This content has been marked as final. Show 29 replies
I have downloaded the Sony A700 and Nikon D300 images, which have been compared, and looked at them. Two notes:
1. The D30 images are 14bit, the Sony 12bit.
2. The ISOs of Sony are fake already at 800, perhaps even lower. Nikon's 1600 appears true ISO, and although the higher ISOs are fake, that is less "disruptive" in 14bit mode. Plus, Nikon alternates the generated levels between the two green channels, while Sonny does not.
All this is no reason for the crappy result ACR is delivering; I mean these only as cautioning, that the comparisons between the images of these cameras are not very helpful.
Btw, what is the reason to shoot with fake ISOs when recording raw data? The fake ISOs are for JPEG! The only positive is to see a brighter image on the LCD.
I told the "ps experts" time and time again they did not need those Nikon D300 images, and they had all they needed with my A700 files that I posted from the beginning.
Finally, I gave them what they wanted but did not need, and told them it was a waste of time in doing so.
I repeatedly posted A700 images for conversion and finally A700 image 4 has been rendered by the two photoshop gurus with acr, and compared to IDC it they failed miserably. File 1a also has both conversions...and Ken had them do one as well. All cases showed ACR falling down, as we knew it would since pixel peepers, reviewers, pros, and photo mag publishers have been documenting and moaning about this A700 ACR problem for months.
The Nikon and Sony are using the identical Sony sensors. Both Nikon and Sony say they are ISO200-3200 boost 100/6400. Phil at DPReview says ISOs are as claimed.
We Sony A700 shooters shoot both Raw+JPG as JPG's are best in the business after FMV2:
A700 Dynamic range is killer and also DRO circuit can be a miracle maker...no sense not to do Raw+Jpg. DRO is done in raw before jpg is written and can not be duplicated after the fact in the converter.
IF WB is on the money, no since opening the raw file in many cases. When shooting in this mode DRO and other adjustments to the image are accurate on rear display.
The #4 coin came from this A700 ISO6400 series:
The fish #1a is from this high ISO series:
***Note: If you recieved this post via email, the links are now correct.
Note - "identical" was too strong "same Sony 12MP CMOS sensor" with a slightly different implementation is what I meant to say in the post above. I tried to correct it, but it was too late.
> The ISOs of Sony are fake already at 800, perhaps even lower. Nikon's 1600 appears true ISO ...
G Sch: Out of curiousity, what is a "fake ISO", and how can you tell?
Don: Have you compared ACR results with other 3rd -party developers? The reason I ask is because I can also find fault with ACR and my Oly ORFs, no no other developer can do better, and in fact they're a lot worse. I attribute it to manufacturers softwares doing it better just because they know how, and also because they refuse to collaborate with Adobe.
That's my own guess --- but it may also be because Adobe cannot provide a suitable "one algorithm to fit all". Which do you suppose it is? In either case, it's not entirely Adobe's fault.
my CA$0.02 :)
Yes, and in that thread I posted is also a comparison against ACDPro 2 KH did against a "ps expert", and two other links are posted that links to articles that compared another half dozen converters by one one of the most respected A-Mount pros on the planet.
The head to head conversions have been done in forums since September showing the ACR flaws against numerous converters by dozens of photographers.
That is the problem...it is just ACR mangling the A700 files. With the other converters the grain stays tight and small...with ACR as ISO rises to about 1600, smearing and blotches start taking over the image.
C1V4, Bibble, RT, ACDSeePro 2, etc, etc, all give a nice tight grained A700 high ISO file while ACR4.3.1 is the only one giving us the splotched watercolor blotches and detail smearing.
I think ACR is not recognizing something properly in the A700 files that the others are. Tone curve reproduction seems off at times, and some sort of NR/Blotches seem to be in the image before you even get started with ACR. Those two things don't happen when you use any of the other converters.
I don't know why it's happening, I just know it is happening. And since I have 200K files from a half dozen cameras it really sucks to have to use another processor on the A700 files...really throws a hitch in your giddy-up.
PS...Here is a link toy one of a couple articles by DK, quite possibly the most respected A-Mount guy on the planet. His test is not perfect, but it clearly show the splotches and artifacts (confetti, he calls it) that no other processors in the land gives us.
These "watercolors" absolutely ruin shadow areas and backgrounds, and as I mentioned before everyone else (third party) in the business gives a fine grain pattern that can be easily dealt with after the fact.
Those ACR splotches are horrendous to deal with and that is the main reason we need the converter tweaked for A700 files...nobody wcan deal with those "watercolor" effects at high ISO!
I prefer IDC, most all other shooters are using RT, ACDPro2, C1V4 and others...
