+1 hehe...
1 and 2 are horrid. And 2 might be a 7D.
But 3, 4 and 5 look best and about the same, but 4 and 5 look best, but 5 is the winner IMO.
Another test -- getting to the bottom of resolved detail
- Thread starter Barry_Green
- Start date
+1 hehe...
1 and 2 are horrid. And 2 might be a 7D.
But 3, 4 and 5 look best and about the same, but 4 and 5 look best, but 5 is the winner IMO.
Barry_Green
Moderator
Okay, well, here's the deal -- this was an attempt to find out where the actual resolvable detail level lies, apart from and ignoring and even removing the aliased "false detail". Since some people seem to get their panties all in a bunch whenever something like a res chart (or, apparently now, other things) are used, I tried to use something where there is no way that false detail would help make the image appear sharper; instead, you either can read it or you can't.
By my viewing, we all can read the 5-point line on all of them, and we can't accurately make out the 4-point line on any of them. Yet it's clearly visible on the printed page. The printer very clearly and easily resolved the 4-point type and it's quite legible.
There were two cameras involved:
1) a JVC HD100, which is a native 1280x720p chipset recording to a 1280x720p recording format.
2) a Canon 7D, recording in 1080/24p mode, using a Zeiss ZF 25mm lens.
Why the HD100? Because it's the most pixel-to-pixel camera I could find. No Bayer pattern, no de-mosaic'ing, no pixel shifting, no downrezzing or uprezzing or cross-converting, it should be the most completely alias-free example I could find.
We got several comments about contrast and detail. Check out Ian-T, hitting the nail on the head:
#1 is the JVC HD100, with a liberal dose of "detail" applied. It is, I think, slightly lower res than the 7D versions, but the detail circuit caused the blacks to be darkened and white edges around the text, so it's far contrastier and caused some folks to pick it as the easiest to read.
#2 is also the JVC HD100, but with the detail circuit turned off. You'll notice that the actual resolved information is the same, but the detail circuit makes #1 look a lot "sharper" than #2. It isn't, they're the same, and you can crank up the sharpening in photoshop to make #2 look just like #1.
#3, 4, 5, and 6 are all the 7D with the same lens focused at the same point. And all seem to be very close, if not identical, in resolved detail. I think if any of them is slightly softer it's #4.
All shots were with the "neutral" scene file setting at default settings (except #4)
So what's different?
#3 is the 7D with a 25mm Zeiss ZF at wide open. The ZFs get flat, muddy, and chromatic aberrations start popping up when they're used wide open. As Dan Kanes pointed out, yep, there's chromatic aberration on this shot. But, in terms of resolved detail, even used wide open, it's resolving just as much as the other shots, and a hair more than the 720p HDV camera.
#4 is the same configuration, same wide open, but with a Caprock anti-moire filter installed. The particular version of the Caprock is 1.0, which is the appropriate strength for a 25mm lens. I think it looks very close (in resolved detail) to #3, maybe just a tiny bit less. Which means that using the Caprock 1.0 (ONLY on a 25mm focal length lens, or 35mm) will strip out all the false detail, eliminate the aliasing, but still allow you to see probably 98% of the "true" detail that's in the shot. Interesting. And you sure wouldn't think it by looking through the viewfinder! With the 1.0 installed, the 25mm lens looks like you smeared a layer of vaseline over the lens, but by the time it gets recorded in the footage, it's delivering the same detail but cleaned up. But the Caprock 1.0 washes out all contrast terribly; it's like using a 35mm adapter on a conventional video camera, you really do have to boost the sharpness and contrast to make a usable image. So for this shot only, I pushed contrast and sharpness up three notches. I reset them back to normal for the others.
#5 is the exact opposite situation -- that's the 7D with the sharpest, clearest, highest detail you're gonna find. That's the 7D with the 25mm ZF stopped down to a 5.6/8 split, where the lens is breathtakingly sharp, ultra-hyper-crisp. The difference in the viewfinder is startling -- the comparison between #4 and #5 in the viewfinder is like looking at VHS vs. IMAX. But in the recorded footage -- the only real difference is a massive amount of aliasing and moire. In terms of resolved detail, they look pretty much the same. Maybe a tiny bit clearer on the "naked" lens, but the difference in aliasing and moire is utterly incredible.
