Barry_Green
Moderator
Just to be pedantic, that's not color space. 4:2:0 is color sampling. The color space is Rec 709.
C100: Canon Log. Wow!
- Thread starter Erik Naso
- Start date
Barry_Green
Moderator
A standard step chart will not be of any use for testing gradient banding, that is definitely true. You need a chart with a true gradient band on it. The DSC Billups VF/X28 has two nice true gradient bands, one gray and one blue, which are useful for testing banding.
Older frontlit reflected charts are definitely limited in how much DR they can represent. But the newest category of self-illuminated backlit charts can well exceed cameras' abilities in respect to DR. DSC has backlit charts that can exceed 18 stops.
So, when testing these attributes, you definitely need the right tool for the job. Using charts that were designed for the original 5-stop video cameras, to test 12+ stop modern cameras, is definitely pointless...
cpc
Well-known member
You can see from the diagram that the WideDR has pretty much the same DR as Canon log. What is different is the DR distribution. While Canon log is near-flat (it is not flat in the blacks, which is really a smart thing to do for a 8-bit camera), WideDR keeps lots of resolution in the mids and compresses the highlights without clipping them. This is exactly what a good film transfer to digital does to the upper part of the curve. This way you retain good tonal precision in the mids (which means good non-plasticky skin), and get the highlights soft rolloff.
Using Canon log only makes sense if there is crucial detail in the highlights that you need to expand (that's rare, and usually means bad exposure), or if there is detail in the shadows you need brought up (the more common case). Even then, AVCHD is hardly the right choice. You need a low compression low contrast friendly codec, so better use an external recorder + 4:2:2.
Scott Freestyle
Well-known member
I would never shoot a flat profile using 8 bit. However, when I first saw these images on my iPad with the Wow statement, they looked impressive... On my monitor at home there is some artefating.. But that could also be further image web compression.
I will always shot 8 bit footage as close to what I want the final product to be, but if this codec is standing up to some rigorous grading... That's another big plus for me as I grade everything..
There is a lot to be positive about with this camera, I'll let all you guys work out it's idiosyncrasies and enjoy reading about them..Definitely will be having one early in the new year
I will always shot 8 bit footage as close to what I want the final product to be, but if this codec is standing up to some rigorous grading... That's another big plus for me as I grade everything..
There is a lot to be positive about with this camera, I'll let all you guys work out it's idiosyncrasies and enjoy reading about them..Definitely will be having one early in the new year
Yes, you're right.
morgan_moore
Major Contributor
I dont think film does that at all, because with a flat top, instead of plastic skin you get plastic sky - which if you (secondary) grade back (burn in) bands instantly
Film blows highlight, but it does it without artifacting on the transition to blown, another IMO widely misunderstood and undertested evaluation in checking out these cameras.
Most of them look fine under a simple situation like the OPs images, which are nice, but not a conclusive test.
Barry
Great and logical sounding tools you suggest - I agree with all you say about them!
S
Liam Hall
Veteran
You need to understand there is more compression in the highlights in log mode. It's not a true linear gamma curve. Everything above 60% IRE is compressed further than the mids and lows.
morgan_moore
Major Contributor
So?
Liam Hall
Veteran
So, (depending on your shot) you don't need to a full histogram.
morgan_moore
Major Contributor
Im not getting that - I mean surely you want to capture as much data as possible as this will ultimately lead to the best image or is that over simplifying?
To me DR seems to be a trade off for micro detail given a fixed 'bucket size'
No free lunch?
log works when the bucket size (bit depth) is bigger than the delivered bucket (monitor gamut) but when they are the same it has to be a trade off?
Some people like that trade - fair enough
S
To me DR seems to be a trade off for micro detail given a fixed 'bucket size'
No free lunch?
log works when the bucket size (bit depth) is bigger than the delivered bucket (monitor gamut) but when they are the same it has to be a trade off?
