Jump to content
Interceptor121

Is the GH6 a worthy upgrade for a GH5/GH5s/GH5M2 user

Recommended Posts

For underwater use, the short battery life is a real issue with the current generation of houses.

But if that wasn't enough, it seems that white balance performance is nothing to write home about. No better than the Nikons and has been leap-frogged by Sony's latest (A7S III/A1/A7 IV)

https://www.thedigitalshootout.com/bonaire-2022/gear-tests/Panasonic-Lumix-GH6-Camera-Underwater-Review.php

Share this post


Link to post
Share on other sites
For underwater use, the short battery life is a real issue with the current generation of houses.
But if that wasn't enough, it seems that white balance performance is nothing to write home about. No better than the Nikons and has been leap-frogged by Sony's latest (A7S III/A1/A7 IV)
https://www.thedigitalshootout.com/bonaire-2022/gear-tests/Panasonic-Lumix-GH6-Camera-Underwater-Review.php

White balance under 50 feet? Is that a joke?
There is no red color left at those depths the camera may produce something but with an RGGB bayer sensor is likely to be off whatever the case
Some canon cameras white balance deeper that doesn’t make the colours any good
Am really not sure about the understanding of sensor construction some of those people have
Nevertheless the housing of the GH6 costs more than sony full frame
Not sure is nauticam is secretly supporting sony but today their only housing that are competitive are sony and Olympus the rest is way over what anyone would want to spend
I see how second hand market is the way to go assuming you find anything you like


Sent from my iPhone using Tapatalk

Share this post


Link to post
Share on other sites

Notwithstanding that there's no reds to speak of, if you look at the Digital Shootout videos from the Sony and Canon cameras, they still produce more 'neutral' looking results at depth than the GH6. 

In my view, underwater white balance for video is not really about reintroducing reds and yellows that aren't there at depth.. It's about creating a neutral rendering where shades of grey remain un-tinted and the water column stays blue (even if desaturated), not purple. When filming below 50 feet with natural light on the Canon and Sony cameras, you end up with a desaturated but neutral image dominated by shades of grey and brown with a desaturated (but still recognizably blue) water column.

Is it anything like shooting with artificatial light? Nope. But I rather like the desaturated grey-brown-blue color palate. 

  • Like 2

Share this post


Link to post
Share on other sites
11 hours ago, dreifish said:

Notwithstanding that there's no reds to speak of, if you look at the Digital Shootout videos from the Sony and Canon cameras, they still produce more 'neutral' looking results at depth than the GH6. 

In my view, underwater white balance for video is not really about reintroducing reds and yellows that aren't there at depth.. It's about creating a neutral rendering where shades of grey remain un-tinted and the water column stays blue (even if desaturated), not purple. When filming below 50 feet with natural light on the Canon and Sony cameras, you end up with a desaturated but neutral image dominated by shades of grey and brown with a desaturated (but still recognizably blue) water column.

Is it anything like shooting with artificatial light? Nope. But I rather like the desaturated grey-brown-blue color palate. 

White balance is a multiplication it simply takes the reading of RGB and applies a multiplier

The issue is that when a value is zero even multiplied it is still zero

When it comes to blue which are prone to clipping you decrease saturation yourself and then you get the effect you want if that is what you want 

I don't know how those people make those tests but if you test one in log or raw the others have to be tested the same

In VLOG you do get that desaturated look that you talk about again those 'tests' do not explain anything of what they do and as such they are not test they are just opinions

The issue is of course the fact that conditions can't be reproduced unless you dive with the two cameras. What does 15 feet colors mean one day is not another

Those short articles are really underwhelming they may as well not post anything

Share this post


Link to post
Share on other sites

Thanks for the (as usual) informative comments. I am wondering if white balance is purely a linear transformation on the RGB, or whether some (possibly contextual, in the case of AWB) non-linear model can be used, introducing some bias? I expect if some camera company uses some ML or AI model, they can introduce a bias variable?

Thanks,

Ajay

Share this post


Link to post
Share on other sites

I have to agree with Interceptor on this one when it comes to testing.

Any meaningful comparative result can only be made under the same conditions.