#6 is probably my favorite shot, because it's the no-budget way to solve the problem. #6 is the same 7D with the same 25mm ZF lens with no filters, but it's stopped down to f/22. I actually had to raise the ISO up to 800 to be able to get the lens all the way down to f/22. You'd think the picture would go straight to hell due to diffraction, but actually diffraction turns out to be our friend here! The diffraction strips out probably 80% of the aliasing and moire, while leaving ALL the resolved detail intact! Now, I wouldn't recommend doing this indoors, but outdoors it's heavenly. You can shoot at 1/50th, at 100 ISO, and get razor-sharp images with reduced aliasing, and it costs nothing. You don't have to buy a Caprock, and furthermore you don't have to buy a Fader ND! I've been using the Fader to get the f-stop more open up to 5.6, and the shots SUCK. But get rid of the Fader and stop down to F/22 and the shots are sooooo much better... anyone want to buy an 82mm Fader ND?
(and to Dan, who pointed out chromatic aberration in #6 -- I don't think it is. I think it's chroma noise introduced by going up to 800 ISO. I should probably re-shoot outdoors where I could get the full sunlight and get f/22 without having to gain up to 800. I also think that the shot looks a tiny bit softer than the others, again caused by going to 800 ISO; the other shots were all at 160).
What conclusions can we draw?
1) The 7D's 1080p mode's true video resolution is in the ballpark of 720p. That's better than I was expecting and is actually a pleasant surprise; my prior estimation was that it was between SD and 720p, but it looks to deliver a good solid 720p -- at least as compared to a 1/3" camera. I don't know if its 720p would be as sharp as a 2/3" camera, but -- hey, it's better than I was expecting and it's good news.
2) The 7D's core image (once you strip out the aliasing) beats the hell out of the HD100. The HD100's image has a noisy grainy texture to it, whereas the 7D's is smooth as silk. Much more pleasing.
3) There are many ways to defeat the aliasing. Just about any way that you employ will help, and it's hard to go wrong. The worst thing you can do is crisply sharply focus your images; that will bring the aliasing into play. But a little bit of defocus, a little bit of blur filter, a specific anti-moire filter, or relying heavily on diffraction will all preserve the image detail but reduce the alias contamination.
4) I'm much happier with the 7D now. I've found a way to actually be able to take a wide-angle deep-focus shot without it looking like complete @$$. You just have to stand conventional cinematography logic on its head -- avoid the sweet spot of the lens, embrace diffraction, and defocus a little -- and you can get shots that are on par with what an HD100 can do, but with a cleaner/smoother texture to them. That ain't bad at all for $1699.
By my viewing, we all can read the 5-point line on all of them, and we can't accurately make out the 4-point line on any of them. Yet it's clearly visible on the printed page. The printer very clearly and easily resolved the 4-point type and it's quite legible.
There were two cameras involved:
1) a JVC HD100, which is a native 1280x720p chipset recording to a 1280x720p recording format.
2) a Canon 7D, recording in 1080/24p mode, using a Zeiss ZF 25mm lens.
Why the HD100? Because it's the most pixel-to-pixel camera I could find. No Bayer pattern, no de-mosaic'ing, no pixel shifting, no downrezzing or uprezzing or cross-converting, it should be the most completely alias-free example I could find.
We got several comments about contrast and detail. Check out Ian-T, hitting the nail on the head:
#1 is the JVC HD100, with a liberal dose of "detail" applied. It is, I think, slightly lower res than the 7D versions, but the detail circuit caused the blacks to be darkened and white edges around the text, so it's far contrastier and caused some folks to pick it as the easiest to read.
#2 is also the JVC HD100, but with the detail circuit turned off. You'll notice that the actual resolved information is the same, but the detail circuit makes #1 look a lot "sharper" than #2. It isn't, they're the same, and you can crank up the sharpening in photoshop to make #2 look just like #1.
#3, 4, 5, and 6 are all the 7D with the same lens focused at the same point. And all seem to be very close, if not identical, in resolved detail. I think if any of them is slightly softer it's #4.
All shots were with the "neutral" scene file setting at default settings (except #4)
So what's different?
#3 is the 7D with a 25mm Zeiss ZF at wide open. The ZFs get flat, muddy, and chromatic aberrations start popping up when they're used wide open. As Dan Kanes pointed out, yep, there's chromatic aberration on this shot. But, in terms of resolved detail, even used wide open, it's resolving just as much as the other shots, and a hair more than the 720p HDV camera.
