GH2 Incoming Canon shooter: I think I want a GH2, but I have a question.

[Aki_Hartikainen]

Not to change the subject but this comment is total nonsesne.

Had the lens been better in HV30 with less chromatic aberration, I probably would not have switched and would still be filming with the HV30. Only using the HDMI output to record onto external recorder. It was the CA that forced me to switch cameras.

 

[Aki_Hartikainen]

Here is how GH cameras deal with the over exposure:

Values end at 100. If I lower the levels, nothing above 100 becomes visible. Whereas HV30 would have revealed values that were stored all the way to 108. Compression works so that the higher brightness levels have more possible values. Brightest f-stop gets 69 different values of 256, 2nd brightest 50 possible values and so on (for jpeg anyway). By recording values "hotter" Canon HV30 has more possible values to choose from when dithering to 8 bits.

This is the problem that we try to solve on the GH2. It records values less hot than HV30, thus having lower possible amount of levels to choose from. Hence banding and posterization, which was really not a problem with HV30.

 

[DPStewart]

I've seen some REALLY surprisingly good stuff shot with the HV30.
I was all set to get one a couple years ago...
Killer bang for the buck that you could run-and-gun with or point-and-shoot or whatever.

 

[Matthew P]

Aki, you might want to try "vivid" or "cine" film modes on the GH2, and expose so that nothing in the sky is clipping. You'll be surprised at how nice sky shots come out then.

 

[Aki_Hartikainen]

Thanks a lot for the suggestion, I will try those film modes next. I was spoilt by the HV30 that allowed some head room for the exposure, meaning that I did not need to be quite as careful with the exposure. I always knew something would show up from the high lights.

 

[Aki_Hartikainen]

If you look at this image, several things become clear:

Firstly, the scale reaches from -20 to 120. Not 0 - 100. Secondly, the steps should be equally spaced apart from each other, but they are not. I made the image in Photoshop so that I added always the same numerical amount of brightness, +5 or so. But on the scale that truly linear image appears to have larger jumps at the bright end.

This is how good the good engineers can sometimes be; that scale is made in the same way human eye reacts to changes in brightness. Human eye does not react to changes in brightness in linear fashion. We see changes more easily in the bright values. Hence the engineers created image compression so that bright values have more latitude. They have more possible values to choose from.

The conclusion can only be, that the hotter you can expose and record for the camera, the better quality you will get. Because there are more possible values to choose from when dithering to 8 bits for brighter values than darker values.

 

[Kholi]

The conclusion can only be, that the hotter you can expose and record for the camera, the better quality you will get. Because there are more possible values to choose from when dithering to 8 bits for brighter values than darker values.

Supporting that it's definitely better to go brighter versus darker in camera, especially putting it to the AVCHD (hacked or not).

Slightly underexposing has sort of been a norm but I'm not agreeing with it at this time. Maybe if we were working with 10 4 2 2.

And sorry guys, been sick the last week... awake now.

 

[Aki_Hartikainen]

That's why we need to focus more on the high light latitude of the GH cameras. Perhaps the hack department could help in this and see if there is a way to allow the camera to record values outside the 0-100 range, like the HV30 could do. By my calculations the HV30 could record values up to 108, and that has to be half stop or possibly one f-stop more high light latitude. The ability to record super-white values is more common in production video cameras, and I am sure Lumix conforms to 0-100 in-camera to avoid complaints from consumers for over exposed images, who are not specialists in post processing.

Those 100 - 108 values from HV30 had to be conformed to 100 maximum, but the difference is that I could do that in 32-bit color space, thus increasing quality and latitude of the end result.