That means you all the cameras with you on the same dive and you shoot on the exact same conditions otherwise is not objective.

But I agree with dreifish that Sony has moved a lot in terms of white balancing underwater. That being said I am pretty happy with the GH5s and the result are comparable to an A7s3

  • Like 1

Share this post


Link to post
Share on other sites
Thanks for the (as usual) informative comments. I am wondering if white balance is purely a linear transformation on the RGB, or whether some (possibly contextual, in the case of AWB) non-linear model can be used, introducing some bias? I expect if some camera company uses some ML or AI model, they can introduce a bias variable?
Thanks,
Ajay

The sensor conversion of photons into electrons follows a linear model
Which means each pixel is read and stored for what it is and the values of R G1 G2 and B are set
At this point a white balance read applies a coefficient to each of those to bring it to a ‘correct’ value however if there is no red left at all in the scene the red channel will stay low and brining it up will only add noise
On the other hand you can decrease the value of blue and green channel but ultimately they will stay high
This is why when you white balance at depth there is a washed out look you may as well desaturate the blue and green channels yourself
The most critical part is to avoid clipping of the blue channel as other wise you may also have banding or other nasty effects
I looked into this few years ago and had a very detailed discussion with daniel keller at keldan
Filters reduce the amount of light true but mostly avoid the blue channel to get too high and therefore make white balance more effective
A 2 stop filter reduces light 75% and 4 stop filter 93.75% but works mostly on the blue and red channel putting things in check and allowing a camera to do a better job at white balance
My approach is to use a filter until I see that I am needing to go too high with ISO. Generally I avoid using the camera without filters unless am doing macro and using 100% lights or am at the surface


Sent from my iPhone using Tapatalk
  • Like 1
  • Thanks 1

Share this post


Link to post
Share on other sites
On 8/3/2022 at 5:11 AM, Interceptor121 said:


The sensor conversion of photons into electrons follows a linear model
Which means each pixel is read and stored for what it is and the values of R G1 G2 and B are set
At this point a white balance read applies a coefficient to each of those to bring it to a ‘correct’ value however if there is no red left at all in the scene the red channel will stay low and brining it up will only add noise

 

Massimo, I imagine your explanation of how the sensor readout is amplified by differing coefficients for R/G/B in order to achieve a custom white balance is technically correct. However, there's two assumptions I would question:

1. There's no red light to speak of below 5 or 10 meters, and thus multiplying 0 by whatever still = 0. I believe that actually the intensity of red light is reduced in some logarithmic fashion. So even at 15 or 20 or 30 meters there will be /some/ red light picked up by the sensor, though many stops reduced.

2. Different manufacturers apply the same amplification formula. I would not be surprised if different camera makers allow the amplification coefficient to be higher or lower, leading to different outcomes.

Finally, even if the CWB function is implemented in exactly the same way by all the different manufactures, it's only one step in the image processing pipeline. The 'color science' each manufacturer chooses also has an impact on the final result. By this I mean that manufacturers aren't trying to give you an accurate renditions of the real world colors. Each has adopted certain deviations in their standard profiles intended to achieve a more 'pleasing' look to the footage. Even in their log profiles/wide color gamut, they've made different choices on how to record the color space. Not to mention the profiles bake in different degrees of contrast and saturation.

Generally, these choices are optimized for making land-based footage look more pleasing to their intended audience. When applied to underwater footage, these choices can lead to odd results, some serendipitously good (as with Canon and the newer Sony cameras), some pretty awful (the older Sony cameras). 

This is I think what ultimately accounts for Canon's pleasing rendition of color in underwater footage -- the deviations in the EOS standard picture profile from what would be a technically "accurate" rendition of the scene colors. Or perhaps it's allowing a wider range for the R/G/B amplification coefficients. Or a combination of the two. I have no idea what the 'secret sauce' is, but I can confirm that the ultimate result is nice, neutral, and very pleasing for skin tones.

Here's an example from the Canon R5 C at 10 meters with ambient light only and CWB done off the sand. 

