Panasonic GH5 10 bit internal recording not good enough

[interceptor121]

2 hours ago, deezid said:

yep, that's exactly how it is.
The 10 bit 150mbit/s implementation is way more effective but also more CPU excessive than the other H264 implementations. And in terms of artifacting it surpasses every 8 bit H264 choice on this camera quite easily in basically every situation (Talking about V-log L here).

It isn’t mor

 

2 hours ago, deezid said:

Just shoot a clear sky in V-log using the 8 bit codecs.

 

And you'll never do it again. Try using the codec instead and do colorgrading in Resove.

 

In terms of actual visible quality the GH5 codecs behave like this:
400Mbit/s Intra and 150Mbit/s IPB 10bit both look visually the same in most situations and grade extremely well, even in extreme low light with lot's of noise and usually don't show any banding artifacts. I actually managed to create banding using these codecs only once since I've bought the camera (gray wall underexposed).

200 Mbit/s 10bit 4:2:0 H265 codec for 6K anamorphic has some slight banding issues and shows some macro blocking.

150Mbit/s 8bit codec doesn't like gradients at all. Workaround is to use CineLike D and the Cinelike D to V-log conversion by @Sage

72Mbit/s 10bit H265 codec shows some slight banding, macro blocking and even some blur.

100Mbit/s 8bit codec doesn't work with gradients nor noise. Comparable to the GH4 H264 implemenation. Don't ever use it.

 

Some showcase of the terrible 150mbit/s 10 bit codec:

 

There isn’t an IPB 150 mbps 10 bit codec, the codec is IP only like Sony xavc. The level of compression is actually less than IPB 100 mbps or IPB 150 mbit 8 bits as you can see from the stream analysis there is less information packed in there 

the reason you can’t decode it easily is because at 10 bit uncompressed your graphic card probably overflows

The all intra has no cabac entropy encoding it just has intra frame interpolation pretty much like prores. Cabac introduces 20-30% efficiency so the codec once you shoot a static scene like yours are not much apart

in terms of the prores comparison the avc intra does not have any technical benefits over prores so maybe the atomos implementation has problems on its own

 

 

[Shirozina]

29 minutes ago, OliKMIA said:

Sure, it's not completely useless and but the bitrate and codec are very limiting. But again, it's nice to have.

Explain exactly how the bitrate and codec are 'very limiting'?

[webrunner5]

2 hours ago, MurtlandPhoto said:

Not a technical comparison, but these guys saved me the time to do my own testing months ago. Haven't looked back since ?

 

LMAO. Sad but true. ?

[jonpais]

Outstanding video for sure. 

[interceptor121]

1 hour ago, Dan Sherman said:

It's better than 99% of the "technical" and "mathematical" reviews I've seen by people who are using pseudoscience. It seems like about 1 in 1000 understand their is more than bit rate, bit depth, chrome sampling, and codec involved in getting a good image.  

It hilarious how many individuals think they more more than a large huge company.
 

 

 

No there isn’t really

you have optical image translated into digital signal and then compressed

at equal optical quality the compression determines the perceived quality 

Atomos are in the business of producing recorders but don’t own any of the codec logic I don’t know if they have a problem themselves with the hardware or software implementation 

for what concerns that video a guy with a black shirt on a black background with a tiny part of the frame moving doesn’t really prove much about internal vs external but it does prove the point that for simple shots the camera works just fine which is in fact the idea behind heavily compressed IPB acquisition 

You need a few more colours and movement to make the case although again there are many people (with a 8 bit monitor) that can take apart the 150 and 400 10 bit codec on such camera but the this could be due to the colors or detail not being there in the first place

if you look on color depth measured off the GH5 it doesn’t reach 24 bits or 8 bits rgb for raw...

So maybe the point is although the internal codecs will not allow for real 10 bits 422 uhd the camera doesn’t capture that detail anyway and therefore external recorders are a waste

still a bigger screen than the gh5 lcd would be nice!

 

[jonpais]

pseudoscience sums it up quite nicely methinks. 

[Dan Sherman]

1 hour ago, interceptor121 said:

No there isn’t really

you have optical image translated into digital signal and then compressed

at equal optical quality the compression determines the perceived quality 

 

Run this simple test with any noisy footage from any camera. 

1. Run the footage through ffmeg and encode it to h264 or h265 in crf mode
2. Run a de-noiser on the footage and then encode it with the exact same settings as you used above.

You will find the de-noised clip is smaller.

This is important imo, because I don't believe the GH series applies the same processing to hdmi out as it does to its internal recording. Also keep in mind NR isn't the only thing the camera is doing internally. The video linked above even shows signs of this.

