Simon Shasha
Well-known member
Shane Hurlbut also added this in response to a comment asking about dynamic-range: "I felt the latitude was about 12.5 stops with Canon glass, if I had used the Leica’s maybe 13.5. This is an impressive camera."
Why the downer on the 1D C?
- Thread starter Graemsay
- Start date
cpc
Well-known member
What's that even supposed to mean? Makes no sense to me. The sensor records as much range as it can. Glass has nothing to do with its DR capabilities.
Also, good luck using 12.5 stops in 8-bit video. Even 10 is pushing 8 bits pretty hard.
cpc
Well-known member
Yes, lenses can change scene dynamic range before it reaches the sensor. They can't make a 12 stop DR sensor into 13 stop DR sensor. And even if the camera has 99 stops of DR, they can't make a 12.5 stop mapping into 13.5 stop mapping.
But yes, I guess he got a bit carried away.
You can put easy 13 stops form an Alexa in 8 bits. I am sure in a year there will be a possibility of 15 stops (that you can put in 8 bits, if you want). And the more useable DR you put in 8 bits, the less actual precision you have in the useable stops. You can't beat Math. I will stay with max 10 stops (and 7-8 useable) in my 8 bits, thank you.Noel Evans said:
squig
Veteran
I got my first Leica-r lens last year on Shane Hurlbut's recommendation (and a bunch of vintage glass freaks) and took a shot of the crap on my desk (it's a nicely lit desk) and then took a shot with a nikon ai prime....the difference was amazing! Contrast, colour, DR, 3D, sharpness, all wow. I replaced all my Nikkors with Leicas. Have I shot charts? no...what for? I don't make chart films. The Leicas just have the filmic characteristics that I like. The nikkors are great lenses and grade well but they don't have the magic fairy dust. If I wanted to shoot the real world I'd buy Canon L zooms.
If the 1D C has an extra 2 stops over the MKIII in the highlights then I'm a lot more interested in it especially if the street price is under 10 grand. the 60p 1080p is another nice addition. 500Mbit/s m-jpeg should look great too.
No glass doesn't add more dynamic range. However, what it does add is greater separation of values depending on how well coated it is. Zeiss glass can have up to as much as 7-8 different coats on each side of an optic within a lens, and some lenses have as many as 13 elements. This cuts down on diffraction and increases sensitivity to different color ranges in the spectrum. Which means that within the same spectrum of values between two different lenses, there may be more detail in one than another. Hence the term microcontrast, as opposed to just generalized contrast. Zeiss's strength is it's coating technology, but sometimes the quality of it's glass isn't great (look at it's lower end offerings on millions of budget cameras). Leica not only has good coating technology, but also their selection (and rejection) of glass is a high priority, because their market is always high end.
cpc
Well-known member
My ~$100 Contax Zeiss 50/1.7 leaves my Summilux 50/1.4 eating dust far behind in terms of both microcontrast and resolving power. Just sayin'.
Btw, separation loss is really a product of veiling glare and not microcontrast. But yes, veiling glare is to a large degree result of coatings (or rather, the lack thereof). Lots of people still prefer single coated lenses (especially for B&W photography) because of the scene DR control that they allow by limiting contrast. In a typical exposing-for-the-highlights approach this increases apparent shadow detail.
Noel Evans said:
Noel, the C300 is undoubtedly a great tool.
But Canon's log gamma (and how it is not really log) was already discussed in this very topic in some detail.
Though I never mentioned the Canon Log gamma setting, clearly whats been shown is numbers dont add up in real world performance.
FWIW my point isnt to try and argue you should use the Canon C300 / C500 / 1DC or any other camera for that matter. My point is, and this is not directed at your per se, but reading numbers online doesn't add up to seeing how it performs in your own hands.
maranfilms
Veteran
+1. I agree, if it did, none of us would bother with 8 bit cams unless we had to. I have heard several people say this cams 8 bit is like grading 12 bit. Don't ask me why, because the math doesn't add up. What does add up is the beautiful image. And at the end of the day thats all that really matters.
I have to say I thought Shane's comments were a bit silly. But then after thinking about it, this guy has shot about everything under the sun, and more. I'm willing to give him the benefit of the doubt and take his word for it. Now if it's proven to be hogwash, then his credability with me is nil, and I will think of him as Canon's trick. Only time will tell. I'm looking foreward to trying out this cam asaic.
cowpunk52
Veteran
Wow! I just revisited this thread for the first time in a week or so - crazy amount of misinformation and incorrect conclusions. Let me just go ahead and address a few: 1) the GH2 did not outresolve the C300 in PB's test; watch the test and you'll see. 2) C-Log is a true logarithmic codec, not Rec.709, it is however less flat than S-Log and, along with the camera's DSP & SNR, tuned specifically for maximum usable DR in 8-bit codec. 3) the C300 handles at least 12 stops of DR extremely well in it's XF codec, the 8-bit spec does not limit it's DR to under 10 in any way shape or form. I've no doubt about the accuracy & robustness 12 - 12.5 stops of DR in the 1D-C.
cpc
Well-known member
Well, you mentioned 12 stops from the c300. And the c-log is the only profile that gets you 12 stops from the c300, says Canon. Other cruves are up to slightly above 10 stops (300% DR in Canon terms).
