Dxomark has A700 v4 up now

[Dusty]

To an extent i also agree with Don in that kudos to DXomark for revising its results to allow for V4. I think what concerns some people here and elsewhere is how level is the playing field? In other words if producers are altering their raw files, how can you make a valid comparison? I may be way off the mark here but if we accept that Nikon and Sony use the same FF sensor then in theory at least shouldnt the RAW results be the same? As it is, my understanding is that different processing is applied to produce the raw file so how do you allow for that? The answer is, i suspect, that you cant so these results are only ever going to be a guide albeit a pretty good one.

No RAW results shouldn't be the same unless you're processing them with the same engine tuned the same way. We're not talking sensor output, we're talking the RAW file, which is the sensor pixel data interpreted by the firmware's implementation of the Bayer pattern. If one gives more weight to a color channel, or decides that R at +4 and B at +2 and GG at +1 = mauve, while another engine says that = lavender, you have different RAW outputs.

The same way that Racer X and Racer Y both buy racing motors from BMW, however Racer X installs a better exhaust and intake systems, resulting in more horsepower. DXOLabs was saying, in effect, "We only test motors with carburetors, not fuel injection, to determine horsepower."

If I came up with a software engine that totally eliminated noise in RAW files at 64000 ISO, that's RAW file, like it or not. RAW output, indeed all digital photography, is a combination of sensor output and it's interpretation by the software.

Dusty

[bossel]

We're not talking sensor output, we're talking the RAW file, which is the sensor pixel data interpreted by the firmware's implementation of the Bayer pattern.

Hmmm, I'd say the RAW file data is not demosaic-ed (i.e.that Bayer thing). RAW contains sensor information before calculating coloured pixels from sensor data.

[douglasf13]

I can't believe I was actually about to back up Don, and then he checked out of the forum! IMO, DxO Mark, while certainly not perfect, is a very useful reference, and is a step in the right direction for testing. Scientific testing, like Dxo Mark, and more anecdotal testing, like Michael Reichmann's, both have value.

Barry, I think the thing to keep in mind about the Dxo Mark DR tests, like just about every DR test on the web, is that you should ignore the the number applied to the DR, and instead use the DR charts to compare digital cameras to one another.

[Vidgamer]

I can't believe I was actually about to back up Don, and then he checked out of the forum!

Haha, I just re-read his last message and saw that! He probably never had so many here AGREE with him, and NOW he checks out?! OH well....

IMO, DxO Mark, while certainly not perfect, is a very useful reference, and is a step in the right direction for testing. Scientific testing, like Dxo Mark, and more anecdotal testing, like Michael Reichmann's, both have value.

Barry, I think the thing to keep in mind about the Dxo Mark DR tests, like just about every DR test on the web, is that you should ignore the the number applied to the DR, and instead use the DR charts to compare digital cameras to one another.

Yeah, I would think of it as an aid to help get a good idea of the relative difference between cameras. If two cameras are "close", I'm not sure you could draw conclusions about one "wiping the floor" with the other, or some such hyperbole, but it should give you an idea of relative performance.

[bfitzgerald]

Well DxO say the A350 has a better SNR ratio than the A200. So that should mean it's better at high ISO right?

The more you dig around this DxO place, the less believable it becomes.

I notice at the bottom of the ist page it says..
"once upon a time"

[Javelin]

One post on DPR claims that according to DXO the A700 at ISO 100 has lower noise and higher dynamic range and colour than at ISO200 (supposed base ISO) this dissagrees with what DPR and lots of other reviewers have said. Some independant guys posted a while back that it's at it's best at ISO 160 but no lower than that.

[Vidgamer]

Well DxO say the A350 has a better SNR ratio than the A200. So that should mean it's better at high ISO right?

When I look at it with the "screen" setting, it shows the A200 SNR as better than the A350 at all ISOs. That's looking at raw pixels. If you choose "print", it shows them basically the same, with the A350 having a slight edge at a couple of points. All that is doing is letting the higher number of pixels make up for some of the additional noise at the pixel level. They have an article about how more pixels can mitigate the noise problem.

So, based on the DXO charts, I'd say that for high ISO, either camera would be about the same unless you were purely interested in noise at the 100% pixel level and resolution didn't matter.

The more you dig around this DxO place, the less believable it becomes.

I wouldn't say that raw numbers are everything either. But if you can't make some measurements and at least get an idea of performance, well, I don't know what to say. What site do you recommend for camera comparisons? DPR?

I notice at the bottom of the ist page it says..
"once upon a time"

I don't see that...???

