Highlight Roll-Off: GH5, GH5s, A7III and X-H1 and ???

[mercer]

At second glance, I stand corrected. I thought he was doing a respective ISO test going from each camera’s base iso but he was actually doing an even iso test which is a strange DR comparison test. I think I am also tired. 

[webrunner5]

3 hours ago, jonpais said:

Apparently, dynamic range is directly related to bit depth.

I tend to agree with that statement. Pretty much the reason cameras that shoot Raw Video are usually 2 to 3 stops higher in DR when shooting Raw. Raw is usually @ 12 bit, or higher.

Even Photos have higher DR in Raw than JPEG. 14 bit is the average Raw in Photo these days.

[Emanuel]

@mercer LOL Better to keep that screening on hold... Hell of a curse! :-D

[mercer]

So it’s obviously a flawed test because the GH5 starts off at a disadvantage being a stop over its base ISO.

[Emanuel]

Mate, go figure if it was the GH5S... hehe

Never mind, convinced now. No publicity is bad. First video to watch back to work Tuesday morning :-D

[Robert Collins]

2 minutes ago, mercer said:

So it’s obviously a flawed test because the GH5 starts off at a disadvantage being a stop over its base ISO.

My feeling is that Log isos are fake.

I know for sure that with Sony while log states iso800, it actually records at its base iso of 100 (which is why you have to over expose +2 stops). I suspect that with Vlog while it states iso400, it actually records at the base iso of 200 (which is why it adjusts the metering one stop in camera.)

Having said that, the fact that the Sony has a base iso of 100 gives it a DR advantage at base iso over Fuji/Panasonic with a base iso of 200 because it allows the sensor to collect more light without clipping (much the same way that the D850 has a DR advantage over Sony at base iso because its base iso is 64.)

[mercer]

4 minutes ago, Robert Collins said:

My feeling is that Log isos are fake.

I know for sure that with Sony while log states iso800, it actually records at its base iso of 100 (which is why you have to over expose +2 stops). I suspect that with Vlog while it states iso400, it actually records at the base iso of 200 (which is why it adjusts the metering one stop in camera.)

Having said that, the fact that the Sony has a base iso of 100 gives it a DR advantage at base iso over Fuji/Panasonic with a base iso of 200 because it allows the sensor to collect more light without clipping (much the same way that the D850 has a DR advantage over Sony at base iso because its base iso is 64.)

Well wouldn’t the camera with a higher base iso have the advantage especially when considering lowlight and DR?

But I think overexposing sLog by 2 stops has to do with noise in the curve and not DR?

[Robert Collins]

2 minutes ago, mercer said:

Well wouldn’t the camera with a higher base iso have the advantage especially when considering lowlight and DR?

But I think overexposing sLog by 2 stops has to do with noise in the curve and not DR?

There is a trade off between your base iso setting between dynamic range and lowlight. The lower you set your base iso the more dynamic range you can have on the sensor because the more light you can get on it without clipping. However, the downside is that your lowlight performance will tend to be worse because you have to have more 'gain' to reach any given higher iso.

A good example of this is the D850 v A7riii

The D850 has better DR because of its lower base iso but the A7riii has better lowlight helped by having a higher base iso.

https://www.dxomark.com/Cameras/Compare/Side-by-side/Sony-A7R-III-versus-Nikon-D850___1187_1177

BTW just to show Sony's fake sLog iso...

I know the same applies to the A7riii.

In all likelihood Panasonic is doing the same thing with the GHH5/s.

[Mark Romero 2]

14 hours ago, Don Kotlos said:

With linear encoding like in RAW files, the bit depth determines the upper limit of dynamic range.  jpeg or log use non linear encoding which can map higher dynamic range in a limited bit depth. Thats how jpegs can typically have ~11 stops of dynamic range and log profiles >12. 

The problem is that the more you compress the dynamic range of an image into limited bit death, the more you under-sample the color space and the more limited tonalities & increased banding you get. 

Bitrate has nothing to do with dynamic range. 

OK, I can understand that. Although it brings up a question.

The (soft) 1080p of the D750 has a boatload of dynamic range at base ISO 100, and there doesn't seem to be many complaints about the color or the D750 footage (even though it is less than 50mbs and it is 8-bit 4:2:0), nor complaints about banding.

So how does the D750 pull off the high DR without ending up with colors that look like slog? (I honestly don't know.)

Is it because the lack of sharpness leaves "room" in the codec for more DR and for better colors? (Meaning, if it were 4K and had lots more "information" would the DR and the colors all take a hit???)

Is it just a matter of 8-bit, 4:2:0 being enough for 1080p but not enough for 4K?

[webrunner5]

12 hours ago, Robert Collins said:

BTW just to show Sony's fake sLog iso...

I know the same applies to the A7riii.

In all likelihood Panasonic is doing the same thing with the GHH5/s.

