Tech radar 5D3 results?

Joesiv said:

rwbaron said:

Joesiv said:

rwbaron said:

It would seem to me that a Tiff really shouldn't make any difference as it's just a container for an uncompressed file so why would the results from a Tiff be different and not valid?

Click to expand...

Tiff isn't a container for RAW, it is fully processed including NR, contrast, sharpness, saturation, etc.. not to mention the debayering which is what most people consider "raw". TIff is closer to JPG than it is RAW.

Click to expand...

I did not say a TIFF was specifically a container for a RAW file, you did. A TIFF is what you make it. If you convert a RAW file to a TIFF without NR or any changes to contrast, saturation, sharpness, etc. the TIFF will be just that. It's simply the characteristics of the source file which is nothing at all like a JPEG which applies tone curves, sharpening, compression, etc. You can also choose the level of compression including zero.

http://en.wikipedia.org/wiki/Tagged_Image_File_Format

Please explain where I'm wrong.

Click to expand...

You're right, sorry I misread what you said.

However, you're still mistaken to think you turn down your dials in camera and still expect to actually have things turned off. Depending on what the manufactuer does with the firmware, there likely isn't an option to actually turn "off" things like NR and sharpening. This is easily seen using Dpreviews noise widget, where it lets you select NR setting. Canon set to "off" still removes chroma noise, and maybe even some luminance noise. Also if you "bake in" a really bad white balance, you're going to lose a lot of the tonality of certain color channels even with 16bit per channel, but that likely won't affect DxO marks.

yep Tiff is what you make it, but you are at the limits of the tool used to create the tiff. That in its self makes it different than RAW for analyzing for things like noise and dynamic range.

Click to expand...

I still don't think we're communicating.

My understanding is TechRadar is using DXOMark software and equipment to capture RAW files with cameras which are then converted using the camera manufacturers supplied converter (Digitial Photo Professional or Nikon's version of it). They convert the RAW file at the standard default settings in the software and then use the software to convert to a 16 bit Tiff for analysis. If done properely in-camera settings should not play a part nor does the file ever become a JPEG. It stays as pristine as a usable converted RAW file can be with the exception of the NR applied by the conversion software. TechRadar would prefer to zero out the NR but apprently Nikon's software does not allow for this as Canon's does (but who really knows). This would be no different if they used ACR as it too applies different default settings for different cameras which as reported by others are not completely defeatable.

As mentioned above what can you do with an unconverted RAW file from any camera? Can you perform processing in LR or PS? Can you send it to a printer? Can you display it on the Web?

I think there is value in the method TechRadar is using because they are analyzing the file in a generic usable form.

Bob

--
http://www.pbase.com/rwbaron

rwbaron said:

I still don't think we're communicating.

Click to expand...

lol, yeah I don't think we're doing too well, probably my fault

I have no problem with them using tiff, I was just trying to relay the difference from using native RAW that's all.

The results seem to be quite different, that's for sure, which is why we're discussing this in the first place.

--
Cloverdale, B.C., Canada
Currently shooting: Nikon D3S, D700
http://www.joesiv.com

David Franklin said:

As far as I know, the 16 bit uncompressed Tiff format is the least "lossy" of those output formats and is certainly the most common high quality one. Therefore, in order to judge the output of any camera shooting raw files, it would absolutely seem to be most appropriate to judge it using the 16 bit Tiff output from a camera's raw file, as converted in the camera makers' own conversion software, using their own designed settings and parameters. This seems to me to be only the most elementary common sense. Would you judge the performance of a car by bench testing its engine on a dynonometer and suspension on a stationary dynamic response platform, or would you do so by driving it on the kinds of roads it was designed to run on? The answer is probably both, but the latter would be much, much more convincing and authoritative.

All this is a long-winded way of saying that Tech Radar's 16 bit Tiff test is at least as good, if not a better, way to test just about every outcome of a camera's photo performance, as a whole system that delivers image results, rather than the theoretical expression of the level of performance of a sensor's raw image potential that DXO chooses to employ.

Click to expand...

16-bit TIFF may be the least lossy of the output formats but the raw conversion is the most lossy of them all. There are several irreversible decisions made by a specific raw processor, including some that are unique to a particular camera model. To extend your car analogy, judging a camera/sensor by a specific raw-processor's rendering is like judging a sport car by driving it only on gravel roads.

Again, considering that no one but testing engineers, and technno enthusiasts who emulate them, actually make direct use of raw files, but all actual photographers who seek to create any actual pictures must use 16 bit Tiffs or lesser picture files, to continue your anology, all roads everywhere are gravel roads, except those that are worse.

