Luminous Landscape Forum
Equipment & Techniques => Cameras, Lenses and Shooting gear => Topic started by: The View on October 30, 2008, 05:53:31 pm
-
Dpreview has the review of the 50D out.
http://www.dpreview.com/reviews/canoneos50d/ (http://www.dpreview.com/reviews/canoneos50d/)
Main points: no real detail advantage over the 10MP 40D, but 50% larger files.
Worse high ISO performance than the 40D.
The 50D seems to be a victory of the marketing idiots over the photographers and the engineers by pumping unnecessary 15MP into an APS-C sensor.
-
They've made a few changes to the review since this morning. (ACR 4.6 instead of ACR 4.6 RC.) People are still fighting over it. One claim is that DPP is much better with 50D files than ACR. I dunno. I've seen samples where the 50D is markedly better than the 40D but those weren't controlled tests. So I think I'm hanging on to the 40D for a while. I would like the better LCD and the improved seals.
-
They've made a few changes to the review since this morning. (ACR 4.6 instead of ACR 4.6 RC.) People are still fighting over it. One claim is that DPP is much better with 50D files than ACR. I dunno. I've seen samples where the 50D is markedly better than the 40D but those weren't controlled tests. So I think I'm hanging on to the 40D for a while. I would like the better LCD and the improved seals.
Oh yes, every Canon SLR performs much better in DPP than in ACR.
There might be something true about that.
Maybe dpreview should not use ONE RAW converter for every camera but the best suited RAW converter. For Canon SLRs it's nonsense not to at least include DPP processing (I completely stopped using ACR as DPP results are so much superior).
Anyway, the basic truth of the review might be that engineers and photographers need to have more power than the marketing people.
Marketing people should market what's there in products, not have the power to get products that are easy to market, as it defeats the purpose of marketing: winning customers.
I love Canon SLRs, the character of their images, the colors, the high ISO performance. I hope they'll get back on track and reduce marketing people back to what they are: hired help. And if they can't market good products, they should get the hell out of the business and have smarter people do their job.
-
hehe, dynamic range (Section 19 (http://www.dpreview.com/reviews/canoneos50d/page19.asp)): ISO100: 8.3 EV ISO1600: 8.5 EV
One day the DPreview guys will find out that the Earth is flat.
The RAW noise comparision 40D vs 50D (Section 18 (http://www.dpreview.com/reviews/canoneos50d/page18.asp)) seems strange. Despite the higher resolution of the 50D sensor, if the test was done properly the results for the 50D are very disappointing.
BR
-
One day the DPreview guys will find out that the Earth is flat.
BR
If they step too close to the edge, they'll find out there's no railing.
-
The DPReview crew have proven again, that they don't have even a rudimentary understanding of "dynamic range".
At ISO 200 = 400 = 800 = 1600 = 8.5 EV. The three stop loss in the highlights is totally made up by higher ISO gain - is that not great? (forget about the rest)
-
The DPReview crew have proven again, that they don't have even a rudimentary understanding of "dynamic range".
At ISO 200 = 400 = 800 = 1600 = 8.5 EV. The three stop loss in the highlights is totally made up by higher ISO gain - is that not great? (forget about the rest)
hey Gabor, and they managed to accurately quantify the DR improvement with HTP, not just 1EV, but a fantastic 1.1EV! perhaps you should start using HTP in your panos to prevent the skies from blowing
-
You can't properly measure dynamic range without looking at the raw sensor data in linear light. To try and do it off a JPEG will lead to vastly misleading results, and doing it of RAW in ACR trying to get a "best curve" is nearly as bad. At least they use a test wedge.
Graeme
-
You can't properly measure dynamic range without looking at the raw sensor data in linear light. To try and do it off a JPEG will lead to vastly misleading results, and doing it of RAW in ACR trying to get a "best curve" is nearly as bad. At least they use a test wedge.
That's right; it needs to be measured on the pure raw data (even if the light happens to be curved :-).
The result is, that the noise characteristics is the same as that of the 40D.
-
I really appreciate the camera reviews completed by dpreview.com. They are very comprehensive and informational. There are several issues with the measurement systems and conclusions, however.
First, the use of 100% crops to compare with other cameras and draw conclusions about resolution and noise is a mistake, as it magnifies each camera to a different degree, like holding one print under a loupe and another at arm's length. If the reviewer were to resample images from both cameras to the same resolution (or just print them), the conclusions about noise and resolution would be far different.
The other issue is the software and settings used for RAW conversion. All the popular converters apply vastly different conversion styles, behind the scenes, to different camera models, even from the same manufacturer. Therefore the only fair way to analyze what the camera is capable of is to use the un-popular software that treats RAW data the same, and to measure that RAW data. Using dcraw and IRIS would allow authoritative measurement of the true noise performance.
If dpreview were to correct these two mistakes, I believe they would quickly realize how their crusade against pixel density is based entirely on a false premise.
-
Dpreview has the review of the 50D out. ...
Main points: no real detail advantage over the 10MP 40D, but 50% larger files.
Yes, it does seem that the 50D would have been a "better" camera had it also used an updated version of the 10 MP sensor in the 40D, rather than jumping to a 15 MP sensor. Cleaner 10 MP files that can be interpolated if need be, and are inherently able to be sharpened more smoothly, would seem to be preferable to me. I wonder how sales for Canon would go if it ever released a 10 MP edition of the 50D? However, I estimate that that's rather unlikely.
-
You can't properly measure dynamic range without looking at the raw sensor data in linear light. To try and do it off a JPEG will lead to vastly misleading results, and doing it of RAW in ACR trying to get a "best curve" is nearly as bad. At least they use a test wedge.
Graeme
There is an over emphasis on jpeg comparisons at dpreview, from my perspective because I don't use jpeg mode. However, it's probably true that most users of DSLRs do shoot jpegs, and therefore it is only sensible for Dpreview to cater to those people.
For the same reasons, most people who shoot RAW probably also use ACR to convert their RAW images, therefore does it not seem sensible that Dpreview should use ACR to compare the DR of different cameras?
It's quite likely that there are more objective methods of measuring and comparing DR, using, for example, DCRAW, IRIS, Rawnalize etc, but how do results from such 'possibly more objective' analyses help users of ACR? If you are a user of ACR, you want to know what the DR of your new camera will be like in ACR. If you think DPP does a better job, then that's reason enough to use DPP instead of ACR. However, why should there be any reason for a DCRAW based analysis to be any more relevant to DPP users than ACR users?
You can't satisfy all of the people all of the time, ya know!
-
It's quite likely that there are more objective methods of measuring and comparing DR, using, for example, DCRAW, IRIS, Rawnalize etc, but how do results from such 'possibly more objective' analyses help users of ACR?
By showing the difference between the capability of the camera and the limitations of the software. It's one thing to say "the camera stinks" and another to say "the camera is good, but Adobe's raw processor does a poor job on this one."
-
Yes, it does seem that the 50D would have been a "better" camera had it also used an updated version of the 10 MP sensor in the 40D, rather than jumping to a 15 MP sensor. Cleaner 10 MP files that can be interpolated if need be, and are inherently able to be sharpened more smoothly, would seem to be preferable to me. I wonder how sales for Canon would go if it ever released a 10 MP edition of the 50D? However, I estimate that that's rather unlikely.
I actually expected something like 12 MP, which is considered as the limit for the APS-S sensor.
But I agree, 10MP would even have been better.
-
most people who shoot RAW probably also use ACR to convert their RAW images, therefore does it not seem sensible that Dpreview should use ACR to compare the DR of different cameras?
I don't think that "most people" use ACR. Many, yes, but most?
A camera can't be tested with what "most people" use, but what the most knowledgeable people use. And that's often not ACR.
The best test is done in a way that is best for a certain camera.
The best tests are done in a way that matches actual shooting situations.
And do the box shots of dpreview match those?
Of course not.
Also: those close crops often hide how a sensor represents light quality. How colors are represented.
Test sites have a big influence on camera manufacturers.
In a worst case scenario it could give us cameras that perform well under those test conditions - not in real shooting situations.
This is why I personally prefer subjective reviews of photographers, that do their work and give reasons what the strengths and weaknesses of a certain camera are.
Then I don't need any charts.
That said, the results of this dpreview test don't seem to sustain our knowledge, that pixel packed sensors are not something we really want. They seem to have gotten the right results with the wrong methods.
-
So you think that interpolating sparse data is more correct that sampling denser data?
I don't think that interpolation ever is a replacement for real data. I definitively would suggest that there are advantages in increasing resolution, but it is well possible that the limits for diminishing returns have been reached. The resolution needed is dependent on the size of the enlargement. There is some advantage to having higher resolution than the resolution of the lens, namely that a much weaker anti aliasing filter can be used.
The advantage of higher pixel density would only be visible in:
- Large prints (or pixel peeping)
- With the very best optics
- At the best aperture
So if not all of the above are fulfilled no or little advantage would be seen.
Best regards
Erik
Yes, it does seem that the 50D would have been a "better" camera had it also used an updated version of the 10 MP sensor in the 40D, rather than jumping to a 15 MP sensor. Cleaner 10 MP files that can be interpolated if need be, and are inherently able to be sharpened more smoothly, would seem to be preferable to me. I wonder how sales for Canon would go if it ever released a 10 MP edition of the 50D? However, I estimate that that's rather unlikely.
-
Hi,
There is no real good reason to talk about a 10, 12 or 15 MPixel limit. The linear difference between 15 and 12.5 MPixels is 9.5%. In terms of resolution if a 12.5 MPixel sensor needs 50 lp/mm a 15 MPxel sensor would require 55 lp/mm, that is a small difference. Also I'm pretty sure that DR will not be reduced more than half a stop and I cannot see that the speed advantage of a 10 MP sensor over a 15 MP sensor would be more than a half stop.
Essentially, I don't think that there is a limit but we have diminishing returns when increasing MPixels if other factors are kept fixed.
Erik
Note: the 50 lp/mm example is contrived. A 12.5 MPixel sensor has a pitch of 5.5 microns and has an inherent resolution of 91 lp/mm. Resolution on it's own is a meaningless figure, however. The only way resolution makes sense is in the combination with the MTF. So we could say an MTF of 50% at 50 lp/mm or an MTF of 20% at 90 lp/mm. Very few lenses for 135 format would achieve something like 20% at 90 lp/mm but it is definitively possible to design a lens having that capability.
I actually expected something like 12 MP, which is considered as the limit for the APS-S sensor.
But I agree, 10MP would even have been better.
-
I don't think that "most people" use ACR. Many, yes, but most?
A camera can't be tested with what "most people" use, but what the most knowledgeable people use. And that's often not ACR.
The best test is done in a way that is best for a certain camera.
The best tests are done in a way that matches actual shooting situations.
Sorry! Can't agree with your statements. Perhaps my statement, "Most people use ACR', should be modified to 'Most people who are concerned about good image quality, use ACR.' Photoshop is not the only image processor, but it appears to be the best, if you can afford it.
A camera can be tested with what most people use. Dpreview do it.
The best test is done in a way that is best for a certain camera.
Such a test is not always feasible. If a reviewer had the time and resources, he/she could compare all available converters that could handle the new camera's RAW files. How many converters are there, currently, that can convert 50D RAW files? If there were an obscure converter available that somehow could drag more DR out of the RAW files, people who are concerned about image quality would of course be interested.
If anyone knows about such a converter, please speak up now. Show us the results so we can see how much better 50D DR is with this new or obscure converter.
-
I was rather pleased DP called for an end to the megapixel race.
Please repeat after me:
"There is more to image quality than megapixels alone"
;-)
-
I was rather pleased DP called for an end to the megapixel race.
