GH3 Photo Samples Specs and Comparison

[galenb]

[quote name='Julian' timestamp='1350741223' post='20040']
This is where the GH3 starts to shine, although I'm curious how a hacked GH2 bit bitrate similar to the GH3 would hold up.

This GH2 footage was around 20Mbit. The GH3 is 72Mbit All-i (thats what i'm shooting in all the time btw).
[/quote]

OMG! I totally forgot that you were comparing the GH3 to an un-hacked GH2! So Yes, since the GH2 image is only "falling apart" due to the low bitrate of the codec, a hacked GH2 will probably look similar at the same bitrate. And it can only get better from there since the GH2 can go way beyond 75Mbps. My GH1 is currently hacked at around 75Mbps.

[quote][color=#282828][font=helvetica, arial, sans-serif]Dialling down the sharpness on the GH3 to -5 has a pretty big effect. More so than -2 on the GH2 it seems.[/font][/color][/quote]

Yes indeed! So it looks like Panasonic decided to give us a lot more control range with -5 to +5. It's looking more like -10 though. ;-) This is good though. I have a feeling this is going to stump a lot of people who get the GH3 and dial everything down and get mushy images. I'm really glad you were smart enough to double check this. I don't know that I would have thought of that. I'd like to see the lamp test again with the sharpness dialed to 0 too. Actually if you could pan slightly and post a video that would give us a better perception of any moire issues. I have a feeling it might come down to dialing the sharpness to some optimum level to avoid moire.

[quote]Vibrant, really? I thought Smooth/Nostalgic were the best modes to get the most out of the GH2 (dynamic range etc).[/quote]

Well, it's obviously pretty subjective. I first saw this latitude trick on youtube done by a guy named Drew.

[media]http://www.youtube.com/watch?v=vPHIEU1X1Bo[/media]

[size=2]I don't recommend watching this video right now though. It's really long and he kind of rambles on a bit as he meanders through the topic.[/size]

He recommended "Vibrant" as the profile that captured the most latitude. I think for these tests you just have to make sure they are set to the same profile. Otherwise, there will always be some question as to wether the test is fair. We are just trying to figure out how the cameras compare to each other. Basically, Moire issues aside, I think a hacked GH2 (@75Mbps) and the GH3 are in a dead heat with latitude. I have a feeling that the "Wide dynamic range" listed in the marketing literature is just that: Marketing literature. I have a feeling that they just decided to mention it because they didn't realize it was as important as it is to people's buying decisions till now.

These are awesome test by the way Julian. I totally appreciate you taking the time to do these tests. really answering a lot of questions here. Thank you so much!

[Julian]

No problem, I'm curious myself as I'm going to make the choice between the GH2 and GH3.. In case of the GH3 I could probably ditch my Canon 5D (mark 1) for photos, the photo quality of the GH2 isn't really good enough.

I don't think wide dynamic range is a marketing term. Don't forget the GH3 is a photo camera more than a video camera. I didn't shoot much pictures yet and there is no proper raw support yet, but I have the idea Panasonic jumped the Sony wagon with this camera. That would mean massive dynamic range, like the OM-D, in raw. They don't claim wdr for video I think... Sadly this new sensor and good photo quality seems to have some downsides for video. I also miss the multi-aspect sensor..

I took the GH3 out into the real world today for some shots. Kept the sharpening at 0. Moire is an issue with bricks sadly... But overall I'm pretty pleased with the image. It is really sharp and full of detail, but then again, the GH2 also is.. The body is much better than the GH2 though. I'm not too happy about the controls yet because a lot changed from the GH2 but that's probably easy to get used to and there are a lot of Fn buttons so it should be fine.

Also did some shots in 1080p50, which is pretty nice. Overall I have a good impression of the colors/toning. It was overcast today so not a lot of contrast, but the image looks good to me. Maybe better than the GH2, but It's hard to compare since the color profiles are so different. I'll put a video of the shots together ASAP.

I'll reshoot the lamp, but I'm afraid moire is here to stay. In moving images it's pretty noticeable.

[Chrad]

How is banding?

[Julian]

What is a good way to test for banding?

[Axel]

[quote name='Julian' timestamp='1350813593' post='20065']
What is a good way to test for banding?
[/quote]

You did it already. The lamp shots show it, both cameras have the problem. Clipping was to be expected, but the gradient on the wall is not even. To reliably get banding, the room should be completely dark, with only one bulb, and you pan up and down the gradient. Then the borders of the bands will move above the surface and can't be ignored. Also of course the remedy against banding: To avoid such situations. The GH3 has a softer fall-off, but no wonder, because the whole image is too soft.

