GH5 Prototype

[zetty]

2 hours ago, Cary Knoop said:

10 bit 4:2:2 is 8 times as much information that needs to be encoded!

How do you figure? 10 bit certainly isn't that much more from the point of view of how information is stored digitally -- and according to this even saves bandwidth for the same quality result: http://x264.nl/x264/10bit_02-ateme-why_does_10bit_save_bandwidth.pdf

And 4:2:2 would only double the vertical colour resolution, so it should only be a fractional increase?

[Cary Knoop]

34 minutes ago, zetty said:

10 bit certainly isn't that much more from the point of view of how information is stored digitally

I beg to differ. 

If we take a standard 8 bit 4:2:0 video and discard for the sake of simplicity video levels, transparencies and gamut constraints we have:

3 x 2^8 bits of information per pixel subsampled to 4:2:0 which is: 384 values per pixel

For 10 bit video this becomes:

3 x 2^10 bits of information per pixel subsampled to 4:2:2 which is: 2048 values per pixel

That's an effective factor of 5!

34 minutes ago, zetty said:

And 4:2:2 would only double the vertical colour resolution, so it should only be a fractional increase?

I would not call that fractional, it is twice as much as 4:2:0 in an absolute sense and it is 25% more with respect to the 4:4:4 baseline.

You seem to grossly underestimate the sparsity of 4:2:0

Your document talks about saving bandwidth when the destination is going to be 8 bit.

Now do encoders need 5 times as much bandwidth for 10 bit 4:2:2? 

Of course not, the more data the more effective the compression (one would hope) but I think it should be somewhere between 250 to 300 Mbps for H.264 and about half of that if it is H.265.

 

[zetty]

38 minutes ago, Cary Knoop said:

I beg to differ. 

If we take a standard 8 bit 4:2:0 video and discard for the sake of simplicity video levels, transparencies and gamut constraints we have:

3 x 2^8 bits of information per pixel subsampled to 4:2:0 which is: 384 values per pixel

For 10 bit video this becomes:

3 x 2^10 bits of information per pixel subsampled to 4:2:2 which is: 2048 values per pixel

That's an effective factor of 5!

I would not call that fractional, it is twice as much as 4:2:0 in an absolute sense and it is 25% more with respect to the 4:4:4 baseline.

You seem to grossly underestimate the sparsity of 4:2:0

Your document talks about saving bandwidth when the destination is going to be 8 bit.

Now do encoders need 5 times as much bandwidth for 10 bit 4:2:2? 

Of course not, the more data the more effective the compression (one would hope) but I think it should be somewhere between 250 to 300 Mbps for H.264 and about half of that if it is H.265.

 

Thanks for the insights.

From the point of view of how the data is stored, 10 bits is only 2 bits extra over 8 bits -- thus, while providing immensely more values to work with, it doesn't take as much space when in memory. So it doesn't matter if there are more possible values for each pixel, what matters is how much space does it take to store the actual value(s) for the pixel.

From your chart it appears that 4:2:2 indeed doubles the chroma information, however the amount of luma information still stays the same.

As about the compression doc I linked, it's talking about two things, first about 10 bit source with 10 bit compression being more efficient in attaining the same level of quality than having to pass through an 8 bit codec, and then about 8 bit source being more efficiently compressed in 10 bits due to higher threshold of acceptable error -- but not necessarily about 8 bit output. But indeed, it isn't saying much cause the colour resolution would be lost using an 8 bit encoder anyway, so it's unclear against what quality standard are they comparing this to.

I am by no means expert in video processing and codecs, just have a coding background (assembler) and those things interest me (even if not anally so ).

 

[Cary Knoop]

12 minutes ago, zetty said:

From the point of view of how the data is stored, 10 bits is only 2 bits extra over 8 bits -- thus, while providing immensely more values to work with, it doesn't take as much space when stored. So it doesn't matter if there are more possible values for each pixel, what matters is how much space does it take to store the actual value of the pixel.