I have a question here: if we assume the Canon 50D with a 4.5 MP/cm² pixel density has reached the limits of the Mpx race as DP tests seem to demonstrate, how are there people around (including Ken Rockwell) managing to take nice pictures even at 100% crops with their Canon G10 and 34 MP/cm² pixel density?
If I see a limit for Mpx not to go to 20Mpx or above, is more related to the practical usefulness and the storage and processing power requirements of such huge resolutions, than to physical limits.
-
Perhaps next we'll see the MTF race as well as the dynamic range race? There's always an aspect of camera design you can put a number against and race on it.
-
Also: those close crops often hide how a sensor represents light quality. How colors are represented.
If you think that's a problem, then you can download the full image and judge for yourself.
The only image they don't provide the full version of, is the one they base their noise measurements off of (the GretagMacbeth color checker chart, but there you can compare the various colour settings, and with a different camera, too).
-
I have a question here: if we assume the Canon 50D with a 4.5 MP/cm² pixel density has reached the limits of the Mpx race as DP tests seem to demonstrate, how are there people around (including Ken Rockwell) managing to take nice pictures even at 100% crops with their Canon G10 and 34 MP/cm² pixel density?
If I see a limit for Mpx not to go to 20Mpx or above, is more related to the practical usefulness and the storage and processing power requirements of such huge resolutions, than to physical limits.
I think there are a number of points here.
Ist off, is that we are already at fairly high levels of resolution anyway, the increases we get nowadays are far less dramatic than years before. For example going from a 3/4 mp camera to 10mp odd, was a fairly decent jump. However, at 10mp levels, a few extra here and there, are pretty much insignificant. At 15mp, frankly you are going to have to have a big increase to be of any use at all.
2nd point, limitations of lenses, which is of course tougher on smaller sensors..there will be a point that adding more equals nothing extra. You could argue that lenses can be improved, well to a point they can, but it is still a valid issue.
I think now we are at 24mp for a FF DSLR, really, do we need any more than this??? Take it to 30mp and it is not really worth bothering with. The game is up really...big numbers are not enough. I suspect over time the industry may move to FF, as APS-C is clearly running out of pixels for them to slap on there. I just hope they think of more interesting areas to explore. But cost is a factor, expect FF to reach affordable levels in the next few years..maybe APS will be reserved for lower end bodies only, who can say.
The G10 looks decent enough, but I doubt these 15mp are noticeable over the previous model's 12mp odd, and I would imagine it is not very good over ISO 400.
I think digital will really come into it's own, when we get rid of bayer sensors (I don't think they are up to the job colour/tonal wise), and when dynamic range is really much better than today. Maybe your average joe buys cameras on megapixels alone, I cannot say that I do.
-
Perhaps next we'll see the MTF race as well as the dynamic range race? There's always an aspect of camera design you can put a number against and race on it.
For around 10 years, all the Japanese car manufacturers agreed not to release any model with more the 280 hp. They focused on enhancing the torque and improving fuel efficiency of high power enginers.
My view is that camera manufacturers should agree on an explicit resolution cease fire, say at 25 MP for FF, 15MP for APS (12 would have been better but Canon wouldn't want to back down) and 8MP for compacts, and work on enhancing the other variables.
As long as the MP race goes on, photographers will not get the tool they could be getting because the designs will keep being too heaviliy biased towards more MP.
Cheers,
Bernard
-
I'd think you'd want more megapixels. Get enough to fully capture the lens diffraction.
-
Main points: no real detail advantage over the 10MP 40D, but 50% larger files.
It seems to me that the issue of extra noise becomes moot if there is no extra detail, and if there is no extra detail then there is no value in larger file sizes. Judge for yourself from the crops below:
(http://photos.imageevent.com/tonybeach/mypicturesfolder/sharing//Untitled-3%20copy.jpg)
One is from the 50D (100%), one is from the 40D and another is from the D300 (both resized to match the 50D). Can you pick which one is which?
-
(http://photos.imageevent.com/tonybeach/mypicturesfolder/sharing//Untitled-3%20copy.jpg)
Can you pick which one is which?
I would bet left to right: 40D, 50D and D300!
Anyway it's not only the sensor that is enough to capture a certain degree of detail. A bad lens or focusing on a perfect sensor would ruin the image. To increase resolution is a necessary condition to achieve more detail than a given balanced system, not a sufficient condition (or whatever you call it in English).
And in any case the difference in resolution of the 50D with respect to the 40D is moderate, not huge, so no huge improvement can be expected.
(http://img50.imageshack.us/img50/8839/40dvs50dbisty7.gif)
-
I would bet left to right: 40D, 50D and D300!
You got one out of the three. Anyone else care to venture a guess?
Anyway it's not only the sensor that is enough to capture a certain degree of detail. A bad lens or focusing on a perfect sensor would ruin the image. To increase resolution is a necessary condition to achieve more detail than a given balanced system, not a sufficient condition (or whatever you call it in English).
I believe the problem is the AA filter Canon is using. Notice that at DPR there is no moire detected at any level in the 50D resolution numbers. Their reason was obvious, by applying a stronger AA filter they could reduce noise.
And in any case the difference in resolution of the 50D with respect to the 40D is moderate, not huge, so no huge improvement can be expected.
It is bordering on insignificant, certainly less than it should be considering the differences I can readily detect between a 10 MP D200 and a 12 MP D300.
-
I'd think you'd want more megapixels. Get enough to fully capture the lens diffraction.
If that didn't come at the price of something else becoming worse, yes sure.
As far as I am concerned give me more DR and less noise at 25 MP. That will impact my photography a lot more positively than an additional 5 or 10% of detail.
Cheers,
Bernard
-
I have been considering buying the 50D as an upgrade for my 30D (perpetually a tweener, I am), so I have been following this closely. As a matter of fact, I was unfortunate enough to jump into the DPReview forum scrum. I had a couple of observations:
The results comparing resolution crops between DPP, ACR, Zoom Browser and jpg were a big part of the problem. The moire effect on the original ACR version was very bad. I happened to still have a browser window open to the old image after they updated the images, so I saved copies. Here they are:
Original ACR 4.6 RC Version
(http://www.trailpixie.net/temp-image/dpr50d/50DRes_acr-001.jpg)
Updated ACR 4.6 Version
(http://www.trailpixie.net/temp-image/dpr50d/50DRes_acr-001new.jpg)
This would have set off alarm bells with me when I saw it. They claimed that they updated all the RAW images in the review, but I am skeptical.
DPReview also observed that the Dynamic range performance was worse than the 40D. They based this on a magenta shift in the highlights although the 50D's actual range was wider than the 40D's. I am very curious if that same magenta shift in highlights would be observed in conversion using DPP. I would really like it if they released their RAW images. That would seem to put this all to rest.
As for resolution at 15MP, I am a bit disappointed. I agree that it probably would have been better to stick with 12MP. It does appear that the 50D gets a bit more detail in some circumstances. I spent some time reviewing their test shots and found that focus was a bit off in places, so the resolution improvement could be seen more clearly in parts of the image that they didn't enlarge. Most notable difference was visible in the text that runs vertically on the side of the middle bottle on their standard test image.
Arghh...but I am getting into the peeping. The 50D seems to have marginally the same image quality performance as the 40D with some areas that are a gnats @ss better and others that are the same amount worse, all with the extra gift of larger file sizes (CF card manufacturers and hard drive manufacturers will rejoice). But, the usability features included in the 50D are pretty nice upgrades, particularly the screen and ability to magnify on the screen to check focus.
As for the disappointment over the lack of improvement in image quality (high ISO, resolution, dynamic range) I suspect that we haven't heard the end of improvements to RAW file handling for this camera.
What I think what we have is a decent 12MP camera stretched out onto a 15MP sensor. I will need to do my own at-home tests to decide how large I can print with this camera. I may not be able to apply my simple algorithm where I take the pixel dimensions and divide by 240 to decide my maximum image size.
-
Can you take it into the shower with you and let me know if the weather sealing is improved?
-
A couple issues if most people use ACR why are there so many other RAW software converters out there? ACR is written by a company whose bread and butter is doing post processing if they converted RAW files that required very little or no post processing they would be in a world of hurt. I prefer a company that doses nothing but RAW processing and if it is done correctly very little or no post processing is required. I do very little post processing and in many instances there is no improvement in doing it, that is good RAW conversion. I believe the major issue with the 50D besides the testing procedures used is software related and it will be resolved.
-
By showing the difference between the capability of the camera and the limitations of the software. It's one thing to say "the camera stinks" and another to say "the camera is good, but Adobe's raw processor does a poor job on this one."
To determine that ACR does a poor job with 50D RAW files would involve an additional set of procedures and a new ball game. Dpreview is in the business of reviewing cameras, not RAW converters.
I own both the 40D and 50D and could find no significant difference between conversions of 50D raw images in DPP and ACR 4.6 beta, with NR and sharpening turned off completely in both converters. There were very subtle differences to be observed at the extreme pixel-peeping level, but in my opinion so small as to be totally irrelevant in the context of all the processing adjustments one makes before an image is ready for display or printing.
The resolution difference between the 40D and 50D are smaller than I expected. Having done a few tests using an Australian $50 bill as a target, I get the impression that the increase in resolution is comparable in magnitude to the decrease in resolution as one stops down one stop from a lens' sharpest aperture to an aperture where the lens begins to be affected by diffraction.
For example, with a good prime lens like the Canon 50/1.4, the 40D at F5.6 is about the same as the 50D at F8. The 40D at F8 is about the same as the 50D at F11 and the 40D at F11 is about the same as the 50D at F16, in terms of total detail captured at the plane of focus.
What this means for me, is that I would have less hesitation using F16 with the 50D when I want maximum DoF, because I know that at the plane of focus I will get results on a par with the 40D at F11, but away from the plane of focus I will get a significantly sharper image than the 40D can produce at F11.
However, the reasons I bought a 50D were not centred around the possibility of significantly higher resolution. I was having trouble getting accurate autofocussing with my 40D so I was attracted to the 'autofocus micro adjustments' feature of the 50D as well as its higher LCD screen resolution. My decision was also affected by the plunging value of the commodity-based Aussie dollar. It's going to make future imports more expensive and I would expect future shipments of the 50D to be more expensive.
-
Having done a few tests using an Australian $50 bill as a target
...
My decision was also affected by the plunging value of the commodity-based Aussie dollar
Next time make your tests with Yen. Anyway, if the value plunged, then you would not mind uploading some raw files from the tests, would you? (Pls different ISOs). I will reprogram Rawnalyze for the new exchange rate.
Seriously, I have too few raw files for my tests; I don't even know the saturation levels with certain ISOs.
Thanks
-
You got one out of the three.
great, then I got the best possible result right after getting three out of the three
-
A couple issues if most people use ACR why are there so many other RAW software converters out there?
First of all, I would say that most people don't shoot in RAW mode in any case. If one shoots in RAW mode and therefore has a use for a RAW converter, it's a sign that one is particularly concerned about subtleties of image quality, obtaining maximum detail in highlights and shadows etc and, in my opinion, quite likely that one will be a Photoshop user.
Of course, without extensive market research it's impossible to say whether greater than 50% of all people who shoot in RAW mode use ACR or Lightroom. Some people will use, or at least try out, a number of different converters and find that a particular converter produces more satifying result with certain types of scenes than ACR. I used to prefer RSP myself, for a while until it was bought out by Adobe. However, I found that ACR always did a better reconstruction job of blown highlights.
Is the actual percentage of ACR users really important? Photoshop and accompanying programs like Bridge and ACR are leading industry standards. That's the essential point. It would get very complicated to run through the whole gamut of available converters on the market every time a new camera was tested.
