Panasonic GH5S 14.6 Stops of Dynamic Range

[DBounce]

According to EBU 3335 assessment the GH5S has 14.6 Stops of DR in HLG mode and 13.5 in VLogL.

https://www.newsshooter.com/2018/04/27/panasonic-gh5s-ebu-assessment/

[sanveer]

"The sensor has about 11 megapixels according to the specification and is a Micro Four Thirds format (sensor size 17.3x13mm which means 4:3 aspect ratio). The specification says that only “10.28 Megapixels are active”, making the resolution about 3704×2778 if it’s 4:3. So the individual photo-sites must be about 4.7μm square, about the same as a conventional 3-sensor HDTV camera. Alan found that neither of these sets of dimensions was enough for 4096×2160 recording unless his assumptions are over-simplifications – which seems likely from the measurements above.

...

The camera achieves HD Tier 1 because of the sensor size, while smaller than the preferred 1” size, is greater than the minimum-acceptable 2⁄3” size. The wording of R.118 is not completely consistent on this because testing is required whatever the sensor size, but the camera passes the tests anyway.

... 

It is also important for users to know that they can obtain more dynamic range shooting in the HLG or CineLike-D picture profiles than V-LogL. In saying that, V-LogL and HLG have a lot more dynamic range in the stops allocated above middle grey than CineLike-D. The amount of stops isn’t as important as where they are located above or below middle grey. Just because Cinelike-D has more DR than V-LogL doesn’t mean it’s a better picture profile to shoot in."

 

 

The edifice of his research itself is wayyyyyyy off.

It is NOT a 4/3 ratio sensor, and this is NOT the aspect ratio for 10.23 MP.

They Are actually these:

3680 x 2760 (10.2 MP, 4:3),
3840 x 2560 (9.8 MP, 3:2),
4016 x 2256 (9.1 MP, 16:9),
2752 x 2752 (7.6 MP, 1:1),
2592 x 1944 (5.0 MP, 4:3),
2736 x 1824 (5.0 MP, 3:2),
2816 x 1584 (4.5 MP, 16:9),
1936 x 1936 (3.7 MP, 1:1),
1824 x 1368 (2.5 MP, 4:3),
1920 x 1280 (2.5 MP, 3:2),
1920 x 1080 (2.1 MP, 16:9),
1360 x 1360 (1.8 MP, 1:1),
3328 x 2496 (8.3 MP, 4:3),
3504 x 2336 (8.2 MP, 3:2),
3840 x 2160 (8.3 MP, 16:9),
2880 x 2880 (8.3 MP, 1:1)

 

I am guessing the GH5s could have about 1/5 stop more dynamic range than the GH5 for raw, and less than that for video.

Also, I doubt ANY camera can do 14.6 Stops in a 10-bit codec. I doubt even the highest grade ProRes 10-bit will reach 14.6 Stops. As a matter of fact, after this review, I am curious as to how many Stops of Dynamic Range the Alexas and REDs Achieve on their ProRes Codecs. 

Also, if this sensor were So Capable, Blackmagic would have advertised the 14.6 stops as the USP, above the Dual Native ISO.

Also he says Cinelike-D has more dynamic range than Vlog. How?

Take this review with a ping of salt. Or something stronger. 

[Mmmbeats]

The author admits that his resolution calculations are speculative.   That won't have affected the outcome of his tests though.

[no_connection]

I don't see 4096x2160 in the table above. So the assumption isn't that off?

*edit*

The dynamic range also makes sense, since I read it that it is done in HD. and scaling 4k to HD would gain one stop of DR from noise being reduced.

What I don't like about the noise test is "All measurements were made in HD mode (1920x1080). Noise levels in UHD and 4K modes ought to be similar."

[Luke Mason]

They measured absolute dynamic range, which they did not do for Alexa: https://tech.ebu.ch/files/live/sites/tech/files/shared/tech/tech3335_s11.pdf If they used the same method from the GH5S test to evaluate Alexa, it would have easily reached 15+ stops of dynamic range.

[Mmmbeats]

1 hour ago, Luke Mason said:

They measured absolute dynamic range, which they did not do for Alexa: https://tech.ebu.ch/files/live/sites/tech/files/shared/tech/tech3335_s11.pdf If they used the same method from the GH5S test to evaluate Alexa, it would have easily reached 15+ stops of dynamic range.

Any chance you could explain in simple terms what the difference between the two methods is?

[Marcio Kabke Pinheiro]

"The camera achieves HD Tier 1 because of the sensor size, while smaller than the preferred 1” size, is greater than the minimum-acceptable 2⁄3” size. The wording of R.118 is not completely consistent on this because testing is required whatever the sensor size, but the camera passes the tests anyway."

Huh?