1126085555_CanonR5CCWB_1.2.1.thumb.jpg.2437188e68231fe773f24b7545b9f04c.jpg

On 8/3/2022 at 5:11 AM, Interceptor121 said:


On the other hand you can decrease the value of blue and green channel but ultimately they will stay high
This is why when you white balance at depth there is a washed out look you may as well desaturate the blue and green channels yourself
The most critical part is to avoid clipping of the blue channel as other wise you may also have banding or other nasty effects
I looked into this few years ago and had a very detailed discussion with daniel keller at keldan
Filters reduce the amount of light true but mostly avoid the blue channel to get too high and therefore make white balance more effective
A 2 stop filter reduces light 75% and 4 stop filter 93.75% but works mostly on the blue and red channel putting things in check and allowing a camera to do a better job at white balance
My approach is to use a filter until I see that I am needing to go too high with ISO. Generally I avoid using the camera without filters unless am doing macro and using 100% lights or am at the surface
 

How the CWB algorithm handles the blue and green channels and clipping may also differ from manufacturer to manufacturer, and thus the need for a red filter. So far, I haven't seen any point or need with the new Canon R5 C. Whereas a red filter certainly was necessary below 8m with the GH5 to achieve best results.

Share this post


Link to post
Share on other sites
49 minutes ago, dreifish said:

Massimo, I imagine your explanation of how the sensor readout is amplified by differing coefficients for R/G/B in order to achieve a custom white balance is technically correct. However, there's two assumptions I would question:

1. There's no red light to speak of below 5 or 10 meters, and thus multiplying 0 by whatever still = 0. I believe that actually the intensity of red light is reduced in some logarithmic fashion. So even at 15 or 20 or 30 meters there will be /some/ red light picked up by the sensor, though many stops reduced.

2. Different manufacturers apply the same amplification formula. I would not be surprised if different camera makers allow the amplification coefficient to be higher or lower, leading to different outcomes.

Finally, even if the CWB function is implemented in exactly the same way by all the different manufactures, it's only one step in the image processing pipeline. The 'color science' each manufacturer chooses also has an impact on the final result. By this I mean that manufacturers aren't trying to give you an accurate renditions of the real world colors. Each has adopted certain deviations in their standard profiles intended to achieve a more 'pleasing' look to the footage. Even in their log profiles/wide color gamut, they've made different choices on how to record the color space. Not to mention the profiles bake in different degrees of contrast and saturation.

Generally, these choices are optimized for making land-based footage look more pleasing to their intended audience. When applied to underwater footage, these choices can lead to odd results, some serendipitously good (as with Canon and the newer Sony cameras), some pretty awful (the older Sony cameras). 

This is I think what ultimately accounts for Canon's pleasing rendition of color in underwater footage -- the deviations in the EOS standard picture profile from what would be a technically "accurate" rendition of the scene colors. Or perhaps it's allowing a wider range for the R/G/B amplification coefficients. Or a combination of the two. I have no idea what the 'secret sauce' is, but I can confirm that the ultimate result is nice, neutral, and very pleasing for skin tones.

Here's an example from the Canon R5 C at 10 meters with ambient light only and CWB done off the sand. 

1126085555_CanonR5CCWB_1.2.1.thumb.jpg.2437188e68231fe773f24b7545b9f04c.jpg

How the CWB algorithm handles the blue and green channels and clipping may also differ from manufacturer to manufacturer, and thus the need for a red filter. So far, I haven't seen any point or need with the new Canon R5 C. Whereas a red filter certainly was necessary below 8m with the GH5 to achieve best results.

At 10 meters any camera works

White balance works like this Sum of RGB = 1

Initial coefficient a b c scene is aR + bR + cB = 1

if R is very small and the white balance operation changes a to a large number to compensate you get still a small number and the result is increased noise on the red channel

Camera manufacturer change the maximum amount of color correction that is possible hence the 10000 k or 15000 k limit although the camera may balance with higher 15000K the effectiveness of it may be reduced as noise can be very high

The rest which is really important is the photo style which is normally gamma compressed or log compressed together with bit depth and bit rate. If the photo style color science is not well implemented and the files are small they are harder to correct so having a better white balance in camera becomes more important

 

 

Share this post


Link to post
Share on other sites

Sponsors

Advertisements



×
×
  • Create New...