-
From Eric Chan (madmanchan) in a post over at Naturescapes:
the early support of the 50D in Camera Raw 4.6 RC did not estimate the white point correctly; this led to suboptimal treatment of the highlights during highlight recovery in CR. There was also an issue involving the balance of green channels that led to very suboptimal resolution. Both addressed in the final version of CR 4.6, as well as CR 5.1.
Apparently DPR used the beta version of the camera profile for doing their tests
-
DPReview also observed that the Dynamic range performance was worse than the 40D. They based this on a magenta shift in the highlights although the 50D's actual range was wider than the 40D's. I am very curious if that same magenta shift in highlights would be observed in conversion using DPP. I would really like it if they released their RAW images. That would seem to put this all to rest.
(...)
As for the disappointment over the lack of improvement in image quality (high ISO, resolution, dynamic range) I suspect that we haven't heard the end of improvements to RAW file handling for this camera.
It seems to be one of the key flaws of dpreview that they set testing standards just as they feel like it.
Why use ACR as a standard? It makes no sense.
It's a fact that many Nikon and Canon photographers prefer Capture NX 2 and DPP to ACR.
Now, which RAW converter would a manufacturer try to optimize his camera for? For his own RAW converter or a third party product?
Hard guess.
For the 50D, many photographers will use DxO, DPP, C 1 Pro, Bibble, ACR, etc.
Is it too much to ask to process the test images in all major RAW converters, decide in which it performs best, and then give us this best shot.
Of course, that's more work.
But how much work would it be to develop a camera, and see some cowboys ride it to death with poor riding skills on the test circuit?
-
I suppose I should reconvert the following images now that it seems ACR 4.6 beta was so lousy with respect to the 50D.
The following shots (converted with 4.6 beta) show comparisons between the 40D at F11 and the 50D at F16, at 200% enlargement. The Canon 100-400 zoom was used, which is not as good as a prime I know, but in my experience all lenses ranging from moderately good to excellent tend to have very similar perfromance at apertures from F11 downward.
For the purpose of the comparison, the 40D image was interpolated with bicubic to the same size as the 50D image. The conversions have had no processing except default sharpening in ACR and negative EV adjustment with respect to ETTR. The 50D image does appear to have slightly more noise.
In order, from left to right; (1) the full scene with no interpolation of the 40D image; (2) 200% crops of the focus area. LiveView was used for manual focus. (3) the DoF benefits of using F16 with the 50D.
[attachment=9362:full_sce...parisons.jpg] [attachment=9363:point_of...mparison.jpg] [attachment=9364:F11_40D_..._F16_50D.jpg]
Edit: Just for the record, the 50D shot has more noise because I used ISO 400. The 40D shot is at ISO 100. I intended to equalize shutter speed, but made an error, thinking the 40D shots, taken first, were at ISO 200.
-
It seems to be one of the key flaws of dpreview that they set testing standards just as they feel like it.
Why use ACR as a standard? It makes no sense.
It's a fact that many Nikon and Canon photographers prefer Capture NX 2 and DPP to ACR.
Some people might prefer one converter over another for a whole range of reasons that have little to do with fundamental image quality in respect of DR. It might simply be that one particular converter produces an effect which is easier to get, or another converter has better noise reduction algorithms, yet another converter produces a more accurate 'as shot' white balance. I used to prefer Raw Shooter Premium mainly because of its 'vibrancy' and 'detail enhancement' controls which ACR lacked at the time but now has.
One can argue till the cows come home about which converter is better, but such discussions are quite meaningless without comparison images.
If you want to make a point that a particular converter is better than ACR with respect to a particular model of camera, then please show some comparison images so we can all benefit. I'm sure all of us want to get the best out of our RAW images.
The last time I saw a comparison between a DPP conversion and an ACR conversion on this site (apart from my own comparison) the poster had clearly used a lot of noise reduction with the DPP software and very little or no noise reduction with ACR. Now clearly that's not a fair comparison.
-
great, then I got the best possible result right after getting three out of the three
Yep, I had thought of that too; but you could have gotten any one of the three right or none of them too -- obviously, if you had gotten two than you would have gotten the third by default.
I'll help you out here, you got the first one right.
-
A couple issues if most people use ACR why are there so many other RAW software converters out there? ACR is written by a company whose bread and butter is doing post processing if they converted RAW files that required very little or no post processing they would be in a world of hurt. I prefer a company that doses nothing but RAW processing and if it is done correctly very little or no post processing is required. I do very little post processing and in many instances there is no improvement in doing it, that is good RAW conversion. I believe the major issue with the 50D besides the testing procedures used is software related and it will be resolved.
What?!?
-
One can argue till the cows come home about which converter is better, but such discussions are quite meaningless without comparison images.
There are many raw converters. Each raw converter uses different conversion behind the scenes for different camera models. Each converter has many, many options that may be changed, none of which correspond with other raw converters. Noise reduction: off in one converter is noise reduction: medium in another. New versions are released. It leads to an infinite combination of possibilities for comparison.
Thankfully, there is a very simple and straightforward method for measuring camera performance: look at the RAW data itself using dcraw or IRIS, where every camera is on a level playing field.
-
Edit: Just for the record, the 50D shot has more noise because I used ISO 400. The 40D shot is at ISO 100. I intended to equalize shutter speed, but made an error, thinking the 40D shots, taken first, were at ISO 200.
I'm afraid that kind of negates the ability to compare them. At first I looked at the middle two images without looking at the title bar, and I couldn't decide which was which until I saw the noise in the roof and knew that was ISO 400. I would be least concerned with shutter speed differences, particularly if it was on a tripod.
-
Yep, I had thought of that too; but you could have gotten any one of the three right or none of them too -- obviously, if you had gotten two than you would have gotten the third by default.
I'll help you out here, you got the first one right.
hey, that means I didn't manage to distinguish which camera had the highest pixel resolution (50D), but I managed to distinguish which one had the lowest resolution (i.e. worst result the 40D), and managed to distinguish which camera of the same brand (Canon) provided the highest detail and it happened to be the one with higher resolution.
Anyway, I think it was not a fair test to the 40D and D300 (even if it won) since they were rescaled and that means detail loss. The only right test I can think of would be using a good enough lens in all cameras so the resolution limit is set only by the sensor, and then shoot over a specific chart to measure lpm at 100% crop each camera provides.
As you mention, a stronger AA filter could make a better sensor in principle capture less detail than a worse sensor in favour of the AA capabilities, rarely needed. I wish the AA filter was a choice of the user or could be removed easily. I have a feeling camera makers are conservative on this, and prefer to reduce sharpness in the RAW data than risking to have a bunch of users complaining because one day they made one shot over a particular subject with a particular high spatial frequency and developing the RAW file with a particular RAW developer they got strange artifacts, and showing this all around Internet.
BR
-
I'm afraid that kind of negates the ability to compare them. At first I looked at the middle two images without looking at the title bar, and I couldn't decide which was which until I saw the noise in the roof and knew that was ISO 400. I would be least concerned with shutter speed differences, particularly if it was on a tripod.
I wasn't addressing noise issues in that test. The lens was on a tripod, but just to be sure, I thought it would be better to use the same shutteer speed.
I have no reason to suppose the 50D has more noise than the 40D at any ISO. I'm a firm believer in comparing equal size images.
Comparing the 50D using a good prime at F11, with the 40D at F5.6, I was surprised to find that the 50D is better. In the crops below, any detail you see in the 40D crop that isn't present in the 50D crop is pure aliasing artifact. The lens is the Canon 50/1.4, both shots were at ISO 100 and the converter was ACR 4.6, default sharpening.
From left to right: (1) the uncropped image, (2) 200% crops, (3) a close-up of the banknote showing the full detail.
[attachment=9371:full_scene.jpg] [attachment=9370:non_beta...1___F5_6.jpg] [attachment=9372:0923_full_detail.jpg]
-
hey, that means I didn't manage to distinguish which camera had the highest pixel resolution (50D), but I managed to distinguish which one had the lowest resolution (i.e. worst result the 40D), and managed to distinguish which camera of the same brand (Canon) provided the highest detail and it happened to be the one with higher resolution.
No doubt the 50D does better than the 40D, but it should do better than the D300 and it doesn't. If Canon had stuck with 12 MP they could have gained fps (6 fps could have become 7.5 fps) and they could have improved the noise performance. We all know that Canon chose more megapixels because it looks better to an unsavvy consumer sitting on the shelf next to a camera with fewer megapixels.
Anyway, I think it was not a fair test to the 40D and D300 (even if it won) since they were rescaled and that means detail loss.
No loss of detail; just less detail available to be enlarged.
The only right test I can think of would be using a good enough lens in all cameras so the resolution limit is set only by the sensor, and then shoot over a specific chart to measure lpm at 100% crop each camera provides.
In terms of resolution, I've been arguing that APS-C and DX formats top out 10-12 megapixels, beyond that additional resolution is just not nearly as useful as improving other image quality characteristics would be.
-
You're turning things around. A software maker tries to make its converter optimized for the cameras served, not the other way around.
It has been argued that the camera maker knows their own product better than third party SW makers. But OTOH one might argue that a specialized SW developer is better at this game than a hardware maker. At least for workflow and UI Adobe is better. They are the reasons 'many' don't use DPP. For image quality there are so many variables that comparisons are difficult.
Now, which RAW converter would a manufacturer try to optimize his camera for? For his own RAW converter or a third party product?
Hard guess.
-
I have no reason to suppose the 50D has more noise than the 40D at any ISO. I'm a firm believer in comparing equal size images.
Comparing the 50D using a good prime at F11, with the 40D at F5.6, I was surprised to find that the 50D is better. In the crops below, any detail you see in the 40D crop that isn't present in the 50D crop is pure aliasing artifact. The lens is the Canon 50/1.4, both shots were at ISO 100 and the converter was ACR 4.6.
I can see merit in your idea of comparing at the same size, but your two images from the 40D and 50D cannot be compared due to not using the same F stop.
-
A software maker tries to make its converter optimized for the cameras served, not the other way around.
More specifically, the software maker makes a profile -- or profiles -- that are (ideally) optimized for the converter to use on a camera's files.
It has been argued that the camera maker knows their own product better than third party SW makers. But OTOH one might argue that a specialized SW developer is better at this game than a hardware maker. At least for workflow and UI Adobe is better. They are the reasons 'many' don't use DPP. For image quality there are so many variables that comparisons are difficult.
I don't think the specialty of the company is as critical as the competence and number of people assigned to the task. For Adobe the problem is that they are seeing new cameras from many companies coming out that they have to profile and update ACR for ASAP; whereas the camera companies (theoretically) know the spectral characteristics of their latest cameras and have had adequate time to prepare profiles using their own software to get the maximum quality from those files, and they have fewer cameras to profile for. My problem with ACR has to do with this image quality gap, and with very costly upgrades to the entire software package to handle the latest cameras -- otherwise you can get the DNG converter, but that does not facilitate a seamless workflow.
-
Some people might prefer one converter over another for a whole range of reasons that have little to do with fundamental image quality in respect of DR. It might simply be that one particular converter produces an effect which is easier to get, or another converter has better noise reduction algorithms, yet another converter produces a more accurate 'as shot' white balance. I used to prefer Raw Shooter Premium mainly because of its 'vibrancy' and 'detail enhancement' controls which ACR lacked at the time but now has.
One can argue till the cows come home about which converter is better, but such discussions are quite meaningless without comparison images.
If you want to make a point that a particular converter is better than ACR with respect to a particular model of camera, then please show some comparison images so we can all benefit. I'm sure all of us want to get the best out of our RAW images.
The last time I saw a comparison between a DPP conversion and an ACR conversion on this site (apart from my own comparison) the poster had clearly used a lot of noise reduction with the DPP software and very little or no noise reduction with ACR. Now clearly that's not a fair comparison.
Why don't YOU show us some tests?