[sanveer]

20 minutes ago, Márcio Kabke Pinheiro said:

"The camera achieves HD Tier 1 because of the sensor size, while smaller than the preferred 1” size, is greater than the minimum-acceptable 2⁄3” size. The wording of R.118 is not completely consistent on this because testing is required whatever the sensor size, but the camera passes the tests anyway."

Huh?

EXACTLY.

A 1 inch sensor is 16mm Diagonal, whereas the GH5s sensor is 21.63 mm Diagonal. How could it then, possibly, be smaller than a 1inch sensor???

The guy who has reviewed it, obviously did no research at all, and despite almost every fact being incorrect, he arrives at a conclusion which he believe to be of great accuracy. He doesn't even know the specifications of the sensor, leave alone the exact number of pixels on it in different modes (which Panasonic itself shared in various literature).

Suddenly, after seeing this one, one would wonder how off the mark, the rest of his results are (with regard to the other cameras mentioned). It is one thing to get a conclusion incorrect, but a Totally Different thing to start on a Completely Erroneous Footing. 

This is the site that came up with:

https://tech.ebu.ch/publications/tech3335

 

 

This reminds me of one of Murphy's famous laws (on experts):

"An expert is a person who avoids the small errors while sweeping on to the grand fallacy."

                                      -   Weinberg's Corollary

[elgabogomez]

Well considering that the alexa is barely passing their tests (1440x810 real resolution, 1db more noise at mid gray, some aliasing, some ir pollution) ... their tests are specifically designed for bbc qc and standards... So as the rebel punks anti establishment shooters that we here are, of course we must dismiss their findings.

[Neufeldt]

10 hours ago, Mmmbeats said:

Any chance you could explain in simple terms what the difference between the two methods is?

"How do you define dynamic range?" is the question that underlies and undermines most of these discussions. We could measure it as the difference between maximum signal and the point at which signal:noise reaches 1:1 (how DXOMark measures RAW still dynamic range). We could decide that 1:1 is too noisy, and that "usable" dynamic range starts at 8:1 meaning 3 fewer stops of dynamic range. We could measure from any signal at all, adding one or more stops of dynamic range relative to the 1:1 reading. We could weight RGB in various ways, perhaps considering the beginning of dynamic range to be when any one channel starts to have signal and then terminating only when all three channnels are saturated (this gives the highest possible rating).

 

The ARRI linked pdf only tested out to 11.75 stops and didn't find dynamic range limitations, if I'm reading it correctly. It didn't test for the limits.

[elgabogomez]

Neufeldt, that’s the problem of this test. I guess there is logic in how he (Alan Roberts) decides when to end his tests and why... There are certain things he looks the camera to achieve and once they are met that’s enough, but methodology wise, they are all over the place from one camera to the next... ie: why test a6300 and GH5s and not gh5 or a6500 or a7s/r/xyz or bmcc or pocket... is it project specific? Somebody asks if they can use such camera to the bbc then bbc asks mr. Roberts to test it? We don’t know so it almost feels as a fight between Superman, Goku and hello kitty, that’s how random it looks from the outside. They (EBU) are testing them as “television cameras” and passing them or failing them as such. What television camera is better than the alexa??

[Kisaha]

EBU is testing for high end television or cinema cameras, Alan Roberts does that for European Broadcasting Union (that exists since 1950), and used to work for BBC, 2 of the most accurate and respectful organizations in the whole world, and does that for a lot of decades, maybe more than your dads (some of you).

This is not a camera review, or a vlog, is a technical paper that is important for some of us that we must follow some of these "findings" for specific projects, he is certainly not God (some will doubt that God, is God anyway, ) but I trust he is doing the best he can, or he is getting so old, he can't really comprehend new cameras (joke)!

https://tech.ebu.ch/cameratests

Alexa was tested as a 2K camera if I remember right, and he said in his newest reports that it is not certain that the full dynamic range is truly usable in the field. 

All in all, great findings for the GH5S, I believe it raises the camera near the top, at Tier 2! Seriously, excellent results for the GH5S, I believe is a truly remarkable camera, if it wasn't for the waiting game of the Pocket4K.

The a6300 review was a specific project, not under EBU https://www.newsshooter.com/2016/03/31/newsshooter-sony-a6300-ebu-lab-test-by-alan-roberts-surprising-results-not-suitable-for-use-in-general-broadcasting/

Edit: He indeed, did the a6300 paper after all, sorry, I missed that one. Interesting that they have 5 places reserved (5 other cameras?) 

https://tech.ebu.ch/camera_reports_tech3335

[jonpais]

Alan Roberts’ assessment of the color of the GH5s is that it is quite good, while he had this to say of the Ursa Mini Pro 4.6K:

Overall, the colouring is rather lurid. Skin tones are rather too pink, while all other colours with significant red content (orange, violet, salmon pink, purple, light orange, red and yellow) are all too bright and pink. The cyans are too bright. The grey scale is correct, so this is not a colour-balancing problem. Clearly, all pictures from this camera will need post-production grading to improve the colour performance.

