Today I introduced the Canon 7D to the greatest res chart in the world, the DSC Labs MegaTrumpet 12-SW. This is a chart that was designed because conventional video camera res charts weren't good enough to deal with new cams like the Red One. And, in testing my HMC40 against my old res chart, yeah, that little cam bested what the chart could do -- so, I got myself the greatest res chart in existence.
When I saw how the Canon performed on this chart, in 720p mode, I couldn't believe my eyes, so I went and got the HMC40 (the only other 1920x1080 camera I have right now; the GH1 is out on loan) and put it in 720 mode and shot it too, and then split-screened the results here.
Check it out:
And, for good measure, I of course also shot 1080/24 on both.
Wow. As "The Contintental" would say, wowie wow wow wow.
The little HMC40 blows the doors off the 7D when it comes to sharpness. In 720p it's embarrassing; the Canon puts up a much better fight in 1080, but it's still subject to some bizarre color aliasing and vertical aliasing as far out as 500 lines! I mean, keep in mind that this chart is perfectly black and white, there's no gray, no color, no anything, but look at the bizarre detail extinction on the horizontal trumpets at about 800 lines, and the color moire happening on the 45-degree angles.
By my readings of these charts, vertically, the Canon doesn't even register on the chart in 720p mode. The chart starts at 500 lines, and it's already screwed up with bizarre color patterns right where the chart starts. The HMC40, on the other hand, delivers pretty clean up to 700 lines. Horzontally, the Canon looks to be delivering about 550 lines before detail extinction, and the HMC40 starts aliasing about 550 lines before complete detail extinction at about 720. And what's with the weird aliased color-moire pattern in the central zone?
In 1080 mode, the HMC40 is clean as a whisper in vertical resolution, easily hitting 900 lines and complete detail extinction at 1000. Horizontally there's some alias pattern but still pretty clean res down to about 900 lines. And look at the 45-degree-angle trumpets -- crystal clean and perfect down to about 1100 lines, where the detail becomes completely extinguised. The Canon, on the other hand, what the... ??? Look at the straight horizontal trumpet, measuring vertical res. There's nothing good about it. You can see the aliasing wants to take over starting early, and the aliases actually cross over the lines at about 700. On the vertical trumpet (measuring horizontal resolution) it's a much cleaner story, but still low-res. It aliases (crossing lines) as early as 700 lines, with complete detail extinguishing at about 780. The HMC40 doesn't hit the detail-extinguish point until about 1200 lines!
It's weird all the way through. The HMC40's 45-degree is clean to 1100 lines, the 7D is heavily aliased at 600 lines and has completely failed by 800. Look at the 7D's 292-degree trumpet (the one above the horizontal trumpet) -- it's actually rendering those just-off-horizontal lines as vertical lines! I have never seen anything like that, ever. Biiiiiizzzaaarrrreee.
In another thread, Glenn Przyborski said "I don't want to get 7D members on my butt, but others at our edit facility who are used to screening high-end HD, thought the 7D 1080p footage had the sharpness and image quality of 720P, blown up to 1080." I think I'd have to agree.
Now, res charts don't mean everything, they're only one way to evaluate an image, but -- yeah.
I'll put the GH1 through the same tests when it arrives. I frankly don't expect it to perform any better than the 7D, I mean -- DSLR to DSLR, same basic technology in each, I expect similar results.
Results 1 to 10 of 381
10-09-2009 04:39 PM
- Join Date
- Sep 2003
10-09-2009 04:46 PM
Thanks for putting this together, Barry.
Can you tell us more about how you had the cameras set up? Lens on the 7D? Aperture and focal length? Mode settings?
I haven't shot any res charts like this, but my experience with the 7D was much the same. Take it out of the box. Point it at the same thing as another camera. Be disappointed at the softness of the image.
Then I played around with some of the in-camera settings, and the results were a definite increase in "perceived" sharpness (again, to my eye, with no res charts).
10-09-2009 04:47 PM
Last edited by ydgmdlu; 10-09-2009 at 04:52 PM.Current cameras: Panasonic LUMIX DMC-GH2 | Sony HDR-SR11 (infrared modded)
10-09-2009 04:50 PM
Good thing we don't shoot res charts on a regular basis!
- Join Date
- May 2009
- Manchester, U.K.
10-09-2009 05:43 PM
- Join Date
- Apr 2009
On one hand, I'm glad I haven't been mistakenwhen I said 7D has some pretty wacked out chroma and luma aliasing.
On the other hand, I wish I had been.
Barry, what software settings were used for this? Have you tried shots with the contrast/sharpness all the way down?
Thank you very much for taking the time to post this up and take the time to really look into it.
10-09-2009 05:50 PM
- Join Date
- Mar 2009
My god...it's full of moire!!
I wonder how much this has to do with the sensors, vs the way the sensor is being read out for video? If you have the time, Barry, maybe you could trigger the 7d via USB and get a shot of the chart and a regular still picture of the chart - which I predict will be perfectly crisp (duh, Captain Obvious). Scaling the video up to match the resolution of the still image might yield clues about what kind of pixel binning or skipping Canon are using.
I still very much like the footage I'm seeing out of this camera under most conditions, particularly night footage, but this is certainly an unwanted annoyance.
10-09-2009 05:53 PM
- Join Date
- Jan 2006
Well I'm not shooting res charts but I can't get an image out of this camera that I'm happy with, in any settings. It is aliasing hell. I thought a lot of the footage posted online looked great, and when you magnify the image looks amazing on the camera, then when you get it on the computer.... well not so much. I think it might be ok for half scale web video, and shallow dof shots, but deep dof on the 10mm seems like a non-starter.