I see you talking about it a lot, but if you were able to prove ACR's superiority over any other RAW converter, you'd have long done it.
There are many who prefer Nikon Capture and Canon DPP to ACR and they don't need to do any heavy testing to see the difference.
It's pointless to argue on this level, no matter where they cows are, if they stay home, go out, or come home.
-
There are many raw converters. Each raw converter uses different conversion behind the scenes for different camera models. Each converter has many, many options that may be changed, none of which correspond with other raw converters. Noise reduction: off in one converter is noise reduction: medium in another. New versions are released. It leads to an infinite combination of possibilities for comparison.
Thankfully, there is a very simple and straightforward method for measuring camera performance: look at the RAW data itself using dcraw or IRIS, where every camera is on a level playing field.
Good point.
As photographers, we should stay away from brand fidelity and fanboy-ism.
Until good testing is in place, we have to rely to real photographers and their reports on how cameras perform and under what conditions, and how the images are processed.
-
I don't think the specialty of the company is as critical as the competence and number of people assigned to the task. For Adobe the problem is that they are seeing new cameras from many companies coming out that they have to profile and update ACR for ASAP; whereas the camera companies (theoretically) know the spectral characteristics of their latest cameras and have had adequate time to prepare profiles using their own software to get the maximum quality from those files, and they have fewer cameras to profile for. My problem with ACR has to do with this image quality gap, and with very costly upgrades to the entire software package to handle the latest cameras -- otherwise you can get the DNG converter, but that does not facilitate a seamless workflow.
Good point. It should be easier to specialize a RAW conversion software to one particular kind of files form one company than to have to serve all cameras.
Also, Canon and Nikon do not fully disclose their files, so it's a bit of a guessing game for third party manufacturers.
-
What we all want is testing that doesn't as much show the preference of the tester, but data: how a particular camera performs a particular task.
Daniel Browning has mentioned that it is possible to get unobstructed data from the sensors.
Or, as mentioned, use at least the original RAW converter for the testing, too, for Canon, Nikon, Leaf, Hasselblad.
Because we are interested how the camera is doing, not what a third party RAW converter is doing with it.
Other than that: a real, flesh and blood photographer who takes a new camera through his own way of working and reports his findings will always be important for judging if a certain camera is be right for you.
Isn't this forum based on exactly this: that we need to get other people's personal opinions?
-
What we all want is testing that doesn't as much show the preference of the tester, but data: how a particular camera performs a particular task
Well, the "particular task" is a nonsense. A DSLR camera's only task is to deliver pixels; the rest is called post processing. If you don't understand this, then buy a P&S, the task of which is to shoot, print and frame a picture and hang it on the wall, no matter what you were doing while pressing the button.
As to dealing with the data, i.e. with the unadultered raw data, as far as it is possible: I have been dealing with the measurement of measurable. I compiled a serie of noise measurements, which show that the 50D has the same noise characteristics as the 40D. Not bad, regarding the 50% higher pixel count:
Layered TIFF with captures (http://www.panopeeper.com/Demo/DR_Comparison_Canon50Dvs40D.tif)
I find the concept of "image noise" a nonsensical excercise and do not engage in such speculations.
-
Why don't YOU show us some tests?
I see you talking about it a lot, but if you were able to prove ACR's superiority over any other RAW converter, you'd have long done it.
There are many who prefer Nikon Capture and Canon DPP to ACR and they don't need to do any heavy testing to see the difference.
It's pointless to argue on this level, no matter where they cows are, if they stay home, go out, or come home.
I do show tests. Haven't you noticed? Whenever I make a point and am able to demonstrate that point with tests, because I have the equipment, I usually do so. When I tried RSP a few years ago and found it produced sharper results than the then current version of ACR, I demonstrated the fact with test images on this forum.
Isn't this the whole point of the forum? Seeing is believing. As I mentioned before, there might be many reaosns why someone prefers a particular converter. It might simply be a quicker and easier way of getting the same results, or the controls might be more intuitive for one particular user, but not necessarily another.
Of course, I don't need to do this. I sometimes wonder why I bother. But the fact is, the very process of organising my own test results into a demonstration format, tests which are initially made for my own edification, clarifies the issues in my mind and helps me discover flaws in my methodology. I learn from the process. If I didn't, I would no longer bother.
There's no reason for me to prove ACR's superiority. I'm not making any claims for ACR other than it's probably, on balance, as good as any other converter on the market but has the advantage of being more of an industry standard than any other converter. If someone wishes to make the claim that another converter does a better job than ACR, then let them demonstrate it so we can all benefit. You don't go through life believing everything that everyone says, do you?
-
has the advantage of being more of an industry standard than any other converter
ACR is my preferred raw converter at the moment, but I do not find any single factor supporting this argument.
-
I can see merit in your idea of comparing at the same size, but your two images from the 40D and 50D cannot be compared due to not using the same F stop.
We should all know by now that the resolution differences between the 50D and 40D are very slight. My test results in this thread are an exploration of how such slight differences might be of practical value to me.
It has been suggested many times that such high-pixel-density cameras have no benefit at small apertures and that any resolution advantage will only be apparent with exceptionally good lenses at their sharpest apertures.
My own tests are suggesting, some of which I'm sharing with you, that the real benefit of this very slight resolution advantage of the 50D, is that one can confidently stop down an additional stop without sacrificing any resolution (compared with the 40D at the plane of focus), yet gain the more significant advantage of the greater DoF that stopping down affords.
-
ACR is my preferred raw converter at the moment, but I do not find any single factor supporting this argument.
Of course not. It's a conglomeration of factors which might support this argument. As I mentioned, on balance, taking all factors into consideration. There might also be quite a few amateurs who shoot in RAW mode and who don't use ACR because they simply can't afford to buy Photoshop, but would nevertheless like to use ACR.
-
Well, the "particular task" is a nonsense. A DSLR camera's only task is to deliver pixels; the rest is called post processing. If you don't understand this, then buy a P&S, the task of which is to shoot, print and frame a picture and hang it on the wall, no matter what you were doing while pressing the button.
As to dealing with the data, i.e. with the unadultered raw data, as far as it is possible: I have been dealing with the measurement of measurable. I compiled a serie of noise measurements, which show that the 50D has the same noise characteristics as the 40D. Not bad, regarding the 50% higher pixel count:
Layered TIFF with captures (http://www.panopeeper.com/Demo/DR_Comparison_Canon50Dvs40D.tif)
I find the concept of "image noise" a nonsensical excercise and do not engage in such speculations.
A DSLR is a device to deliver pixels?
And the rest is post processing?
I'd say, the rest is silence.
Are you actually a photographer, or just a pixel peeper?
The task of an DSLR you could use for would be night photography, low light photography, portrait photography. And you can test for he quality of file that you get from a night shot, a portrait shot, etc.
And then you use different RAW converters to see where you get the best file.
Regarding your posting style:
Anybody could use the formula "you don't understand".
An intelligent person just would not use it. Not only because it's so easy and unreasonable. But because it would be much more worthwhile to deliver an argument, show connections and correlations, give a great image of understanding himself.
Not just tap a disconnected comment into a keyboard.
-
I do show tests. Haven't you noticed? Whenever I make a point and am able to demonstrate that point with tests, because I have the equipment, I usually do so. When I tried RSP a few years ago and found it produced sharper results than the then current version of ACR, I demonstrated the fact with test images on this forum.
Isn't this the whole point of the forum? Seeing is believing. As I mentioned before, there might be many reaosns why someone prefers a particular converter. It might simply be a quicker and easier way of getting the same results, or the controls might be more intuitive for one particular user, but not necessarily another.
Of course, I don't need to do this. I sometimes wonder why I bother. But the fact is, the very process of organising my own test results into a demonstration format, tests which are initially made for my own edification, clarifies the issues in my mind and helps me discover flaws in my methodology. I learn from the process. If I didn't, I would no longer bother.
There's no reason for me to prove ACR's superiority. I'm not making any claims for ACR other than it's probably, on balance, as good as any other converter on the market but has the advantage of being more of an industry standard than any other converter. If someone wishes to make the claim that another converter does a better job than ACR, then let them demonstrate it so we can all benefit. You don't go through life believing everything that everyone says, do you?
If it would be so easy to show what's best, then we'd probably already have an industry standard RAW converter.
But, like with many creative and technical things, there are different approaches to solve each problem, and different solutions can stand next to each other.
So, if someone tests a device, he should test it for all existing approaches. Which would in this case include more RAW converters in the testing.
It's like philosophic systems. There are many of them, and they are contradicting each other, but still the good ones can't cancel each other out. They coexist.
Aesthetic points you can't prove. Judging image quality will always include aesthetic categories.
-
If it would be so easy to show what's best, then we'd probably already have an industry standard RAW converter.
But, like with many creative and technical things, there are different approaches to solve each problem, and different solutions can stand next to each other.
So, if someone tests a device, he should test it for all existing approaches. Which would in this case include more RAW converters in the testing.
It's like philosophic systems. There are many of them, and they are contradicting each other, but still the good ones can't cancel each other out. They coexist.
Aesthetic points you can't prove. Judging image quality will always include aesthetic categories.
Dynamic range, resolution and accutance are not esthetic issues. They are the easiest factors to compare. You seem to have strayed off the point. The objection in this thread has been raised that Dpreview's choice of ACR to compare the 50D with the 40D, with respect to DR and resolution, was not the best choice.
Is it not reasonable to ask that those who claim that there is a better choice, demonstrate the fact, so we can all judge whether or not they are correct?
Dpreview are quite likely not aware that there is a better alternative, just as I am not. If those who are aware of a better choice would share their knowledge and experience and demonstrate clearly and concisely why their choice of converter should be used in preference to ACR, then we could all benefit.
-
My own tests are suggesting, some of which I'm sharing with you, that the real benefit of this very slight resolution advantage of the 50D, is that one can confidently stop down an additional stop without sacrificing any resolution (compared with the 40D at the plane of focus), yet gain the more significant advantage of the greater DoF that stopping down affords.
Now that is interesting.
Have you done any tests with the half size sraw? Any noise advantages?
-
Now that is interesting.
Have you done any tests with the half size sraw? Any noise advantages?
I haven't yet compared 50D noise with 40D noise, but I have compared 50D noise with 5D noise, and the 5D wins, which just goes to show that sensor size is a significant factor regarding noise.
However, there's another way of looking at this. The cropped format has the advantage of greater DoF at any given aperture, and consequent faster shutter speed. When one equalizes DoF and shutter speed, by using a narrower aperture with the 5D and a higher ISO to allow the same shutter speed, then noise appears to be about equal between the two cameras.
-
We should all know by now that the resolution differences between the 50D and 40D are very slight. My test results in this thread are an exploration of how such slight differences might be of practical value to me.
It has been suggested many times that such high-pixel-density cameras have no benefit at small apertures and that any resolution advantage will only be apparent with exceptionally good lenses at their sharpest apertures.
My own tests are suggesting, some of which I'm sharing with you, that the real benefit of this very slight resolution advantage of the 50D, is that one can confidently stop down an additional stop without sacrificing any resolution (compared with the 40D at the plane of focus), yet gain the more significant advantage of the greater DoF that stopping down affords.
I won't argue with what you say above, but your two example images cannot be compared. In general I find your tests interesting, but in this case you have to compare like with like, to eliminate possible causes of error. Lens resolution usually varies significantly with F stop.
-
Lens resolution usually varies significantly with F stop.
Lens resolution does not vary significantly with F stop, within a reasonable range, say F5.6 to F16. It seems to vary roughly to the degree of difference between the 40D's and 50D's resolving capability, for each stop of difference. That's my point, a point which I've demonstrated with examples.
