Home of Professional Cinematography since 1996
class="Paragraph" style="margin-bottom: 0">Sony
Gamma Builder For F900
class="Paragraph" Published : 26th January 2004
I saw this press release at
which that states that the Sony Gamma builder app CvpFileEditor is available for the F900.
so, I'm curious, what exactly can you do with this program,
and what are it's limitations? Could I, for instance, create
a gamma curve that would mimic a logarithmic file, like the
Cineon curve (or something of that nature that would give
a lot of steps to the shadows and not quite as many to the
highlights), so that when I color-correct in post and linearize
the curve, I get a really wide overexposure latitude (kind
of like the Viper)? And if this were possible and I did the
above to the gamma curve on the F900, does this effectively
increase the usable dynamic range of the F900, giving it something
like what the Viper is able to do (especially in overexposure
latitude), or am I basically going beyond the capabilities
of the CCD's at that point, and the results, when trying to
CC them, will be less than spectacular due to other problems
that this approach may incur? At this point right now, I'm
basically just curious what can and can't be done with this
program, and also wondering if it is indeed available for
the F900, or is it only available for the HDC-F950 (or whatever
the model name on that camera is) that records to the 4:4:4
Post Production Artist
Virginia Beach, VA
One possible problem with this approach is that if you try to squeeze too much exposure latitude into the 8 bit F900 HDCAM format, you may end up getting banding in the sky or other subtle gradients. Even in the mid tones.
Especially if the CC is heavily pushed.
But then again if it is recorded to film, the grain may dither the banding enough, for it not to be noticeable.
Worth testing though.
Digital Film Technician
Animal Logic Film
This is just a non-technical viewpoint, but I don't think playing with black gamma and knee compression can radically improve the dynamic range of the HD image -- just slightly improve it. For one thing, you're somewhat limited by the 8-bit recording format of HDCAM or DVCPRO-HD. And playing with these gammas is sort of like slicing up a pie into differently-sized slices -- you don't end up with more pie. Maybe a little. You just rob Peter to pay Paul for the most part, which is useful when you need more info in one exposure region than another.
Of course, given the limiting nature of 8-bit recording, it seems to me a good idea to work with the gamma and knee settings to squeeze a little more info into the tape -- every little bit helps when you start to color-correct later. On the other hand, I also have come to think that some people have placed TOO much importance on these menu settings, thinking that Master Gamma at "0" is going to look radically different than at "-20" (it matters more if you have to nail the look in the field rather than plan on color-correcting later.) The truth is that I think the 8-bit recording is the ignored elephant in the room, the major factor that determines image quality while people futz with fairly insignificant adjustments to their gamma, knee, etc. settings.
Since the Viper sends out a 10-bit signal, it seems to have a wider dynamic range even without image processing (in fact, most types of image processing throw away bits of information more than add any.) Anyway, it seems to me that HD camcorders have to get away from 8-bit recording first, then work on moving up from 4:2:2 minimum to 4:4:4 second, then take the next leap up in resolution third.
Cinematographer / L.A.
David Mullen wrote:
>Anyway, it seems to me that HD camcorders have to get away from 8-bit >recording first, then work on moving up from 4:2:2 minimum to 4:4:4 >second, then take the next leap up in resolution third.
Yes, 10 bit, log, uncompressed, 4:4:4 is the way to go.
Part of the problem with existing HD formats is the drastic filtering done to get the data small enough to fit onto variants of existing tape formats -- i.e. reducing the horizontal res of HDCam down to 1440 pixels, and sampling chroma at 3:1:1 -- meaning that if you lose the ability to do radical color correction in post in a clean manner.
But you know that...
How big would the throughput be on a 4:4:4, 10 bit with, say 3840x2160?
I figure Jeff would know this off the top of his head.
Writer / Filmmaker
Is it possible to download the gamma builder from the internet?
I looked at www.cinealta.com but I have only found the manuals.
As far as I know, you cannot change the latitude of the F900. The black and the white are of fixed distance, you can only change the behaviour between.
There are several issues which are getting very confused here.
Firstly CCD's are analogue devices and are therefore not subject to the bit depth equation. They do however determine the maximum dynamic and colour range available to the rest of the process.
The analogue charge from the CCD's is measured and digitised. This is followed by the signal processing where things like the gamma and knee adjustments are made. In the F900 I understand this is a 10 bit colorspace. After this the image is converted to 8 bit colour and sub-sampled before the compressor.
Also, we should be very careful about confusing bit depth and dynamic range. Yes they are related issues, but they are not the same.
10 bit will give a better rendition of a broad dynamic range and because of this it is usually fed with a very broad dynamic range. In 8 bit if you push too broad a dynamic range into the 8 bits then you can get banding when it is stretched back out.
However, it is quite possible to use the 10 bit processing in the F900's front end to control the shadow and highlight parts of the range before the image is squashed into the 8 bit range.
One thing I have discovered with the F900 is that it is possible to compact more detail into the shadow and highlight areas while keeping a fairly normal gamma slope in the middle of the range (eg.0.45) thus avoiding some of the problems of the 8 bit space.
Remember, bit depth affects the smoothness of the transitions in the tonal scale. Dynamic range is the product of the source (eg. CCD) and can then be reduced by subsequent processing. Sub-sampling is about the special resolution of the colour recordings not the range of colours or the dynamic range.
Ben "how do I get off this soapbox" Allan.