GH4 Lots of tests with the new GH4

Great set of tests, thank you. I share the disappointment with everyone regarding what appears to be quite limited DR.

I just wanted to check though that there isn't some misconfiguration between working with and exporting full/legal range. The highlight roll-off is resembling a lot of the superwhite stuff I've seen on YouTube/Vimeo over the past week in regards to this camera. For instance https://vimeo.com/92613144 which is premier based workflow, but I know Resolve can be a bit funky with YUV/RGB too.
 
RatLabProductions said:
And for those requests you will be paying $8K plus. I mean really its a sub $2K DSLR what were you expecting?

Thanks for the tests Bruce & MacGregor.
Click to expand...

1080p with 12 stops DR and a 10bit or RAW codec is only reserved for cameras costing $8k or more? You haven't done much shopping then because that's waaaaay off the mark.

i may may have missed this info somewhere, but what IRE setting was used when doing the DR test?

Thanks
 
Last edited:
mr bill said:
I'd be very surprised if Panasonic were to perjure themselves with 'conservatively 11 stops'
Click to expand...

Well, they're not under oath, so no perjury, but if they make the same claim in Europe, they'll be subject to false advertising laws. That's the key - not what they say in the U.S., but what claims they make in Europe. Does anyone know? Anyhow, it doesn't matter - there's all sorts of technicalities they may get off on, f.ex., claiming the 11 stops applies only to stills DR etc.

Right now, I'm interested in skin tones. I hated - HATED - the GH2 green cast to everything. I much prefer the GH3 color renditions, but obviously it could be improved (it's a bit magenta) - more neutral would be great.
 
dustylense said:
The reddish cast on her arm??? Has nothing to do with the cameras color profiles.
Click to expand...

Well, if the selected profile had nothing to do with the appearance of the redishness, then I'd expect to see it in all of them. As it is, it is less obvious in the Cinelike-D and Natural, and most/more noticeable in the CineLike-V and Vivid.
 
bruceallen said:
Anyway... if you are testing DR with a grey chart, you will get one number. But if you test DR with colors, you might get a different number. Definitely I would like to find my Stouffer wedge and shoot it not just with greyscale but also with different lighting gels on top of it. But that can come later (and doesn't need an F35 side-by-side).
Click to expand...

The same was true for Film film... Even B&W was 'derated' for Tungsten vs. Sun/Daylight.

The most significant difference these days seems to be that for the DSLR type camera, and perhaps even for more expensive cameras, the manufacturers do not publish how they determined their stated numbers, DR, ISO, or under what conditions. When I went around to the various booths at NAB a few weeks ago, none of the 'booth babes' (male or female...), had any real statement to make on how ISO was determined which was my 'quest' question.

From my tests, I've come to the conclusion that my GH-1 has about 6 stops of 'effective DR'. My Nikon D600 about 7. This is dealing with the camera as a 'black box', and using points on the IRE scale that seem 'reasonable' to me, that is anything under 10% IRE is 'mud/black', anything above 90% is 'whiteout', and then how things fall in between, will yield a 'good image' based on how well lighting and elements are placed.

On the other hand if I looked at the waveform on my tests, sure I could see differentiation of my grey card several 'stops' lower and higher than those points. Hence I could 'rave' about an 9 or 10 stop DR on my GH-1...

But in the case of 'dark', below 10% seems to be 'murk' on most displays( no 7.5% setup...), and unless the object truly is 'white' or nearly so, anything above 90% looks 'melted plastic unsaturated'...

I hope you do recover your test charts to see what they would show.
 
Last edited:
dvbrother said:
Bruce or anyone involved with the tests, have you experimented with adjusting the Master Pedestal setting? By default I assume it is set to 0. I've heard that by lifting it to+1 or +2, you can avoid crushing the blacks so easily.
Click to expand...

RE master pedestal... that is phase 2 - when we deviate more from standard settings. Maybe this cam just has terrible defaults.

I am not optimistic though because in theory if our 0-255 post pipeline is correct, all we will be doing is further limiting usable tonal values. Unless the camera produces "superblack" values that the master pedestal move up will reveal!

itimjim said:
Great set of tests, thank you. I share the disappointment with everyone regarding what appears to be quite limited DR.

I just wanted to check though that there isn't some misconfiguration between working with and exporting full/legal range. The highlight roll-off is resembling a lot of the superwhite stuff I've seen on YouTube/Vimeo over the past week in regards to this camera. For instance https://vimeo.com/92613144 which is premier based workflow, but I know Resolve can be a bit funky with YUV/RGB too.
Click to expand...

RE 16-235 etc... pretty sure we didn't mess it up. Sadly. But we could be wrong of course. The post pipeline was quick just to get initial stuff out and we will examine further.