I think you are wrong. 

It surprises me, with a coding background you should readily understand that 10bit data requires 4 times as much storage as 8bit data. .

10 bit requires 1024 distinct values as opposed to 8 bit requiring 256 distinct values.  How does that not take four times more space?

 

Edited to add: I am wrong, 10 bit is only 1.25 times the size of 8 bit!

 

[marcuswolschon]

2 minutes ago, Cary Knoop said:

I think you are wrong. 

It surprises me, with a coding background you should readily understand that 10bit data requires 4 times as much storage as 8bit data. .

 

?????

10 bit require 2 bit more storage space then 8 bit and it can encode a value range that is 4 times the size.

[zetty]

6 minutes ago, Cary Knoop said:

I think you are wrong. 

It surprises me, with a coding background you should readily understand that 10bit data requires 4 times as much storage as 8bit data. .

 

Maybe I am wrong in absolute data increase cause I don't really know what exactly goes into describing each pixel but certainly 10 bit value doesn't take four times as much storage as 8 bit value. Computer memory is measured in bytes, each consisting of 8 bits. You can read the whole value as well as specific bits, so to store a 10 bit value, you'd use a byte plus 2 bits of another byte, still having 6 bits of that byte available for the next value (of course, you have to keep track of where your bits are as there is no address in the computer memory for separate bits, only bytes). At most, if you just leave a full byte for the extra 2 bits, it will still take only the double of amount.

[Cary Knoop]

11 minutes ago, marcuswolschon said:

?????

10 bit require 2 bit more storage space then 8 bit and it can encode a value range that is 4 times the size.

Wait hold on folks, I apologize I am totally wrong on the bit size increase!

I just cleared up my brain fog!

Of course, while the information quadrupled the required space is only 10/8 more, a factor of 1.25!

 

[zetty]

Computer's binary ways can be confusing -- for a moment there I almost started to doubt myself

I also hope for a bitrate of 200-250mbps if it's x264 as well, should cover all the extra data and still be a decent increase in overall detail and motion smoothness over the GH4.

[studiodc]

You can actually count to 10 bits (or much higher, even) with just your 10 fingers !   https://en.wikipedia.org/wiki/Finger_binary

[webrunner5]

Biggest thing I could hope for from Panny on the GH5 is if they move to the newer H.265 format. That would really free up space for a looser, better Codec.

[Davey]

I am glad that's been cleared up. Thought for a moment we might be heading for a ten page debate similar to the senor crop factor / light gathering fiasco from last year.

This is the most excited I have been about a camera that I have no intention of buying since first taking an interest in video three years ago. The GH5 was on my immediate radar before settling for the a7sii and I might well have bought it had it arrived a year earlier, or a large sum of money I was owed had arrived a year late.

I like Panasonic and hope that the GH5 is a huge success for the company and for those who buy it.

Certainly, 4k 60p will become the standard for prosumer cameras from now on and I expect that the a7riii and a7siii will both need it in order for Sony to remain on the same pitch.

[jonpais]

H265 would be a great option since it will deliver the same image quality at half the bit rate.

[Emanuel]

32 minutes ago, jonpais said:

H265 would be a great option since it will deliver the same image quality at half the bit rate.

My tests haven't proved it at all, but... let it be. Maybe, that's me. I just don't find it in that way :X

[Kisaha]

There are some examples of the hacked NX at the NX sub-forum. Even Samsung's early (and rather "weak") H265 implementation, is great for live performances and other byte consuming jobs.

There is always the option of 160/180 even more, Mbps which grades better and holds more information. The codec is not that great though, because Samsung probably used a very basic one (do not forget that was back in 2014).

In anyway, H265 will be the future for the next multiple K generations (I think it supports up to 8K?) and that's it was created by the same H264 consortium.