Thing is, people tend to get subjective, suffer from wishful thinking and get carried away easily (Shane and his Leicas are the perfect example). Math never lies. FWIW, I'd rather trust numbers until I see a miracle with my own eyes. If I could get my hands on one of these I would certainly test for myself. Until then: samples, common sense, math and Canon white papers are a good start.
2) You say this like if there are two possibilities on Earth: a true log transfer curve (it is not a codec) and Rec709 transfer gamma. Well, guess what? Not so. I don't even need to go out of power-law gamma land to demonstrate that: every single camera curve that applies a knee (that's pretty much all cameras) is already breaking out of Rec709 spec transfer gamma. C-log is what Canon could get out of the camera without screwing tonal precision in the important stops up in the mids. Anyway, you just confirmed what I said a few pages back: it is not as flat as s-log (not that s-log is truly flat, either). It is a compromise. It is what Canon could stuff in 8 bits. Kudos to them for that.cowpunk52 said:
3) You can save the "at least". If Canon thought the sensor was up for more, they would have said it.
8 bits limit you to exactly 256 stops in all shapes and forms. 10 stops is what I like to have without sacrificing precision in mids. Read more carefully.
You can easily prove me (and the numbers) wrong. Considering that around 8 stops is what most cameras will chop and encode in 8 bits in order to deliver pretty pictures, you can underexpose a scene 4 stops and show me a pretty picture after bringing the image up and fixing contrast in post. THAT will mean the camera handles 12 stops "extremely well". Alternatively, just shoot a step chart and show me good tonal precision in the low stops.
I am not here to try to convince you of anything. Simply offering my own real world experience. Your opinion is yours to keep. I shoot professionally and I know what is working for me regardless of what numbers are thrown at me.
A TVC shoot I did was in a white cyc studio. I didnt light it entirely flat as the intention was to place someone unsure of the experience that was to follow. Gradients everywhere. Smooth from one end to the other. If my first AC from that shoot happens along here, I guarantee he'll back me up on that and some other over performance we've noted.
Cosimo Bullo
Veteran
It sure got close... and left a lot of good camera in the wake. And that was before Driftwood's latest masterpiece, Canis Magoris, the latest opf great hacks for the GH2.
cpc
Well-known member
I guess I have to repeat myself. I am sure the c300 is a great camera. And I find your experience valuable; we are not talking about convictions here.
Btw, gradients are not a good example. Gradients can only show artifacts when contrast is pushed in post back to normal and further. Any camera (no matter of the transfer curve and how much DR it encodes) uses the coding space continuously. This means that unless the codec is crap (like most consumer codecs: 4:2:0 subsampling + macroblocks and insufficient bitrate), one will never see smoothness problems in gradients in the unprocessed out-of-camera footage. In other words, nice out of camera gradients are a sign of nice codec quality, but don't carry information about tonal precision because there is no detail in gradients. Washing/protecting subtle detail is what can show issues/strengths in tonal precision. For example, a flat lit face. That's where one can see problems with true (i.e. flat) log curves that handle lots of DR in 8 bits. But I am sure the c300 won't show such problems (or they will be limited) because Canon are smart enough and don't use a very flat curve with the limited 8-bit precision. Which is my point really. From what I've seen, their curve lacks precision in the dark stops (which is a smart decision, I will repeat). Yes, these stops are probably down there and create the perception of rich blacks but they are not useable (as in moveable/stretchable/etc) DR.
A simple step chart shot would make all this debate redundant. I hope someone will do this eventually.
Last edited:
A lot of color correction programs are nowadays in 32-bit float mode. You can grade 8-bit material pretty well if it's shot well. A way bigger concern to image quality is the compression itself.
Judge for yourself. To my eye, I see the GH2 image resolving more detail than the C-300 did, especially when you factor in that the C-300 image was shot under contrasty sunlight ( high contrast gives the impression of more resolution than is actually there ) where the GH2 image was shot under flat overcast sky.
Canon C-300 : 400 Percent Blow-Up
Panasonic GH2 : 400 Percent Blow-Up
The crop is different... They should be framed exactly the same to compare resolution. Quite a lot of sharpening on the C300 image though, I usually see that much in GH2 shots but this is reversed. And it's kinda funny how people are making screencaptures out of vimeos heavily compressed images shot in completely different lighting and comparing them... That's kinda useless. Just shoot a testchart and be done with it.
...In this case it's not as bad as you might think, as the original footage has been artificially enlarged 400 percent in post, so the fine details won't be as effected by the Vimeo compression.
indiawilds
Veteran
These discussions remind me of a few memorable quotes:
Lies, Damn Lies and Statistics! - Mark Twain
......
"He uses Statistics as a drunken man uses lamp posts... more for support than for illumination" - Andrew Lang
Lies, Damn Lies and Statistics! - Mark Twain
......
"He uses Statistics as a drunken man uses lamp posts... more for support than for illumination" - Andrew Lang