Flaat Shootout 2012: 550D vs NEX-5N vs GH2 vs D800

Samuel H

Samuel H

Major Contributor
Here it is:
https://vimeo.com/46037571

This is not a serious shootout like the Zacuto ones, this is just three people meeting up to torture the crap out of their cameras.

My conclusions about each camera are in the comments below the video. Summary of the summary: I'm selling my Canon, and keeping the 5N. The GH2 is lacking DR. And the D800 is great but not perfect.

Also: my computer has some kind of heat-related fever and is creating data corruption at rendering, I'll try to post a fixed version in a few weeks, but I'm not going to have a lot of time for this, so, just in case, there you go :)

And finally, again to Ignacio Sánchez and David Calf for helping out and providing the GH2 and the D800 for the test.
Ignacio Sánchez: vimeo.com/user3247032/videos
David Calf: mirallaycalf.com/index.html
 
Did you try the Smooth or Nostalgic film modes on the GH2? Nostalgic has a huge DR improvement over Standard and the best highlight rolloff, though a slight yellow cast (removable with WB adjustment). Smooth is the next flattest setting with proper color.
 
Well, no, I didn't. It was the first time I landed my hands on a GH2, so I asked Ignacio to set up his camera with his usual settings, which, from the feedback we're getting, don't seem to be geared towards DR.
 
Yeah, Nostalgic and Smooth are great for high DR scenes in good lighting. I find I prefer shooting Vibrant in low light, though. Both Nostalgic and Smooth raise the middle of the curve while extending the ends for DR, which also brings up noise in darker areas (like any other flat profile). Even when intentionally overexposing (depending on the scene) with expectations of lowering the exposure in post, it's hard to get rid the noise of compared to using a more normal film setting and exposing properly. I'll be trying Cinema next time I shoot drifting at night, since it's pretty much the opposite of Nostalgic and Smooth, but that won't be for another 3 weeks. I do have a nighttime car meet in 2 weeks that I'll be messing around at, but the lighting won't be controlled. Should be interesting either way...

EDIT: Dunno if it's worth it, but I can do a DR comparison between my GH2 using Nostalgic and my D7000 using the Alvaro Yus flat profile and the same lens this weekend. I'll even remember to use the right adjustment settings on the D7000 this time. :3 The D7000's results should be close to the 550D's.
 
Last edited:
At lunch, I tested what the guy did above on one of my few ISO 160 videos from this weekend, and it does work. Even when set to Vibrant, a normal looking curve. Anything that looks black will still retain detail at higher bitrates instead of being compressed into black/purple blocks. The noise isn't very intrusive at all, mainly chroma (as long as you're reasonable with lifting the shadows). The hardest part is balancing the overall curve so the image doesn't look awkward. I'll post some screenshots soon.
 
@blue bomber, funny you say that, I been using the vivid setting lately as it has that pani mojo look like the hvx, I honestly dont think shooting ultra flat with this codec is always the best choice. I found I have better success and with less noise baking in the color, for grading. As far as latitude goes, there's not much difference between vivid and smooth. And nostalgic just looks ugly to my eyes. I dont like the look at all, even when tweaking the white balance. You do have to be careful in vivid, because it clips quickly, you have to really control your lighting.
 
Last edited:
OK, so after the initial complaints by GH2 owners that the workflow was not optimal, Ignacio sent me a few framegrabs from the transcoded footage (using 5DtoRGB to get BT709 footage)...


This is the original image:
gh2_1a_original.png


And this is its waveform, just to confirm that the highlight is about to clip:
gh2_2a_wf.jpg


I've checked this carefully, and there's indeed a bit more headroom in the highlights here than with the Canon. But that's half a stop at most: there's two thirds of a stop between that peak and the absolute clipping point, whereas the canon has slightly under a third of a stop between its clip and its absolute clipping point. Point for the GH2. But let's go on...

This is the image transcoded with 5DtoRGB, with BT709 as target:
gh2_1b_bt709.png


And this is its waveform:
gh2_2b_wf.jpg


Looking at the image itself, it may seem like it has more DR, but the waveform is already telling me that it doesn't. It's just better at showing the detail that's recorded in the dark areas, but I was already enhancing that in the shootout.

