Fohdeesha
Well-known member
Thanks for repeating exactly what I said. The sensor is still clipping at the exact same level of light, a gamma curve just changes the way it is recorded.
Quality of EX1 and HPX500
- Thread starter Lor
- Start date
Thanks for repeating exactly what I said. The sensor is still clipping at the exact same level of light, a gamma curve just changes the way it is recorded.
Robert Ruffo
Well-known member
But... Not quite. The conversion from 12 bit color to 8 bit (which is then tough to grade) is based on those very gamma curves. If you lighten dark areas with a high curve in the low range, it will have better signal-to-noise (it will result in greater contrast in the range allotted to darks) and thus a lower noise floor with still usable retrievable details in dark areas - end result: more usable dynamic range in a true, practical sense, even though 000 black and 255 255 255 white remain based on the same light output (are thus unchanged) Who cares what the sensor sees? It's what you are able to record - which in this case is only an 8bit slice of what the sensor saw, a slice either conducive to seeing deep into shadows (and having more dynamic range in a practical sense) or not (where shadows must be crushed down as they contain too little retrievable info to be saved by some good NR).
I must also say your pompous condescending tone is not appreciated.
Last edited:
basspig
Veteran
In a way, us audiophiles think of gamma curves almost like dbx Type 1 compression/decompression--it extends the dynamic range of analog reel to reel tape from 63dB to 127dB. Compressing the highlights before recording to SxS and then mirroring that curve in post should have a useful benefit in noise reduction. Of course it works better if the bit depth is greater.
Fohdeesha
Well-known member
You're missing my point....the camera itself is only capable of around 11 stops of DR, in the very best of situations, this has been tested. With each gamma curve, the recording is clipping whites at the exact same ire. No change in curves allowed for a greater range of dynamic range to be captured, they simply allocate more recording depth to certain ranges of the luminance curve. What you say about rasing darks definitely stands true I would imagine in the real world, although I would assume the change to be minuscule at best, especially when dealing with -3db and high-light situations in the first place. What I'm saying is the 12 and 13 stops of DR some people claim to be getting is absolutely ridiculous.
ala basspigs response, I would compare gamma curves much more to an EQ, changing the response curve, therefore boosting certain frequencies, or levels of light, higher as to take them out of the noise you speak of, or roll them off, to give the effect of "soft clipping". although clipping nonetheless.
I don't mean to sound condescending or anything, I own an ex1r, and I love it. But I just get tired of coming onto forums and seeing all these threads about how "ooh this and that curve gives me so much more highlight range" or "this curve works so well under harsh light, it retains all the highlights!". That's just scientifically false. While some curves lower the exposure altogether, switching from one to the other might give the impression it's capturing more highlights, but once you adjust your exposure to compensate and get your shadows back, it's clipping at the exact same ire.
basspig
Veteran
Your example is an analogy for sharpening. EQ=Sharpening. It works on a particular band of frequencies, in this case, higher spacial frequencies. That might be comparable to Dolby B noise reduction in tape machines. Even better example would be the 50/75uS preemphasis/deemphasis used in FM broadcasting.
Changing the dynamic range over the whole spacial spectrum is akin to compression/expansion, like dbx did for open reel tape. The gamma curves fall into the category of dynamic range, not sharpening.
Robert Ruffo
Well-known member
No, not about clipping. You set your exposure to protect highlights, then you raise gamma and black gamma to get more differentiation between the darker tones in 8 bit. As such, your shadow areas will have better real info (you'll be able to better see a real difference between what was very dark, and very, very dark) simply because you stretched those areas of the curve using a high gamma before committing to an 8bit recording of your images. Dynamic range is all about noise floor. Period. Its measurement is, as such, subjective, as how much noise is acceptable? NR increases dynamic range a bit (in practice), as do gamma curves that optimize the 8 bit translation of the 12 bit info only the sensor sees. You don't have to adjust your exposure to get your shadows back, you just have to optimize the recording of their information via gamma curves that lift the 8 bit recorded values of near-black so that your NR can give you back what you need to rise above black crush. In Red, in-camera gamma is irrelevant because virtually all the info the sensor sees is recorded and can be fully affected in post.
This is fact, and more importantly, fact in practice.
Last edited: