Gh2 skew

B

Barry_Green

Moderator
I frankensteined together a device that would let me measure repeated, controlled, absolute measures of skew (and therefore wobble and jello, etc), across cameras regardless of chip size or whatever.

I sold my GH1, so I can't compare them side by side. But I can say this -- the GH2 and the AF100 most definitely do not perform the same.

The AF100 performs extremely consistently in all its modes (1080 @ 24p, 30p, 60i, 720 @ 24p, 30p, 60p). The GH2 performs consistently in three of its modes (720/60p, 720/30p, and 1080/60i). But its 24p is much skewier. It shows about 60% more skew in 1080/24p mode than it does in 1080/60i or 720p modes. The AF100 and GH2 are about the same in their other modes, they're within about 5% of each other which to me probably falls within the margin of error for framing the shots, but ... the GH2's 1080/24p is a lot rubberier than the AF100's that's for darn sure.

Wish I still had the GH1 so I could see how it compares to that. I'll try the HMC40 next to see how it compares.

Mind you, I'm not saying the GH2's 24p skew is bad! I'm saying its 60i/60p performance is excellent, but its 24p performance isn't quite as good. I don't know how it compares to a 7D or other such product until I can get my hands on one and run it through ... (wait for it...) The Skewer.
 
Last edited:
I suspect the GH2's skew will be on par with the 7D (not good, not bad, just is what it is). At least that's what I've seen in test videos thus far.
 
Update -- for those of you who were wondering if the skew would be lessened in Extended Tele-Conversion mode, the answer is:

YES!

Much less skew in the ETC mode. The GH2's ETC mode cuts the skew down to 52 pixels, vs. 88 for the normal mode. A lot less. And it's hugely helpful in the 24p mode, it cuts the skew from 152 pixels down to 82, so -- not quite in half, but almost. That does mean that the 24p mode is still skewier than the 60i/30p/60p modes even when ETC mode, by the same margin (about 60% more) but in ETC mode it's really very very good.

Somebody get me a 7D and a 5D and a GH1 and an EX1 and a Red One so I can get baselines on all those, so we have some context to work with here. :thumbsup:
 
loving it. knowing comparatively where the AF100's skew/jello resides is the final "unknown" for me before buying one. or two.
 
Ran the HMC40 through The Skewer, to get a bettter idea. The HMC40 performs very much like the AF100 in 1080/60i and 720/60p; not quite as well in 30p, and not nearly as good in 24p. But it's still a darn sight better than the GH2's 24p.

The HMC40 skews by about 90 pixels in 60i, by about 94 in 30p, and 108 in 1080/24p (and oddly enough, a little more in 720/24p, it's up to 116). But those numbers are still a lot better than the GH2's 24p. I knew the HMC40 got a little rubberier in 24p than it is in the other modes, so I thought it might be comparable to the GH2, but no, the GH2's 24p skew is really quite a bit more.

Now I need to find a 7D/5D to test to see how the GH2 compares to its DSLR competition.
 
Wow, this is cool. I would love to know if there is a significant difference between Gh1 and Gh2 in 108024P. Without referencing the Skewer, do you feel like there is a significant difference between the two Gh's? I'm not a pixel peeper. I could use the Gh1 for documentary work (hand held). Is the Gh2 worse enough to make that impossible?
 
Couldn't possibly answer at this point. But Ben probalby could, as he has both right now. Ben, howsabout putting the two top to bottom, exact same focal length/framing, each on 1080/24p, with a 1/250 or faster shutter speed, and then doing a whip pan past a fence or some other vertical object, so we can compare how the GH2 skews compared to the GH1? Just for a seat-of-the-pants comparison, until I can run a GH1 through The Skewer.
 
Barry_Green said:
Couldn't possibly answer at this point. But Ben probalby could, as he has both right now. Ben, howsabout putting the two top to bottom, exact same focal length/framing, each on 1080/24p, with a 1/250 or faster shutter speed, and then doing a whip pan past a fence or some other vertical object, so we can compare how the GH2 skews compared to the GH1? Just for a seat-of-the-pants comparison, until I can run a GH1 through The Skewer.
Click to expand...

