GH4 'Lab Review Reveals 10bit Flaws in Vlog + FREE GH5 LUT'

[stoneinapond]

......
IMHO it's a good thing they compare those cameras - if anything it shows how GOOD the GH5 is.

You make a very good point. It's the headline to the article that disturbs for me. It's pure click bait, which is a shame because some of the content is valid.

I've noticed a bewildering behavioral trend surrounding the release of a new camera, espetially one like the GH5.

First there is speculation. Then anticipation. Which eventually gives way to anxiety, and then to full-blown fear as the various opposing opinions fly around the Internet at the speed of light.

A resounding chorus rises from the ranks with chants of cancelling pre-orders, is the manufacturer blind, they didn't listen to what I want. People head for the hills before the onset of the apocalypse.










Then everybody rushes out and buys the camera anyway. :grin:

Please, stop with the bogus headlines. It just muddies the waters.

 

[8string]

just saw this.a good rebutal....http://www.eoshd.com/2017/03/cinema5d-slate-panasonic-gh5-call-v-log-10bit-unusable/

 

[Ron Evans]

The inclusion of the Canon C700 in this lab test is only for demonstration purposes so people can see how 10bit should have been looked like, so of course there is no a direct competition between the 2 cameras.

Thank you.

Johnnie

I understand your position Johnnie but for completeness recording externally to an Atomos Inferno from the GH5 may have completed the picture for a lot of us. As it is you proved that the equivalent of a Porsche Carrera is faster than a small Kia. If the measure is can you get a speeding ticket then they both comply !!! There has to be a cost/ performance measure and for most this is what is important. How much performance can one get for the money and for most the GH5 is a winner even if the performance is little changed from the GH4 the extra features are of real value.

 

[Thomas Smet]

I think C5D makes the best out of testing things. Any test will get questioned but as they build a library of tests which can be used to compare different cameras with a constant testing formula one can at least see how things compare.

Any popular camera with killer features soon to be released isn't to welcome to criticize and buyers will complain. Because who wants to admit their new and shiny camera has flaws?

Except they are not very realistic tests because it puts details under a microscope instead of actually looking at what the camera produces visually. The articles focuses solely on these few flaws and rarely discusses the advantages of the camera. It is typical one sided journalism that picks and chooses the perspective to present the facts.

Even the title of the article is all doom and gloom and immediately makes people feel like the GH5 is an epic failure. It is irresponsible journalism at its finest and the kind of attention seeking headlines one would find in the Enquirer.

Honestly I would take real world V-log examples from Luke Neumann as an example of what the camera can do vs a tunnel vision unrealistic lab test by Cinema5D with no understanding of the test process. The Gh5 or any camera isn't just about lab tests and benchmarks. It is about capturing images in the real world. Now I am a geek and love tests more than I should but this was such a narrow minded view of what makes the GH5 a good camera and it mainly focused on hunting for whatever flaws it could find.

The flaws the article did point out are kind of obvious or at least they should be and in all reality I doubt anybody really expected miracles when it comes to those issues. Noise, dynamic range and trying to cram 10bit V-log color into a file 1/27th the size is not going to be spectacular but in the scope of what that actual means in the real world was never really addressed.

For example lets look at the subject of external recorders vs internal 8bit GH4 recording. There are dozens on video and text reviews of external recorders that claim they cannot see a visual difference between the internal codec. Of course there is a difference but the point is that in the real world such nitpicking things are not nearly as important as some may think they are. While the Cinema5D article focused on the flaws on the GH5 in the scope of what that actually means to 99% of users it may only affect 1% of the overall quality. Is that 1% worth passing on buying a GH5 or does this article over inflate that 1% to make it seem like 50% in the minds of the reader?

In the end it is the content that ultimately matters and articles like this only make people feel as though they need to keep obsessing over cameras and their microscopic flaws.