[Julian]

Interesting, EXTC only works at lower bitrates. 50mbps mov 24 and 25p. Not at 72mbit and not in any 1080p50 mode either.

[Julian]

Also, ISO 125 and 160 (extended) are selectable in video mode but the exposure is exactly the same as iso 200. It shows ISO 125 during recording it doesn't seem to be any different than ISO 200.

[see ya]

[quote name='galenb' timestamp='1350771074' post='20053']
Well, it's obviously pretty subjective. I first saw this latitude trick on youtube done by a guy named Drew.

[/quote]

So how does this differ from using 5DToRGB and picking full range for GH2 which it isn't and therefore stretching levels, which inturn alters gamma and brightens the image, retorical question I think as it's exactly the same.

All he's doing is compensating for the initial interpretation by the NLE going YCC to RGB for RGB preview, the usual coloquelisums, crushed blacks etc due to bad handling by the NLE or media player first off, notice he's using a 32bit project so he can pull the levels around and bring detail up without it getting clipped and lost.

I'd be interested to see if he gets same results at 8 or 16bit. I've asked him and await he's reply.

And I'd also suggest that this has nothing to do with increased DR of a GH2 but just the same as any player mishandling the video files and Canon MOV's are no different, they can also be mishandled by an NLE or media player to appear crushed.

Again here's a test file, [url="http://www.yellowspace.webspace.virginmedia.com/fullrangetest.zip"]http://www.yellowspa...lrangetest.zip[/url] to show a simple mishandling of levels going to RGB, stick the unflagged file in a 32bit project like he did, see it shows a black and white bar which is the same as his blown highlights to the water and supposedly underexposed trees, adjust levels and miraculously see detail appear in the shape of a 16 & 235 text. Use a pipet tool and sample the 16 grey (16 RGB) and the 235 grey (235 RGB).

Honestly don't think there's any more to it than that.

If using QT for this on a mac your milage may vary.

[Julian]

Took the GH3 with me today and put it on 1080p50 fps (50Mbit MOV).
ISO 200, 1/100s, 12-35mm X stopped down to f/8-f/11 most of the time.

I shot pretty flat, (left sharpening at -2 instead of -5 though) and it was a dull grey day so I decided to play with the colors in DaVinci Resolve. Here's the result, there are some ungraded/graded examples at the end.

By accident I was working in a 24p project in resolve and I couldn't export as 25p. I didn't see anything strange in particular, if you do, that might be it...

[media]http://www.youtube.com/watch?v=UlogTKjm5Fw&[/media]

Tomorrow I'll probably get the chance to mount the GH3 and GH2 together, both with 14-140mm for a side by side video.

[Julian]

Some real life bricks, walls and fences.. and moire.
1080p25 @ 1/50s

http://www.youtube.com/watch?v=hcbNJUDMEPY&

[galenb]

[quote name='Julian' timestamp='1350854826' post='20094']
Some real life bricks, walls and fences.. and moire.
1080p25 @ 1/50s[/quote]

Hmm. You know I was looking at this video and noticed something strange about the motion. There's this ghosting going on. I thought it might be interlacing at first but the more I look at it, the more stumped I am. I think this making the moire even worse. Do you remember what mode this was shot in?

[img]http://www.eoshd.com/comments/uploads/inline/20711/5084df7876dba_Ghosting.jpg[/img]

[Julian]

Did you notice it in the other video as well? It could be something with the exported video. It looks weird too. I'm using Sony Vegas Pro 11, struggling a bit with the export options as it won't export to H264 for some reason (crashes..).

I'll reupload a single clip straight from the camera. It was shot in Manual, 1080p25 1/50s

/Edit: maybe I f*cked up the framerate settings in the Moire videoproject.. It could be that I put the 25p files in a 24p project. If you look at the Fair video, the original cuts in the end show the same problem (it's 25p on a 24p project).

[galenb]

Okay... That's just weird... If you look at the fare video, it's the un-graded clips that come from the camera that are messed up but the ones from Resolve seem just fine... Usually it's the post processing that can botch up footage. And even stranger is that those clips are 50fps right? if they are being conformed to 24, the results should simply be slow motion. Maybe it is the framerate conversion and it's just that Resolve knows how to properly handle it but Vegas doesn't know what to do. I guess it's possible that Vegas is trying fix the framerate conversion by frame blending instead of just conforming the rate. Maybe Vegas thinks you are time-stretching or something?