One post on DPR claims that according to DXO the A700 at ISO 100 has lower noise and higher dynamic range and colour than at ISO200 (supposed base ISO) this dissagrees with what DPR and lots of other reviewers have said. Some independant guys posted a while back that it's at it's best at ISO 160 but no lower than that.

Why would the A700 have more noise at ISO 100? Less dynamic range, I could understand, but I'd expect less noise. (Although, DxO shows it as having a tad more DR at ISO 100, too!) Maybe if I had an A700 I'd see if I could make my own determination...........

[Javelin]

[quote="Vidgamer]

One post on DPR claims that according to DXO the A700 at ISO 100 has lower noise and higher dynamic range and colour than at ISO200 (supposed base ISO) this dissagrees with what DPR and lots of other reviewers have said. Some independant guys posted a while back that it's at it's best at ISO 160 but no lower than that.


Why would the A700 have more noise at ISO 100? Less dynamic range, I could understand, but I'd expect less noise. (Although, DxO shows it as having a tad more DR at ISO 100, too!) Maybe if I had an A700 I'd see if I could make my own determination...........

Maybe not more noise but the story was that the base ISO at 200 ad better IQ because ISO 100 was derived through processing rather than real.

[Vidgamer]

Maybe not more noise but the story was that the base ISO at 200 ad better IQ because ISO 100 was derived through processing rather than real.

But are they guessing or did they actually see that? I should be able to take any ISO 200 photo, reduce it by a stop of exposure, and get a less noisy photo.
I'll have less DR, but I don't think Sony would add ISO 100 unless it did something. ("Ok, here, you can select it, but it'll just look worse AND be slower, but we'll give it to you anyway.") Well, I suppose you might want a slower mode to help with waterfalls or some such.

[Javelin]

here is part of the DPR review where they talk about ISO sensitivity and DR most review sites said variations of the same thing

http://www.dpreview.com/reviews/sonydslra700/page20.asp" onclick="window.open(this.href);return false;

there was much discussion on this at DPR when the review came out and people settled on ISO 160 at the time as being the best

[Greg Beetham]

I'm no expert on in camera processing and PPP I'll admit, but going on what I've seen written around here in the past, if you were to underexpose an ISO200 photo by one stop too emulate ISO100 and then "lift" it afterwards it's going to generate noise, to my mind it's going too be more noisy than if you actually had a real ISO100 in camera.. it would be like exposing too the left and then dragging detail out of much darker shadows than you would have had otherwise.
Greg

[Greg Beetham]

There is still something I don't "get" with DXO, they claim too measure the sensor generated output before any "in camera" processing takes place...what is not made clear so far is, how do they intercept the output directly from the sensor? before any onboard camera systems get to deal with it, that's a very neat trick, if that is what they are in fact doing, you would have too think that somehow they are able to uncouple the sensor from the system board and take a full channel dump into their own buffers and do their evaluations from there....hmmm might not be good for the camera...I think I read somewhere that they use calibrated light sorces for some tests....maybe that's it, they use a known measured strength of light to expose in, and then evaluate what the camera did in response after an exposure....
Greg

[Greg Beetham]

Well that sort of opens up a new can of worms, a camera and lens combo is a balanced exposure system, the Minolta/Sony system relies on metering through the lens and feedback from the lens to a certain extent...I thought, so now I'm wondering how much the nofocussed light without a lens is affecting the exposure system, as for absolute sure the body would "know" that there is no lens fitted...ie. I would have thought the body would need a lens too establish "on" exposure value for a given ISO at a given ambient light level @ 18% with that particular lens, and from there the +/- exposure compensation table...from that previously established baseline exposure value.
All I'm saying is, how can one be absolutely certain that the behavour of the exposure/sensor/processing system is the same, both with and without a lens.
Greg

[douglasf13]

I have decided to come out of retirement for an occasional post, and today my post is to confirm that many of you guys are still in need of some work on this stuff. Seriously, a few of you need to at least read (and/or remember) the people that know what they are talking regarding the a700 and noise instead of making shitzki up, and/or posting random shitzki that is wrong or does not make sense.

#1 DxOMark does not measure usable DR range, they measure sensor max DR.

#2 Several discussions have taken place about ISO 100 being less noisy, and instead of talking bunk junk, next time just take a few pictures for yourself, or check Dave's or other raw files to see for yourself, or do a forum search on the subject, or just be quiet on things that you don't have a clue on and leave the serious lifting for those that do. Please do not post further nonsense to try to back up your own previously posted nonsense.

#3 sNr does not alone make up high ISO image quality...not even close. High ISO IQ is a combo of everything, including actual measured sensitivity, DR, color accuracy, tonal range, size of the grain, type of grain, wb control, effective NR, metering, curves, etc, etc....and of course af systems, viewfinders, signal processing and a ton of other things also make a camera good at high ISO.