There was no error in my test that I'm aware of. Camera was on tripod and focussing was manual with Liveview. The aliasing artifacts at F5.6 are an indication that focussing was very accurate. You don't get aliasing artifacts with misfocussing.
-
By special request from Slough, the same banknote at F5.6 with both cameras, this time converted with DPP, no sharpening or noise reduction etc.. Notice the significant aliasing.
[attachment=9393:DPP_F5_6...mparison.jpg]
-
I haven't yet compared 50D noise with 40D noise, but I have compared 50D noise with 5D noise, and the 5D wins, which just goes to show that sensor size is a significant factor regarding noise.
I would expect that. Very curious how the 5D mk II fares.
However, there's another way of looking at this. The cropped format has the advantage of greater DoF at any given aperture, and consequent faster shutter speed. When one equalizes DoF and shutter speed, by using a narrower aperture with the 5D and a higher ISO to allow the same shutter speed, then noise appears to be about equal between the two cameras.
I believe that is the equivalency theory. The question is always can you find fast enough lenses for the crop camera to shrink DOF and is your high ISO clean enough on FF to give you large DOF when you need it.
-
Lens resolution does not vary significantly with F stop, within a reasonable range, say F5.6 to F16. It seems to vary roughly to the degree of difference between the 40D's and 50D's resolving capability, for each stop of difference. That's my point, a point which I've demonstrated with examples.
Then I think you are misinformed. The expected variation in lens resolution is going to be greater than the small differences that you are looking at. And the difference in going from F11 to F16 is often very significant, as is the drop in contrast, though it does of course depend on the lens. The 50D would exaggerate that difference.
-
By special request from Slough, the same banknote at F5.6 with both cameras, this time converted with DPP, no sharpening or noise reduction etc.. Notice the significant aliasing.
[attachment=9393:DPP_F5_6...mparison.jpg]
Thanks. The above comparison proves my point.
Actually something very odd is going on as the two 40D F5.6 images (one from each comparison) are quite different, the second being much worse than the first. Just look at the name at bottom right. In the second case it is quite blurred. I don't mean to be rude but I have lost confidence in your comparisons.
-
Then I think you are misinformed. The expected variation in lens resolution is going to be greater than the small differences that you are looking at. And the difference in going from F11 to F16 is often very significant, as is the drop in contrast, though it does of course depend on the lens. The 50D would exaggerate that difference.
In photographic matters I am informed by what I see. I don't see much difference, or anything I would describe as very significant, at the plane of focus moving from one aperture to the next one, either up or down one stop, within the range where I find most lenses are acceptably sharp and at F stops most often used (by me). It's not a matter for objective argument because the statement 'don't see much difference' is a personal and subjective statement based upon many tests over the years photographing real-world scenes as well as line charts.
Whilst I find there's a significant difference between F5.6 and F16, I find little difference between F5.6 and F8, between F8 and F11 and between F11 and F16, with the Canon 50/1.4. With other lenses and at other apertures, the situation may be different. With the Canon 100-400 at 400mm the difference between F5.6 and F8 is greater than the difference between F8 and F11. The lens is less sharp at F5.6 than I would like.
However, if you are serious and genuine in your assertion that going from F11 to F16 can be very significant, then please show us all a comparison so we can get an idea of your idea of significant.
Edit: I should add, in case anyone is confused by the above, that I concede it might be possible for a lens to have a very significant difference in performance at F11 and F16, but I don't own any such lenses. I imagine such a lens would be an old-fashioned design for large format film cameras, optimised for sharpest results at F16 or F22 and having a maximum aperture of F11.
-
Actually something very odd is going on as the two 40D F5.6 images (one from each comparison) are quite different, the second being much worse than the first. Just look at the name at bottom right. In the second case it is quite blurred. I don't mean to be rude but I have lost confidence in your comparisons.
Nothing odd there. The earlier comparisons have been converted using ACR 4.6 with default sharpening. For the latest comparison above, I used DPP and turned off all sharpening. I mentioned this fact in the text. You've made an inappropriate comparison. Even Dpreview advise against comparing current tests with previous tests, possibly because different versions of ACR have been used and at different settings, and general testing procedures might vary slightly over time.
All my comparisons (done at the same time) use the same settings, except for levels and exposure adjustments to get contrast and tonality as similar as possible.
Comparing two different converters is a different exercise. The controls and sliders will tend to produce different effects. One converter's default sharpening might be another converter's 'no sharpening' etc etc. I haven't tried to compare converters here.
-
First of all, I would say that most people don't shoot in RAW mode in any case. If one shoots in RAW mode and therefore has a use for a RAW converter, it's a sign that one is particularly concerned about subtleties of image quality, obtaining maximum detail in highlights and shadows etc and, in my opinion, quite likely that one will be a Photoshop user.
Of course, without extensive market research it's impossible to say whether greater than 50% of all people who shoot in RAW mode use ACR or Lightroom. Some people will use, or at least try out, a number of different converters and find that a particular converter produces more satifying result with certain types of scenes than ACR. I used to prefer RSP myself, for a while until it was bought out by Adobe. However, I found that ACR always did a better reconstruction job of blown highlights.
Is the actual percentage of ACR users really important? Photoshop and accompanying programs like Bridge and ACR are leading industry standards. That's the essential point. It would get very complicated to run through the whole gamut of available converters on the market every time a new camera was tested.
We are in the digital age that means hardware and software work together one cannot function without the other. You cannot test hardware without optimized software, Canon designed the hardware and they know how to get the most out of it with the software they also created. We are no longer in the film world were you test the hardware alone the software also has to be tested in the digital age. To properly test any camera it should be tested with the software that was written for it no matter how awkward the interface is, that is the only way to determine how good it is. It is irrelevant how many people use ACR since people are wiling to pay for other programs and choose not to use it. It is also designed to work with different manufactures RAW files it may be the jack of all converters but it is a master of none. dpreview needs to get their act together and start testing cameras with their intended software to measure output and compare them to other cameras, they can use ACR later for a comparison of raw converters. Adobe probably would not like it since it may show that ACR needs improvement.
-
We are in the digital age that means hardware and software work together one cannot function without the other. You cannot test hardware without optimized software, Canon designed the hardware and they know how to get the most out of it with the software they also created. We are no longer in the film world were you test the hardware alone the software also has to be tested in the digital age. To properly test any camera it should be tested with the software that was written for it no matter how awkward the interface is, that is the only way to determine how good it is. It is irrelevant how many people use ACR since people are wiling to pay for other programs and choose not to use it. It is also designed to work with different manufactures RAW files it may be the jack of all converters but it is a master of none. dpreview needs to get their act together and start testing cameras with their intended software to measure output and compare them to other cameras, they can use ACR later for a comparison of raw converters. Adobe probably would not like it since it may show that ACR needs improvement.
I understand your argument but I see it as flawed. Whatever procedure Dpreview adopts (probably without exception) there will always be disaffected people whose ego is slighted, who don't like the results and who will therefore criticise the methodology and suggest another converter should have been used. This is the very reason why Dpreview have attempted to standardise their procedure by always using ACR if it supports the camera.
If we had a situation where it was widely accepted that the software that shipped with the camera always gave the best results when converting the RAW images from that camera, then your argument would have merit and ACR would probably not exist. Who would want to use ACR if better results could always be obtained from free software that shipped with the camera.
Now, it might well be the case that Canon have lifted its game in recent years with regard to DPP and that the latest versions of DPP have certain advantages over ACR. But one can't assume that such a situation automatically applies to all manufacturers' own converters.
What happens if one particular converter applies a default level of sharpening which one can do nothing about? What happens if another manufacturer's own converter really is poor software and doesn't do as good a job as ACR? Should the hardware then be judged by the performance of the poor software when other software, such as ACR, can do a better job?
In this particular review of the 50D and its comparison with the 40D, it was unfortunate that initially the version of ACR used was a beta. But they corrected that, didn't they? As I understand, the beta version of ACR 4.6 produced a very marginally less detailed result at the extreme pixel-peeping level, didn't it.
-
I understand your argument but I see it as flawed. Whatever procedure Dpreview adopts (probably without exception) there will always be disaffected people whose ego is slighted, who don't like the results and who will therefore criticise the methodology and suggest another converter should have been used. This is the very reason why Dpreview have attempted to standardise their procedure by always using ACR if it supports the camera.
If we had a situation where it was widely accepted that the software that shipped with the camera always gave the best results when converting the RAW images from that camera, then your argument would have merit and ACR would probably not exist. Who would want to use ACR if better results could always be obtained from free software that shipped with the camera.
Now, it might well be the case that Canon have lifted its game in recent years with regard to DPP and that the latest versions of DPP have certain advantages over ACR. But one can't assume that such a situation automatically applies to all manufacturers' own converters.
What happens if one particular converter applies a default level of sharpening which one can do nothing about? What happens if another manufacturer's own converter really is poor software and doesn't do as good a job as ACR? Should the hardware then be judged by the performance of the poor software when other software, such as ACR, can do a better job?
In this particular review of the 50D and its comparison with the 40D, it was unfortunate that initially the version of ACR used was a beta. But they corrected that, didn't they? As I understand, the beta version of ACR 4.6 produced a very marginally less detailed result at the extreme pixel-peeping level, didn't it.
If the manufacturer cannot write better software than ACR then they should perhaps be looking at another business. This refers to the output only not user interface which usually leaves a lot to be desired. The point I am attempting to make is the cameras should be tested with the software that was provided as a total system, the output can then be compared to ACR as a so called standard (not my standard) RAW processor. That would be a truer test. The 50D test just brought out the problems in testing in the digital age, you need to test both hardware and software provided by the manufacture to measure camera performance, you cannot pick software provide by a third party to compare output to standardize it, what are you standardizing ACR performance, it does not come with the camera. When you purchase a camera you are also paying for the software the development and manufactures tweak their software and not just cannon Nikon also does it. If you use ACR thent Adobe sets the standard not the manufacture, in essence you are testing ACR output.
-
If you use ACR thent Adobe sets the standard not the manufacture, in essence you are testing ACR output.
Writing RAW conversion software is Adobe's business. They are probably more experienced in such matters than camera manufacturers. It is true that in using ACR you are testing to a certain extent the way ACR handles that particular model of camera. That's unavoidable. No matter what converter you use, the same principle applies. The way the software has been written affects the results.
An example of just one reason for using ACR with all cameras would be that one could expect a certain standardisation of the treatment of sharpening of the RAW file. For example, no sharpening means no sharpening, or at least, if there is some degree of built-in sharpening that is not user controllable, it's likely to be consistent across different models of cameras.
Supposing Dpreview adopted the policy of using only the manufacturer's software to convert RAW files, and came across one particular brand of packaged software that appeared to be doing some basic sharpening, like in-camera sharpening of jpegs that couldn't be completely turned off. What do you do then?
I've noticed a few times on this forum in the past, posters asking Michael why a particular RAW file looks so soft in ACR. The answer tends to be, because the manufacturer's converter does not allow all sharpening to be turned off, whereas ACR does.
-
Writing RAW conversion software is Adobe's business. They are probably more experienced in such matters than camera manufacturers. It is true that in using ACR you are testing to a certain extent the way ACR handles that particular model of camera. That's unavoidable. No matter what converter you use, the same principle applies. The way the software has been written affects the results.
An example of just one reason for using ACR with all cameras would be that one could expect a certain standardisation of the treatment of sharpening of the RAW file. For example, no sharpening means no sharpening, or at least, if there is some degree of built-in sharpening that is not user controllable, it's likely to be consistent across different models of cameras.
Supposing Dpreview adopted the policy of using only the manufacturer's software to convert RAW files, and came across one particular brand of packaged software that appeared to be doing some basic sharpening, like in-camera sharpening of jpegs that couldn't be completely turned off. What do you do then?
