7D. What is the problem with 720 60P????

G

Gene Sung

Veteran
I've read a numer of comments regarding the antialias issues when shooting the in the 720 60p mode. However, many of the clips I've seen posted on Vimeo shot in 720 60p look really good (like the ones linked below). no banding, artifacting, etc...

So my question is, how are they shooting 60p to avoid the anti-aliasing (the correct ISO, Fstop, etc..)? Are they just doing something different to get it to look good? I did notice that a majority of the shots are locked off. Does the anti-aliasing become more apparent when shot hand held?

Could someone pass me a link where the issues of anti-aliasing are very apparent?

I need to shoot a lot of 60p stuff (surfing and b-roll elements for motion graphics) and was thinking about getting the GH1. But the 7D colors just look so nice, so it's giving me second thoughts.


------- 720 60p clips that look good.

BMX
http://www.vimeo.com/6786060

Aiicia in the Park
http://www.vimeo.com/6956706

Philly at Night
http://www.vimeo.com/6846396
 
The 7D's colours are indeed better than on the GH1, but the GH1's 60p mode is free of most of the GH1's own image quality foibles, as well as the 7D's (moire, aliasing).

I believe that is the only difference. Apparently moire is worse in 720p mode on the 7D than in 1080p. I have yet to see with my own eyes what the difference is though but on the 5D versus GH1, the 5D certainly has more moire.

If moire concerns you, maybe you are best off with the GH1. Otherwise I think the 7D shoots perfectly fine 50p or 60p footage - and yes, you get the choice between PAL or NTSC frame rates unlike on the GH1 which is either one or the other.
 
gene_can_sing said:
Could someone pass me a link where the issues of anti-aliasing are very apparent?
Click to expand...
It's not anti-aliasing that's the problem, it's aliasing and color moire. Which are pretty much related in the 7D (and 5D); when the detail gets too fine, it doesn't blur to mush, it blurs to a color mishmash.

So where you're going to see it is on fine detail, especially repeating fine detail. A chain-link fence is probably an ideal example -- at certain focal lengths and distances it'll look perfect, but then if you get just too close or too far away, it'll turn from being a chain-link fence into being a big orange/purple smear across your shot. That's the kind of thing we have to dig into and research to find out how much of a problem it is (or isn't) in real-world shooting scenarios.

The other thing about the 720p is that it just isn't very high-res. It's probably somewhere halfway between standard-def and 720P.

But, you asked for examples, here are two:

In this one, the waves are just at the right size/frequency to cause the color aliasing all over them:
http://www.vimeo.com/6546960


--------------------------

In this second example, I'm using a composite of two still photos, shot by Barlow Elton, and originally posted in this thread:
http://www.dvxuser.com/V6/showthread.php?t=185934

I've taken the original photos and highlighted where the aliasing is causing problems in the image. The top photo shows what it's supposed to look like, the bottom photo shows when the aliasing has kicked in. You can see that the (presumably) vinyl siding of the building is just fine enough and at just enough of an angle to cause a serious aliasing interaction. Also, look at the horizontal crossmember of the power pole -- it's supposed to be a solid black, but it's aliasing and, because of the way the 7D reacts at the limits of its resolving power, it doesn't just become stair-steps, it becomes chunky color-blob stair-steps. In the larger picture you can see what both items are supposed to look like.

7D-Elton-Alias.jpg



In still photos you might even be able to get away with this, but in motion the aliases will move across the surface of the image. And the thing is, the aliasing will only happen when the detail hits beyond a certain point. So the distance between the shots, or perhaps changed focal length, whatever it was that Barlow changed between those two shots, was just enough to get the alias to kick in. So you won't really know for certain that it's happening unless you're monitoring on a 1080p monitor (which, actually, that won't help either because the 7D puts out a smaller embedded image to the monitor, so the resizing will be different and that'll change the aliasing again. Hmmm...)
 
I don't think the only question is "what is aliasing" but also why aren't we seeing more of it in the clips linked to by the original poster? There seems to be a good bit of fine detail in those pieces, but very little if any aliasing.
 
stephenmick said:
I don't think the only question is "what is aliasing" but also why aren't we seeing more of it in the clips linked to by the original poster? There seems to be a good bit of fine detail in those pieces, but very little if any aliasing.
Click to expand...
Because it's not only fine detail, it's repeated patterns of fine detail that really cause the interference.*

Aliasing is a funny and weird thing, because what happens is that the sampling is not high enough to accurately resolve what's being sampled. So you can get false patterns and contamination in your signal. If what you're sampling has no repeated pattern, like a BMX track with flying dirt, you'll probably never notice it. But if there's a repeated pattern, like the siding on the building, or the texture of the waves, then it can become massively noticeable.

