garyvot wrote:
Yes of course, the 7D is not a D3s. However, it seems to be widely acknowledged that it has set a new benchmark for high ISO performance froim APS-C, at least until the appearance of the D7000 (and even then it's not clear that the D7000 is better, merely as good).
Interestingly Pop photo found the 7D was still better at high ISO than the D7000 and remarked the D7000 wasn't great at ISO 3200.
It's kind amusing to read your comments regarding ISO performance between 7D and other cameras. People above told me that NO NO you cannot compare 7D to 1D3 because they are different cameras (sensor/MP). Now you compare 7D to other cameras (Canon FF/D3s) and people agree with you. What's the point then? Is 7D suddenly the same camera as Canon FF/D3s so you can do the comparison? If not, what is the criteria to say that one is XX stops better than the other? Using the same logic, may I challenge you that you should enlarge D3s photos to 18MP and do a comparison?
As said, if you shrink the 7D photos for web posting, ISO 6400 may seems perfectly fine. But that is sort of cheating from my point of view. The pixel level quality is the easiest thing to compare without any manipulation.
SoundHound wrote:
There are always, processed and unprocessed, exceptions of image material that will work at the extremes of several parameters. It's hard to compare up loaded images for IQ, unless of course, this is your ultimate use. Reproduced image size and media are critical.
So the question might be "Will it work for me @ xxxx ISO."
The 7D is populated with 18Mp on a sensor that is 39% of FF. That means that, at higher ISOs, ultimate, resolution will be limited by noise. Indeed, ultimate resolution at 18Mp (approx 47Mp FF pixel density) and any ISO, is difficult to realize under most conditions.
The question might also be "Compared to What:" From these posts it seems that the 7D lags the FF Canons by at least one stop in hi-ISO performance and, it seems, extra internal NR (over the FF Canons) is necessary to achieve that. That might imply a gap of 1.5 stops. That makes sense since the sensor area so much smaller than FF.
And, I can tell you from some experience, that means the 7D is at least 3 stops behind Nikon's D3s. FYI I rate the FF Canons for my use at ISO 3200, the D3 at 6400 and the D3s @ 12,800. I would hesitate to advance any cropped sensor's ISO above ISO 1600. YMMV...Show more →
terminator wrote:
So, sounds like that you do admit that 7D's images are flawed? I have a 1Ds3, why cannot I complain about 7D's image quality? I never expect 7D's image quality to be as good as those contemporary or even older cameras with lower pixel density. I just think the road Canon took, sacrificing pixel level image quality for high pixel count, is not right
I think that whether it's right or not is second to the fact that this is what they've done, like it or not. In addition, they appear to have incorporated behind-the-scenes noise reduction even to RAW files in their latest hi-rez cameras, though this has yet to be 100% substantiated. Just from peering over and over again into my 5D II RAW files, I have little doubt this is the case. Again ... whether or not this is right is second to the fact that this is what Canon has done ... like it or not.
I, for one, don't like it. So I went backwards, sold my 5D II and now enjoy for the second time a 5D Classic, a 40D for travel (which is about as close to a 5D in IQ and noise control as a crop body can get, the t2i notwithstanding) and a 1D II for that very special set of 1-Series qualities. The only thing I'm contemplating at the moment is perhaps returning to a 1Ds II. That's about as much pixel density as I can tolerate.
I think the majority of 7D images I'm seeing here look absolutely terrific, including the ISO 3200 samples. Not being a bird shooter, I am more impressed with the sharpness, color and detail you folks can capture than I can express.
But processing for small JPEGs on the web isn't the same as processing for large prints, and large prints is what I do 95% of the time. My average size print is 16 X 20. I can't help but wonder what those print sizes and larger might reveal from those same ISO 3200 shots. Even my 5D II's high ISO shots didn't look quite right in large sizes, and I mean ISO 800 and 1600, not 3200. ISO 200-400 from the 5D II also revealed some unwanted grain at large sizes as well as some other really odd stuff that I can't even describe.