 

[philiplipetz]

Johnny,

It is a question of balance. In your XC10 review you were careful to hint at the positive before you said what you did not like, "Although this camera can produce at times gorgeous looking images, the way to get those are in my opinion very painful – but before I share my concerns, here is what I liked:"

This review hides the positive at the end, and the headline talks about flaws. And then you ignore the solution to the flaw, using an external recorder as James Miller demonstrated. https://we.tl/wC2RIj8TjI



I am currently shooting a project near you, in Munich, and I can appreciate that English is not your primary language; however, the way you organized the article unfortunately makes it seem like you want only to talk about how you found a flaw in a highly anticipated camera.

 

[joe12south]

philiplipetz,

No one is trying to "make a reputation by destroying cameras as they are introduced". PR and marketing became an important tool to the respected companies in our industry and while occasionally the end user might find him/herself overwhelmed with that information, we knowingly decided to take the unpopular role of doing those lab tests (along side our user reviews), and by using a constant methodology put on the table how a camera behaves. The routine is similar. First we are being accused for "what ever", then the camera gets into the market and customers starts to see its shortcomings like we previously reported.

In this particular case, there was/is constant communication with the manufacturer. In fact, this lab test was scheduled for 2 weeks ago but we delayed it until getting the latest firmware. It is also important to highlight that most if not all manufacturers whom we have been in contact with, are using those lab test to improve their products.

There is much more to write but as the FUJIFILM GFX, X-T20 and Canon 700 are on our table for testing, I'll keep it short for now.


Thank you!

Johnnie

Whether intentional or not, the language you use to describe the issue...which I agree is real...is technically wrong, and worse, misleading. "h264 compression of LOG images is problematic, regardless of the bit depth." End of story.

You also shot resolution tests through an adapter, and spent almost zero time talking about the usability of the camera. All of these factors combine to impact the credibility of the review.

 

[joe12south]

The inclusion of the Canon C700 in this lab test is only for demonstration purposes so people can see how 10bit should have been looked like, so of course there is no a direct competition between the 2 cameras.

Thank you.

Johnnie

No, you're comparing two completely different codecs and two different bitrates...again, what does this have to do with 10-bit or than a headline?

 

[Alpha Beta Omega]

Why don't just chill and read C5D review just like any other review. Some are always positive and some are negative. Stop bitching about it and take it for what it is.

 

[hyalinejim]

It's entirely possible that Panasonic has been aware for some time now that 10bit 150mbps is not enough to prevent chroma macroblocking in VLog, and that this is why they're introducing the 400mbps All-I codec later this year.

 

[OPHERBA]

...and spent almost zero time talking about the usability of the camera.

Not sure how much you are following our site but we always try to do an additional "visual/usability test". This time we released the lab test first and the visual/usability test is on the making. As I don't just go out, wave the camera and call it a "test/review", it takes a bit of time.

Thank you.

Johnnie

 

[Lpowell]

Did anybody on here honestly expect 150mbps 4:2:2 10bit H264 to look as good as external ProRes422 HQ? While the color space enhancements are there with internal 10bit the encoder is still a H264 based encoder that needs to group similar colors together to get the video down to only 150mbps.

To 99% of the GH5 users the internal 10bit format will be phenomenal and they will fall in love with it. If a colorist will be touching any of the material however then I have always highly recommended using an external recorder and a format like ProRes422 HQ.

Thomas is absolutely correct, and this test confirms what many have been saying from the start of the V-log L controversy: For 10-bit resolution, use an external recorder.

Inadequate bitrate has been a persistent pitfall with internal H.264 recording on Lumix cameras, ever since the GH1. There's an inherent trade-off between SD card recording capacity versus image quality, and with Panasonic, mainstream consumer satisfaction always takes priority. Just as the GH4's 100 Mbps 8-bit 4:2:0 4K bitrate was barely adequate, so is the GH5's 150 Mbps 10-bit 4:2:2 4K bitrate. To capture those extra two bits of precision, the H.264 encoder has to preserve at least twice as much detail in both horizontal and vertical dimensions, making it far more susceptible to bitrate degradation from artifacts and low-level noise contamination. At 150 Mbps, most of those 10-bit details will inevitably be sacrificed in shots that involve significant motion or panning.

