GH4 Gh4 in camera 4k down scaling

walkaboutcamera

Active member
Does anyone know diffinatively, as in not guessing, whether the 4k hdmi down scaled output is whole sensor or bined and line skipped, or worse pixel skipped and line skipped? If no one knows definatively, does anyone have any advice on how to test for it?
 
I think it is the real down sampling without tricks because camera says that "autofocus performance is reduced when 4k to 2k HDMI". It consumes so much processing power that normal use suffers. It also looks very sharp, sharper than normal fullhd. It is also an evidence that fullhd mode is not real good whole sensor readout. Camera must use different tricks to sample to fullhd video and have power left for other tasks too.
 
Unfortunately I think that Panasonic used a much simpler pixel binning scheme for in camera 1080 than what was used on the GH2. It's not bad on the GH3/4 but GH2's 1080 was spectacular detail wise (had to try to get moire.) I think that they did this to reduce rolling shutter. Use half the pixels and you can cut the readout time in half, instead of taking all the pixels (as I believe the GH2 did) and down sample them.

So 1080 on the GH4 is better in color (I believe) and faster readout than the GH2, but less detail and more moire.

Back to the OP's question, I have no idea, but I would test by:

Film a locked off shot that has something in the frame that will cause moire in 1080/100IPB, 1080/200All-I, and a UHD in camera (UHD probably won't be able to get moire, mainly the 1080 shots)
Assuming you have something to record the GH4's HDMI output, record the HDMI out in 1080 and UHD modes.

Things to look for:
When recording the HDMI in 1080 mode:
Is the output from the HDMI the full sensor size (nothing cropped off around the edges, unlike UHD that's internally recorded)? I would certainly assume so.
Does it have less moire than the 1080 internal shots or the same? Less moire would mean somehow a better scaling coming out of the HDMI. Unlikely, IMO.

When recording the HDMI in UHD mode:
Is the output from the HDMI the full sensor size (nothing cropped off around the edges, unlike UHD internally recorded)? This will be the first sign of what processing it's using.
Does it have less moire than the 1080 internal shots? Less or more than the UHD clip?

This should give you a good idea of just how it's working. I honestly don't know and don't have anything to record the HDMI output, otherwise I'd do this test, because I'd like to know too.
 
@vesku and @blazer003 thanks guys. i do have a ninja blade so that makes things easier. I had thought of the moire root but was looking for something easier. I had not thought about comparing sharpness of the two ways of getting 1080 out of the HDMI port. Sorry there are 3 ways, etc mode as well. I have the 12-35 so I should be able to resolve the difference in sharpness. I will give it a go.
 
Look forward to the results! I'm actually looking at going back to 1080p so this would be interesting for me...
 
In my opinion, 1080p looks worse than GH2/3 even with the All-I 200 Mbps mode. The best choice is to shoot in 4K UHD (to mantain 16:9) and downscale to 1080p.
 
"shoot in 4K UHD (to mantain 16:9) and downscale to 1080p."
Is this still how folks are approaching this? Anyone done comparisons between in cam and Software(s) downscale?
 
"shoot in 4K UHD (to mantain 16:9) and downscale to 1080p."
Is this still how folks are approaching this? Anyone done comparisons between in cam and Software(s) downscale?
4K downscaled with any reasonable NLE looks better than 1080p. There's no question. Less moire, more detail. Only downside is a slightly tighter field of view and more horsepower needed to edit.
 
same thing,downscaling is best done with premier pro, no rendering needed. just set settings and downscale to size...
 
Did anyone get to the root of this?

Now that Atomos have fixed the Ninja2 so it auto triggers, I'm actually seriously considering shooting with the camera set to 4k, but have the hdmi do the downsample to the Ninja...

If not, looks like I'll be spending some time testing! :)
 
Seems like a waste of time, not to mention you risk compressing your image twice.

No. If HD is the end format than this makes all the sense. It actually the reason why we bought the camera. Converting 4k to HD makes a 10 bit 444 image. This is the reason why we dont use an external recorder if HD is the finall format.
 
That was another potential reason, from what I saw, Sanjin. Do you shoot in the UHD mode for more 1 to 1 downscaling?
 
For 1:1 scaling? You mean for simple 1:4 scaling? If so then yes, I do. Usually I remove noise in UHD and then downscale to HD. I also crop and stabilise in UHD before downscaling.
So for bigger projects it's proxy for edit, then 4k for reframing and stabilising and then 444 HD for online. But now that DNxHR finally arrived it's 4k DNxHR Standard quality for edit and all else. Depending on the project type I grade in HD or just keep it at 4k.

Definitely record in UHD with this camera if you want the best IQ. In camera scaling is not the best nor is the codec. It's better to have 4k/UHD luma and 2k/HD chroma samples in a lesser codec then half that in a slightly better (more bandwith) codec that also produces aliasing.
 
Was referring to aspect ratio. I have a client that thinks cinema on GH4 is the way to go (even) when outputting HD..but I see potential drawbacks. Thanks for the walk through on your workflow.
 
Back
Top