[IronFilm]

7 hours ago, Davey said:

I am glad that's been cleared up. Thought for a moment we might be heading for a ten page debate similar to the senor crop factor / light gathering fiasco from last year.

This is the most excited I have been about a camera that I have no intention of buying since first taking an interest in video three years ago. The GH5 was on my immediate radar before settling for the a7sii and I might well have bought it had it arrived a year earlier, or a large sum of money I was owed had arrived a year late.

I like Panasonic and hope that the GH5 is a huge success for the company and for those who buy it.

Certainly, 4k 60p will become the standard for prosumer cameras from now on and I expect that the a7riii and a7siii will both need it in order for Sony to remain on the same pitch.

I half expect we will need to wait for the a7S/a7R mk4 before we see 4K 60fps and 10bit.

As the mk3 version from Sony I fully expect will skip one or both of those features.

[sanveer]

Like I said before the Olympus EM1 Mark ii and the GH5 will most probably share the Same Sensor. The agreement between them has prevented Olympus from using the full potential of the Video features, and Panasonic will probably reduce the burst speeds and other photography benefits of the Sensor.

I am guessing that the VLOG on the GH5 will also be vastly improved. The auto focus for video will also be noticeably improved.

http://www.43rumors.com/ft3-panasonic-gh5-has-20mp-mft-sensor/#disqus_thread

[webrunner5]

4 hours ago, IronFilm said:

I half expect we will need to wait for the a7S/a7R mk4 before we see 4K 60fps and 10bit.

As the mk3 version from Sony I fully expect will skip one or both of those features.

Yeah I really doubt the GH5 will have 10bit 60fps 4k either, will probably be 8bit for that. Might be able to external though?? It sounds like it is going to be hard to beat for the money, but I am just not into non camcorder bodies much these days. For stealth they are great, but to use them like they need to be they look like some Frankenstein contraption.

  The Canon C100, C300 is probably as good as it gets on the way video cameras ought to be designed for run and gun and semi pro stuff. Probably be my next purchase with a small external recorder doing 10bit ProRes for my main camera. My Panny G7 is good enough for having in the glove box sort of stuff for spur of the moment or casual things.

For as few of bodies Panasonic sells with the GH4 compared to the others they sell, I have no clue why they are sticking again to the same body shape. Just admit it is a Video camera and make it look like a Video camera! Hardly anyone buys a GH4 to take photos 90% of the time and videos 10% of the time. For 500 bucks more everyone would still buy one if it was way better shape and operation wise to use.

[liork]

43 minutes ago, webrunner5 said:

Yeah I really doubt the GH5 will have 10bit 60fps 4k either, will probably be 8bit for that. 

You dont have to guess, Panasonic official info says that. Up to 4K/30p is 10 bit internal. 4K/60p is 8 bit.

[zetty]

1 hour ago, webrunner5 said:

For as few of bodies Panasonic sells with the GH4 compared to the others they sell, I have no clue why they are sticking again to the same body shape. Just admit it is a Video camera and make it look like a Video camera! Hardly anyone buys a GH4 to take photos 90% of the time and videos 10% of the time. For 500 bucks more everyone would still buy one if it was way better shape and operation wise to use.

Many of us prefer that form factor and modularity -- in fact being a documentary filmmaker, I don't need any rigs at all, just use it on monopod or gimbal most of the time. Being inconspicuous often is the only chance to get your shots and it's always a convenience; plus size matters a lot for a single/small crew production, especially when shooting abroad.

Sure, there's no built in NDs, decent audio preamps and useable AF that would improve its run and gun capabilities, but that's not really related to form factor (although it is to the size). Handling wise, it's as convenient as it gets. My previous camera was EX1 and I've never looked back since getting GH4 in 2014.

There are plenty of camcorder-type video cameras, why do you want every one of them be like that?

[Neumann Films]

If you could see one test with the GH5, what would it be? Standalone test, as in, not comparing it to another camera (unfortunately).