For comparison, I've shot a great deal of 8-bit 4:2:0 4K footage at 150 Mbps on the JVC LS300 in J-log1 mode. At that bitrate, the effective color bit-depth of this camera is excellent, but in extreme cases, 150 Mbps is barely enough to preserve smooth under-exposed gradients. For 10-bit resolution, it would definitely need more than 200 Mbps.

 

[hyalinejim]

Thomas is absolutely correct, and this test confirms what many have been saying from the start of the V-log L controversy: For 10-bit resolution, use an external recorder.

Inadequate bitrate has been a persistent pitfall with internal H.264 recording on Lumix cameras, ever since the GH1. There's an inherent trade-off between SD card recording capacity versus image quality, and with Panasonic, mainstream consumer satisfaction always takes priority. Just as the GH4's 100 Mbps 8-bit 4:2:0 4K bitrate was barely adequate, so is the GH5's 150 Mbps 10-bit 4:2:2 4K bitrate. To capture those extra two bits of precision, the H.264 encoder has to preserve at least twice as much detail in both horizontal and vertical dimensions, making it far more susceptible to bitrate degradation from artifacts and low-level noise contamination. At 150 Mbps, most of those 10-bit details will inevitably be sacrificed in shots that involve significant motion or panning.

For comparison, I've shot a great deal of 8-bit 4:2:2 4K footage at 150 Mbps on the JVC LS300 in J-log1 mode. At that bitrate, the effective color bit-depth of this camera is excellent, but in extreme cases, 150 Mbps is barely enough to preserve smooth under-exposed gradients. For 10-bit resolution, it would definitely need more than 200 Mbps.

What's your opinion of the potential for the 400 Mbps All-I?

 

[Lpowell]

What's your opinion of the potential for the 400 Mbps All-I?

Depends on how much shadow detail you're looking for. The main advantage of All-I encoding is consistency of frame-to-frame image quality, regardless of motion or panning. Since each All-I frame is separately encoded, it doesn't matter how much motion occurs between frames. With 150 Mbps IPB encoding, recording of relatively static scenes is very efficient, resulting in high image quality, until things start moving.

For fast-action, high-contrast outdoor scenes, I'd definitely use the 400 Mbps All-I encoder. For static, low-contrast indoor scenes, I'd expect the 150 Mbps IPB encoder to capture finer shadow details.

 

[filmguy123]

I agree the article headline is buzzfeed-esque and over dramatized. I appreciate the in-depth lab test! But I don't feel the conclusions drawn from such are well balanced and useful like some of the posts here from Barry, Thomas, and others.

Saying "Vlog has serious problems" is simply not a useful or accurate analysis; going into depth like Thomas regarding bitrates and compromises, etc. IS.

Anyway, I'm thankful for the lab tests for what they are, but think the journalism and analysis could be greatly improved.

Bottom line for me is, lab tests and pixel peeping is helpful, but I know what my eyes are telling me. The properly shot images out of the GH5 are YEARS better than the GH4, they really phenomenal, and I could tell how much better the downloaded footage I graded in post held up than what I had been working with for years on GH4. At the end of the day, the rest is just numbers.

 

[Thomas Smet]

What's your opinion of the potential for the 400 Mbps All-I?

Thats a tough one. We have had all-I formats on the GH3 and GH4 that left a lot to be desired. A codec can never really be measured by just bitrate alone. There are so many other factors that make 400mbps either look like garbage or make it look stunning.

The 200mbps HD mode on the GH4 should be more than enough for 8bit 4:2:0 HD but alas it is not and the codec falls way short of what one typically expects of formats like ProRes. That 200mbps is a ratio more like 800mbps for 4k so 400mbps all-I could be severely bit starved. With that said Panasonic's ACV-Ultra format also uses 400mbps and looks great. It all really comes down to how well does the encoder handle detail and areas of subtle flat color. GH encoders tend to do very well with detail but ironically fall apart in subtle graduations of color.