Some people like that trade - fair enough
S
Liam Hall
Veteran
There's more "bits" of data for each stop below 60%, as it's compressed above that point, so you are capturing more data using log. It's more about protecting highlight detail than else else. It's certainly not something to lose sleep over. Though, when shooting very fair skin it pays to keep exposure on the WFM below 50% IRE. I aim lower...
morgan_moore
Major Contributor
If its more compressed above that point you are capturing that data less well - so as far as I can tell you are not capturing more data, but different data
S
S
cpc
Well-known member
Not necessarily. This depends on the luminance of the sky and the knee point in the curve. A high enough knee will keep detail in high mids and only roll highlights. If you look at film transfers, that's what happens above mids. You need to commit to highlight compression at some point because of the limited tonal resolution of 10-bit files (in the case of the original Cineon/DPX scans), even though there is pretty decent resolution in higher densities of the film gamma curve. And because otherwise your image will look weird on a high gamma display such as a computer monitor (2.2 gamma) or digital cinema projection (2.6 gamma). Film scans do blow/compress skies if necessary. Better that, than plastic faces.
Liam Hall
Veteran
As I understand it. You are throwing away less data in the highlights, but the trade off is potential increase in the noise floor. The Abelcine advice I posted earlier is good advice.
morgan_moore
Major Contributor
It seem you are throwing away different data, not throwing away less data.
When you flatten a curve you reduce the ability to resolve between luminance levels.. which is throwing away some form of resolution.. data
If a scene has 300 steps of tone and you only have 255 levels to record it you must throw some tones out, just choose which ones, maybe blow the highs, crush the blacks, or do some weird s*%t that makes video look like video by tinkering in a region, in this case above 60
I guess people either see it or they dont
Everyone has their own preference and look
S
When you flatten a curve you reduce the ability to resolve between luminance levels.. which is throwing away some form of resolution.. data
If a scene has 300 steps of tone and you only have 255 levels to record it you must throw some tones out, just choose which ones, maybe blow the highs, crush the blacks, or do some weird s*%t that makes video look like video by tinkering in a region, in this case above 60
I guess people either see it or they dont
Everyone has their own preference and look
S
Last edited:
Liam Hall
Veteran
True. The point is people need to test. Test different gamma curves and exposure settings in real world scenarios - not just test charts - then spend some time in the edit/grading suite. I see a lot of C300 footage (and other cameras) with plastic skin and horrible compression artefacts in the highlights. Yet, this is easy to avoid.
Erik Naso
Veteran
The frame grabs are jpg. Keep that in mind. On my playback monitor it looks great. Showed it to a couple producers and they like it as well. But what do I know. I like to grade footage. I always have.
I have the C100 for a week and when I was reading the manual and found the log settings I was surprised and thrilled. Sorry I didn't go out and shoot the right sample for everyone. I was impressed with how easy it was to get a good looking image. Nothing more. Please disregard my happiness. Nothing more to see here.
I just wrapped a long all outside shoot in crazy contrasty conditions. Scrims, bounce boards and LED panels where a big help. I did shoot it in Log. Im happy I did. But what do I know.
I have the C100 for a week and when I was reading the manual and found the log settings I was surprised and thrilled. Sorry I didn't go out and shoot the right sample for everyone. I was impressed with how easy it was to get a good looking image. Nothing more. Please disregard my happiness. Nothing more to see here.
I just wrapped a long all outside shoot in crazy contrasty conditions. Scrims, bounce boards and LED panels where a big help. I did shoot it in Log. Im happy I did. But what do I know.
The frame grabs are jpg. Keep that in mind. On my playback monitor it looks great. Showed it to a couple producers and they like it as well. But what do I know. I like to grade footage. I always have.
I have the C100 for a week and when I was reading the manual and found the log settings I was surprised and thrilled. Sorry I didn't go out and shoot the right sample for everyone. I was impressed with how easy it was to get a good looking image. Nothing more. Please disregard my happiness. Nothing more to see here.
I just wrapped a long all outside shoot in crazy contrasty conditions. Scrims, bounce boards and LED panels where a big help. I did shoot it in Log. Im happy I did. But what do I know.
I for one am happy you posted the grab, but knew that the pixel peepers would pounce on it.
Would love to hear your thoughts on the C100 in your contrasty day shoot and some grabs if possible.
Erik, you're one of the few people I see on this forum making good consistent work with the AF and as I am also looking to upgrade to the C100 I'll value your thoughts and opinions very highly.
Cosimo Bullo
Veteran
I can't be the only one who wants to see a C100 vs C300 internal codec test. Yes, I know, the bitrate is doubled in the C300, but I'm just astounded by the quality coming from the 24mbps C100 codec and have a feeling we'll be surprised.
Yes, I too would love to see some non-compressed screenshots from such a comparison.