You may feel that it is somehow cosmically unfair that the one cannot actually produce picture files that perfectly reflect the characteristics that your testing procedures indicate in raw files; however, current realities are what they are, and you must learn to deal with them on their own level, as they really exist.

David
--
Keep learning; share knowledge; think seriously about outcomes; seek wisdom.

David Franklin said:

Again, considering that no one but testing engineers, and technno enthusiasts who emulate them, actually make direct use of raw files, but all actual photographers who seek to create any actual pictures must use 16 bit Tiffs or lesser picture files, to continue your anology, all roads everywhere are gravel roads, except those that are worse.

You may feel that it is somehow cosmically unfair that the one cannot actually produce picture files that perfectly reflect the characteristics that your testing procedures indicate in raw files; however, current realities are what they are, and you must learn to deal with them on their own level, as they really exist.

Click to expand...

Your description seems to imply there is only one way to generate a TIFF from a RAW. Two photographers using even the same raw processor will generate entirely different looking TIFFs based on their artistic and aesthetic preferences. And some of those preferences rely on a sensor's abilities more than others. TIFF is simply a container for the rendered image - it's the rendering where the rubber meets the road.

My description actaully doesn't imply anything - rather it states quite plainly that, while there might be in some wonderful bright new future, a raw converter that would be able to more reflect the specific findings made by DXO and others, from their chosen parameters and testing procedures, about the characteristics of various cameras' raw files' potentials to possibly create a file more reflecting their found raw file virtues, at present there is no such thing in the known universe. There are only the real picture files that we actually have, and the real converters that actaully make them from raw data. Therefore, when one judges the real measuarble and useful picture files (e.g., 16 bit Tiffs) in the best way known to be possible (those recommended by the camera manufacturers themselves are the obvious candidates), then this must be the truest test of what real - not potential - photographers creating real - not potential - useable picture files, can expect. This is all quite elementary and obvious, and I'm having an extremely difficult time understanding your reluctance to accept this as true. You may keep and extoll the results of your raw file testing, but if they don't show up in real picture files the way you have described them, then you have two real-world options: 1) find or create for yourself a new picture file which perfectly reflects your testing, or 2) rework your test procedures and parameters to better reflect what people can actually expect from using the actual camera. This is really the essence of the scientific method. You don't complain that reality doesn't prove to coincide with your testing methods, but, instead, sometimes you must revise your tests to coincide with reality. The only thing that DXO-type testing can confirm, if it doesn't conform with the reality of current raw-to-useable picture fille conversion, is that they have perhaps done a very good study of the capacities of a sensor, but not described what a photographer can expect when taking pictures with the camera that uses that sensor.

Regards,
David
--
Keep learning; share knowledge; think seriously about outcomes; seek wisdom.

As photographers we all work with intermediate file formats to develop our pictures, be it explicitly TIFF or implicitly in Photoshop before we've confined the render to a specific file format. If your interest is proving that DxoMark's raw measurements don't apply to real-world usage via intermediate formats then you can easily construct a test to prove/disprove that. Is there a specific DxoMark measurement which you feel doesn't reflect reality? If so, do you have evidence to the contrary?

Look at the Tech Radar results, the whole point of this thread. You may find some flaw in it or not, but it is the very result you challenged me to provide, as it does not coincide with the raw file potential results from DXO and other testing. Still, you are merely talking over what I've said and not directly addressing it. It is obvious that you just can't get over the idea that since 16 bit Tiff is some intermediary iteration that it doesn't tell the story of a camera's picture making IQ performance, but rather that the true results are those that are theoretical only. This is plainly not true. I hope I don't have to go back to Plato's cave to clarify this. Is the idea of the picture more real than the picture itself? I don't think so.

Regards,
David
--
Keep learning; share knowledge; think seriously about outcomes; seek wisdom.

David Franklin said:

Look at the Tech Radar results, the whole point of this thread. You may find some flaw in it or not, but it is the very result you challenged me to provide, as it does not coincide with the raw file potential results from DXO and other testing. Still, you are merely talking over what I've said and not directly addressing it. It is obvious that you just can't get over the idea that since 16 bit Tiff is some intermediary iteration that it doesn't tell the story of a camera's picture making IQ performance, but rather that the true results are those that are theoretical only. This is plainly not true. I hope I don't have to go back to Plato's cave to clarify this. Is the idea of the picture more real than the picture itself? I don't think so.