 

Edit:
Also take note of the fact that even if  the camera does the exact some processing to the internal and hdmi streams, the hdmi stream still has to go through a transform process so it is complaint with the hdmi protocol. Then the recorder takes the stream and then translates it and then encodes it. So, the hdmi output goes through a lot of extra processing and we all know what that can lead to.

[interceptor121]

23 minutes ago, Dan Sherman said:

 

Run this simple test with any noisy footage from any camera. 

1. Run the footage through ffmeg and encode it to h264 or h265 in crf mode
2. Run a de-noiser on the footage and then encode it with the exact same settings as you used above.

You will find the de-noised clip is smaller.

This is important imo, because I don't believe the GH series applies the same processing to hdmi out as it does to its internal recording. Also keep in mind NR isn't the only thing the camera is doing internally. The video linked above even shows signs of this.

 

 

 

The camera does not apply any codec to HDMI out it simply passes the buffer before encoding for the recorder to acquire code and save

The test you mention is not appropriate as the size of the file depends on many other things not just bitrate H264 as a bunch of flags that squeeze the file but when encoding real time most of those are not active as otherwise the processing cannot keep up. Furthermore the codec is lossy so compressing it over and over again makes it smaller

Noise reduction sharpening and all that comes in the picture profiles and the rest are applied before encoding which generally only makes things worst as it can transform sharpening and noise artefacts into other errors when compressing

I think the improvements that you see over the 422 10 bit are not due to the 10 bits but to 422 vs 420 sub-sampling very well explained here

http://www.red.com/learn/red-101/video-chroma-subsampling

[Shirozina]

Time the OP puts up some samples and examples to illustrate these issues or goes away and diverts their evident spare time and persistence into something more worthwhile and constructive for both the film making community and mankind in general! The GH5 has been out for quite a while now and if any of the OP's 'concerns' were manifested in actual problems that users and testers could see we would have heard about them (ad nauseum) by now.....

[anonim]

4 hours ago, webrunner5 said:

LMAO. Sad but true. ?

Using both solution for extensive period - I'd say that conclusion of this video is very arbitrary, aside from pretentious pathetic tone for my taste.

For example, people all around are lamenting because of Panasonic's sharpening even at lowest sharpen-level - but in this video "sharp" details are almost exclusive criteria of judging - but Atomos power is neutralisation of NR and oversharpening.

Secondly, if we compare final or OOC front-"look" just in a such way - then there'll be not difference even between well compressed 8-bit image. Real difference is in level of grading freedom - I found that Atomos file is much more solid in that sense - higher HQ profiles are so good that make necessity for raw footage very rare... but this is not the same with internal codec's result. .. It is also very flexible but not so much as Atomos result. But that is not at all tested in this video... concentration just in the OOC "look" and even at greenscreen capability don't cover all sides of usability and differences.

Actually, I'm pretty indifferent - I'm very content with both solution. But, for the sake of the truth and based on my experience - test is not completely correct, i. e. without faults and too fast generalizations for the sake of - it seems to me - little bit of bombastic effect.

(Not to say anything that writing on SSD is so better and cheeper in comparison to a card - but of course, this fact is  not strict topic, although it is very important in decision to so "desperately" and "angry" brake Atomos.) 

[Dan Sherman]

 

3 hours ago, interceptor121 said:

The camera does not apply any codec to HDMI out it simply passes the buffer before encoding for the recorder to acquire code and save

The test you mention is not appropriate as the size of the file depends on many other things not just bitrate H264 as a bunch of flags that squeeze the file but when encoding real time most of those are not active as otherwise the processing cannot keep up. Furthermore the codec is lossy so compressing it over and over again makes it smaller

1.  If a device can keep up or not, depends on what preset you use. Even a camera could keep up with the lower end presets as they are not that intensive, but a camera for sure isn't going to be able to handle veryslow or placebo.
2. who said compress it over and over again? I said take clip A and trans code it to clip B using CRF mode.  Then take clip A again, and run NR on it before transcoding it to clip C. Then compare clip B & C. C will be smaller, because less noise means the codec can generate the same image quality with a lower bit-rate. In other words noise has a direct effect on how good a codec like  h.264  works.

3 hours ago, interceptor121 said:

The camera does not apply any codec to HDMI out it simply passes the buffer before encoding for the recorder to acquire code and save

 

3 hours ago, interceptor121 said:

Noise reduction sharpening and all that comes in the picture profiles and the rest are applied before encoding which generally only makes things worst as it can transform sharpening and noise artefacts into other errors when compressing

 

 

Picture profiles effect hdmi output, if it didn't v-log wouldn't work. The question is how much of the picture profile is applied to the hdmi output.