Bruce Allen
www.boacinema.com
 
In terms of the dr and overall shot quality, the problem I'm having is massive noise in any dark area. The following screen cap is from my gh4, cinelike D, 200 ISO, sharpness -5, all other settings default. I don't really understand why the image is so noisy. Is there a way to suppress noise in the darks? I would understand at a high ISO, but at 200, and worse at higher ISO's, it seems far noisier than my GH1. Any ideas?

Screen Shot 2014-04-27 at 9.25.07 PM.jpg
 
doccutter said:
In terms of the dr and overall shot quality, the problem I'm having is massive noise in any dark area. The following screen cap is from my gh4, cinelike D, 200 ISO, sharpness -5, all other settings default. I don't really understand why the image is so noisy. Is there a way to suppress noise in the darks? I would understand at a high ISO, but at 200, and worse at higher ISO's, it seems far noisier than my GH1. Any ideas?

View attachment 85493
Click to expand...

My question is are you sure it's not the monitor? And I'll point out that Cinelike-D on nearly all Panasonic cameras is usually very noisy.
 
Many thanks for the reply, Mike.

I don't think it's the monitor, because the noise moves when the clip is playing and stops when I pause the video.

I didn't know that about Cinelike D, is there a better all around mode?
 
itimjim said:
Great set of tests, thank you. I share the disappointment with everyone regarding what appears to be quite limited DR.
Click to expand...

It really is probably 8 usable stops and around 10 total stops. 11+ stops is most probably raw stills. It would be interesting to see a side by side with other dslrs like 5dm3, etc... Maybe we are being a bit unrealistic here. I really doubt that GH4 has a lower DR than GH3, what I don't doubt is that the DR was improved coming from the GH3. DR is probably better by half a stop or maybe 1 stop. So everybody that was happy with theirs GH3 shouldn't really be scared by the results from these tests...
 
I also think that 8 stops+ a little of real usable dynamic range is believable. Shian of Color Gears rated the GH2 at 7 stops. Ryan Walters on vimeo also concurs with +3 over and -3 under. The GH3 brought about 1/2 stop more and the GH4 brought another 1/2 stop.
 
Hi

itimjim said:
Great set of tests, thank you. I share the disappointment with everyone regarding what appears to be quite limited DR.

I just wanted to check though that there isn't some misconfiguration between working with and exporting full/legal range. The highlight roll-off is resembling a lot of the superwhite stuff I've seen on YouTube/Vimeo over the past week in regards to this camera. For instance https://vimeo.com/92613144 which is premier based workflow, but I know Resolve can be a bit funky with YUV/RGB too.
Click to expand...

A good tutuorial.

I think the vast majority of cameras capture superwhites, the exception now is probably the GH4 as if set specifically to 16-235 it will adhere to that range, whereas the vast majority of cameras that have no exposed settings for the RGB range will virtually always go from 16-255.

For Sony Vegas the Colour Curves filter seems to be a good way to map superwhites into the legal range and recover otherwise lost highlights. I've recovered whole distant mountain ranges that were a shadow in a white heat haze, made clouds appear in what looked like cloudless skies, and various other details that would otherwise be lost. The colour curves can also be adjusted to map superblacks back into range, which will be important with the 0-255 range (if used) of the GH4 when wanting to recover lost shadow detail rather than perhaps favouring lost highlights.

For Sony Vegas I use the project examples from http://www.glennchan.info/articles/vegas/color-correction/tutorial.htm to save the Colour curve settings for ease of use.

Regards

Phil
 
goodgoings said:
I also think that 8 stops+ a little of real usable dynamic range is believable. Shian of Color Gears rated the GH2 at 7 stops. Ryan Walters on vimeo also concurs with +3 over and -3 under. The GH3 brought about 1/2 stop more and the GH4 brought another 1/2 stop.
Click to expand...

I think Shian rated the GH2 as having 6 usable stops, with a recommendation to keep everything with 4 if you can.
 
I just wanted to third the idea of further tests and experimentation with the 0-255 vs 16-235 levels issue. I'm not an expert on the subject and can't say if that is what is causing the DR to appear more limited than it should be, but while working on posting my latest video online I ran into some issues with DR which led me to do quite a bit of reading and testing on this stuff. I realized that these levels and/or the associated gamma settings are completely all over the place depending on what player and even OS/computer you play the video on. My tests indicated that the DR looked very noticeably different on all 3 desktop players I used on one computer (Quicktime, Windows Media Player, VLC). Then the same player would look entirely different (not just a monitor difference) when viewed on another computer (both Windows).

I found that the web players (youtube and vimeo) definitely reshaped the level curves of the original video, to the best of my recollection. Would you be able to upload a raw clip of one of the tests and make it available to download? That would be helpful to determine if the post process is introducing any problems.
 
You must log in or register to reply here.
Back
Top