
DXO has no credibility at all ... without any interest for me

Then I would say that your thought process is irrational and unscientific. If you read their web site, you will see that those fellows are quite adept technically. Dynamic range involves two measurements: the maximal and minimal useful signals. Their dynamic range measurements at the low end assume a signal:noise of 1. Now a S:N of one will not give a very good image and a decent photo should have at least a S:N of 10. A Nikon D3 may have a higher S:N than the 45 MP medium format camera, but the 45 MP camera has many more pixels and the noise will be much finer grained and less objectionable. Read here on the DXO site. Noise has character as well as quality, and some of these characteristics can be expressed as a frequency spectrum. However, ultimately the how much noise in the shadows is acceptable for a photographically useful DR calculation has to do with perception and is rather subjective. Here is where you need to supplement the numbers with observation and judgment, but don't argue with the scientific technique.

Bill
Logged
lisa_r
Guest
« Reply #19 on: February 03, 2009, 08:18:18 PM »
ReplyReply

Some very impressive D3x landscapes:

http://forums.dpreview.com/forums/read.asp...essage=30863275

Logged
Pages: [1] 2 3 ... 6 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad