GH5 GH5s dynamic range tests

Jim Arthurs

Jim Arthurs

Well-known member
So I did some tests for my own use, then thought I'd share... Your mileage may vary, etc.

I'm prepping for some green screen work with the GH5s and have had great luck with the Natural picture profile in the past, so I wanted to compare it with VLOG-L at the low base ISO at two different ISO settings.

The methodology is as follows; I prepped a back lit cutout on milk plexiglass, illuminated it to right at the white clipping point, and used shutter speed and my f-stop range to move through 15 stops of decreasing exposure.

I recorded the screen of my Odyssey7Q+ with the waveform full screen, then in post cut out the dead spots and lined all the "slices" up side by side.

http://files.datausa.com/imageshoppe/outgoing/GH5s/GH5s_DR_tests_01.mp4

My take-aways... I believe Panasonic said the best dynamic range for VLOG-L in the lower range is 800ISO, however I don't see any loss in dynamic range using 320ISO, but I DO see the noise level easily cut in half! I see 11 stops free and clear, with the 12th stop touching the noise floor at 320ISO. On the monitor, I can continue to distinguish the high contrast shape on the camera original file out to 15 stops, but of course that's not the same as having that as usable range. I believe DPReview had VLOG pegged at 11.3 stops, and I sort of feel it's between 11 and 12...

The Natural profile was interesting... a good clean 10 stops! Natural is what I use when I have to have something "ready to edit" for a client, or pass off footage. The 200ISO Natural is of course cleaner than the "native" 400, but doesn't seem to impact dynamic range, and of course that profile can take advantage of the full IRE range in 10 bit if you want to do so.

Interesting camera, lots of fun...

What about HLG? I did test it, but had an issue with the way the waveform was displayed so I couldn't include it in this chart. I did see 10 clean stops before it was interacting with the noise floor.
 
Jim this is great, thank you.

Did you ever do this with GH5? Or have recollection of how those results compare to the 5?

From what I am reading here it sounds like your observation is a solid 11.5 or so, with a very smooth roll off in either direction - not necessarily "usable" but also definitely contributing to a smoother appearance as the values disintegrate VS a hard stop?
 
No, just the GH5s... It would be interesting to compare, as the setup is easy to replicate. I would like to try this with the EVA1 at some point.

Yes, I agree with your summation of what I said... :) Or put another way... that 11th stop is a hard working, still productive senor citizen stop. The 12th stop has one foot in the grave, and the rest are merely shades and spirits best revealed through seances and visits with spiritual mediums...
 
Hi Paul, your stills posted in another thread from the feature you shot using HLG look great!

As I mentioned, I did include HLG in the round of testing, but didn't include it in the line-up because I found an issue with the GH5s and the Odyssey7Q+ in how HLG is represented in the waveform. It should be legalized, but on the Odyssey scopes it shows going from -10 IRE to 109, which doesn't match how it appears on the GH5s internal scope.

Anyway, I will link to it here, independently, but I'm only seeing 10 stops free and clear of the noise floor. That's my personal baseline for usable dynamic range... how many stops can you count on that are not interacting or touching the base noise.

http://files.datausa.com/imageshoppe/outgoing/GH5s/GH5s_HLG_dynamic_range.mp4

However, I will redo optimistically when I include all the VLOG high native ISO tests and check it on an NLE's scopes, and I will include HLG in my actual green screen tests when I'm pointing at something other than a scope. :)



Regards,
 
Interesting. HLG really should be holding onto as much DR as V-LOG L. It's certainly capable on paper of doing so. Given the "hybrid" part of HLG, I wonder if this testing methodology might influence results. At first blush, I can't see any flaw in how you did it, though.
 
Yes, and I certainly have no bias against it, to the contrary, if it can eek out similar DR to VLOG and HLG can better fill that 10bit bit bucket I'm all for it.

I will retest where I not only have the recodring waveform but put the original camera footage on the NLE and see how the NLE scopes handle it. Also this weekend I'll do some green screen tests and certainly include HLG with the Natural and VLOG...
 
Hey guys!!!!

for 8bit modes like 50P/60P 4K, did any of you try using Natural profile with -5 contrast and the curves at -3 highlights and +3 shadows to have an inverted S shape to help with encoding of lower IREs and to help a little with highlight details???

A sort of REC709 loglite for 8bit he he??
 
Not specifically. One thing I did discover in my green screen tests is that the Natural profile is a bit more noisy than I thought it would be until you were down around 160 ISO or lower. 80 ISO Natural was of course very, very clean, but VLOG 320 or HLG 320 "beat" Natural in noise until you were 160 ISO or lower with Natural.

In truth, the noise of both HLG and VLOG compared at 320 and 2500 are surprisingly close indeed. I would have no reservations shooting a green screen interview at 2500 ISO and that is a "new" experience for me.

For 8bit work I would suggest experimenting with the slope and point functions of the LIKE709 profile... you should be able to preserve a fair amount of highlight detail. I know this is on my short list to test.

Regards,
 
Thanks Jim!

the problem is when I rent the GH5s I often shoot outdoors @ 60P 4K so 8bit and skies are a huge no for V-LOG in 8bit...banding is insane!!!

What about Cinelike-D? Maybe it's the right one for maximizing DR in 8bit??? It seems better than LIKE709 from user opinions around...

What's your take on this?

Thanks a lot!!!
 
I never really grooved on Cine-D (and REALLY disliked Cine-V) from past work with either the GH4 or my little bang around G7, but that's a personal issue and I haven't done even remotely scientific tests. I know Paul's LUT's make good use of Cine-D and tame it nicely. I will probably re-visit all of these in context of the GH5s over time.

I am very happy with how banding is under control with the GH5s and VLOG in 10bit though... shot some very flat pale blue skies this weekend that survived a good deal of creative grading under the Pana conversion LUT with no issues... nice!
 
dia3olik said:
Thanks Jim!

the problem is when I rent the GH5s I often shoot outdoors @ 60P 4K so 8bit and skies are a huge no for V-LOG in 8bit...banding is insane!!!

What about Cinelike-D? Maybe it's the right one for maximizing DR in 8bit??? It seems better than LIKE709 from user opinions around...

What's your take on this?

Thanks a lot!!!
Click to expand...

If you like the colors in Cinelike-D, then yes go for it. I've always found that profile to be problematic especially for skin. It's better on the GH5 than GH4, and GH5s seems to improve it a bit more, but I still don't love the CineD colors.

But yes, seems to get the best DR in 8 bit without as many banding issues as the Log profile.
 
CineD is the best for Dynamic range when you can't use V-log. But it requires more color correction then Like709. Easy fix with leeming lut tho.

dia3olik said:
OMG wait a sec...

check out this video...from the official Panasonic channel...

https://www.youtube.com/watch?v=Z19KR5chOio

It seems that shadows and highlight IS indeed increasing DR...mmmmmm
Click to expand...
Yes but the results would be the same if you exposed for highlights and adjusted the shadows in post.
 
What is with Panasonic claiming 14 stops of DR for the GH5S. Isn't VLOG limited to 11-12?
 
npm said:
What is with Panasonic claiming 14 stops of DR for the GH5S. Isn't VLOG limited to 11-12?
Click to expand...
Its for stills. Its been a long time tradition of DSLR/SLR companies to list the raw photograph Dynamic range and not the video.
 
You must log in or register to reply here.
Back
Top