Click to expand...

Both Tech Radar and DxoMark's results are based on the same analytical measurements of image data (they both use the same DxoMark software). If you believe the analytical results of a post-gamma rendered TIFF file with a predefined set of raw conversion parameters baked into it is a more accurate representation of sensor performance for real-world usage than the raw "clay" from which raw-shooting photographers start from then we are just too far apart in our opinions to make further discussion useful to either of us.

Look, perhaps we are too far apart to discuss this further, but I'll try one last time right now. After this, we'll probably have to agree to disagree.

You said:

If you believe the analytical results of a post-gamma rendered TIFF file with a predefined set of raw conversion parameters baked into it is a more accurate representation of sensor (my emphasis) performance for real-world usage than the raw "clay" from which raw-shooting photographers start from then we are just too far apart in our opinions to make further discussion useful to either of us.

Click to expand...

It seems to me that you either refuse to consider what I've been saying here, over and over, or you really don't understand what I've repeatedly explained in detail. I don't necessarily believe that TR tests are necessarily accurate tests of the best possible potential results of the capacities and theoretically possible output from any SENSOR. But, it is clear that they may indeed represent the really, actually, existing (at the present time and in our present universe) results reflecting what a real photographer can expect, creating a real useable picture file, using the presently best possible format, converted from raw file data using the best case scenario settings from the raw converter software best suited to his particular camera. Why does this matter? Because I can't presently publish a raw file. Even software previews of raw files in converter programs are not views of raw files themsleves but usually hastily constructed Tiff files, because raw data is not discernable to the human eye in any coherent way. You seem to only care about sensor measurements. Photographers mostly care about photo images. Measuring the best possible photo image parameters in converted useable fiiles is vastly more appropriate than measuring sensor data capacities when trying to figure out how a given camera system, including the necessary raw image converter, will be able to yield an image suitable for actual photographic uses.

Regards,
David

--
Keep learning; share knowledge; think seriously about outcomes; seek wisdom.

You've said a few times I do raw measurements and only care about sensor measurements but neither is the case. I only care about DxoMark's measurements because they correlate to my real-world usage, in particular their SNR, DR, and color selectivity measurements. I've posted many photographic comparisons on dpreview and youtube in the past few years, all of which roughly matched DxoMark's relative measured differences. If you don't believe that their measurements translate into meaningful differences of IQ for real-world usage then we'll just have to agree to disagree.

A generated tif file has fixed ends ,how the tif looks like depends on who has generated the Tif files and how with which parameters

Do you understand the difference between sensor data and a fixed Tif file ?

David Franklin said:

Look, perhaps we are too far apart to discuss this further, but I'll try one last time right now. After this, we'll probably have to agree to disagree.

You said:

If you believe the analytical results of a post-gamma rendered TIFF file with a predefined set of raw conversion parameters baked into it is a more accurate representation of sensor (my emphasis) performance for real-world usage than the raw "clay" from which raw-shooting photographers start from then we are just too far apart in our opinions to make further discussion useful to either of us.

Click to expand...

It seems to me that you either refuse to consider what I've been saying here, over and over, or you really don't understand what I've repeatedly explained in detail. I don't necessarily believe that TR tests are necessarily accurate tests of the best possible potential results of the capacities and theoretically possible output from any SENSOR. But, it is clear that they may indeed represent the really, actually, existing (at the present time and in our present universe) results reflecting what a real photographer can expect, creating a real useable picture file, using the presently best possible format, converted from raw file data using the best case scenario settings from the raw converter software best suited to his particular camera. Why does this matter? Because I can't presently publish a raw file. Even software previews of raw files in converter programs are not views of raw files themsleves but usually hastily constructed Tiff files, because raw data is not discernable to the human eye in any coherent way. You seem to only care about sensor measurements. Photographers mostly care about photo images. Measuring the best possible photo image parameters in converted useable fiiles is vastly more appropriate than measuring sensor data capacities when trying to figure out how a given camera system, including the necessary raw image converter, will be able to yield an image suitable for actual photographic uses.

Regards,
David

--
Keep learning; share knowledge; think seriously about outcomes; seek wisdom.

Click to expand...

Answer: yes, I very well do. Your question: if you read my back and forth with Horshack, you would know that that is almost irrevelevant to our discussion, as this was treated as a given by both of us in our conversation.

