C300: Cinema 5D Test Claims 12.3 stops of DR for the C300 MKII!
- Thread starter puredrifting
- Start date
FrameFarmMedia
Veteran
Thanks for the links Dino. I hadn't seen the Boyle test before. The Radiant one I've seen but they messed up a few cameras pretty bad. Is the FS7 & A7S even shot in Log?
Dino
Veteran
Only the F55 looks like Log mode of the Sony's. The Radiant Images test shows the Mark II is a shadow detail leader, competitive with all in highlights except Arri. Geoff Boyle has tested most every camera for DR. When he says he's happy with a camera, I tend to take him at his word--he doesn't pull punches and the Mark II makes the cut.
i want hear more about C5D's Canon XFAVC codec impressions(as well as others), they seem to feel it's superior to XAVC. I would also like to have pro colorists grade the codec in 10 and 12 bit settings to see if highlight roll off is improved. This is where Arri still rules.
I think we'll be getting lots more feedback soon. My favorite reviewer is Adam Wilt, but don't know if he's doing the Mark II. Think I'll ask him!
FrameFarmMedia
Veteran
I also found that interesting as I had assumed that XFAVC and XAVC where just versions of the same code. Since they are both originated from the same base I'm curious to see what the differences are. I haven't compared data rates so I wonder if its just more data for less artifacts.
jm.gardiner
Well-known member
According to the BH pages. C300mk2 almost doubles the data rate of the FS7.
Joshua Cadmium
Well-known member
I don't have the time to really dive into it, but I just wanted to say a few things about this (and I have no connection to C5D.)
One, C5D is using a $3,500 DSC Labs Xyla chart which is probably the best test chart out there for Dynamic Range. The chart is also self lit. All they have to do is put it in a pitch black room, turn the thing on, and make one of the white chips just begin to clip on the camera. I don't think you can really argue with what the chart images show. The interpretation of those images are, of course, up for debate.
Second, they are using imatest to measure DR numbers from the chart data (http://www.imatest.com/docs/dynamic/). The whole 5D has more DR than the C300 is really due to the imatest software. imatest is using the standard formula for calculating DR, which needs noise to be preserved. So, it works best with raw, since noise is not obscured by further processing.
The formula for Dynamic Range is somewhat easy (see here: https://en.wikipedia.org/wiki/Signal-to-noise_ratio_(imaging) ). You just divide the full well capacity of the sensor (the amount of photons it takes before it clips) by the read noise (square root of the average variance from the background, AKA Root Mean Squared, AKA RMS). Then you take the 20 log of that number to determine the dB rating. Then you would basically divide that number by 6.06 to get the actual f stop number.
(Looking at an off the shelf sensor [in this case the one in the Blackmagic Pocket - http://www.fairchildimaging.com/files/data_sheet_cis_1910f_0.pdf] you can do the math: 30,000 full well / 1.2 RMS read noise = 25,000 ; 20log(25,000) = 88dB ; 88dB / 6.06 = 14.5 f stops of DR. But 14.5 stops might seem a little high for the Blackmagic Pocket. Blackmagic themselves only claim 13 stops. So it comes down to what is usable, which is going to be subjective.)
So, because read noise is measured as RMS, you are most likely going to get higher DR numbers if you are not dealing with the raw data coming of the sensor. A weaker codec combined with more noise reduction is going to get rid of the noise you need to properly calculate DR.
For video not in raw form, you are going to have a harder time calculating DR.
C5D is really using the right tools for the job here, though.
One, C5D is using a $3,500 DSC Labs Xyla chart which is probably the best test chart out there for Dynamic Range. The chart is also self lit. All they have to do is put it in a pitch black room, turn the thing on, and make one of the white chips just begin to clip on the camera. I don't think you can really argue with what the chart images show. The interpretation of those images are, of course, up for debate.
Second, they are using imatest to measure DR numbers from the chart data (http://www.imatest.com/docs/dynamic/). The whole 5D has more DR than the C300 is really due to the imatest software. imatest is using the standard formula for calculating DR, which needs noise to be preserved. So, it works best with raw, since noise is not obscured by further processing.
The formula for Dynamic Range is somewhat easy (see here: https://en.wikipedia.org/wiki/Signal-to-noise_ratio_(imaging) ). You just divide the full well capacity of the sensor (the amount of photons it takes before it clips) by the read noise (square root of the average variance from the background, AKA Root Mean Squared, AKA RMS). Then you take the 20 log of that number to determine the dB rating. Then you would basically divide that number by 6.06 to get the actual f stop number.
