Day 4 field testing: notes on aliasing, camera support & meters

[Josh Dahlberg]

Aliasing.... with the default settings, aliasing and moire patterns can be nasty (you don't need to hunt for artifacts, they pop out when complex geometric patterns or horizontal lines are in focus). However, after turning sharpness down a few notches, I didn't notice any problems at all with today's footage (shot in a variety of contexts). It would be interesting to learn how much perceived or actual resolution is being lost by decreasing sharpness, but from today's experience I'm going to keep it turned down.

Camera support.... today I shot using a monopod and lenses ranging from 24mm to 135mm. With my XH-A1, I frequently shoot with a monopod (or indeed handheld) to good effect, even at much longer focal lengths. With the 5d it's a different story. Anything over 50mm IMHO requires a tripod or steadycam as rolling shutter bites hard with even the slightest jerk. Having said that, rolling shutter is not an issue with smooth, controlled movements.

One caveat - I haven't tried an IS lens, which may improve the situation somewhat. But with primes, I won't be using a monopod beyond 35mm, or 50mm at a pinch. Also note (someone else mentioned this earlier and I can confirm), Smoothcam (the FCP filter) will not work with 5d footage as even minute rolling shutter causes havoc in the render.

So we have a very portable unit that demands very controlled movement.

Light meters.... I've never used a light meter, having always relied on my camcorder's internal exposure info. However, I've just ordered a Sekonic as I think it would be very difficult to expose correctly with the 5D without one. In order to set the camera to a desirable ISO, shutter speed and aperture, one has to perform any number of tricks (well documented elsewhere) to get the camera to behave. In the course of which, the camera ends up giving us exposure data out of context, unaware that there won't in fact be a torch in frame during the shot, or your hand covering the frame, or that you're about to stack ND's on the lens, or whatever. I am missing something, or do we all need light meters?

[Robin Lobel]

Do you have video frame captures with and without sharpness, showing the problem ?

[Josh Dahlberg]

I'll try and get you something tomorrow. Jim Jannard posted a chart over on the Red user forum demonstrating the effect (I can't find it just now but there is a link to it on this forum). Of course, there's no issue with soft textures, natural environs, skin and the like, but when fine/horizontal lines enter focus it's not pretty, and the moire is much worse than anything I've previously encountered in video.

Don't let it put you off though. This camera has a lot more strong points than shortcomings - I'm really enjoying it so far.

[Steve Mullholand]

Great read. Thanks for the info.

[Robin Lobel]

BTW, do we know how the cam perform the video scaledown ?
I mean, from the full sensor resolution down to 1920x1080, does it gently smooth using some bilinear/bicubic/.. filter, or a simple "nearest neighbor" method (which would explain the problems with thin horizontal/vertical lines).

I was a bit disappointed when I read Red was simply cropping into the sensor when shoting 2K.

[Don Miller]

What's fairly clear is that alternating lines per frame are used when reading the sensor frame to frame. That seems to cause the problems with horizontal lines. In the future if they can read every line and downsample noise should be less. Since they apparently are reading every other line the downsampling algorithm is probably non standard. Add to that how they might build the different frame types and it gets very complicated.

One advantage of DOF control is allowing blurring of non critical elements. This caqn certainly help with both artifacts and perceived sharpness.

Beyond exposure control I'm not going to be working hard to get this camera to do things it doesn't do well.The testing being done here is very helpful in understanding the limitations of this model.

Considering even the sub $1000US single chip cameras now produce excellent images in good light, I'm thinking about the HF11 to shoot with the 5DII and go completely tapeless. I'm still unclear if AVCHP plays nicely with FCS, however.

[Evan Donn]

Originally Posted by Josh Dahlberg

In order to set the camera to a desirable ISO, shutter speed and aperture, one has to perform any number of tricks (well documented elsewhere) to get the camera to behave. In the course of which, the camera ends up giving us exposure data out of context, unaware that there won't in fact be a torch in frame during the shot, or your hand covering the frame, or that you're about to stack ND's on the lens, or whatever. I am missing something, or do we all need light meters?

Not sure how you would make use of a light meter with this camera in video mode - at best with nikon lenses you can get manual aperture control but you are still leaving shutter and iso up to the camera. You might be able to use the light meter to guess at what shutter speed the camera is going to use given a particular aperture, but even that would require a known ISO which I don't think is possible.

Even if we had full manual control I don't know that there's much use of a light meter in terms of setting camera exposure for video - I mean you can see exactly what you're getting on the monitor. The only thing I'd use one for is checking ratios for lighting - although even then final decisions are probably going to be made based on what the camera sees.

[Josh Dahlberg]

Originally Posted by Evan Donn

...at best with nikon lenses you can get manual aperture control but you are still leaving shutter and iso up to the camera.

Once you have aperture control, it's not too difficult to trick the camera into using the ISO and shutter speeds you want (it's just a bit fiddly, and then, there's always the issue of whether one can trust the exposure information provided).

Originally Posted by Evan Donn

"Even if we had full manual control I don't know that there's much use of a light meter in terms of setting camera exposure for video - I mean you can see exactly what you're getting on the monitor.

That's actually my reasoning behind using a light meter: the camera exposes as it sees fit > the exposure may be balanced, but it's not at a desirable shutter/iso/aperture > so we add/remove light by varying degrees to alter and lock down these parameters. Then, when one frames to shoot (the actual scene having less/more light than was exposed for), the LCD shows us the under/overexposure. I figure by using a light meter we would have a better target combination to work towards for a given shot, rather than taking a stab at it. Then again, one could always aim to slightly overexpose, then use the LCD as a reference for slotting in more layers of ND until exposure is correct.

[Josh Dahlberg]

Originally Posted by Don Miller

One advantage of DOF control is allowing blurring of non critical elements. This can certainly help with both artifacts and perceived sharpness.

This is a very good point. While the aliasing/moire can be nasty, in a typical shot it's fairly easy to keep troublesome elements out of play controlling DOF. When problems do occur, they show up clearly in the viewfinder and you can deal with it on the spot. It's just something to be aware of.

[Don Miller]

The camera is a light meter before we start messing with the video processes. Using manual exposure in photo mode the target exposure values for the scene can be determined. That's all a separate light meter is going to do anyways.
For people who have shoot film, it a normal process to decide how to get to the exposure value needed. It doesn't take too long to do this easily.

[Josh Dahlberg]

Originally Posted by Don Miller

Using manual exposure in photo mode the target exposure values for the scene can be determined.

But of course you are right Don, how silly of me. Still, the meter will come in handy for incidence readings.