In other words, quite contrary to the experience of most online reviewers who’ve tried matching the GH cameras to the Ursa - not the other way around! So it will be interesting to see how the colors of the Batcam compare to those of the GH5s. Assuming that BMD would want to allow footage from the Pocket to be intercut seamlessly with the Ursa, I’d expect the color science to be similar, if not identical.

Mr. Roberts also seems to believe in-camera noise reduction is best left at the factory default setting, which again goes against received wisdom. 

 

[elgabogomez]

I understand the credentials, the institutions represented and the goals of such tests. And I’m not disputing the GH5s findings. But I (person not institutional member of a board protecting standards) believe it’s assessments like the alexa one, are responsible of its dismissal for a project on Netflix and similar projects... Not a 4k native camera... I’ve seen tests made by creative people where ump and alexa are tested next to each other and they can be made to look 99% the same. No poor color on ump, no poor resolution on alexa... Then on a6300 tests, there are several things not tested (including external recording and picture profiles)... so it maintains the upressed 4k of the alexa and the downconvertion of 4k 4:2:0 8bit to near 1080 4:2:2 10bit an internet myth, I strongly believe it’s these experts quasi obligation to do such extra steps to push the envelope.

[jonpais]

24 minutes ago, elgabogomez said:

 I’ve seen tests made by creative people where ump and alexa are tested next to each other and they can be made to look 99% the same. 

Care to share those tests? Because I've also seen tests where the GH cameras can look 99% the same as the UMP.  So in theory at least, the GH5s should be able to achieve .9801% of the Alexa.  

[Kisaha]

GH5S is not just a GH camera. In all the reviews and tests we have posted here it has the best color science of them all.

[jonpais]

1 hour ago, Kisaha said:

GH5S is not just a GH camera. In all the reviews and tests we have posted here it has the best color science of them all.

First, let me correct my .9801% statement. I meant to type 98.01%, though the first figure might be more accurate.  As far as GH* cameras go, the GH5 has outstanding color, even if the GH5s has better color. It's really all about the skills of the operator, color correction and grading. Lots of ppl talkin' few have examples of their own work to show.  

 

[jonpais]

And here's a screen grab from a color correction tutorial I worked on this morning. Far from shitty color science. (click twice to enlarge)

[sanveer]

 

I realised that while this test may be borderline ludicrius regarding actual dynamic range, it does highlight certain very interesting facts and even maybe give us a hint or two about the present and future developments in the GH line of cameras and their capabilities.

 

"Like 709 10.3

HLG 14.6

V-LogL 13.5

CineLike-D 13.9

... 
Note that these figures are a little speculative and subjective: it is up to the user to decide whether the performance at the extreme ends of the dynamic range are acceptable, thus there is an uncertainty of about 1 stops in all these results.


Neither of these sets of dimensions is enough for 4096x2160 recording unless my assumptions are over-simplifications – which seems likely from the measurements above."

He claims the HLG has 14.6 stops of dynamic range. The CameraStoreTV guys (who are now on DPReview) said that Panasonic reps told them that HLG has a little more dynamic range than VLog, which could make sense since Vlog L, the one used on the GH cameras is capped at 12 stops and below. I also remember someone mentioning (in a DVD Forum?) about 11.63 or so stops, measured for Vlog on the GH5s.

If that is the case, then the guy is off by approx 1.87 stops or so (13.5 - 11.63). If that is the case then 14.7 - 1.87 (the stops by which it is off) is about 12.73 stops (or a little less). 

Wide Dynamic Range is a common rant mentioned in all the Sony Sensor Literature on  the Sensor. Also, BMPCC 4k mentions 13 stops of dynamic range (I am guessing it should be around 13 stops -/+ a third of a stop).

If that be the case, then it is time Panasonic redo the Vlog L (and push it for 13 stops, instead of capping it at 12 stops). The 12 stops limit, may be the sole reason the GH5s doesn't seem to have more dynamic range in 10-bit 4-2-2.

Panasonic were probably too lazy to improve the Vlog, apart from wanting to save their more expensive cameras. And since the GH5 according to DXO does 13 stops in RAW photo, the Vlog seems to be a similar version of the GH4 which hasn't been improved in dynamic range since then.

 

[Don Kotlos]

Yeah the dynamic range numbers are of little value. He measured the absolute dynamic range and since you divide by something extremely small, it is very sensitive to measurement error as well. Something like the Xyla 21 would be more useful. GH5s should be ~13stops with a usable dynamic range even less than that.