(Looking at an off the shelf sensor [in this case the one in the Blackmagic Pocket - http://www.fairchildimaging.com/files/data_sheet_cis_1910f_0.pdf] you can do the math: 30,000 full well / 1.2 RMS read noise = 25,000 ; 20log(25,000) = 88dB ; 88dB / 6.06 = 14.5 f stops of DR. But 14.5 stops might seem a little high for the Blackmagic Pocket. Blackmagic themselves only claim 13 stops. So it comes down to what is usable, which is going to be subjective.)
So, because read noise is measured as RMS, you are most likely going to get higher DR numbers if you are not dealing with the raw data coming of the sensor. A weaker codec combined with more noise reduction is going to get rid of the noise you need to properly calculate DR.
For video not in raw form, you are going to have a harder time calculating DR.
C5D is really using the right tools for the job here, though.
cpreston
Veteran
That's pretty bad. Like atrocious. Of course, that's just what youtube does with that type of footage, so who knows what to make of it?
What is the formula for calculation of the usable region based on the noise? IOW, what is the noise cutoff criteria?
Ebrahim Saadawi
Carbonite Member
PB 120p 2K samples are 10x higher resolution/quality.
remsyatassi
Well-known member
Yeah, I don't exactly think it's reasonable to make a purchase for a $16,000 camera based on one poorly shot image of waves crashing back and forth at who-knows-what compression.
Usable stops, according to Ansel Adams, are the stops with texture. So you cannot measure these stops with pure tones(as with the 5D test), since there is no texture. The 5d test is a measure of the absolute dynamic range. They get the camera to clip on the first chip and measure from there. That is why the Alexa was opened up an extra stop, it has about an extra stop of DR in the highlights. The test doesn't give over and under middle gray values.
Dynamic range measurement is relative. It is generally accepted that the Arri Alexa has 14 stops of Dynamic range. Using that as the measure then the FS7 and C300 mark II have about one less stop, or 13 stops of dynamic range.
Dynamic range measurement is relative. It is generally accepted that the Arri Alexa has 14 stops of Dynamic range. Using that as the measure then the FS7 and C300 mark II have about one less stop, or 13 stops of dynamic range.
Last edited:
The white sign with the stop value is causing horizontal cmos smear on the side of the female model from -3.5 stops to -2.0 stops(darkens background and models face). This seems like a bigger concern than dynamic range.
mraaron895
Well-known member
Seeing as how the sensor smear only shows up when the entire image is incorrectly exposed I wouldn't worry too much about it.
The sensor smear showed up in the cinema5d test as well. It shows up at 4 to 5 stops under middle gray. This may be a problem in high contrast scenes, such as a subject inside near a bright window, night near bright signs or lights ..etc
The red dragon has the same type of artifact.
Last edited:
How Canon Measured 15 Stops of Dynamic Range on the C300 Mark II
How Canon Measured 15 Stops of Dynamic Range on the C300 Mark II
How Canon Measured 15 Stops of Dynamic Range on the C300 Mark II
https://www.cinema5d.com/canon-measured-15-stops-dynamic-range-c300-mark-ii/
"In summary it looks like the testing methods of Canon and cinema5D are very similar if not mostly identical, with the only difference of where to apply the threshold in order to determine how far into the noise the dynamic range is counted. cinema5D’s evaluation software did not count the 13th stop, sitting inside the noise floor, as another stop of dynamic range, while Canon did, as they did for stop 14 and 15."
How Canon Measured 15 Stops of Dynamic Range on the C300 Mark II
How Canon Measured 15 Stops of Dynamic Range on the C300 Mark II
https://www.cinema5d.com/canon-measured-15-stops-dynamic-range-c300-mark-ii/
"In summary it looks like the testing methods of Canon and cinema5D are very similar if not mostly identical, with the only difference of where to apply the threshold in order to determine how far into the noise the dynamic range is counted. cinema5D’s evaluation software did not count the 13th stop, sitting inside the noise floor, as another stop of dynamic range, while Canon did, as they did for stop 14 and 15."
ReelWorksMedia
Well-known member
Thanks for posting this Terry. Thats exactly what I thought.
Dino
Veteran
I have to agree with C5D, those last stops are buried in the mud. I'm curious as to why the Mark II seems to be slower, with Thorpe quoting a base ISO of 400 for C-Log, when the Mark I was 850? It's supposed to have a much higher signal to noise ratio, but it seems like the preamps were gained up to make a base ISO of 800 for C-Log 2.