We discussed the relative value of the measurement of a photographically useless file (raw) to a useful one (Tiff) and the applicattion of those measurements to relevancy for real world photogtraphy. He finally, in the end, claimed that his own tests of the only files that matter to photography, the finished picture files, yielded by conversion of the raw files, showed a correlation to the theoretical findings of pure sensor performance found in tests like those of DXO. I pointed out, over and over, that the findings along those lines by Tech Radar do not corroborate those same findings regarding both the 5D3 performance and the relative performance of the D800 compared to it. Your pointed question is only an elementary matter tangental to the heart of the previous discussions, but not approaching the crux of our discussion. Of course, the Tiff will be converted with the equivalent of tone curves and many other applied parameters. So what? The point is, at the end of the process, you will have what you have - very good, better, best, whatever - in all cases and in the only form that is of critical use to actual photo results. This is the difference between the perspective of the Tech Radar measurements and the DXO measurements, the actual subject of this thread.

Regards,
David
--
Keep learning; share knowledge; think seriously about outcomes; seek wisdom.

David Franklin said:

He finally, in the end, claimed that his own tests of the only files that matter to photography, the finished picture files, yielded by conversion of the raw files, showed a correlation to the theoretical findings of pure sensor performance found in tests like those of DXO.

Click to expand...

Btw, here are a few DR tests I've posted. And they're TIFFs ;-)

http://forums.dpreview.com/forums/read.asp?forum=1032&message=37516824
http://forums.dpreview.com/forums/read.asp?forum=1032&message=36267286

If you set DXO to "screen" (100% resolution), the results aren't that much different than Techradar's.

But DXO defaults to "print" which compares them more fairly at equalized sizes.

--
Rick Krejci
http://www.ricksastro.com

David Franklin said:

Again, considering that no one but testing engineers, and technno enthusiasts who emulate them, actually make direct use of raw files, but all actual photographers who seek to create any actual pictures must use 16 bit Tiffs or lesser picture files, to continue your anology, all roads everywhere are gravel roads, except those that are worse.

You may feel that it is somehow cosmically unfair that the one cannot actually produce picture files that perfectly reflect the characteristics that your testing procedures indicate in raw files; however, current realities are what they are, and you must learn to deal with them on their own level, as they really exist.

David
--
Keep learning; share knowledge; think seriously about outcomes; seek wisdom.

Click to expand...

A raw file can be shared so everyone can view & manipulate the data directly & come to their own conclusions.

A good example is Imaging Resource making the raw files for the 5DMKIII & D800 available for direct comparision.

I would much rather have a raw file's to compare cameras than Tiff's , the IR raw files came in real handy exposing the manipulated nonsense Legion5 was posting in his threads.

Since everybody can view the raws directly on their monitor it is the raws that should be measured.

To use the previous analogy, if the only roads in your world are gravel roads, the blacktop performance of a car is irrelevant. So just maybe techradar is more cognizant of the gravelly reality than dxomark.

5tve said:

I would much rather have a raw file's to compare cameras than Tiff's , the IR raw files came in real handy exposing the manipulated nonsense Legion5 was posting in his threads.

Since everybody can view the raws directly on their monitor it is the raws that should be measured.

Click to expand...

Actually, no - I don't think anyone can view RAW files directly on their monitor. Perhaps you meant something else?

Best,

--
'There is no un-suck filter.'

• David duChemin

Which is too bad, because I could sure use one.
http://www.flickr.com/photos/jasper_rubin/

Yes, you are correct. Everybody cannot "see" a raw file on their screen - no one can; they are actually looking at a preliminary raw conversion (probably a Tiff or similar file) that completes when you save it to disk. And when any site like IR offers to allow you downloads and you do so, you must somehow use whatever software you choose to preview and or convert them fully to another viable picture format, such as Tiff or Jpeg. Your choices in setting the conversion software and its parameters is what will determine your results at least as much as the content of the raw file itself. Apparently, this must not be widely understood, judging from some of the comments I've read on this site.

Regards,
David
--
Keep learning; share knowledge; think seriously about outcomes; seek wisdom.

Thank you for the response/confirmation. While my technical knowledge is limited,
I can appreciate the basic logic of how things things work, for the most part.

Unfortunately, the introduction of two new cameras from competing manufacturers has been an apparently dizzy-making affair for this forum, and analytical reasoning seems to have has suffered a bit. I'm sure things will right themselves soon enough, and folks will be back to explaining things like back-button focusing to new-comers.

Best,

Jasper

--
'There is no un-suck filter.'

• David duChemin

Which is too bad, because I could sure use one.
http://www.flickr.com/photos/jasper_rubin/