GH5 Prototype

[Don Kotlos]

6 minutes ago, marcuswolschon said:

However PLEASE, never ever meassure dynamic range in bits.

You are correct of course. Bits are frequently used for dynamic range instead of stops wrongly, and that also confuses people with the color depth. 

[Cary Knoop]

58 minutes ago, Don Kotlos said:

No that is definitely not true. 10bits per channel offer a much denser space that can improve color reproduction and tonal differences while getting rid of artifacts like banding. Especially v-log with its

10 bits per channel only has a benefit if the noise floor is sufficiently low. 

Otherwise we just have more random bits which can be easily reproduced by simply dithering the signal from 8 to 10 bits.

And like the others stated the bit depth of a signal is not the same as the dynamic range.  However it is a limiting factor, if you want a full dynamic range of 14 stops you need a noise free minimum bit depth of 14 bits but you can certainly compress it to 10 or even 8 bits.

 

[liork]

Well, we all know it will be cleaner than GH4, as we see with the G85.

[Don Kotlos]

1 hour ago, Cary Knoop said:

10 bits per channel only has a benefit if the noise floor is sufficiently low. 

Otherwise we just have more random bits which can be easily reproduced by simply dithering the signal from 8 to 10 bits.

No that is not correct. You are confusing dynamic range with bit depth. While the bit depth can potentially limit the dynamic range the opposite does not hold true. Sure if you are at noise levels then bit depth might not offer an actual gain in the number of colors, but for the range of values that the noise is less than your signal (the ~12 stops with the GH4 for example) 10 bits per channel offer more color tones that you otherwise throw away. 

See the link in the original response for an actual example if you are still not convinced. If you still don't think there are significant differences then you should accept that others do find differences important enough and don't think 10bits is a gimmick with 12 stops of dynamic range. Again cameras with noisier sensors give better colors, and bit depth is one of the reasons. 

Quote

And like the others stated the bit depth of a signal is not the same as the dynamic range.  However it is a limiting factor, if you want a full dynamic range of 14 stops you need a noise free minimum bit depth of 14 bits but you can certainly compress it to 10 or even 8 bits.

I am not sure what you are trying to say here, but in the original response there was no confusion between bit depth and dynamic range which frequently but wrongly is expressed also in bits. 

[Cary Knoop]

17 minutes ago, Don Kotlos said:

No that is not correct. You are confusing dynamic range with bit depth. While the bit depth can potentially limit the dynamic range the opposite does not hold true. Sure if you are at noise levels then bit depth might not offer an actual gain in the number of colors, but for the range of values that the noise is less than your signal (the ~12 stops with the GH4 for example) 10 bits per channel offer more color tones that you otherwise throw away. 

See the link in the original response for an actual example if you are still not convinced. If you still don't think there are significant differences then you should accept that others do find differences important enough and don't think 10bits is a gimmick with 12 stops of dynamic range. Again cameras with noisier sensors give better colors, and bit depth is one of the reasons. 

I am not sure what you are trying to say here, but in the original response there was no confusion between bit depth and dynamic range which frequently but wrongly is expressed also in bits. 

10 bits as opposed to 8 bits does not give any guarantee of better quality video, it depends on the noise floor.  You can sample those pixels on the sensor with 1000 bits of resolution but it won't do you any good, it is limited by the noise floor.

It is the same as with sound, you can record at 16 or 24 bits but if the noise level is not low enough 24 bits does not gain you anything.

It seems we simply have to agree to disagree with it because you think what I am saying is wrong.

 

 

[webrunner5]

I am not too sure noise has that much of an effect with less or more color. It is just harder to see it for the noise! 8bit is limited to so many colors, and 10bit has a ton more colors. Sure both will be noisy if the sensor sucks, but there is still a LOT more data to deal with for 10bit compared to 8bit. The noise floor is lots lower now with modern sensors that unless you are pushing ISO way up on small sensors, having a fast lens solves even those problems a lot, and you are allowed to add lighting!

[marcuswolschon]

You are still  confusing a lot here.