As I said before, converting the internal data to be hdmi compliant has an effect on image quality. The wiki has a good overview of the actual spec. Keep in mind when you are doing transforms like this, rounding issues can cause problems as your camera isn't running at 64 or even 32 bit.
https://en.wikipedia.org/wiki/HDMI

 

Quote

To ensure baseline compatibility between different HDMI sources and displays (as well as backward compatibility with the electrically compatible DVI standard) all HDMI devices must implement the sRGB color space at 8 bits per component.[6](§6.2.3) Ability to use the Y′CBCR color space and higher color depths ("deep color") is optional. HDMI permits sRGB 4:4:4 chroma subsampling (8–16 bits per component), xvYCC 4:4:4 chroma subsampling (8–16 bits per component), Y′CBCR 4:4:4 chroma subsampling (8–16 bits per component), or Y′CBCR 4:2:2 chroma subsampling (8–12 bits per component). The color spaces that can be used by HDMI are ITU-R BT.601, ITU-R BT.709-5 and IEC 61966-2-4.[6](§§6.5,6.7.2)

The gh5 is most likely working in YUV/YCbCr internally, as that's what its internally recorded files are. HDMI out is most likely 10bit sRGB to maintain maximum compatibility. Prores will support  sRGB and Y’CbCr, but I don't know what color space recorders like the atomos line are using.

Thus you have at-least one color space transom and if you are unlucky maybe 2. You will have rounding errors, and thus image quality degradation.

 

Saying internal or external recording is better than the other based on a handful of non scientific tests is idiotic. Even if you had full transparency (Intellectual property level knowledge ) of what Panasonic and Atomos was doing, something as simple as cable interference could heavily skew the results. 

 

 

 

 

 

 

[no_connection]

I think a lot of these tests kind of miss the point that, a certain bitrate will look fine until it does not. Flicking hair in 1% of the frame does not really stress the amount of data needed so it can be at almost lossless level, but when everything starts to move around it will at some point run into a brick wall, and if you have external recorder that can be set much higher. Think lossless.

On the flip side, where everything is moving and it's really pushed, so much will be happening anyway that unless you freeze frame, you will not ever manage to see it, provided the codec is sufficiently good at piking what parts to trash. I would assume someone have done studies and found what you can get away with for most usage and then used that to balance the bitrate.

 

[interceptor121]

1 hour ago, Dan Sherman said:

 

1.  If a device can keep up or not, depends on what preset you use. Even a camera could keep up with the lower end presets as they are not that intensive, but a camera for sure isn't going to be able to handle veryslow or placebo.
2. who said compress it over and over again? I said take clip A and trans code it to clip B using CRF mode.  Then take clip A again, and run NR on it before transcoding it to clip C. Then compare clip B & C. C will be smaller, because less noise means the codec can generate the same image quality with a lower bit-rate. In other words noise has a direct effect on how good a codec like  h.264  works.

 

 

Picture profiles effect hdmi output, if it didn't v-log wouldn't work. The question is how much of the picture profile is applied to the hdmi output.


As I said before, converting the internal data to be hdmi compliant has an effect on image quality. The wiki has a good overview of the actual spec. Keep in mind when you are doing transforms like this, rounding issues can cause problems as your camera isn't running at 64 or even 32 bit.
https://en.wikipedia.org/wiki/HDMI

 

The gh5 is most likely working in YUV/YCbCr internally, as that's what its internally recorded files are. HDMI out is most likely 10bit sRGB to maintain maximum compatibility. Prores will support  sRGB and Y’CbCr, but I don't know what color space recorders like the atomos line are using.

Thus you have at-least one color space transom and if you are unlucky maybe 2. You will have rounding errors, and thus image quality degradation.

 

Saying internal or external recording is better than the other based on a handful of non scientific tests is idiotic. Even if you had full transparency (Intellectual property level knowledge ) of what Panasonic and Atomos was doing, something as simple as cable interference could heavily skew the results. 

