UCSB wrote:
NO. That is only in ARC, I am not seeing that in C1. We know that ARC is not working correctly and the 7D is not fully supported. It is a waste of time to use defective software to perform tests.
while maybe ACR is not handling the 7D properly it is a bit extreme to call it defective software since it is using the same methods that worked 100% fine for all previous DSLR bodies....
And even Canon's own DPP shows some degree of problems, even if you tone settings down enough to largely hide the strong mazing effect, there are still linear patterns of dots or stray bits of noise left behind, of the sort that I don't see with DPP with old 20D/40D/50D/5D2/1D3/1D2/1D2n/10D files.
gfiksel wrote:
I don't think it's compression. First, it was level 7, not 6
Second, I see the same in uncompressed images as well. Here's the same at compression 10:
I understand your argument about the mRaw format. However, its an attractive shooting option for many.
one thing to keep in mind is that the exact shade can make a difference in how noticeable the mazing is, on my CCC some squares show it strongly some only minorly
another thing is what settings did you use?
I used ACR settings for pre-sharpening and NR of 20 sharpness, 0.5 radius (to maximize the finest details in my images), 25 detail, 0 masking, 0 luma NR, 9 Chroma NR (for ISO100) and then did a multi-pass sharpening in CS4 (it shows using both my normals methods the agressive and slightly more toned back settings, although more apparent with the former).
Again, with an old 20D/50D color checker, if I do that, ALL squares simply looks silky smooth and noisless as to be expected of a smooth, slightly OOF surface taken at ISO100.
Of course maybe the 7D does need some special thigns to be done, so maybe it is not fair to use ACR 5.5, but at least it is fair to note that the 7D wont play well with quite a few demosaic algorithms that had worked fine for almost any camera before.
Even Canon's own DPP definitely shows left over signs of the problem that do not appear when I use it on files from other cameras.
C1v5 perhaps does the best, especially at default settings (although at default it doesn't bring as much fine micro-contrast and detail as it did with older cameras or can be gotten otherwise form a 7D, that said it might be possible to turn luma NR down and boost sharpening to get it to a reasonable level without bringing back too many issues, still need to play around with it more, a side note is that I don't like that C1 doesn't filter out, at all, any of the extreme black or white dots, in fact with the 7D it seems to work an awful lot like DPP only it allows many more fine tunings of settings).
anyway i'm waiting on ACR 5.6 now, DPR pics appeared promising, but it is hard to say really.
skibum5 wrote:
although to be honest with that red patch above, in that one, I simply used the same sharpening settings throughout that I do for actual photos I'd take with my old 20D, the sharpening is getting towards the high side but it's not at insane levels or anything
anyway as for why UCSB's suddenly got better, did he maybe use masking and NR in ACR? change the ACR pre-sharpening settings? not sharpen at all in CS4?
Yes, I did use ACR NR (including masking) and sharpening. Nothing in CS4 because there are basically unlimited options. I did crank up SmartSharpen all the way ... nothing but a lot of over sharpened noise.
I haven't done any of the raw analysis tests but you certainly still can see it with ACR 5.5 style-demosaic, I think it doesn't happen as much as with mine though, but it is still not gone or as smooth as from a 20D-50D.
ACR 20 sharpness, 0.5 radius, 25 detail, 0 masking, 0 luma NR, 9 chroma NR
and then an agressive sharpening of first pass smartsharpen with radius 0.6 at 80% and then second pass with 100% and 0.3 and then 70% at 0.1 or with a bit less agressive 0.6 at 50%, 0.2 at 90% and 0.1 at 70% (less evident but still there a bit on a few patches). Doing the same with a 20D or 50D gives one no mazing whatsoever AND smoother looking patches (granted the 20D is only giving 8MP so that is not fair but the 50D's 15MP are close to the 18MP of the 7D).
I haven't tried it with C1 yet.
It doesn't appear to show up quite as easily as with either of my two copies, which is a little odd if the whole thing is by design, that almost hints more towards QC issues? Even more odd is if his camera really did suddenly change in the column to column variance, it is hard to understand how it could spontaneously do that.... sure wish mine would . Maybe one of those one in a google to a google to a google to a google random sets of quantum fluctuations redesigned his sensor in one fell swoop . Although that said, his copy still has more fluctuations, by a noticeable degree than my 20D,40D,50D,5D2.
Anyway, even his copy still doesn't really play as well with typical standard demosaic that were used before.
Hopefully ACR 5.6 works some magic.
Of course if you just want fast speed and better AF and better performance and don't care about every last ounce of reach then even my copies do a lot better than a 40D pretty much every which way, since certainly by 10MP the artifacts are gone and it gives a little better detail and less noise and banding than a 40D, etc.