Take a well lit scene with some shadowy parts.

The darkest 5% of the image will be noisy (unless you add more light or open the aperture or increase shutter angle and are thus able to reduce gain/ISO).

The brightest 1% has one or two colour channels clipping but not yet the third one.

Dynamic range tells you how dark the black (with details remaining) and how bright the white (without clipping) can be.

 

Going from 8 bit 4:2:2 to 10 bit 4:2:2 doesn't change any of that.

 

It DOES change that you suddenly can distinguish 3 additional tones for each of the 3 colour channels. (That's not 3*3*3 but 3*2^10 - 3*2^8 additional colours.)

* So a very gradual fog or sky will have less visible steps.

* You may get a much better green screen (although that's mostly by going from 8bit 4:2:0 to 10bit 4:2:2).

The darkest 5% of the image are still noisy. That doesn't change.

The brightest 1% still clip one or two colour channels.

What does change is that the 94% of the image that are well exposed suddenly resolve a lot more colours that have formerly been indistinguishable.

What does change is that any colour grading that expands a part of the brightness, saturation and/or tonal spectrum will have a lot more steps to work with and can thus be done more boldly without introducing ugly steps in gradients. This is important to a lot even when delivering the final result in 8bit 4:2:2 .

These people are already doing 10bit recording with the existing GH4/GH4R and they do it for very good reasons. So obviously it is important even without increasing dynamic range and/or decreasing thermal noise from what the GH4/GH4R offers. They just can't do it INSIDE the camera yet and have to use external recorders. Having a clean 4:2:2 10bit output was one of the major improvements of the GH4 and GH4R.

 

[Emanuel]

Well, I guess the whole point here is (also) the standpoint that the new model upgrade shan't be the mere introduction of 10-bit 4:2:2 internal recording as 'the' new feature corpse. There must be something more to add it up.

Fellas, in a way or another, he is right :-)

[Cary Knoop]

1 hour ago, marcuswolschon said:

It DOES change that you suddenly can distinguish 3 additional tones for each of the 3 colour channels. (That's not 3*3*3 but 3*2^10 - 3*2^8 additional colours.)

* So a very gradual fog or sky will have less visible steps.

* You may get a much better green screen (although that's mostly by going from 8bit 4:2:0 to 10bit 4:2:2).

The darkest 5% of the image are still noisy. That doesn't change.

The brightest 1% still clip one or two colour channels.

What does change is that the 94% of the image that are well exposed suddenly resolve a lot more colours that have formerly been indistinguishable.

What does change is that any colour grading that expands a part of the brightness, saturation and/or tonal spectrum will have a lot more steps to work with and can thus be done more boldly without introducing ugly steps in gradients. This is important to a lot even when delivering the final result in 8bit 4:2:2 .

These people are already doing 10bit recording with the existing GH4/GH4R and they do it for very good reasons. So obviously it is important even without increasing dynamic range and/or decreasing thermal noise from what the GH4/GH4R offers. They just can't do it INSIDE the camera yet and have to use external recorders. Having a clean 4:2:2 10bit output was one of the major improvements of the GH4 and GH4R.

 

10 bits of data obviously can resolve more than 8 bits but just because you output 10 bit does not guarantee it will be any better.

When photons fill the well of a pixel a particular value is read.  The AD bit depth is no measure of the accuracy of the value.  Compare it to using a bad lens, if the bad lens can only resolve 2M pixels having a 36M pixel camera is not going to do you any good.  By the way raising ISO above base ISO does not reduce noise as you seem to suggest.

There are basically four kinds of noise due to:

1. The discrete nature of light ("the less light the more noise")

2. The limited efficiency of the sensor to trigger an electron to react to a photon ("% of photons not catched by the sensor").

3. The noise of the sensor ("temperature and other things").

4. AD quantization error.

In addition while noise appears at all luminance levels more or less evenly (except for item 1) for rec 709 or 601 like outputs it appears stronger at lower levels because the output is converted from linear to gamma curved output.