Let's add a simple curve to both images, like the one I used in the Shootout. It only has two points, (0,64) and (191,255). This curve "adds 64" to all the RGB channels of the image; detail and gradation are preserved (except in the highlights: anything above 191 is utterly destroyed), and it's now easier to see what's in those shadows.

This is what the original image turns into:
gh2_3a_original.png


And this is what the transcoded image turns into:
gh2_3b_bt709.png


To my eyes, there's actually less detail in the tree leaves in the transcoded image. This is probably because of the flip-flop 8-bit to 10-bit to 8-bit conversion, and if that's so then that detail would still be available if we were working with the transcoded footage on a 10-bit-deep color space (instead of showing an 8-bit png image). In any case, what seems clear is that there's no more shadow detail in the transcoded image than in the original one.

The fact that with the transcoding it may look like there's more DR comes just a combination of two factors: a not-so-great monitor will clip the darkest part of the image (I know mine does), and even if it doesn't, your eyes are not as good at differenciating very dark tones as they are at differenciating midtones. This is why I used that "add 64" curve in the shootout, to get a better look at the detail that's actually recorded.

Now, back to the highlights, it's true that the GH2 does have a bit more headroom in the hightlights in this test, but it's around half a stop, and I already took that into account in the chart at the end of the shootout, which computed the distance between the absolute clipping points in the highlights and the shadows. The numbers from that chart came from the computation of the full light response function, which looks like this on the GH2 (with the settings we used, Standard -2):
gh2_4a_gh2.jpg

And like this on the T2i with Flaat_11:
gh2_4b_550d_flaat_11.jpg



(Of course, there may be some problems in my arguments above. Please pick up on them, but don't be too harsh, I'm just trying to learn here.)
 
This is an interesting test as at the moment I'm debating what's the best camera to buy to use with an external recorder and S35 lenses. It needs to be under $3,000 and I cant wait for the Blackmagic.

The D800 seems to have the most going for it with better resolution, more dynamic range, plus uncompressed 10bit 4:2:2 recording. But I would need to use it in crop mode with the S35 lenses. Does it have a resolution loss in crop mode like Philip Bloom said the D4 does? According to him, the D4 is very soft in crop mode and not suitable.

But the GH2 also looks interesting, although the smaller sensor doesn't please me with the deeper DOF and by making my lenses longer and effectively depriving me from a wide angle lens in my set. Also, it has no clean uncompressed 4:2:2 right?

I have a 7D myself and it's nice. Better than the 550D but quite similar in many ways. But it would be great if it's HDMI output was full HD 4:2:2. But I guess it's not? I have heard that it is uncompressed 10bit 4:2:2 full HD, just not clean. But I have also heard it's compressed. Honestly, since I'm cropping to 2:35 if all it's left on screen is the recording red dot it may be fine, as long as it's uncompressed 4:2:2. Which is it?

The wild card seems to be the 5N. What kind of HDMI output does it have? The VG20 seems to be the same but with a few extras. The other thing is how much the lack of image control really matters. Image seems quite contrasty.

Thanks.

 
Last edited:
J.F.R. said:
Wa the GH2 hacked for this shoot? If it wasn't I wonder if the Dynamic Range would be different?
Click to expand...

The hack improves the compression. That may have a slightly noticeable impact on colored details, but I doubt it would actually create more DR.
 
M. Gilden said:
The hack improves the compression. That may have a slightly noticeable impact on colored details, but I doubt it would actually create more DR.
Click to expand...

It would not create more DR, but it would better record what is there, compared to the low bit rate Canon that would just get rid of it. How much it is etc, that is the aim of this discussion.
 
* The GH2 was hacked (Driftwood Sedna AQ1-A); a different picture profile could have helped DR, but the GH2 operator suggested the one we used as a good general-purpose option.
* I don't even know about the HDMI-out in the 5N, but I would expect it to be horrible and unusable. This camera is great, but expandability is an issue (no mic-in, even in 3.5mm jack form, no hot shoe mount, no nothing).
 
You must log in or register to reply here.
Back
Top