UCSB wrote:
I think getting the second body was a good experiment. After the dust of this product introduction settles, I will pickup another one myself.
Not that you implied anything to the contrary, but just to reiterate: I got the second body not because of the IQ issue, but because the first body had a big spot on the sensor that wouldn't blow away.
I had been somewhat loosely following the Low-ISO-noise/G1,G2/Mazing stuff (I'm not implying that they're necessarily intertwined in any particular fashion...), and always intended to look for such artifacts, but it was luck, in a sense, that I have had two bodies that I could do a precise comparison between.
I will be sending the first body back to the vendor soon, and I may do some more comparison shots before then, but my time is finite, so I may not do much. Until and if I post any careful comparisons, here's a 200% crop of a real-world shot from the first body that hasn't been pushed wildly - the processing parameters are my default Picture Style (Standard with -1 Contrast, -1 Saturation; sharpening is the default 3). It is at ISO 1600, i.e., not "low-ISO". Indoor lighting of unknown type.
Do we think these artifacts are reasonable for a "normal" ISO 1600 image (& no noise reduction)? (I am asking earnestly; I haven't gone back to compare with my prior x0D body ISO1600 images.)
cameron12x wrote:
No offense intended, but I would not be "smiling" or happy with the results of the processing on that image.
Sorry that I can't just post the raw file; no "model release." In any event, I only posted the crop for informational purposes; no suggestion of quality or lack thereof.
Is the exif metadata available? What lens/aperture/shutter speed combination did you use?
35mm f/2, 1/160 @ f/2, handheld. I don't purport that it is sharp; really just posting it so folks can see the noise artifacts/details at 1600 for this particular body.
How are you sending your original body back? I may be doing the same at some point.
The body is going back to the internet seller from whom I bought it; I let them know about the sensor spot w/in two+ days of receiving the camera. They sent out the second body as a replacement immediately and just need me to return the first w/in a specified time frame.
keithreeder wrote:
And for the umpteenth bloody time - no observable maze artifacts whatsoever from Raw Therapee, Bibble 5, UFRaw and others...
Not so. Bibble 5 shows it quite clearly, and even DPP shows artifacts, though not as mazing and to far lesser degree.
Edit: I have to retract my characterization of Bibble 5. It doesn't show mazing on these images. My own testing has been noise in skies, and for those images Bibble 5 looks quite like Lightroom.
After Darwin's review, I wanted to do my own sharpness test shooting a $20 bill. F/11 still looks sharp to me, f/16 the softness creeps in. I used Live View with manual focus using the 180L. Raw files also included for DL. No capture or output sharpening applied
That is quite a large difference. I hadn't looked at your first sample before now and so hadn't realized how bad the mazing could be. In fact I think that I've been mischaracterizing my camera, as it seems to be much closer to your second copy than the first. There's no doubt that there's still something going on in your second copy (and mine), but the question is, how much (if at all) is image quality affected. After taking quite a lot of photos now, I'm beginning to think that the answer is not very much.
theSuede wrote:
I've tested 13 cameras so far - and TWO has been "good", eight has been "nah...." and three have been "OMG! that's BAD!"...
Where would Paul's two copies fit in that group? Does a "good" 7D have no green imbalance at all, and if so does that imply that a camera with it could be adjusted to remove it?
System MTF, the SUM of all the parts of the system is what counts as you say - when you try to find your maximum reproduction size. And the "maze" effect lowers this (if the rest of your system is up to the task!) by a noticeable margin. That is not acceptable, if you buy an 18MP camera you want 18 usable MP, not MP's that you have to downsample. Having to downsample to get "pixel-perfect" results is kind of "ok" in a 150$ compact, but not a 2500$ flagship APS DSLR.
If a camera doesn't show obvious "mazing" as with Paul's second copy, but still shows green imbalance as shown by Emil, what will be the effect on MTF? How can we ever quantify this?