I've noticed a few times on this forum in the past, posters asking Michael why a particular RAW file looks so soft in ACR. The answer tends to be, because the manufacturer's converter does not allow all sharpening to be turned off, whereas ACR does.
When you purchase a camera does ACR come with no it does not, manufactures want to tweak their hardware, why do thing DNG has not been accepted by all the manufactures, if it was, Adobe would control the output, only manufactures accepting DNG are the ones that do not want to invest in software development which by the way is very costly. As far as control goes the raw file coming out of the camera has already been tweaked by the manufacture, you do not believe that you are actually dealing with raw data do you? The RAW processor just tweaks more the already processed data and the manufacture knows how better to do that then Adobe since they know the entire process. Have you seen software provided by the manufacture that was worst than ACR, I certainly have seen much better conversions with factory software for both Nikon and Canon than Adobe and C1 certainly does a better job than ACR. The point I am making is when you purchase a product you test the product hardware and software not just half of it otherwise you are not testing the product you purchased.
-
Perhaps we should flog dcraw instead.
-
There are two different categories of SLR users (of course this is oversimplified):
1) Those who shoot jpg and send their images to a finisher for printing. (80%-90% of users)
2) Those who can explain what a RAW workflow is and who perform post-processing. (most of us at this board)
For people in category 1, the evaluation is simple. Compare jpgs out of the camera.
For people in category 2, there are a multitude of divergent approaches to evaluation.
For this reason, I don't think we can expect to come to consensus on methods of evaluation to determine the 'BEST' camera. On the other hand, a more productive line of discussion would be to discuss how to get the most out of our camera using specialized tools and workflow.
There is no such thing as an "Objective Best Camera." So, let's stop arguing over that kind of stuff. For that, go to DPReview forums, and please stop ruining these great Luminous Landscape Forums with these ideological debates.
-
Writing RAW conversion software is Adobe's business. They are probably more experienced in such matters than camera manufacturers. It is true that in using ACR you are testing to a certain extent the way ACR handles that particular model of camera. That's unavoidable. No matter what converter you use, the same principle applies. The way the software has been written affects the results.
Maybe ACR was not the right choice. Adobe is a team with a lot of people working on it, but still is too much work for them to get the best from every camera. I was recently surprised to find out that ACR seems to develop Fuji RAF's with two issues:
1. A wrong consideration of the real saturation point of the camera (R sensor) which produces magenta artifacts in the highlights
2. A wrong consideration of the white balance for the highlights (R sensor again)
In this sample I developed the same RAF file with ACR and -4EV and -2.5EV exposure correction, and compared results to DCRAW. While in the shadows (sensor S) they were almost identical, in the highlights (sensor R), ACR was not optimum in preserving the neutrality and detail of the highlights which appeared magenta, and also the white balance was very greenish.
Bottom left: DCRAW developing sensor R with white balance calculated for that sensor.
Bottom right: DCRAW developing sensor R with the white balance calculated for the S sensor (it gets closer to ACR result but still not so greenish).
(http://img513.imageshack.us/img513/6384/compvv0.jpg)
-
In photographic matters I am informed by what I see. I don't see much difference, or anything I would describe as very significant, at the plane of focus moving from one aperture to the next one, either up or down one stop, within the range where I find most lenses are acceptably sharp and at F stops most often used (by me). It's not a matter for objective argument because the statement 'don't see much difference' is a personal and subjective statement based upon many tests over the years photographing real-world scenes as well as line charts.
Whilst I find there's a significant difference between F5.6 and F16, I find little difference between F5.6 and F8, between F8 and F11 and between F11 and F16, with the Canon 50/1.4. With other lenses and at other apertures, the situation may be different. With the Canon 100-400 at 400mm the difference between F5.6 and F8 is greater than the difference between F8 and F11. The lens is less sharp at F5.6 than I would like.
However, if you are serious and genuine in your assertion that going from F11 to F16 can be very significant, then please show us all a comparison so we can get an idea of your idea of significant.
Edit: I should add, in case anyone is confused by the above, that I concede it might be possible for a lens to have a very significant difference in performance at F11 and F16, but I don't own any such lenses. I imagine such a lens would be an old-fashioned design for large format film cameras, optimised for sharpest results at F16 or F22 and having a maximum aperture of F11.
Ray: As I know you are a fan of Imatest MTF plots, I suggest you take a look at Photo Zone tests. All lenses drop in performance before F16 due to diffraction. It is a physical property which cannot (yet) be avoided. I also see this even in my 5 Nikon micro lenses but Photo Zone tests have the advantage that you are more likely to take notice. There can also be quite large changes away from F16, depending on the lens, and not just old large format lenses.
-
Nothing odd there. The earlier comparisons have been converted using ACR 4.6 with default sharpening. For the latest comparison above, I used DPP and turned off all sharpening. I mentioned this fact in the text. You've made an inappropriate comparison. Even Dpreview advise against comparing current tests with previous tests, possibly because different versions of ACR have been used and at different settings, and general testing procedures might vary slightly over time.
All my comparisons (done at the same time) use the same settings, except for levels and exposure adjustments to get contrast and tonality as similar as possible.
Comparing two different converters is a different exercise. The controls and sliders will tend to produce different effects. One converter's default sharpening might be another converter's 'no sharpening' etc etc. I haven't tried to compare converters here.
Nothing like consistency eh? Now you present photos of the same test subject, at the same ISO and F stop, with each camera. This time you change the processing. And the results look completely different from the first set. In the second case (same F stop) the two images look the same to me.
Confused, I am.
-
Adobe is a team with a lot of people working on it, but still is too much work for them to get the best from every camera.
I have tried a number of converters for Nikon D200 RAW files and none were a patch on the official Nikon software. None could get the colours right. So no, I don't think these companies necessarily do better than the camera maker.
-
I have tried a number of converters for Nikon D200 RAW files and none were a patch on the official Nikon software. None could get the colours right. So no, I don't think these companies necessarily do better than the camera maker.
I had the a similar experience with ACR and the Canon 40d.
With RAW processing with ACR the 40d looked almost as bad as my old Pentax, but using DPP it made the camera fly. For example: a path in the hills, rocks on the path. In ACR, the rocks were blending in with the dust, with DPP I could see clearly their sharp edges. the reflections of light from those edges. The light was much more defined. The difference felt like you wore dirty glasses, which ruined clarity and contrast, then cleaned them.
The colors were mute and without depth in ACR, and became gorgeous and brilliant with the manufacturer's software. I can get a clarity and definition to images in that software that is unsurpassed, and I have the impression I even lose a little bit of this by transferring to Photoshop. Colors, shapes, all is well defined.
Part of the problem for third party manufacturers may be, that Nikon and Canon do not disclose everything about their RAW files. And, as GLujik said, it's just an insane amount of work to get the best for so many cameras. Also, if a manufacturer can design the software from the ground up to specialize for their sensors and the way their sensor data are processed.
So, even though the interface is clunky (I wish there was an improved work flow edition of it - I'd pay for it), I do all my RAW processing with the manufacturer's software.
-
Ray: As I know you are a fan of Imatest MTF plots, I suggest you take a look at Photo Zone tests. All lenses drop in performance before F16 due to diffraction. It is a physical property which cannot (yet) be avoided. I also see this even in my 5 Nikon micro lenses but Photo Zone tests have the advantage that you are more likely to take notice. There can also be quite large changes away from F16, depending on the lens, and not just old large format lenses.
Aren't you confusing diffraction effects on the projection surface (film or digital sensor) with the "lens performance"?
The problem with diffraction is that smaller apertures leads to larger airy disks, which lowers the resolvable contrast and resolution.
Even if f/16 isn't necessarily where the lenses perform best, it can still be within a performance range where a high density sensor medium can resolve more than a low density sensor medium.
-
I had the a similar experience with ACR and the Canon 40d.
The colors were mute and without depth in ACR, and became gorgeous and brilliant with the manufacturer's software. I can get a clarity and definition to images in that software that is unsurpassed, and I have the impression I even lose a little bit of this by transferring to Photoshop. Colors, shapes, all is well defined.
Part of the problem for third party manufacturers may be, that Nikon and Canon do not disclose everything about their RAW files. And, as GLujik said, it's just an insane amount of work to get the best for so many cameras. Also, if a manufacturer can design the software from the ground up to specialize for their sensors and the way their sensor data are processed.
So, even though the interface is clunky (I wish there was an improved work flow edition of it - I'd pay for it), I do all my RAW processing with the manufacturer's software.
I'm on the Nikon side of the house but for the most part agree with you. Most of the Nikon users I know are not ACR fans - while I have no valid statistical evidence being I don't have the time to conduct a scientific poll , from observing forums over the years and talking to other Nikon users, I'd say a slight majority of them don't use ACR as their choice of converter. I would say amongst most of the Canon users I know, ACR seems to be the majority converter there, but certainly not for Nikon.
I also am in total agreement with you that the manufacturer knows it's sensors spectral response curves and other data the best and can optimize the converter for the sensor better, IMO, than most anyone else. But this industry is still relatively young, and I'm sure down the road more third party converters will challenge the manufacturers converters and might in some cases be the best option for conversion. Soon I almost expect it (the choice of raw converter) to be a "best tool for the job" scenario instead of blindly proclaiming that one and only one converter is the best.
I currently am using Capture NX2, and in my tests it does pull out more detail than ACR, and it's a converter that allows for sharpening to be switched totally off as well, so Rays argument doesn't hold any water for this one either. More importantly, it gets the WB and color/density closer to what I shoot in the studio without a lot of mucking around.
However, I do have to give Adobe a lot of credit. Their very latest verions (4.6 and 5.1) do a LOT better with Nikon files than any previous version, by quite a significant margin. With the new profile editor and the option to roll your own, the previous problems with color and tone have been mitigated quite a bit. So the advantage of the other converters has been cut back drastically with the new versions, and for once I can say that if I was forced to, I could make do with ACR. However, there is still something about the conversions from Capture NX2 that make the image more realistic and alive, and thus while it's a pathetic converter in terms of operational stability, UI design, and speed (and I do mean absolutely pathetic), it still provides the best images.
****
In terms of the dpreview discussion, I think it would be nice if they ran three raw converters on the cameras - the manufacturers converter, ACR, and Capture One, but I can totally understand the vast amount of additional time this would take. No single review is perfect, and thus for a while I've been advocating looking at the consensus of opinions, both from reviews and from forums (and other shooters) and then possibly trying something yourself. Unfortunately, some folks seem to want the quick and easy answer - the perfect MTF test that encompasses all about lenses, the single review that is perfect and without flaw to determine which camera is best and so forth, but this isn't a reality at this time, and may never be.
-m
-
Aren't you confusing diffraction effects on the projection surface (film or digital sensor) with the "lens performance"?
The problem with diffraction is that smaller apertures leads to larger airy disks, which lowers the resolvable contrast and resolution.
Even if f/16 isn't necessarily where the lenses perform best, it can still be within a performance range where a high density sensor medium can resolve more than a low density sensor medium.
No, I am referring to the effect of diffraction on lens resolution and contrast, which for most lenses is significant at F16, though the lenses I have still perform well at that aperture. I am of course referring to 35mm lenses. The actual IQ will of course be the combined effect of sensor and lens. Terms such as 'high density' and 'low density' are too lose for me to agree or disagree. Chose your definition of high and low appropriately, and you are right.
By the way, as a Nikon user I can set the actual aperture. One problem is that unless I am mistaken Ray is a Canon user and can only set the effective aperture. If he has tested on close ups, then that might explain his not seeing the performance hit.
-
However, there is still something about the conversions from Capture NX2 that make the image more realistic and alive, and thus while it's a pathetic converter in terms of operational stability, UI design, and speed (and I do mean absolutely pathetic), it still provides the best images.