Where it might start to matter is if one of the two prime purpose of the cam is as an extra body for heavy cropping of distant subjects at low to moderate ISO. I'm not sure, but I wonder if a 50D wouldn't do that a bit more smoothly looking with the current RAW conveters at least.
And regardless of anything, and it probably won't be the end of the world in the end, personally I don't like this sort of compromise and I think they'd be better off giving back a tiny bit of whatever, if anything, this gained them and going back to old style CFA/AMPs for future sensors. I could be wrong, but I don't see how this is gaining them anything in the real world, maybe on pre-demosaic tests like DxO do maybe it can gain them though, but what does that matter in the end.... So I don't mind raising a stink if they go back to a hardware system that makes RAW processing easier to deal with.
Where are all those who at one time used to sing how silky smooth Canon images could be?
Interesting. The difference of the G1 and G2 channels shows the least banding of any example I've seen so far in the RAW data. So maybe it is a QC issue, or some calibration that was omitted on a batch of cameras. Dunno why it would appear initially and then disappear on this camera, though.
The difference of greens does seem to indicate a different response, though, similar to but not identical to past examples:
The scale here is -500 to +500 RAW levels in the difference, and the file has been downsized by a factor of four. In most patches the response is about the same (yielding neutral gray), but in the brighter tones there is a substantial mismatch (particularly the white patches, for some reason).
EDIT: And, yes, the flesh tone patches show a difference, as skibum has discovered.
brainiac wrote:
I just looked at that file for mazing in DPP and could find none, even with sharpening and contrast at maximum. Amazing.
I looked also and I see mazing in almost every patch, most apparent at the edges of the patch. Unless I'm interpreting this differently than what others are considering mazing. Whatever, it looks exactly like my 7D.
> Of all the cameras we have ever used, we loved the handling of the Canon 7D the best. What a little sports car of a camera! We so much wanted to love this camera. But in test after test we constantly were disappointed in the quality of the files.
Sums up my centiments about this camera.
If Canon could only stay at 12mp . But I guess fake 18mp + fake high ISO = better sales after all.
keithreeder wrote:
Yeah, just like we take, process and view real photos.
I despair...
In general I agree with you completely that peeking at pixels in this way makes little sense. When I got my 7D one of the things I noticed was that skies looked a bit noisy at ISO 200. When I compared it to my 20D, I discovered that it is indeed noisier all the way to ISO 800. This was a big surprise to me after seeing how good the 7D is at high ISOs. Other people have discovered this "mazing" characteristic, and that is probably the source of the perceived low ISO noise. I am committed to the 7D. It is a fantastic camera. I just want to know if a) my camera represents the norm, in which case I'll just wait for an update to ACR and hope that this condition is tamed, or b) my camera is outside the norm, in which case I need to get it exchanged.
Fred Tedsen wrote:
I looked also and I see mazing in almost every patch, most apparent at the edges of the patch. Unless I'm interpreting this differently than what others are considering mazing. Whatever, it looks exactly like my 7D.
What software are you using? This just does not sound right. I converted with both C1 and DPP.
I have been processing in DPP with no noise reduction or sharpening.
then output at 16bit TIF
open in Photoshop CS4 64bit
-- remove noise with Neat Image 64bit v6.1
-- sharpen with "smart sharpen" in CS4
save a PNG or JPG
Fred Tedsen wrote:
In general I agree with you completely that peeking at pixels in this way makes little sense. When I got my 7D one of the things I noticed was that skies looked a bit noisy at ISO 200. When I compared it to my 20D, I discovered that it is indeed noisier all the way to ISO 800. This was a big surprise to me after seeing how good the 7D is at high ISOs. Other people have discovered this "mazing" characteristic, and that is probably the source of the perceived low ISO noise. I am committed to the 7D. It is a fantastic camera. I just want to know if a) my camera represents the norm, in which case I'll just wait for an update to ACR and hope that this condition is tamed, or b) my camera is outside the norm, in which case I need to get it exchanged....Show more →
I am seeing light noise in the sky at ISO 100 and greater at 100% view. At this point in time, I would not worry about it. One thing you should realize is that coming from the 20D to the 7D you are really magnifying the contents of the file ... including the noise when viewing at 100%. Just adjust your processing to the new camera.
On the long shot you are using ARC or LR, just stop ... and wait for official 7D support.
UCSB wrote:
What software are you using? This just does not sound right. I converted with both C1 and DPP.
I used Lightroom. I just now tried your image in C1 and the result is very different. There are some strange looking artifacts, but only a few hints of mazing. This is quite different from the experience I've had with some of my images, where looking at skies have shown pretty similar amounts of mazing. C1 does seem to handle the appearance of noise better than ACR 2.5, as does Lightroom 3 beta, so I'm hopeful that all this will fade away eventually.
: 
. But I guess fake 18mp + fake high ISO = better sales after all. 