Yeah, will do tomorrow sometime. Only if you post a pic of The Skewer.

If I pick a complicated enough scene this could be a good codec whip pan comparison as well.
 
Barry, here is some ideas why ETC is less screwed. :-)
I think that center 1:1 crop mode is really not center crop.
It is horizontal band crop from the sensor (and later center of this band is encoded).
So, as band is smaller compared to whole sensor, readout speed is faster.

As for AF100. It either uses different readout mode (like GH1 MJPEG that uses native 60p sensor mode) or differencies are even larger.
 
Maybe I´m being stupid but, is the skew also present in still pictures? :) For example when shooting a highway from a tripod, do cars look faster than they are? :D
I´ve never had a CMOS camera so I don´t really have a clue
Thx
 
So I'm thinking that if ever I had to shoot action requiring minimized skew, I would obviously use the ETC. But what lens might I want to use to make the ETC mode shoot wide or normal? Presumably this is some kind of 16mm camera lens with adapter? And could you use a wider M4/3 lens to create the same effect or would that cause too much distortion, at the same time noting that you would be only using the center portion of that lens?

RBG
 
Vitaliy Kiselev said:
Barry, here is some ideas why ETC is less screwed. :-)
I think that center 1:1 crop mode is really not center crop.
It is horizontal band crop from the sensor (and later center of this band is encoded).
So, as band is smaller compared to whole sensor, readout speed is faster.
Click to expand...
I think that's exactly what it's doing. It's scanning only a small 1/2" or 2/3"-sized section, and not having to do any skipping or binning or whatever, and as a result it scans in about 60% of the time. Which helps it avoid aliasing and moire too. It's a great feature, with the tradeoff being that you lose your shallow DOF.
 
Daniel Krnac said:
Maybe I´m being stupid but, is the skew also present in still pictures? :) For example when shooting a highway from a tripod, do cars look faster than they are? :D
I´ve never had a CMOS camera so I don´t really have a clue
Thx
Click to expand...
No, because the Gh1/GH2 use a physical curtain shutter when taking stills. There's no rolling shutter used during stills shooting.
 
RBG said:
So I'm thinking that if ever I had to shoot action requiring minimized skew, I would obviously use the ETC. But what lens might I want to use to make the ETC mode shoot wide or normal? Presumably this is some kind of 16mm camera lens with adapter? And could you use a wider M4/3 lens to create the same effect or would that cause too much distortion, at the same time noting that you would be only using the center portion of that lens?
Click to expand...
Well, that depends on how much of the sensor the ETC mode is reading. It varies depending on whether you've set the camera in 720p mode or 1080p mode. The ETC in 1080p mode is more wide-angle than it is in 720p mode. And you'd probably want a 2/3" or 16mm lens to cover the ETC sensor area; I'll try that later to see what 16mm lenses I have and how they cover in ETC mode.
 
Just tested using a 16mm (not Super16) zoom lens, Zeiss 10-100 T3. Lens fully covered the frame in both 720p and 1080p ETC mode (of course there was massive unacceptable vignetting without ETC!) But with ETC on, it covered even at the 10mm size. So yes, I would think you would be safe with any 16mm c-mount or PL-mount lens.
 
Did a test, results soon. One thing I did learn is that the new Class 4 SanDisk Ultra (I don't think these are Ultra "I" I think they just rebranded the line) cards I picked up that claim to be the same 15MB/s as the Ultra II cards I have, bought from Costco (same place for both) and what they gave me when I brought them the Ultra II labeled cardboard thing from the store, are not fast enough for GH13 for high shutter whip pans, but the old card works just fine without Card Write Speed errors. Both cards seem to work perfectly fine on GH2 (lower bitrate.)
 
In video mode GH2 has 4076x2800 pxl in a 18.8x10.6mm sensor area (atention: only 3.5mm horizontal less than a 7D). The ETC area, with 1920x1080 native pxl (exactly x2.59 crop), must be 7.26x4.10mm area... which is really similar diagonal as a 1/2" video sensor (6.4x4.8mm). All the lenses from this format (1/2", 2/3", 16mm...) should cover this area.
Hope this helps.
 
You must log in or register to reply here.
Back
Top