>Both Nikon and Sony say they are ISO200-3200 boost 100/6400
I don't understand the above notation; however, it does not matter, as the top true ISO of the D300 appears to be 1250, and the Sony's is much lower.
>what is a "fake ISO"
A permanent adjustment of the native raw values, like an intensity correction (called "exposure" in ACR). The point is, that this occurs before the in-camera raw conversion in JPEG.
For example the Canon 40D offers 1/3 stop ISOs. However, the +1/3 and +2/3 are simply adjusted pixel values: the ISO 100 values get multiplied by 1.26 (corresponding to 1/3 stop) and the result is called ISO 125. When ISO 160 is selected, then ISO 200 is applied and the pixel values are divided by 1.26.
ISO 3200 is "fake" on most cameras, it is simply the doubling of the ISO 1600 values.
This is useful for in-camera JPEG images. Don is recording raw plus JPEG, so it is reasonable to use this feature.
>how can you tell?
Already the histogram indicates it. See the following three histograms:
The red and blue histograms of the D300 are somewhat "hairy"; the green not, because Nikon made it smarter than Sony. The Sony histograms show much more of that. However, this is only indication. The proof is in pixel-level analysis, which is much more difficult to describe, than to do.
For example the ISO 1600 image of the D300 reveals, that after every fourth or fifth pixel value there is a "jump"; there is no pixel with the corresponding value, a clear indication, that those values have been created by multiplying the native values. In the ISO 6400 image from the Sony A700, there are gaps 3 pixel values wide, i.e. only 1/4 of the possible pixel values are present.
I found a very interesting aspect; it may explain the reason for the bad rendering.
The DNG converter (and then ACR as well, I guess) sets the black level at 128 for these images (higher with 14bit recording). I don't know where this comes from, I was blindly mimicking it.
However, there is valuable information under this level (the thin line in the histogram posted above shows the location).
Following crop demonstrates, which areas are affected by this cut-off:
G. Sch. - Thanks for doing that, much appreciated.
The info identified on the panopeeper crop are the exact areas of the image(s) we always complain about. This is where the weird watercolors and blotches always occur in ACR that do not occur with the third party processors on A700 files.
So, this seems easy enough for Adobe to fix, huh?
Don, don't be so sure, this may be useless. I did the same test with the coins image, see http://www.panopeeper.com/Demo/SonyA700_ISO6400_ACRissue_Coins.jpg and I guess the effected areas are here much less than you are complaining about.
Pls download and try http://www.panopeeper.com/Download/A700BL0.DNG
This is your AA700INv2I6400fine.ARW converted in DNG and black levels modified from 128 to 64 (I don't know how high they should be). This change shifts the colors as well, you need to pick WB. Try it if you get a better result with it.
G - the coin image should have almost no problems...if there is not shadow areas the problems is much less evident. Easy, bright images ACR problem does not rear it's ugly head nearly as bad.
I gave them four images to do, and then asked them to do that one last, knowing it will have detail removal problems, but not a lot of large blotching. They opted to skip the tougher ones, and do the easy one instead.
Let me get the fish and a couple others, give me a couple minutes.
I uploaded _DSC0369.DNG, _DSC0440.DNG, _DSC0515.DNG and _DSC0710.DNG, black level set to 64, except in _DSC0440, where it is 50. All from http://www.panopeeper.com/Download/
Let's see if you find them better.
G - re:(A700BL0.DNG)
That makes for a TON better file...Splotches are gone and the detail level is good compared to IDC, wb adjustment makes the file look nice and after tweaking probably better than IDC.
Something similar to that should work.
Wow, what a change...good going, what do we do next?
PS...Downloading the other files now.
>what do we do next?
I am speaking only for myself. I don't know what I will do, for I trusted these values in my program. It is obvious, that there must be *some* black level setting, but I don't know how to find them.
In the other cameras, which my program supports the black levels are either available directly or can be evaluated, if this applies at all; for example Nikon raw images come "black levelled".
G - Not sure what sort of feedback you need.
Awesome grain/noise structure, very good, worlds better than it was.
64 makes the nicer appearing structure I think. This "fine-grain" structure is what everyone is looking for and if the program defaults can give this structure and a "correctish" looking colors and exposure from near default settings this will solve the problem for everyone.
This is where we want to go - so this plus whatever it takes to get rid of the washed out - tinted look appears to me as if will solve A700 owners problems with ACR.
That's good stuff...no more blotches, and nice fine grain.
>Not sure what sort of feedback you need
Nothing, what you could help with. Manufacturers don't publish the structure and meaning of the proprietory information they add to the raw image (called MakerNote). Though much of that has been analyzed/reverse engineered by interested persons, I too am doing it; however, I did not find any useful info re Sonys.