Fred Tedsen wrote:
That is quite a large difference. I hadn't looked at your first sample before now and so hadn't realized how bad the mazing could be. In fact I think that I've been mischaracterizing my camera, as it seems to be much closer to your second copy than the first. There's no doubt that there's still something going on in your second copy (and mine), but the question is, how much (if at all) is image quality affected. After taking quite a lot of photos now, I'm beginning to think that the answer is not very much.
if a copy is as good as his second copy it probably won't matter too much when actually going out and shooting, still likely not quite all it could have been, but with ACR 5.6 or DPP etc. it would probably be a rare photo that was affected to the point anyone would easily notice it, but I haven't used such a better copy or acr 5.6, looked into it in detail, etc. just my guess. I still have a hard time seeing how they gain more than they lose. I hope no future bodies are built this way. But with a copy like his it should be ok I think/
the copies like mine though, of which there seem to be quite a few to say the least, I don't feel are really acceptable even with 'real' images and processing.
If you mean paulfeng - that's very hard to say. I don't have the raw, so I'll have to go from very uncertain already processed samples... The problem is (in a simplified way) twofold. One part is in the shadows near black (this the black-point imbalance), and the other is visible all the way up through the picture - all the way to saturation (this is the gain-imbalance).
Both seem to be lowered in "pictorial" effect, i.e. visible effect in a real picture as you increase signal amplification (ISO). All 7D's seem to have very good high-ISO behaviour compared to the competition.
A "good" 7D has very low black-point imbalance, and a channel amplification imbalance that's at least a decade lower than the shot-noise at mid-grey (inherent uncertainty in measured light amounts).
*To put this in perspective, gain imbalance in the Nikons I've rechecked after this whole "7D-maze" thing started have a gain imbalance in THE WORST sample that's five times lower than the BEST 7D that I've measured. I've no idea how Nikon achieve this, as it certainly isn't hardware calibrated to that level of precision - but still. It's possible to do. This in part answers Emil's previous question about how Adobe handle the 12 output channels of the D3 with only eight channels of "repeat" calibration. Short answer - They don't, because there's no problem to "handle".
The effect on MTF is very easily quantifiable. MTF is a "contrast" vs "resolution" weighing that can be done in two ways - contrast at a certain resolution, or a maximum resolution to get a certain contrast. This means that picture detail with low contrast will be affected by green imbalance, detail with high contrast less so. The MTF has to be weighed against noise and/or "digital artefacting" in real-world scenarios. More noise - lower MTF.
So "MTF50", which is the highest resolution where the camera/lens will still preserve 50% of the real-world resolution will take almost no damage at all until we start measuring at quite low exposure levels (where it effected by shadow noise, less light > more noise). "MTF20" on the other hand will suffer a bit. And detail with a contrast of only 10% - "MTF10" (that's about 1/6Ev) will only be visible in parts of the image that's bright enough.
You can use a log/log contrast/resolution chart to measure things like this. The chart is very "scientific" and many may scoff at it as being of "very little value to real photography". This is totally wrong. You can very accurately measure a camera/lens combination's ability to resolve fine detail like surface structures of very natural things like for example human skin. This is not a bad thing to be able to quantify.
This (low contrast detail), a perceived increase of mid-tone noise, and that you can't sharpen the pictures as much as you might want without the "nasties" showing their ugly faces are the effects of channel imbalance.
theSuede wrote:
*To put this in perspective, gain imbalance in the Nikons I've rechecked after this whole "7D-maze" thing started have a gain imbalance in THE WORST sample that's five times lower than the BEST 7D that I've measured. I've no idea how Nikon achieve this, as it certainly isn't hardware calibrated to that level of precision - but still. It's possible to do.
How do the 7D's you've examined compare against other current Canons -- say the 50D, 5D2, and 1D3? It would be good to know whether and how the 7D is different from its siblings.
Thank you Paul. I'll look at the raw's later tonight. And try to gather some other cameras as well... I already have the data, but not in a uniform presentation format.