AVC-Ultra takes some hefty processing power to get the level of efficiency it has with only 400mbps. I'm not sure a compact body like the GH5 can really handle that format. I hope I'm wrong but my best guess is the 400mbps mode will be similar to the 200mbps HD mode on the GH4 and leave a lot to be desired. 400mbps could still be a very nice format but by the time you use a 400mbps format you might as well just go with a 800mbps ProRes format and be done with it. The extra gear is an issue for some people but if one is that concerned about pixel quality why are they also not concerned about shooting 4k on a 960x540 resolution 3" LCD screen?

 

[Cary Knoop]

Did anybody on here honestly expect 150mbps 4:2:2 10bit H264 to look as good as external ProRes422 HQ?

The target bit rate for ProRes 422 HQ is about 220 Mbps, that is obviously more than 150 Mbps, however intra and inter-frame compression together is more effective than intra frame compression only, so if we take that into account ProRes should at most be marginally better.

The announced 400 Mbps intra option available this summer should be much better than ProRes.

However, for me the big question is did Panasonic pull a fast one on the 10 bit, do we really get this resolution or are the top bits hardly anything more than pure noise and do we effectively have an 8 or 9 bit signal dithered to 10 bits?
More testing obviously has to be done, but the above mentioned test is to me rather discouraging.

 

[Barry_Green]

The target bit rate for ProRes 422 HQ is about 220 Mbps, that is obviously more than 150 Mbps, however intra and inter-frame compression together is more effective than intra frame compression only, so if we take that into account ProRes should at most be marginally better.

Whoa - hold on. ProRes 422 HQ is 220Mbps in HD. In 4K/30p it's 880 mbps. In UHD/24p it's about 734 mbps.

The announced 400 Mbps intra option available this summer should be much better than ProRes.

For UHD/4K? No possible way. I just wrote a big thing up in the other thread, but it says pretty much what Thomas Smet said. I am not optimistic in 400mbps ALL-I looking all that great for UHD/4K 10-bit 4:2:2. You're talking about somewhere around 6x as much data as HD 8-bit 4:2:0, but only giving it twice the bandwidth of the already-mediocre 200mbps ALL-I. We may be surprised, but I sincerely doubt it.

 

[Cary Knoop]

You are right about the 4k/UHD bitrates.

But I am not convinced that the shortcomings coming out of the above mentioned test are solely attributable to the encoding process.

 

[Thomas Smet]

You are right about the 4k/UHD bitrates.

But I am not convinced that the shortcomings coming out of the above mentioned test are solely attributable to the encoding process.

Trust me it is. Plenty of us have recorded externally with the GH4 and have seen first hand how much better V-log can be with a format like ProRes. Plenty of examples of V-log on the GH4 recorded to ProRes that eliminate most of the color issues.

H264 is H264 no matter how you skin it. No matter if the H264 encoder is using 50mbps, 100mbps, 200mbps or 400mbps it still uses the same process to reduce and convert the image into digital information in blocks. A higher bitrate means the blocks stay small but many of the other size reduction methods are still taking place.

 

[Cary Knoop]

H264 is H264 no matter how you skin it. No matter if the H264 encoder is using 50mbps, 100mbps, 200mbps or 400mbps it still uses the same process to reduce and convert the image into digital information in blocks. A higher bitrate means the blocks stay small but many of the other size reduction methods are still taking place.

Are you saying that given the same bitrates ProRes is better than H.264?

As far as I understand it ProRess like H.264 is a DCT based encoder, it is not any better than H.264 if we compare it will similar bit rates.

One can encode H.264 all the way to lossless if necessary (but anything close or lower than CRF 10 will become indistinguishable from the original) and unlike ProRes H.264 allows inter frame compression as well, but one easily can compress intra frame only.
H.264 also supports higher bit depths and chroma sub sampling all the way to 4:4:4.

I think the fact that most H.264 encodings are 8 bit 4:2:0 inter frame with low bitrates does not imply it is an inferior technology compared to ProRes.