 

 

 

 

 

 

HDMI outputs sensor signal that is affected by debayering and aliasing but nothing to do with coding errors

The GH5 outputs REC709 and REC.2020 colour spaces otherwise would not support HDR so obvously it does not just output sRGB which is synonymous of rec709 8 bits

also colour spaces and bit depth are totally different things

in terms of colour depth the GH5 outputs 8 bits and 10 bits colour 

When you record output signal from the 150 mbps modes the camera sends YCB 422 10 bit colour signal to an external recorder colour space will be rec.2020 for HLG REC709 for the rest

if the camera was sending only srgb it would not record 10 bits at all 

In terms of saying if internal or external recording is better is then down to codecs which come after the uncompressed data clearly we need to assume panasonic is not producing faulty HDMI ports and you use decent cables

now I agree that this conclusion is not something I can make so I have taken it out from the post

 

now going back to wolfcrow point if you take the all intra out a static capture of 10 bit 422 should be 1.67 the size of 8 bit 420 but here is the opposite the 420 I frames are bigger than 422 because the 422 format spends data for motion interpolation in a less efficient manner

now can you see it or not and how will the footage withstand grading is a different story most people report similar performance between all intra and 150 422 If you believe panasonic claim that cabac makes compression 20-30% more efficient and precise this would make total sense and the gap between the two smaller

however not many compare 8 bit and 10 bit straight out of camera those who do say 8 bit is better and my calculations are aligned 

[jonpais]

Film isn’t meant to be looked at one frame at a time. ?

[fuzzynormal]

On 8/13/2018 at 7:06 AM, interceptor121 said:

real bad surprise with the 150 Mbps 422 10 bit codec.

It's good 'nuff for me.  

Serious question: are people complaining about this tech actually trying to make anything "for reals" with it?

If this is strictly a technical discussion, I'll disappear --but I'm operating under the assumption that some of us are actual creatives. 

Yeah, I'm feeling salty, but I do find it amusing that many shooters prone to making videos of municipal park leaves, or people walking through tourist destinations, care so much about the the barely discernible nuances and intricacies of codecs.

I don't know, maybe I shouldn't be bothered by it.  If people want to grab that tiger by the tail, go for it.  Who am I to say?

[KnightsFan]

2 hours ago, interceptor121 said:

now I agree that this conclusion is not something I can make so I have taken it out from the post

I appreciate your ability to take criticism! The post is much stronger without straying from the primary topic of comparing internal codecs to each other.

I think you make a solid argument, given your assumption of no motion error and Panasonic implementing their encoders perfectly. Both of those are big assumptions, but I can roll with them and follow your logic as it stands--and take the real world implications with a grain of salt because of those assumptions.

Although, if I'm being honest, some of your language is biased, which gives you less credibility. Saying "Generally there appears to be no benefit using the internal 422 10 Bit codec nor the 420 8 bit double frame rate due to the limitations of the GOP structure" is reasonable (again, given your assumptions), but the next part "here Panasonic has created a few options that to be honest appear more a marketing effort than anything else." is unjustified, I believe.

I know I'm editing without being asked to, which is probably annoying, but anyway that's my feedback. Thanks for taking the time to look into this topic and write about it.

 

1 hour ago, fuzzynormal said:

If this is strictly a technical discussion, I'll disappear --but I'm operating under the assumption that some of us are actual creatives. 

I'm pretty sure this is all strictly technical haha. Fortunately, nothing anyone says actually changes what your footage looks like, so if you're happy with it then that's all that matters to you as a creative.

As for the second part, I'd like to think I'm a creative person, but I do enjoy these technical discussions as well. You'll never hear me say I can't work on a project because technically Prores is better than AVC Intra, but I find it fascinating to learn about the differences.

[Dan Sherman]

1 hour ago, fuzzynormal said:

 

Serious question: are people complaining about this tech actually trying to make anything "for reals" with it?

In my opinion, a large number of creatives are malcontents by nature.

Give them a Gh5, and they'll complain because it's not as good as an Ursa. Give them an Ursa, and they'll complain because it's not as good as a vari-cam.  Give them a vari-cam, an they'll complain because it's not as good as a Red. Give them a Red and they'll complain because it's not as good as an Alexa 65.

They are always held back by the preceived limitations of their gear.

[webrunner5]

Well even 10 bit is not a lot to write home to mother about when it comes to damn good footage to shoot with let alone grade hard. But to the average person shooting even 8bit, or seeing it in 8 bit I doubt most know or care.

Like has been said, hopefully you expose it right to start with and all is well.  

[Mark Romero 2]

On 8/13/2018 at 12:58 PM, Deadcode said:

In the last couple of month i started grading in ACES, and this is where i can feel the 100Mbps 8bit  from Sony is not enough.

Do you mean when shooting SLOG? Or with other gammas?

Bonus Question: Have you tried Resolve Color Management?

[Deadcode]

1 hour ago, Mark Romero 2 said:

Do you mean when shooting SLOG? Or with other gammas?

Bonus Question: Have you tried Resolve Color Management?

This belongs to another thread.

Yes, im using ACES and RCM side by side in post processing after shooting SLOG2. ACES gives me beautiful highlight roll-off as you can see on the attached image, RCM handles extreme color better without clipping.