#4 I see more bogus info posted on this thread by a few people than you can shake a stick at.

#5 Most posters at DPR don't have a clue...many are often repeating someone else that also does not have a clue.

#6 DxOMark is best used to compare the sensor potential of cameras....If sensors are rated closely at DxOMark, that means IQ should be close, so make your buying decision on feature sets, or handling, or something else, and not on super subtle IQ differences.

#7 To compare 8x12 print size use normalized at DxOMark, to see how things look at native file size then use the screen option.

#8 DxOMark is a guide and it should be used as one....it should not be used for people to pretend sensor max DR is the same thing as usable DR or film DR. Most everyone on the planet knows this fact already, but obviously there are still a few people that don't.

I quote myself back in Aug. (DIWA tests JPGs using DxO Anylizer)

"Yea, use it if you need slower shutter speeds. Diwa labs shows ISO200 dynamic range and color fidelity is a bit better than it is at ISO100. I used ISO100 on those fireworks shots I posted, and I use it in very bright conditions at times. It looks to me as if ISO100 shadow noise is pretty good, and Diwa does show lower ISO 100 noise and smaller grain size."

And then DK follows...

"I use both ISO 200 and 100. There is a significant improvement in the noise level of blue skies at 100, over 200, and that counts for me. I don't often need to pull in the 2 stops of raw headroom you can get at ISO 200. But whenever it counts, yes, I use 200."

http://www.photoclubalpha.com/forum/vie ... +100#p2692" onclick="window.open(this.href);return false;

I confirmed shadow (darks and dark colors) noise is less at ISO 100, and DK confirmed blue sky noise was less...this matches what the labs show re: noise and should put an end to this nonsensical discussion.

For those who do not yet understand...DxOMark basically shows you sensor max potential and not actual photographic output. Then, to make it easier for average reader to better understand the data, they then score the hard findings based on three key areas of photography to give you a sophisticated, but yet simple and direct means of easily comparing different camera's sensor performance on some sort of level playing field. To dig deeper into what specific performance it is that you need you then MUST seriously study the individual plots to see what camera may serve you best, where and when you need it, you can then rule out specific models that obviously will not perform as good as you would like them to perform in the specific areas that you need the performance.

I leave you with a few quotes from the DxOMark website...

"Thanks to DxO Labs Image Quality Database and the DxOMark Sensor scale, photographers no longer have to rely solely on manufacturers’ published data to evaluate digital camera image quality....

All published measurements are RAW sensor-based and performed prior to any digital image processing (camera- or PC software-based). As RAW format is considered the equivalent of a film negative by analogy to traditional photography, it is the format by which camera body performance is most effectively evaluated, as it does not depend on the camera optics nor on the quality of the RAW conversion applied...

Objective measurements of RAW image quality are an essential component of any analysis of digital cameras. DxO Labs has developed a new scale for digital camera image quality performance, called DxOMark Sensor, to serve as an additional tool to help photographers rank and compare digital cameras...

One of the innovations of DxOMark Sensor scale is that it was conceived and built on three principal photographic uses:

* studio and portrait photography
* landscape photography
* photojournalism and sports photography

DxOMark Sensor measures only the RAW image quality of a digital camera; therefore, DxOMark Sensor is NOT an evaluation of overall camera image quality or performance..."

Talk to you later, fellas.

-Sonolta

I pretty much agree on all points, and well said. I think I'll tag this post in order to link it in the future whenever these DxO Mark talks arise.

[Vidgamer]

I'm no expert on in camera processing and PPP I'll admit, but going on what I've seen written around here in the past, if you were to underexpose an ISO200 photo by one stop too emulate ISO100 and then "lift" it afterwards it's going to generate noise, to my mind it's going too be more noisy than if you actually had a real ISO100 in camera.. it would be like exposing too the left and then dragging detail out of much darker shadows than you would have had otherwise.
Greg

If you're going to emulate ISO 100, you have to overexpose a bit and "drop" it afterwards, which will reduce, not generate, noise. It will work as long as you don't blow too many highlights, needless to say! The main point being that there is generally enough headroom to have some leeway with a DSLR, so even if it is native ISO 200 or even 160, my thinking is that ISO 100 could have less noise, but perhaps a tad less DR.

What you're thinking of is to emulate higher ISO. Like, if I want to emulate ISO 3200, I'd have to underexpose an ISO 1600 shot in order to achieve a faster shutter. When I pull it up in PP, it will pull up the noise as well. It would also look really awful, as ISO 1600 is bad enough as it is. Unfortunately, I've been in situations where I had an f2.8 lens, and even ISO 1600 wasn't enough.