 

[marcuswolschon]

Quote

Please acknowledge there there are multiple, orthogonal metrics that can be "better".

Noise is one of them.

Dynamic range is one of them.

Framerate is one of them.

Number of Colour and subsampling are some.

Codes and Bandwidth are some.

 

[Emanuel]

1 hour ago, Cary Knoop said:

(...)

 if the bad lens can only resolve 2M pixels having a 36M pixel camera is not going to do you any good.

(...)

Well, well. I recall when people were used to infer VHS would limit 35mm footage to 260 lines of resolution as same as any VHS camera back to those 80s. Or as you wish, why to use 4K recording if/when the content is displayed on 1080p. However, we all know the wonders the Nyquist-Shannon sampling theorem has to show on it.

[Cary Knoop]

17 minutes ago, Emanuel said:

Or as you wish, why to use 4K recording if/when the content is displayed on 1080p.

The content may be (predominantly) displayed on 1080p devices now but obviously content providers make footage not just for viewing in the present time only. 

 

[Emanuel]

2 hours ago, Cary Knoop said:

The content may be (predominantly) displayed on 1080p devices now but obviously content providers make footage not just for viewing in the present time only. 

 

Right. On 4K side, a sort of 35mm standard. But also mainly addressed to the distribution channels our whereabouts are home-based and more internet-centered and focused on Pay-TV, VOD and streaming à la Netflix, more than ever before. Media convergence rules.

I'd even dare to actually state 4K is already now. Even though to still be seen widely spread as broadcasting content inasmuch as mainstream today or for future to come. Time will tell and soon. Once 1080p displays are obsolete. Price policies dictate the market. Faster than the advent of HD, we all can count on it.

E :-)

[Emanuel]

Here's an interesting article on topic:

http://interactive.satellitetoday.com/via/april-2016/pay-tv-operators-evolve-to-survive/

4K HDR TV sets, as for instance, are a curious 4K 2.0, not exactly at sea level, IMO.

(PS:) ...and also why your point is pretty accurate, as matter of fact.

[liork]

Pansonic write in their website:

"6K PHOTO" is a high speed burst shooting function that cuts a still image out of a 4:3 or 3:2 video footage with 18-megapixel
(Approx.6000 x 3000 effective pixel count) that 6K image manages.
The name '6K PHOTO' is tentative.

So, do you understand from that the senser has 6000 pixels width?

[Rinad Amir]

Any news on slomo capability?

[liork]

Nothing yet about slomo.

[Andrew Reid]

8 hours ago, liork said:

Pansonic write in their website:

"6K PHOTO" is a high speed burst shooting function that cuts a still image out of a 4:3 or 3:2 video footage with 18-megapixel
(Approx.6000 x 3000 effective pixel count) that 6K image manages.
The name '6K PHOTO' is tentative.

So, do you understand from that the senser has 6000 pixels width?

People were referring to an 18MP or 20MP sensor, or even the 20MP sensor already out in the Olympus E-M1 II as being capable of 6K because they read this on the Panasonic website about 18 megapixel = 6K

I suggest you check the aspect ratio of 6000 x 3000, which is indeed 18MP but heavily letter boxed compared to what is needed for 3:2 or :4:3 photography.

In actual fact you will need to go to at least 24MP in total for 6K video AND 3:2 photos.

Very few photographers will be ok with a 2:1 aspect ratio! So I find it highly unlikey the sensor will be a total of 6000 x 3000 in such a wide aspect ratio as it is also sold Panasonic's flagship 4:3 stills camera! If take the Panasonic 6K photo press release info literally you will be mislead as to the true nature of the sensor and megapixel count.

[Vesku]

Very confusing this 6k mess.

I hope Panasonic finally leaves the stupid idea of not showing exposure info in video modes. They should also add more customation because GH4 has so much limitations and bad ergonomic/functional design. In general it is very good but it has many wrong thought little details which makes using video difficult.

[Neumann Films]

8 hours ago, Rinad Amir said:

Any news on slomo capability?

I'm sure it will be good news, when it's announced.