Tell me about it! I can hardly use NX2 on my PC. It is the most unstable crash prone POS I have ever used. Which is why I did not upgrade from the demo to the full product, and I stick with NC. If they let me get at that software, I could make it stable. (I've fixed a fair few professional products. But it would not happen.)
-
Maybe ACR was not the right choice. Adobe is a team with a lot of people working on it, but still is too much work for them to get the best from every camera. I was recently surprised to find out that ACR seems to develop Fuji RAF's with two issues:
1. A wrong consideration of the real saturation point of the camera (R sensor) which produces magenta artifacts in the highlights
2. A wrong consideration of the white balance for the highlights (R sensor again)
Nobody and nothing are perfect. Maybe now that you have pointed out this flaw in ACR's treatment of RAF files, Adobe will correct it in their next version .
-
Ray: As I know you are a fan of Imatest MTF plots, I suggest you take a look at Photo Zone tests. All lenses drop in performance before F16 due to diffraction. It is a physical property which cannot (yet) be avoided. I also see this even in my 5 Nikon micro lenses but Photo Zone tests have the advantage that you are more likely to take notice. There can also be quite large changes away from F16, depending on the lens, and not just old large format lenses.
There can certainly be large changes in the height of a bar chart representing resolution at MTF 50. How these apparently large changes are reflected in real world images is another matter. Bar charts and graphs are usually constructed in such a way as to make any differences, however slight, as obvious as possible.
As I've mentioned, with photographic matters I can respond only to what I see, and I don't see any significant differences between F11 and F16 at the plane of focus. But I do see significant differences away from the plane of focus, which is why I will always use F16 when maximum DoF is a priority.
However, don't misinterpret my words. I make a distinction between a significant difference and a discernible difference. I certainly can discern a slight softening in an image taken at F16, compared with the same scene at F11. The difference is there, but it's slight and is usually easily counteracted with a touch more sharpening.
The following two images were taken with the 40D and 100-400 at 400mm. The lens was on tripod. MLU and remote cord were used. Focussing was manual using LiveView and the two shots were taken within seconds of each other using the same focussing.
One image is at F11 and the other at F16. The focus point was the brick wall of the house. Can you see any significant difference between these two 100% crops, which I might add, on the current monitor I'm using, represent a print size of 17"x25.5"? Do any of the two look oversharpened, as though to compensate for significant softness?
Have I made my point?
[attachment=9438:F11___F1...00__crop.jpg]
I should also add that because I can discern a slight softness in the F16 image before sharpening, I would always prefer to use F11 if the additional DoF is not required. When detail is a concern, there is no point in using a slower shutter speed or higher ISO than necessary, and one never knows when one might want to blow up to an extreme magnification a small section of an image. Differences that are insignificant at 100%, may become almost significant at 200%, clearly significant at 400% and particularly significant at 800%.
-
Ray: As I have already stated, your two sets of 40D and 50D images are contradictory and in the second set I see no difference between the two suggesting no gain from the 50D. (Not being a Canon user I don't care about the outcome. It is academic to me.)
I cannot really comment on your latest examples (especially given your rather conflicting test methods from earlier posts) except to say that it looks to be somewhat of a miracle lens. My own careful tests with numerous micro/macro lenses agree with online MTF plots and show obvious drops at F16. And take a look at Rorslett's tests to see that he agrees, where he often says to avoid F16. I agree with him. Oh and my Sigma 400mm F5.6 APO macro lens (which is sharp) also softens noticeably at F16. Something is not right in your latest example. BTW I use a solid tripod, Markins ball head, mirror lock up and cable release in my tests.
-
Bar charts and graphs are usually constructed in such a way as to make any differences, however slight, as obvious as possible.
I'm not sure what you mean by that. They give the numeric values on the left scale so you can judge the result accurately. (Ignoring criticisms of MTF plots per se.)
-
Ray: As I have already stated, your two sets of 40D and 50D images are contradictory and in the second set I see no difference between the two suggesting no gain from the 50D. (Not being a Canon user I don't care about the outcome. It is academic to me.)
I cannot really comment on your latest examples (especially given your rather conflicting test methods from earlier posts) except to say that it looks to be somewhat of a miracle lens. My own careful tests with numerous micro/macro lenses agree with online MTF plots and show obvious drops at F16. And take a look at Rorslett's tests to see that he agrees, where he often says to avoid F16. I agree with him. Oh and my Sigma 400mm F5.6 APO macro lens (which is sharp) also softens noticeably at F16. Something is not right in your latest example. BTW I use a solid tripod, Markins ball head, mirror lock up and cable release in my tests.
Slough,
Your comments are very puzzling. What do you find contradictory or conflicting? I provide you with test images I've carefully made for my own purposes. I label each one correctly, at the top of each image (did you notice that?). I tell you which converter I used and I tell you whether or not I applied sharpening (default sharpening in the case of ACR and no sharpening in the case of DPP). What is there to get confused about? I haven't secretly mixed DPP conversions with ACR conversions. I have deliberately mixed F5.6 shots with F11 shots, and F11 shots with F16 shots, in order to demonstrate that the increased resolution of the 50D is roughly (perhaps very roughly) equivalent to the resolution of the 40D stopped up one stop, to a less diffraction limited F stop. That's just to give you an idea of the practical significance of the resolution increase of the 50D, in my view.
To my eyes, the DPP comparison of both cameras at F5.6 (no sharpening other than what's built into the DPP software and not user controllable) shows the 50D crop as having more detail and significantly less aliasing.
The same crops, but at F11 with the 50D and F5.6 with the 40D, and converted with ACR at default sharpening, also show the 50D image as having more detail, but this time the 50D crop has no aliasing whereas the 40D image has similar aliasing to the DPP conversion, but the magenta streaking is more subdued because of the different converter. One might deduce from this particular result that there's little difference in the sharpness of the Canon 50/1.4 at F5.6 and F8, despite the bar graphs at Photozone.
The first comparison of the rural/suburban scene, 40D at F11 and 50D at F16, both converted with ACR at default sharpening, look equally detailed to me depite both having received equal sharpening.
I haven't provided full details of the latest crops of the 40D at F11 and the 40D F16 because I wanted to demonstrate that you would be hard pressed to distinguish between them. The difference is not significant, perhaps even invisible. This is not due to my using a miracle lens but is possibly due to my using a magical sharpening routine.... called 'Focus Magic'.
As I mentioned before, unsharpened one can discern an additional softness in an F16 image, compared with the same scene at F11 (from the same camera & lens). This requires just a little additional sharpening to correct. Both of these images were converted in ACR 4.6 using ACR's default sharpening of 25 at 1 pixel radius and detail at 25. The images were then further sharpened using Focus Magic's default settings for each image. Focus Magic is a deconvolution program which automatically suggests a pixel blur width to use, after a brief analysis of the image. For the F11 shot, it suggested 100% at 1 pixel blur width. For the F16 shot, the recommendation was 100% at 2 pixels blur width. Focus Magic could recognise that the F16 image was slightly soft, just as I could.
You probably know that DoF extends a greater distance beyond the focus plane than it does in front of it. Below are crops of the same images showing the foreground. Are you now able to tell which is F11 and which is F16?
For the record, the histogram of both images is very similar, same ETTR. The exposure of the F11 shot at ISO 100 is 1/80th, and the exposure of the F16 shot is 1/40th at the same ISO.
[attachment=9443:F16___F11_40D.jpg] [attachment=9444:F16___F1...reground.jpg]
Is everything now crystal clear?
-
I'm not sure what you mean by that. They give the numeric values on the left scale so you can judge the result accurately. (Ignoring criticisms of MTF plots per se.)
The bar charts make very clear the difference between (for example) 1800 lines per picture height and 1825 lines per picture height. You'd have to be half blind not to see it.
If I were to show you two large prints, one consisting of 1800 lines and the other consisting of 1825 lines (or even 1850 or 1900 lines), would you immediately be able to see the difference?
-
The bar charts make very clear the difference between (for example) 1800 lines per picture height and 1825 lines per picture height. You'd have to be half blind not to see it.
If I were to show you two large prints, one consisting of 1800 lines and the other consisting of 1825 lines (or even 1850 or 1900 lines), would you immediately be able to see the difference?
Now you are being silly. Anyone would discount 1800 and 1825 as being as near as the same.
Let's take a lens I own, namely the Nikon 105mm F2.8 AFD. According to Photo Zone (chosen as a respected Imatest reviewer), tested with a D200, the resolution at F11 and F16 is roughly 2,000 and 1,700. Now those are rough estimates from the graph, but even allowing for error, the difference in linear resolution is significant i.e. 15%. In practice when used for close ups at F11 and F16 I see a noticeable change in resolution and contrast. In other words, the PZ data (from a different lens sample) correlates to experience. Yes this is only visible on small prints, but that is what we are talking about c.f. 50D isn't it? On an A4 print you probably would not see it, but in the context of your discussion, it is significant. I tried to find other F16 data on PZ but failed. But I have tested with other lenses e.g. Nikon 200mm micro (both), Nikon 60mm AFD F2.8 micro etc.
Maybe you think lens resolution does not vary much with F stop due to reading MTF plots obained with a ~8-10 MP sensor?
On the D200, diffraction starts visibly softening the image from about F13. So at wider apertures the lens is outperforming the sensor except perhaps near wide open on lesser lenses. So the resolution versus F stop would be roughly flat to ~F11.
-
Let's take a lens I own, namely the Nikon 105mm F2.8 AFD. According to Photo Zone (chosen as a respected Imatest reviewer), tested with a D200, the resolution at F11 and F16 is roughly 2,000 and 1,700. Now those are rough estimates from the graph, but even allowing for error, the difference in linear resolution is significant i.e. 15%.
I think you might have misunderstood the significance of the charts. Your Nikkor lens will also be capable of resolving 2,000 lines at F16, but not at 50% MTF; maybe 40% MTF. The additional sharpening mostly takes care of that MTF difference, but probably not quite.
Maybe you think lens resolution does not vary much with F stop due to reading MTF plots obained with a ~8-10 MP sensor?
This seems to be a very odd situation. I'm the one providing the real world evidence, demonstrating a point, walking the talk. You are the one providing no evidence at all, but instead accuse me of being misled by MTF charts. Seems to be something very wrong here???
The following 200% crops of the banknote were taken on a carbon fibre tripod, placed on a tiled floor on a solid concrete slab, indoors away from any breeze, passing truck, train or aircraft that might have been flying overhead. There was no vibration in the air (from heavy rock music, for example), no earthquakes, tremors or bomb explosions. I used MLU, remote cord and focussed extremely accurately with the aid of the high resolution LiveView screen that the 50D boasts. (In any case, at F11 and F16, absolutely accurate focussing would not be critical, but I did it in any case.)
On my 1280x1024 monitor, these 200% crops represent a print size of 105"x70" (or almost 9ftx6ft). If you were to make prints this size, walk right up to them and peer at the small banknote in the centre of each print from a distance of a foot or so, you would see a very small amount of detail in the F11 print that isn't there in the F16 print.
There is no trickery here. These are the results. They speak for themselves.
The previous shots were taken with the 100-400 zoom at a significant distance. These shots below were taken with the Canon 50/1.4 from a distance of around 8ft. The same converter and sharpening amounts have been applied, as in the previous 40D shots at F11 and F16, ie. default ACR sharpening then Focus Magic with a 2 pixel blur width applied to the F16 shot, and 1 pixel to the F11 shot.
[attachment=9445:F16___F1...banknote.jpg]
-
I think you might have misunderstood the significance of the charts. Your Nikkor lens will also be capable of resolving 2,000 lines at F16, but not at 50% MTF; maybe 40% MTF. The additional sharpening mostly takes care of that MTF difference, but probably not quite.