#4 is the same configuration, same wide open, but with a Caprock anti-moire filter installed. The particular version of the Caprock is 1.0, which is the appropriate strength for a 25mm lens. I think it looks very close (in resolved detail) to #3, maybe just a tiny bit less. Which means that using the Caprock 1.0 (ONLY on a 25mm focal length lens, or 35mm) will strip out all the false detail, eliminate the aliasing, but still allow you to see probably 98% of the "true" detail that's in the shot. Interesting. And you sure wouldn't think it by looking through the viewfinder! With the 1.0 installed, the 25mm lens looks like you smeared a layer of vaseline over the lens, but by the time it gets recorded in the footage, it's delivering the same detail but cleaned up. But the Caprock 1.0 washes out all contrast terribly; it's like using a 35mm adapter on a conventional video camera, you really do have to boost the sharpness and contrast to make a usable image. So for this shot only, I pushed contrast and sharpness up three notches. I reset them back to normal for the others.
#5 is the exact opposite situation -- that's the 7D with the sharpest, clearest, highest detail you're gonna find. That's the 7D with the 25mm ZF stopped down to a 5.6/8 split, where the lens is breathtakingly sharp, ultra-hyper-crisp. The difference in the viewfinder is startling -- the comparison between #4 and #5 in the viewfinder is like looking at VHS vs. IMAX. But in the recorded footage -- the only real difference is a massive amount of aliasing and moire. In terms of resolved detail, they look pretty much the same. Maybe a tiny bit clearer on the "naked" lens, but the difference in aliasing and moire is utterly incredible.
#6 is probably my favorite shot, because it's the no-budget way to solve the problem. #6 is the same 7D with the same 25mm ZF lens with no filters, but it's stopped down to f/22. I actually had to raise the ISO up to 800 to be able to get the lens all the way down to f/22. You'd think the picture would go straight to hell due to diffraction, but actually diffraction turns out to be our friend here! The diffraction strips out probably 80% of the aliasing and moire, while leaving ALL the resolved detail intact! Now, I wouldn't recommend doing this indoors, but outdoors it's heavenly. You can shoot at 1/50th, at 100 ISO, and get razor-sharp images with reduced aliasing, and it costs nothing. You don't have to buy a Caprock, and furthermore you don't have to buy a Fader ND! I've been using the Fader to get the f-stop more open up to 5.6, and the shots SUCK. But get rid of the Fader and stop down to F/22 and the shots are sooooo much better... anyone want to buy an 82mm Fader ND?
(and to Dan, who pointed out chromatic aberration in #6 -- I don't think it is. I think it's chroma noise introduced by going up to 800 ISO. I should probably re-shoot outdoors where I could get the full sunlight and get f/22 without having to gain up to 800. I also think that the shot looks a tiny bit softer than the others, again caused by going to 800 ISO; the other shots were all at 160).
What conclusions can we draw?
1) The 7D's 1080p mode's true video resolution is in the ballpark of 720p. That's better than I was expecting and is actually a pleasant surprise; my prior estimation was that it was between SD and 720p, but it looks to deliver a good solid 720p -- at least as compared to a 1/3" camera. I don't know if its 720p would be as sharp as a 2/3" camera, but -- hey, it's better than I was expecting and it's good news.
2) The 7D's core image (once you strip out the aliasing) beats the hell out of the HD100. The HD100's image has a noisy grainy texture to it, whereas the 7D's is smooth as silk. Much more pleasing.
3) There are many ways to defeat the aliasing. Just about any way that you employ will help, and it's hard to go wrong. The worst thing you can do is crisply sharply focus your images; that will bring the aliasing into play. But a little bit of defocus, a little bit of blur filter, a specific anti-moire filter, or relying heavily on diffraction will all preserve the image detail but reduce the alias contamination.
4) I'm much happier with the 7D now. I've found a way to actually be able to take a wide-angle deep-focus shot without it looking like complete @$$. You just have to stand conventional cinematography logic on its head -- avoid the sweet spot of the lens, embrace diffraction, and defocus a little -- and you can get shots that are on par with what an HD100 can do, but with a cleaner/smoother texture to them. That ain't bad at all for $1699.
Last edited:
Dafilman21
Carbonite Member
Wow once again Barry you dazzle, this is prob one of the better tests out there at measuring resolved detail.
TrueIndigo
Well-known member
Excellent forum support as usual, Barry. It's a shame the moire fix involves stopping right down when one of the compelling reasons to get a DSLR for video is the promise of shallow DoF.
Barry_Green
Moderator
There are many fixes. You can stop down (which is a great fix for outdoor wide daytime shots), or you can use the Caprock (but you need to use the right one for each particular focal length), or you can actually de-focus your shot (but that's tricky to find the right balance) or you could put a softening filter in front of the lens... not all the solutions have to involve losing sensitivity!