So, in my mind, there is definitely something to be said and seriously questioned about Canon's current hi-rez cameras and the kind of technology they've been squeezing into an area the size of a postage stamp. The processing aspect within that area is amazing. 8fps for 18MP images is pretty impressive.
But the actual photo capture part has become somewhat destructive and continues to be headed down that path. Doesn't seem right, doesn't feel right, doesn't look right to my eyes ... not when my 1D II files printed to similar sizes look a bit better ... more like real photographs rather than highly detailed, sharp, colorful digital renderings.
How about this test? This was a shot I did of a lens I just sold, with the 7D at ISO 1600. Raw through DPP right to JPG, resized up to a 19x13 at 280 dpi, then the canvas increased to a 16x20. Probably not quite what you do, but this is completely untouched except for the uprezz, no post noise reduction or sharpening, etc. I have a ton of 3200 shots, but this was still in my working area.
TeamSpeed wrote:
Hi Steve,
How about this test? This was a shot I did of a lens I just sold, with the 7D at ISO 1600. Raw through DPP right to JPG, resized up to a 19x13 at 280 dpi, then the canvas increased to a 16x20. Probably not quite what you do, but this is completely untouched except for the uprezz, no post noise reduction or sharpening, etc. I have a ton of 3200 shots, but this was still in my working area.
Uh .... not sure what I'm supposed to do with that photo or what it's supposed to reveal. Am I missing something? As I say, processing for small JPEGs is one thing. Printing big is another. In that context, what am I looking at and what am I looking for?
veroman wrote:
Uh .... not sure what I'm supposed to do with that photo or what it's supposed to reveal. Am I missing something? As I say, processing for small JPEGs is one thing. Printing big is another. In that context, what am I looking at and what am I looking for?
- Steve
Click on it...
Here is a 3200 raw to jpg uprezzed to 16x20, a boring view, but no post processing. I used the crop tool set to 16x20 at 280 dpi. Again click....
veroman wrote:
I clicked. Then I clicked again and got, I guess, the 100% view. Correct? Well ... it didn't look all that good to me. Really strange stuff going on there. Softness. Noise. Stuff I can't even describe. Why do you have to up-size a 7D shot? What's it native size shot at full resolution?
There is no native print size, it depends on your printing equipment. To send to a printer at 280 dpi for a 16x20, you need 5600 on the long side. If you want 300 dpi (which is overkill for a large print due to the viewing distance), you would need 6000 on the long side.
Sorry you don't like the 3200 uprezzed image. Do you have a link to an uprezzed 1Ds? The 1DsII has less resolution than the 7D, so for a 300dpi printer, you would have to uprez it to print it even more. Whether you do it, or you just leave your image alone when you send it to printer, the printer will then do it, one way or another, it is getting upsized to print.
If you just want ISO 3200 shot in low light high contrast and resized down to fit in a post, I can do that as well. Or anybody that has a 7D. I thought you wanted a 100% pixel view of a 16x20 print (which is ridiculous since nobody puts their noise on a 16x20 print to look at it), I apologize for not understanding what you wanted.
Pixel Perfect wrote:
HTP is of no use if you shoot RAW.
Actually, this isn't quite true. HTP is not a JPEG processing parameter (like ALO).
You can duplicate the results of HTP only if you shoot RAW AND you underexpose by a stop and manipulate the image in post to bring up the brightness while retaining highlight detail. (This is what HTP does for you.)
But just shooting RAW without using a special exposure technique will not give the same amount of headroom in your images as you can obtain shooting RAW and using HTP.
You may have *enough* headroom shooting RAW most of the time without HTP, but at least it's there if you need it.
TeamSpeed wrote:
There is no native print size, it depends on your printing equipment. To send to a printer at 280 dpi for a 16x20, you need 5600 on the long side. If you want 300 dpi (which is overkill for a large print due to the viewing distance), you would need 6000 on the long side.