The other way it exemplifies itself is of course on diagonal lines. Shoot mini-blinds and tilt up and down and you'll see ugly stairstepping on an aliasing sensor. Look at the Red One's chart and you'll see that it's cleanly rendering detail right down to the detail-extinguish point -- that means there's no aliasing beyond the sensor's ability to resolve detail. So, if there's fine detail beyond the point of the sensor's ability to accurately sample, you won't see it -- and that means it won't contaminate the image.

A lens that's sharper than the sensor (and its attendant OLPF) can exaggerate aliasing. A softer lens will (in normal circumstances) minimize or eliminate aliasing.

The reason I insert "in normal circumstances" up there is because the 5D and 7D aren't "normal" -- they're 18 to 21-megapixel super-sensors that are not read out using traditional methods. They use some manner of pixel-binning to create the image, and AFAIK, that's not really been done before. So now we have to explore what this really means for real-world images.

*and yes, it can affect any element of fine detail, including non-repeated patterns, but ... you're not normally going to notice that as anything other than, maybe, a sparkle around the edges? As the sampling point picks up different legitimate detail points on the object, it might "dance" around the edge as either the object moves, or the camera moves across the object. I'm not in a position to try it right now, but try something like a really deep focus shot of a newspaper, so that you've got large letters close to the lens and the letters get tinier and tinier as they move away from the lens. Then try moving that newspaper very slightly horizontally or away from the lens, and you should notice that where the detail is finest, it acts "funky". Under a controlled environment, you'll see stuff like that. In a wild and wooly BMX event, you'll never see anything weird because the mud and the dust is all flying in random directions and probably none of it is fine enough to trigger the funkiness.

But, again, this bears testing, because it doesn't just ... well, for lack of a better term, "jump sampling points" as normal aliasing does, instead it changes color. So ... experimentation is in order.
 
Last edited:
Which makes me wonder…

For the most part, in this evolution of new technology (super-sensors) layered over old technology (the SLR), we talk about shallow DOF and fast lenses. But it's fairly common experience that the fastest lenses are, in fact, not at their sharpest when fully open. So, if we're shooting at the fastest aperture, we might actually be softening the shot unintentionally, which could, in turn, be working to the shooter's advantage in terms of aliasing.

Example…

If Simon Cudby (the motocross guy) usually shoots as wide-open as possible, and did so on that footage, perhaps the inherent softness of the image at that aperture is doing some amount of "alias-reduction." (Just thinking out loud here.)

All of which would support what my experiences have been with the 7D: shooting with a super-sharp, super-fast lens, working at it's fastest aperture, with very little obvious aliasing or stair-stepping in places where one would logically expect it.

Maybe?
 
Entirely possible. On any other camera I'd say "yep." But the only reason I don't say "yep" here is -- because the sensor is using some manner of pixel-binning and therefore the downconversion is yielding weird results like massive color mush. That's just so different than anything else, that I don't know quite what to attribute it to.

For example -- aliasing happens frequently when the lens is noticeably sharper than the sensor, and thus can be passing detail on to the chips that's higher than the chips can properly sample. And so a softer lens can eliminate that problem by acting as its own "low pass filter", chopping off the highest frequencies and eliminating that fine detail that would cause visible aliasing.

But -- Canon says that they're using a six-pixel bin. I haven't even tried to work out what the Bayer math on that would mean, but I can only presume that they're reading six green pixels, and six red pixels, and six blue pixels, simultaneously. And there are gaps between those pixels. They're not big 3x2 blocks of pixels. If they were, that'd change everything, but because the sensor is a Bayer pattern, does it mean that we're getting a single pixel read that looks like this?

Normal Bayer:
RGRGRGRG
GBGBGBGB
RGRGRGRG
GBGBGBGB


six-pixel bin?
R.R.R
..... (it would skip this row bc only green and blue are here)
R.R.R

Is that what they're doing? Because, if so, it means that for each red and blue pixel they're taking six pixels together to turn it into one, but each of those pixels would have big gaps -- a full horizontal line in between and two missing vertical colums. What if your fine detail happens to hit on one of those gaps? It'd miss it entirely, whereas a big pixel would have registered it. The overall aggregate size of the pixel would be 15x as big as an individual pixel, but it would need the fine detail in order for it to register anywhere on the pixel... so if you used too soft of a lens you might end up giving away all your detail entirely.

It's odd. It's different. I don't truly know what to make of it yet.
 
Last edited:
Well, I think I speak for all of us when I say…I don't care what you find.


Kidding, of course. I'd be fascinated to find out the hows and whys of the image I see from my 7D, because it's nothing short of spectacular to my eye.
 
Barry Green,

Thanks for the useful info. I ran across this video where they specifically shot in 720 60p (lots of DOF) with specific intention of recording moire and aliasing. I would have to say overall, it's not bad in terms of moire.

http://www.vimeo.com/7002302

With that said, I video a lot of surfing (hence the need for 720 60p slo-mo and lots of DOF). I did notice in the ocean clip and the second moire search clip that bodies of water tends to cause moire problems the most. Obviously this is because of ripples in bodies of water (especially with some wind), tends to create a repeating pattern in the water surface.