You know, you could just export an XML of the edit in Vegas beck to Resolve and do the final output to H.264 from there.

[edit]

Yeah, now that I look at the footage frame by frame I'm thinking you are right. I bet Vegas is treating the raw footage as if you wanted to time-streatch it. I'll bet too that Resolve knows how to properly conform 50 to 24p and that Vegas thinks it's being clever by frame-blending.

[Axel]

@Julian
The ghosts might come from some frame blending during frame rate conversion.

I want to personally thank you for your effort, the information you shared with us is first-hand and precious.

Let me give you a tip in exchange that is hopefully received with good feelings. Your fairground clip is overexposed. You can watch the histogram in the EVF / display and prevent the values on the right side to be cut off by deliberately closing the aperture and / or using an ND-filter, ND-fader. In a situation like this, when the sky is ten times brighter than everything else, this will lead to an overall underexposed image. But with the right film mode (which for the GH3 has yet to be found) and the right dose, the shadows will hide a lot of useable midtones that can be drawn into the middle i.e. in Resolve. Whereas highlights, once having crossed digital Jordan, are gone forever and a day.

[Julian]

[quote name='galenb' timestamp='1350917712' post='20119']
Okay... That's just weird... If you look at the fare video, it's the un-graded clips that come from the camera that are messed up but the ones from Resolve seem just fine...
[/quote]
That is because I mixed 25p footage (originals) with the 24p footage (edited in resolve) into a 24p project in Vegas...

@Axel: Thanks for the kind words! No hard feelings on the tip, although i'm well aware of what you are saying :) I'm always protecting my highlights during recording and looking at the histogram.

Most of the original clips shot at the fairground aren't blown out, there may be one or two.. But I added more contrast in post and more yellows in the highlights, thats why it looks and sometimes is overexposed... I just felt like playing with the clips a bit, I'll show more original footage (properly exposed) for sure!

Actually if you take a screengrab and look at the histogram, the highlights don't clip (much). I agree it looks overexposed though.

Also it's annoying that the Live View on the GH3 shows more detail. I set the exposure and the histogram looks fine, then I press record an BAM, the image starts clipping. I'll have to see if the GH2 shows the same behaviour. If it does no news..

[galenb]

[quote name='Julian' timestamp='1350919321' post='20121']
That is because I mixed 25p footage (originals) with the 24p footage (edited in resolve) into a 24p project in Vegas...
[/quote]

LOL! Yeah, It's quite simple to understand when you write it out like that. I guess I didn't need to write out my whole thought process like that. Sorry. :-)

[Axel]

[quote name='Julian' timestamp='1350919321' post='20121']
@Axel:
Most of the original clips shot at the fairground aren't blown out, there may be one or two.. But I added more contrast in post and more yellows in the highlights, thats why it looks and sometimes is overexposed... I just felt like playing with the clips a bit, I'll show more original footage (properly exposed) for sure!
[/quote]
I apologize. The weather didn't fit.

[mopixels]

[quote name='Julian' timestamp='1350854826' post='20094']
Some real life bricks, walls and fences.. and moire.
1080p25 @ 1/50s

[/quote]

Hi Julian,

Was the moire test was shot at 75Mbps or 50 Mbps?

I'm curious because I noticed on the dancing footage downloads photographyblog posted on vimeo that the moiré and aliasing was quite apparent on the 75Mbps/ALL-I clips, but hardly at all on the 50Mbps clips. The 50Mbps clips looked quite clean in comparison to the higher bit rate ones.

Anyone else notice this? BTW, I downloaded and played out from CS6 through Matrox hardware to a Sony video monitor.

If you have a chance, could you do a moiré comparison between the 75Mbps vs 50Mbps recording modes?

[Julian]

Shot at 72 Mbps. I'll do a comparison between different bitrates and avchd!

[MaxAperture Films]

[quote name='Julian' timestamp='1350919321' post='20121']
That is because I mixed 25p footage (originals) with the 24p footage (edited in resolve) into a 24p project in Vegas...
[/quote]

Hi Julian, in Vegas you can simply right click on the clip in the timeline, select "properties" and click on "disable resampling." This will disable frame blending on all footage with a different native frame rate than the project. The 25P footage will drop a frame here and there and won't be perfectly smooth, but ghosting will no longer be an issue. You must do this for each 25P clip or subclip in the timeline.