Not sure what you're getting at. It has nothing to do with the printing equipment. "Native print size" (where I come from anyway; I've been involved in geclee/inkjet printing for over 30 years) means the size a given file will print at 300dpi, the standard print resolution held over from high quality offset printing.
Yes, you can set the print resolution in the dialog box to 240dpi or 180dpi or whatever in order to increase print size or for other reasons, but has nothing to do with how the RAW file is processed in the first place and converted to TIFF or JPEG. I convert all of my RAW files at 300dpi and adjust from there depending on the kind of output I'm looking for. The photos in the books I create are all 300dpi.
Anyway, since a 14-15mp camera will print "natively" at about 10 X 15, I imagine the 7D will print natively (300dpi) at about 12" X 18" or thereabouts, correct?
Lance Couture wrote:
I also believe that anyone who thinks that the files from a 1D3 or 5D are miles ahead of those from a 7D, at any ISO, needs to re-evaluate their methods of evaluation as well as their post processing.
Indeed. The images from my 5D2 are superior to the images from my 7D in terms of resolution and noise. No surprise. But the degree of improvement is slight.
Lauding praise on the noise characteristics of the 5D2 while calling the 7D unusable is hyperbole. Nothing more.
garyvot wrote:
Actually, this isn't quite true. HTP is not a JPEG processing parameter (like ALO).
You can duplicate the results of HTP only if you shoot RAW AND you underexpose by a stop and manipulate the image in post to bring up the brightness while retaining highlight detail. (This is what HTP does for you.)
But just shooting RAW without using a special exposure technique will not give the same amount of headroom in your images as you can obtain shooting RAW and using HTP.
You may have *enough* headroom shooting RAW most of the time without HTP, but at least it's there if you need it....Show more →
Ok.
Well I shoot to only get minute about of blinkies in RAW, which can mean underexposing anyway and then use fill or tone curve and brightness sliders to bring back luminance - leaving exposure slider alone.
veroman wrote:
Not sure what you're getting at. It has nothing to do with the printing equipment. "Native print size" (where I come from anyway; I've been involved in geclee/inkjet printing for over 30 years) means the size a given file will print at 300dpi, the standard print resolution held over from high quality offset printing.
Yes, you can set the print resolution in the dialog box to 240dpi or 180dpi or whatever in order to increase print size or for other reasons, but has nothing to do with how the RAW file is processed in the first place and converted to TIFF or JPEG. I convert all of my RAW files at 300dpi and adjust from there depending on the kind of output I'm looking for. The photos in the books I create are all 300dpi.
Anyway, since a 14-15mp camera will print "natively" at about 10 X 15, I imagine the 7D will print natively (300dpi) at about 12" X 18" or thereabouts, correct?
Converting the raw at any dpi is really not pertinent, the sensor will provide a set maximum X x Y pixel resolution, divide that by whatever dpi the printer will print at per side, and that is the native maximum print size. This means neither the 1DsII or the 7D can produce a printed 300dpi 16x20, the original image has to be resized accordingly by creating pixels where there are none, whether using bicubic resizing, or the printer does this for you with whatever software is contained in the printer driver.
If you want to see a native 100% pixel view of a 7D 3200, just ask, and we can then compare that to the 1DsII equivalent. I would expect the 1DsII will be a bit better or very close in regards to noise levels (under 1 stop). What I supplied above was a 7D image that was upsized to provide enough pixel data to a printer to print a 16x20 at 280dpi.
The resolution of the 7D sensor is 3456X5184, so if you print at 300dpi, the largest image is just about 17" on the long side. If you print at 260dpi, the largest print size is now 20" on the long side. If you want to print a 20" image on the long side at 300dpi (which is overkill), then you have to resize the image to be 6000 on the long side. This is the only time DPI comes into play at all, regardless of raw (you can set your raw to 300, 400, 500, it won't matter in the resulting image), or photoshop settings that default to 72, etc.