This seems to be a very odd situation. I'm the one providing the real world evidence, demonstrating a point, walking the talk. You are the one providing no evidence at all, but instead accuse me of being misled by MTF charts. Seems to be something very wrong here???
Here are some comments from lens review by B. Rorslett who is very respected:
55 mm f/3.5 Micro-Nikkor-P·C: "Maximum sharpness needs an aperture setting of f/8. Beyond f/11-f/16, sharpness rapidly deteriorated."
AFS Micro-Nikkor 60 mm f/2.8 ED G N: "Images are super sharp already from f/2.8 and keep their bite up to f/11, from which point a graceful decline kicks in. "
AF Micro-Nikkor 60 mm f/2.8: "For those utilising it, however, sharp images are obtained at apertures from f/5.6 to f/11. Beyond that range image quality deteriorates. "
AF-Micro-Nikkor 105 mm f/2.8: "It reaches optimum sharpness at f/5.6 and delivers good results to f/11 or so, thereafter image quality is quickly lost."
I pose the above quotes as I have several of those lenses and I agree with the comments.
I do not post example images as I have no imaging host, and cannot be bothered to repeat tests I perform when I buy a lens. I do not keep the images as I delete unwanted images to reduce clutter in my archives. MTF 50 is usually taken as a measure of resolution. It matches the above quotes and my experience.
The following 200% crops of the banknote were taken on a carbon fibre tripod, placed on a tiled floor on a solid concrete slab, indoors away from any breeze, passing truck, train or aircraft that might have been flying overhead. There was no vibration in the air (from heavy rock music, for example), no earthquakes, tremors or bomb explosions. I used MLU, remote cord and focussed extremely accurately with the aid of the high resolution LiveView screen that the 50D boasts. (In any case, at F11 and F16, absolutely accurate focussing would not be critical, but I did it in any case.)
On my 1280x1024 monitor, these 200% crops represent a print size of 105"x70" (or almost 9ftx6ft). If you were to make prints this size, walk right up to them and peer at the small banknote in the centre of each print from a distance of a foot or so, you would see a very small amount of detail in the F11 print that isn't there in the F16 print.
There is no trickery here. These are the results. They speak for themselves.
The previous shots were taken with the 100-400 zoom at a significant distance. These shots below were taken with the Canon 50/1.4 from a distance of around 8ft. The same converter and sharpening amounts have been applied, as in the previous 40D shots at F11 and F16, ie. default ACR sharpening then Focus Magic with a 2 pixel blur width applied to the F16 shot, and 1 pixel to the F11 shot.
[attachment=9445:F16___F1...banknote.jpg]
What I do not understand is why you had to make it so complex, and add extra uncertainty by using a different F stop. That meant you could not draw a conclusion without making an assumption about the lens resolution versus F stop for which no evidence was provided beyond hand waving. Then when you posted the next set of images now both at the same F stop, you changed the processing for some reason, which added yet more confusion. And back on the ranch ...
Having looked at the second set on my good CRT monitor at home (and not the crap TFT monitor I use at work) I agree there is a slight but noticeable improvement in the 50D image. Not a big deal, but it is there. Is it worth having? I cannot answer that. And that is interesting.
But ... do you think that is due to the greater resolution of the sensor, or the weaker AA filter? I would assume the 50D has a weaker AA filter due to the smaller pixels.
Regarding Moire, they both seem to have obvious Moire, but in different areas, which is consistent with different pixel pitches. After all Moire appears when the spatial frequency of a pattern nearly matches that of the sampling spatial frequency.
-
Slough,
We once had an argument regarding one of Ken Rockwell's provocative and unscientific aricles. I had little problem with his ideas because I realised they were not intended to be scientific and precise, but provocative and rhetorical. But you hung on to the fact that words have (or should have) literal and precise meanings.
With this notion in mind, that words should have precise meanings, let's look at some of the language in B. Rorslett's comments you've quoted. "Beyond f/11-f/16, sharpness rapidly deteriorated." Which is it? Beyond F11 or beyond F16? Does the rapid deterioration begin at F11, or at F13 or at F16? Or is he perhaps comparing one aperture with the next? That is, the resolution difference between apertures F11 anf F16 is small, but such differences rapidly escalate beyond F11-F16. That is, the difference between F16 and F22 is greater than the difference between F11 and F16.
There's one sure way to find out what Rorslett actually means, and that is to do your own tests.
What I do not understand is why you had to make it so complex, and add extra uncertainty by using a different F stop. That meant you could not draw a conclusion without making an assumption about the lens resolution versus F stop for which no evidence was provided beyond hand waving. Then when you posted the next set of images now both at the same F stop, you changed the processing for some reason, which added yet more confusion. And back on the ranch ...
I thought I had explained that already, but here goes again. Differences always have to be compared with a standard, that is, quantified, or it's difficult have a meaningful discussion. 2 millimetres is twice the size of one millimetre. 100 lp/mm at 50% MTF is twice the resolution of 50 lp/mm at 50% MTF. With such language you know precisely what is meant. I was searching for a simple way of describing the resolution differences between the 40D and 50D that would be of practical use to me in the field when making a choice of aperture. If they are of no use to you, I can't help it.
I now know, as a result of my very careful and meticulous testing, that the resolution increase of the 50D, although small, is at least sufficient for me to use F16 and yet get a result at the plane of focus that is very, very close to what I would have got using the 40D at F11. So close in fact that the differences (at the plane of focus) are entirely irrelevant on prints as large as my Epson 7600 can handle (24"x36"). That's what I call useful information.
As regards using DPP on one occasion instead of ACR, I did it because I wanted to see what sort of result DPP would produce. Why do you think I did it? Just to confuse you?
-
Slough,
We once had an argument regarding one of Ken Rockwell's provocative and unscientific aricles. I had little problem with his ideas because I realised they were not intended to be scientific and precise, but provocative and rhetorical. But you hung on to the fact that words have (or should have) literal and precise meanings.
No. Most people argued against Rockwell with a few lone souls such as yourself applauding him. I do not interpret words literally, but I do expect precision when it comes to a technical discussion. If someone says "Your equipment does not matter" I assume they mean that your equipment does not matter.
If you want to go the way of Humpty Dumpty, then fine.
-
"Beyond f/11-f/16, sharpness rapidly deteriorated." Which is it? Beyond F11 or beyond F16? Does the rapid deterioration begin at F11, or at F13 or at F16? Or is he perhaps comparing one aperture with the next? That is, the resolution difference between apertures F11 anf F16 is small, but such differences rapidly escalate beyond F11-F16. That is, the difference between F16 and F22 is greater than the difference between F11 and F16.
There's one sure way to find out what Rorslett actually means, and that is to do your own tests.
1) I have done my own tests as explained in my post. You did read it didn't you?
2) You suggest ambiguity in one comment on one lens and ignore all the other comments on other lenses.
Why don't you just accept that your statement that all lenses - with a few old exceptions - show negligible change in resolution down to and including F16 is, errr, not completely founded in reality?
-
Why don't you just accept that your statement that all lenses - with a few old exceptions - show negligible change in resolution down to and including F16 is, errr, not completely founded in reality?
I think it may have something to do with you not showing that this is the case.
-
Why don't you just accept that your statement that all lenses - with a few old exceptions - show negligible change in resolution down to and including F16 is, errr, not completely founded in reality?
If you have to ask such a question, I can't help you. I'm rather amazed that you should expect anyone to disregard the results of their own testing and accept someone else's word, without evidence, that their results are wrong. We all know that good lenses are not as sharp at F16 as at F11. However, resolution in the recorded image is a combination of both lens resolution and sensor resolution. The additional resolving power of the 50D, combined with a little extra sharpening, produces a result at F16 which is as detailed as the 40D at F11, with my equipment. I have demonstrated this to my own satisfaction. It matters not one whit to me if you choose to believe I've misrepresented the results of my tests. You're not the one taking my photos.
Conversation closed.
-
If you have to ask such a question, I can't help you. I'm rather amazed that you should expect anyone to disregard the results of their own testing and accept someone else's word, without evidence, that their results are wrong. We all know that good lenses are not as sharp at F16 as at F11. However, resolution in the recorded image is a combination of both lens resolution and sensor resolution. The additional resolving power of the 50D, combined with a little extra sharpening, produces a result at F16 which is as detailed as the 40D at F11, with my equipment. I have demonstrated this to my own satisfaction. It matters not one whit to me if you choose to believe I've misrepresented the results of my tests. You're not the one taking my photos.
Conversation closed.
Sorry?
"We all know that good lenses are not as sharp at F16 as at F11. "
Well, going by your earlier posts you don't.
As regards the statement that "The additional resolving power of the 50D, combined with a little extra sharpening, produces a result at F16 which is as detailed as the 40D at F11, with my equipment. " well that is a new point not made by you before. You have a habit of saying one thing and meaning another.
Whatever. You make a good Humpty Dumpty.
-
I think it may have something to do with you not showing that this is the case.
I suggest you read about diffraction.
-
I suggest you read about diffraction.
Eerily enough, that was what I was about to suggest to you, but I considered it too rude to do so to a fellow photographer.
Considering that you brought it up that way, however, I don't, anymore.
Since you don't appear to know much about diffraction and why it isn't a significant effect of a particular lens design, as much as it as to do with the aperture, I not-so-humbly suggest that you re-read the theory, and reconsider that you might be confusing "diffraction" with another technical term related to lens performance.
-
Eerily enough, that was what I was about to suggest to you, but I considered it too rude to do so to a fellow photographer.
Considering that you brought it up that way, however, I don't, anymore.
Since you don't appear to know much about diffraction and why it isn't a significant effect of a particular lens design, as much as it as to do with the aperture, I not-so-humbly suggest that you re-read the theory, and reconsider that you might be confusing "diffraction" with another technical term related to lens performance.
-
Eerily enough, that was what I was about to suggest to you, but I considered it too rude to do so to a fellow photographer.
Considering that you brought it up that way, however, I don't, anymore.
Since you don't appear to know much about diffraction and why it isn't a significant effect of a particular lens design, as much as it as to do with the aperture, I not-so-humbly suggest that you re-read the theory, and reconsider that you might be confusing "diffraction" with another technical term related to lens performance.
You pretend to be polite whilst being rude. I have no idea what point you are trying to make if any.
Ray stated that lens resolution is pretty much constant throughout the aperture range for most lenses apart from old designs. I disagree with that statement. He has now changed his mind.
"Since you don't appear to know much about diffraction and why it isn't a significant effect of a particular lens design, as much as it as to do with the aperture, I not-so-humbly "
Quite why you make such an obvious statement is beyond me. It is the very reason why I disagree with Ray's original statement.
In most camera lenses the resolution is nowhere near the diffraction limit except when stopped well down other aberrations being dominant at wider apertures. Some Schmidt cameras are fast - e.g. F1.0 - and much closer to being diffraction limited. I am sure you can find information on this from Google. Some of it might be accurate.
You might wish to read up on concepts such as the Dawes Limit and Rayleigh limit, which are measures of the theoretical maximum resolution of a telescope, but have relevance to camera lenses when stopped well down.
-
You pretend to be polite whilst being rude. I have no idea what point you are trying to make if any.
I didn't pretend to be polite, I was admitting that I was being rude, quid pro quo.
You might wish to read up on concepts such as the Dawes Limit and Rayleigh limit, which are measures of the theoretical maximum resolution of a telescope, but have relevance to camera lenses when stopped well down.
Here's one for you (http://books.google.no/books?id=HHX4xB94vcMC&pg=PA80&lpg=PA80).
I'm not the one claiming that diffraction has something to do with other effects than the aperture.
-
I'm not the one claiming that diffraction has something to do with other effects than the aperture.
I don't understand what you mean by that sentence. And your point is?