I have 4 questions.

1) Will I be screw if I bought the 7D to film surfing with it in 720 60p? Is there a work around for this? I probably will be using a cheaper, long zoom (hence less sharp image which might be an advantage); but will probably be shooting around f8 to have a bit more DOF.

2) Also, can you spot the moires and aliasing in the small LCD screen? Or do you only later see it on your computer monitor?

3) Is the moire problem fixable through a future firmware update? Maybe a tweaking of the codec or some thing like that?

4) The GH1 is said to have less Moire problems. Is this because it has a smaller sensor?

Thanks again for the good answer!
 
gene_can_sing said:
With that said, I video a lot of surfing (hence the need for 720 60p slo-mo and lots of DOF).
Click to expand...
What do you mean by "lots of DOF"? Do you mean that you want shallow dof or deep dof?

1) Will I be screw if I bought the 7D to film surfing with it in 720 60p?
Click to expand...
Why do you want to shoot it on a 7D? If your primary usage is 720p, there are better 720p cameras out there. If you need deep doF, you're going to be in a lot of trouble with the 7D at long zoom ranges. If you want shallow DOF, the 7D will have that in spades.

2) Also, can you spot the moires and aliasing in the small LCD screen? Or do you only later see it on your computer monitor?
Click to expand...
By the very nature of the way moire comes about, you won't see what's actually in the footage until you display it on a native-resolution display.

3) Is the moire problem fixable through a future firmware update? Maybe a tweaking of the codec or some thing like that?
Click to expand...
Definitely not. No possible way. It's hardware that's doing it.

4) The GH1 is said to have less Moire problems. Is this because it has a smaller sensor?
Click to expand...
I don't know what the exact reasons are. While the sensor is a little smaller, that shouldn't have anything to do with it. Apparently they employ different hardware than the 7D does, and the downscale is a lot cleaner (but the codec is weaker).

If you were wanting to get something cheap for 720p shooting for slow-mo, I have absolutely no hesitation in saying the GH1 will do a better job than the 7D. You're pretty much talking about the GH1's sweet spot -- bright light, full sun, 720p mode... I would absolutely say that the GH1 is the better choice for your particular stated preference.
 
I completely agree with Barry's take here, Gene.

But if you're shooting surfing, stick to the Standard "film mode" settings if you're using the kit lens. Better to be safe than to encounter the "blue shift" issue.
 
Hey Barry and StephenMick. Your insights have be invaluable. I should have stated "Lots of DOF" as Deep DOF.

In terms of shallow DOF, it appears that the 7D can get a shallower DOF, correct? Is this because of the larger sensor and codec?

Also, what exactly is the "Blue Shift" issue?

Thanks again guys.
 
gene_can_sing said:
In terms of shallow DOF, it appears that the 7D can get a shallower DOF, correct? Is this because of the larger sensor and codec?
.
Click to expand...

The codec has nothing to do with it really, but the sensor plays an interesting role.

There still is a lot of misunderstanding about sensor sizes and how they relate to DOF. In the most scientific sense, sensor size doesn't play any role in DOF as DOF is determined by the lens being used.

In the real world, the reason that sensor size is attached to DOF is because if we are comparing two cameras (lets say a FF and a 1.6 crop) with the lens at the same focal length, aperture, and the same distance to subject, the 1.6 crop sensor is going to have exactly the same DOF but a much narrow angle of view compared to the FF sensor.

So what do you do to compensate? You move the crop camera back until the cameras are the producing close to identical FOV. The problem though is that you now have changed your subject distance and made it further away which when you change you focus you will have more of the frame in focus (because again, DOF consists partly of subject distance).

This basic idea runs constant through the comparison of any two cameras with different sensor sizes.
 
gene_can_sing said:
I should have stated "Lots of DOF" as Deep DOF.
Click to expand...
If you want deep DOF, why not look at something like the HMC40? It's a regular professional video camera, with a great codec, 720/60p for slow-mo, 1080/24p etc., and ... well, you'll never run out of deep DOF on it...

In terms of shallow DOF, it appears that the 7D can get a shallower DOF, correct? Is this because of the larger sensor and codec?
Click to expand...
The 7D is capable of about a half-stop shallower DOF than the GH1, when all other things are equal, and it's because it uses a bigger sensor, which means that you have to use a slightly wider focal length lens on the GH1 to match the same field of view on the Canon.
 
Did some tests this morning in 720 mode. And WOW, intense blocking/aliasing. For me, it is only useable if there are no straight lines and fine detail. guess i will not be using any slow motion any time soon. Poo
 
Any chance you could post a clip or two, and the lens and settings you were using? I'd be curious to see what you're seeing.
 
You must log in or register to reply here.
Back
Top