Well this Long article pretty much explains how S log works. This guy knows his stuff.

http://www.xdcam-user.com/2014/08/exposing-and-using-slog2-on-the-sony-a7s-part-one-gamma-and-exposure/

[Don Kotlos]

8 minutes ago, Mark Romero 2 said:

The (soft) 1080p of the D750 has a boatload of dynamic range at base ISO 100, and there doesn't seem to be many complaints about the color or the D750 footage (even though it is less than 50mbs and it is 8-bit 4:2:0), nor complaints about banding.

So how does the D750 pull off the high DR without ending up with colors that look like slog? (I honestly don't know.)

There are few reasons for that.

First, different input-output curves can compress different parts of the histogram. For example you can have the same dynamic range from HLG or Slog or even a normal jpeg gamma. Slog3 is also worse than Slog2 because it compresses a huge chunk of the histogram even more. 

Then you also have the color space. rec709 is a much smaller area than sgammut and in return you get better sampling & tonalities with rec709 that D750 has for the same colors (sky, skin..). 

Finally, compression (and spatial color undersampling) also affects banding but with the same codec there is not so much of a difference as you get with different color space or gamma. 

So Slog by itself is not problematic. Even slog with 8bits is fine as long as you don't use the original sgammut. For example, with the slog & 8bits you can get excellent results with no banding if you use a slightly reduced color space such as sgammut3.cine/pro/cinema (which are still much larger than rec709), and boost saturation (without any color clipping of course). The problem is that most Sony users don't know that. Even myself when I first got the A7rII, influenced by all the slog hype I set my camera to slog2/sgammut only to kick myself few weeks later for ruining the footage. Sony has allowed so much flexibility that is great for professionals, but they should have be more careful when explaining all these to the average consumer. 

[Mark Romero 2]

@Don Kotlos

Thanks for the explanation.

You mentioned getting good results using slog with a slightly reduced color space and boosting saturation. Should the boost in saturation be done IN CAMERA??? Or in post??? Or both????

[Don Kotlos]

In camera. But again, you have to be careful and not clip any colors and probably leave some extra space just in case. So depending on the scene you need different saturation values. For example with cloudy sky you can saturate more, but if you are shooting bright colored lights during the night you want to bring it down. 

So in post you would add minimal saturation and use rec709 LUTs for stylization. 

[Mark Romero 2]

@Don Kotlos

Thanks again!!!

[Asmundma]

Hi

Got the c200 a while ago. Is there any difference to the Sony A7r3 and 1dx2 which I also have. 

The short answer is YES. The image is simply better both in MP4 and for sure Raw. Shooting canonlog3 give quite much better DR and image then my Sony and 1dx2. 

Now I totally understand why people go for a cinema camera.

 

[ajay]

2 hours ago, Asmundma said:

Hi

Got the c200 a while ago. Is there any difference to the Sony A7r3 and 1dx2 which I also have. 

The short answer is YES. The image is simply better both in MP4 and for sure Raw. Shooting canonlog3 give quite much better DR and image then my Sony and 1dx2. 

Now I totally understand why people go for a cinema camera.

 

From the footage that I have seen out of the C200, I agree. The 8-bit files out of that camera appears to have better DR than the Sony A7 series, Canon 1dx2 and 5DIV, and the GH series of cameras. It's only 8-bit yet it has better DR than a 10-bit file out of the GH series of cameras. Edits very nicely from what I have seen.

On a similar note, Dave Dugdale's comparison between the A6500 and the GH5 shows that Sony's 8-bit Slog2 performs (in most cases) as well as Panasonic's 10-bit VLOG-L. No banding in the sky and can it doesn't fall apart in post-editing unless you really push it.

You have to believe the manufacturers are all manipulating their codecs so that they don't take away from their higher-end cameras. And this leaves me frustrated at the end of the day because until you run a camera through all of its paces, you can't rely on the specs.

[anonim]

1 hour ago, ajay said:

And this leaves me frustrated at the end of the day because until you run a camera through all of its paces, you can't rely on the specs.

I think that really there's no serious reason for frustration I've just enjoyed one beautiful samples of snowflakes etc. at 240p (!) which I'd like to try/share with you as consolation...

 

[DBounce]

I think DR is only one aspect of IQ. You simply cannot look at DR in isolation and ignore everything else. From what I have seen the GH5S represents the best combination of features combined with IQ of any of the currently available hybrid format cameras. Save for large stills it’s a top choice with robust codec, organic footage and forgiving low-light performance.

 

 

[salim]

On 4/2/2018 at 9:33 AM, Don Kotlos said:

The problem is that most Sony users don't know that. Even myself when I first got the A7rII, influenced by all the slog hype I set my camera to slog2/sgammut only to kick myself few weeks later for ruining the footage

@Don Kotlos - very interesting. As someone like myself that is looking into buying a A7iii or A7Riii I'm curious about what your learned and what is now your base settings. Would really appreciate to learn from what you've learned. Thanks! 

[ajay]

Salim...Not to speak for Don, but he did state the following in his post. Hopefully he can clarify further.

On 4/2/2018 at 11:33 AM, Don Kotlos said:

For example, with the slog & 8bits you can get excellent results with no banding if you use a slightly reduced color space such as sgammut3.cine/pro/cinema (which are still much larger than rec709), and boost saturation (without any color clipping of course).