Re ACR: if you inferred from my posts that I have anything to do with Adobe beside being one of their millions of customers, then you misread that. I analyzed these images, for I found the issue interesting from a professional point of view. That's all.
I got you, then it is indeed Adobe that is missing the mark because it's quite clear that the other third party processors are giving the nice grain structure and no blotches quite similar to that of the DNGs you have shown.
I guess then Thomas just needs to make a fewer closer guesses then...:-)
C1V4 is the most recent converter released that gives us the proper renders. Many free converters have also got it right so it must not be extremely difficult for need-to-know peeps to reverse engineer what they need to know on this.
Thanks much for your help, it seems as if we may be half-way home on this now.
BTW, do you know if TK reads these threads often?
>BTW, do you know if TK reads these threads often?
Yes he does but to be honest, the tone of subjectline in the original post is not very inviting to reed for him, or in fact for anyone else... :-(
Not much I can do about it at this point.. Unfortunately, the subject line is quite true.
No offense intended, and I guess it's the fact that A700 users have had to deal with this since September.
We waited through one release and no fix, so we are now hoping we don't get bypassed on this again since the entire A700 userbase is affected.
I use ACR on my non-A700 files and I do quite like the product.
The title of this post should be the one he reads first!!! If he has any pride in the software.
the reason people post these messages is because they want the software to work because it has lots of good UI and design work that is undermined by a weak RAW engine.. Weaker in IQ than the ACR 3.4.
Some of us who have spent hundreds if not thousands on Adobe products have concerns when small passionate development houses and open source developers offer more fidelity to the RAW file format.
This was only made worse by my visit to PMA. NO Adobe booth, and when I went to the "office" adobe rented as asked where adobe was the receptionist for that area of PMA said "She has left for the day"
Is it a surprise that I wonder if Adobe has lost interest in the single photographer and is all about suites and expensive video tools.
I would hope that Adobe would get on this quickly, as this would seem to be a fix that would be easy to implement and quite crucial to A700 owners.
Speaking as an A100 owner, I have been considering the purchase of an A700, and would prefer not having to abandon Lightroom for ACDSee 2 (which currently does handle A700 files properly) or switching to Bibble for RAW conversion (likewise).
Bump - We still need this fixed. Thx.
Would converting the AWR Sony a700 files to DNG (generic raw) before doing any editing solve the high ISO problem? Please let me know.
No...that does not help unless the black level is changed within the DNG file. Gabor even wrote a program to *adjust* what appears to be the messed up black level recognition on the ARWs with DNGs. It's a painstaking process.
Check this DNG thread.
That thread is where we have found at some ISOs ACR blacklevel recognition seems off at some ISOs.
This thread is where Gabor made a program to alter black level withing DNGs.
VFH...Uummm...the first DNG thread I linked is this one you posted on! It pretty much identifies for adobe what one of the problems is!
The second thread I link is Gabor's program to change the black level settings inside the DNG files so then can be renders by ACR better than ACR does straight up (due to the BL being different on the original files).
FYI...the program Gabor wrote adjusts the Sony A700 BL inside the DNG before white balancing. It allows BL to be adjusted on individual channels before raw conversion and this can make a serious difference on many high ISO A700 files.
TO Adobe programmers: IF ACR default parameters are tweaked a bit, and this BL thing corrected, this will go a long ways to UN-Watercoloring the ACR/Lightroom A700 high ISO files.
Has anyone tried the 4.5 beta? I would also make sure and e-mail support.
These are user to user forums and their no guarantees about who will or
won't see posts here. This seems like an important issue so please make sure
and e-mail support. With luck they can get this fixed so that the 4.5 when
it goes final has the fix.
The A700 high ISO problem has not been fixed in 4.5 beta. Tons of people have emailed Adobe, and filed official bug reports during the past year.
Adobe has been blasted by Sony users across the web for almost a year on this and they have done nothing other than ignore the issue on several releases. It's becoming apparent that someone does not want to spend the time on this issue, or they have deemed the issue to be a non-issue. I can guarantee you it is an issue, and almost every raw converter on the planet toasts ACR/Lightroom when it comes to A700 high ISO files. ACR/Lightroom does make a blotchy mess of many A700 high ISO ARWs. Many of us Sony users have had to scrap Lightroom/Bridge because of the poor Adobe performance.
Sony has givven us firmware v4...which pretty much clears up the problems with ACR and a700 files...Dunno why ACR was giving the most watercolors in the first place, but all seems resolved now.
Glad to hear it, Don.