Is there a way to see this translated in some sample footage?
If I understood, a shot taken at f3.5 and perfect focus should look worse than the same shot at f22 and slight out of focus.
I am not up to the task but is there anybody able to show that improvement in a side by side video?
My simple mind does not allow me to fully comprehend how a softer image could look better, I really have to see it to understand.
Thanks to whoever volunteers.
If I understood, a shot taken at f3.5 and perfect focus should look worse than the same shot at f22 and slight out of focus.
I am not up to the task but is there anybody able to show that improvement in a side by side video?
My simple mind does not allow me to fully comprehend how a softer image could look better, I really have to see it to understand.
Thanks to whoever volunteers.
Alpha Beta Omega
Veteran
It's fun to read about another 7D test. I guess it's puzzling when you shoot a chart and it looks awful and when you're on set the 7D looks so incredible. Shooting real life seldom has you spot on and take in all life's defects the 7D presents a real cinematic and detailed picture.
JFreshInEffect!
Active member
7d's got shallow DOF on Lockdown. No need to sweat the moire and aliasing in those situations. It's those outdoor wides at f5.6-8 that kill me. This was the most useful 7d test to me so far. It expands the 7d's usefulness.
Luis Caffesse
Admin
Interesting test, Barry - and I love the interactivity with this one.
Always makes me a bit nervous though... I halfway expected you to say:
"oh, you picked #5? Well guess what - that's a fisher price pixel cam!"
But you forgot one very important detail.
What is that text on the 4pt line?
I want to know how close I got.
Always makes me a bit nervous though... I halfway expected you to say:
"oh, you picked #5? Well guess what - that's a fisher price pixel cam!"
But you forgot one very important detail.
What is that text on the 4pt line?
I want to know how close I got.
Barry_Green
Moderator
The 4-point line is:
Barry_Green
Moderator
Sure -- but sometimes it also looks like this:
(download a camera-original file by clicking here.)
I also took a still during the video; if you want to see a still of what this shot was *supposed* to look like, and compare it against the video to see what it *actually* turned out like, you can download the camera-original still photo by clicking here.
I know people go on and on about how they've NEVER had a shot ruined by aliasing. Maybe I'm just gifted? Or maybe I'm the only guy who's tried to shoot a deep-focus wide-angle shot? I got maybe thirty shots in the space of a half hour, and 80% of them are totally ruined like this. This is, of course, the worst of the bunch -- but that's why I have spent so much time trying to figure out just WHAT IN THE HELL IS GOING ON with these confounded DSLRs. They're soooo promising... but then they can turn to total horsecrap too.
Ryan Farnes
Well-known member
Don't care for chart tests,
They annoy my sense of cool,
Haiku for 7D.
They annoy my sense of cool,
Haiku for 7D.
I don't think people should throw away their fader NDs, they are necessary for shallow DOF shots outside too. For deep focus I shot a video recently at f11 with a Tamron 28-75mm and it was pretty good, almost no aliasing. there were other problems that were operator error but nothing the camera was causing.
Jason Ramsey
Admin
4 POINT - Politicians ancypod:a fdporuty utosure qweytewtewyRUGFDAE eqr (can't make out at all) A_mv.n.bvnm J _jlu poiurteu u_m coonkynyd_rn zNNHgkpk
that's my guess
that's my guess
Green Hornet
Veteran
I don't know where Barry Green gets the time to do all of these tests.
I am glad he does them, I just don't know where e gets the time.
I am glad he does them, I just don't know where e gets the time.
I'm thankful he does it so we don't have to.
Actually, to those of you (or just Barry) who have used a Caprock filter How can you tell what filter strength to use with what focal length? out side of just getting all 8 and testing them all, I mean?
Too bad they're all 67mm diameter. Can't really use them on a lens like the Canon 16-35mm L or 24-70mm L, or the 70-200mm.... or basically all the desirable high end Canon zooms.
Too bad they're all 67mm diameter. Can't really use them on a lens like the Canon 16-35mm L or 24-70mm L, or the 70-200mm.... or basically all the desirable high end Canon zooms.
Jason Adams
Section Moderator
Barry thanks for the detailed reports. There are as lot of camera kings out there that just shoot pretty pictures. I love that Barry green is a codec breaker. Always giving us camera facts instead of camera sex appeal. Every time Barry purposefully pushes a shot, codec, or a camera past its breaking point it keeps me from inadvertently doing it with a paying client. This helps a lot.
Keep up the good work Mr. Green.
Keep up the good work Mr. Green.