I resized the two images above at varying degrees, one was cropped first to try to get to a cropped, then resized 16x20, the other was resized up just a little (Tokina lens) and then recanvased to be a 16x20.
David Baldwin wrote:
"I feel comfortable with the 7D up to 1600 and the 5DMKII up to 3200"
+1
+ 2 - I have them both and agree. Its not the grain, its the dynamic range that you lose that you notice far more. 3200 on 7d and 6400 or 5dii are still better than not getting any picture but the composition would have to be powerful to offset the extra noise and loss of dynamic range.
Scott Stoness wrote:
+ 2 - I have them both and agree. Its not the grain, its the dynamic range that you lose that you notice far more. 3200 on 7d and 6400 or 5dii are still better than not getting any picture but the composition would have to be powerful to offset the extra noise and loss of dynamic range.
.... or the final sizes small enough to obliterate the noise and properly adjust for losses in DR and contrast. However, none of this addresses the issues (if you want to call it that) of noise at lower ISOs.
I just downloaded some sample images from the 7D (first time I've done this; don't know what took me so long), and I do find evidence of noise at ISO 200 to ISO 400. From there to about ISO 1600, noise control appears really good, but then it picks up again. I would consider ISO 3200 my limit on the 7D as well. On the 5D II I set my limit at ISO 5000. My 5D II also showed evidence of noise at lower ISOs, though nothing at all objectionable and only at the pixel-peeping level. My 40D, on the other hand, is super-clean to ISO 800 with little evidence of noise below that. It's quite usable at ISO 1600 even for large-ish prints, but it does go a little bonkers at ISO 3200, but still only a bit worse than my 5D Classic. It's better than most other 10MP cameras of that era as well as some current ones ... and it's significantly better at ISO 3200 than a M4:3 camera.
Here's a 1Ds II shot at ISO 3200 under very poorly lit conditions AND high contrast. Shot with the 24-105 @ f/4 handheld. Focus is on the tip of the guitar. If the 7D can shoot like this under these conditions, I'd like to see samples of that.
Steve, that photo of the guitar player is full of BG noise. I can easily get an image of that quality from the 7D at ISO3200 (and, in fact, this thread already has examples of high ISO 7D images that match the IQ of the image you posted).
BluesWest wrote:
Steve, that photo of the guitar player is full of BG noise. I can easily get an image of that quality from the 7D at ISO3200 (and, in fact, this thread already has examples of high ISO 7D images that match the IQ of the image you posted).
John
John: I see background noise in that shot, but I don't see that it is "full" of BG noise. This is always an issue when posting to the web. Some monitors ... maybe most ... are way too bright, exacerbating whatever noise might exist. On my Mac 30" I see acceptable levels of noise. The monitors brightness is turned all the way down 100% of the time.
Regarding the 7D: I believe it's an excellent camera, one that provides high-resolution files and high ISO noise control that would have been unthinkable in an AP-C camera just 5 years ago. I just don't particularly care for its overall output. I favor the way cameras with larger pixels (photosites) render images, and that includes the 1Ds II, among others. The trade-off between super-fine detail and overall photographic integrity is one I'm willing to make.
So I was shooting my daughter's 3rd birthday with the 7D and took this shot of her with her new toy, and immediately, this thread came to mind. I was also shooting at 3200, so it just seemed too coincidental.
Not quite the same impact as the guitar player image, but still fun!
This turned out to be an outstanding thread with lots of knowledge and experience sharing.
I have ordered my EOS 7D and I thank everyone who have contributed.
What this thread clearly discloses is that the 7D is capable of very good performance IF (and that's a big if) the photographer also knows how to properly process the image. Some of the shots look horrible and others are stunning. I don't chalk that up to camera variations - it's a result of properly exposed images when taken AND post-processing skills.
I have a 1Ds3, why cannot I complain about 7D's image quality? I never expect 7D's image quality to be as good as those contemporary or even older cameras with lower pixel density. I just think the road Canon took, sacrificing pixel level image quality for high pixel count, is not right 
