Canon C300 Mark III Lab Test (Pre-Production): Dynamic Range, Rolling Shutter and Latitude

April 21st, 2020
Canon C300 Mark III Lab Test (Pre-Production): Dynamic Range, Rolling Shutter and Latitude

Shortly after the publication of our Canon C500 Mark II lab test last week, we learned that we will gain early access to a pre-production Canon C300 Mark III in our headquarters in Vienna. And of course we had to take an immediate “lab test” look at Canon’s latest Super 35 addition to their line of cinema cameras.

After 5 weeks of home office, leaving my space mainly for only a quick walk with the dog and food shopping, I had to jump at this chance. So I picked up my face mask, greeted Nino from a safe distance and after a quick coffee together (and washing our hands thereafter) we headed straight into this.

Please note that Canon mentioned this to be a pre-production unit, the firmware used was 1.8.1.1 – the final image may change when the camera is released. Nevertheless, we were super interested in getting a lab benchmark for this new camera so here we go!

Time was very limited, which is not the best precondition for a thorough lab test. However, as we just ran the C500 Mark II through the lab (see our article here) and we quickly saw that the C300 Mark III shares the body, menu structure and a lot of other features including codecs with the C500 Mark II, we were fast in setting up the cam and decided to focus on 10bit XF-AVC, 4K DCI image acquisition for now (internal raw recording mode is available as well).

Canon claims in their press release that this new Super 35 sensor in the C300 Mark III uses a dual gain circuit to capture increased highlight and shadow detail, which should translate into a very good dynamic range.

Dynamic Range of the Canon C300 Mark III at ISO800

As there was no specific information on the native ISO which should yield the best dynamic range results, we choose ISO800 to run our tests. As mentioned above, the 10bit internal all – I XF-AVC codec with Canon Log2 / C.Gamut (color matrix “neutral”, noise reduction “off”) at a resolution of 4096×2160 at 25 frames per second was used.

Canon claims their dual gain architecture to be active from 24 – 60 frames per second, therefore we tested 25, 60, 75 and 120 frames per second.

And here is the result for 25 -60 fps, see figure 1 below: a strong 12.8 stops at a signal to noise ratio of 2 is calculated. This dynamic range result is the same for 60 frames per second –  a really good result for a Super 35 sensor. For comparison, the BMPCC6K shows 11.9 stops in our tests, the FUJIFILM X-T3 11.2 stops.

Fig. 1: IMATEST dynamic range results for the Canon C300 Mark III at 25fps, ISO 800, CLOG2 / C.Gamut, 4K DCI. Very strong 12.8 stops at a signal to noise ratio of 2 are calculated.

The waveform plot of the Xyla 21 chart confirms this result, see figure 2 below:

Fig. 2: Waveform plot of the Xyla 21 stepchart for the Canon C300 Mark III at 25fps, ISO 800, CLOG2 / C.Gamut, 4K DCI. About 13 stops can be identified above the noise floor.

At 75 frames per second (4K DCI) the situation changes, as the dual gain architecture of the sensor is no longer active. This can be seen in the waveform plot in figure 3 below – the image becomes much noisier, and around 12 stops can be identified above the increased noise floor:

Fig. 3: Waveform plot of the Xyla 21 stepchart for the Canon C300 Mark III at 75fps, ISO 800, CLOG2 / C.Gamut, 4K DCI. The dual gain architecture is no longer active – about 12 stops can be identified above the noise floor.

IMATEST confirms this and calculates 11.6 stops (at a signal to noise ratio of 2, medium value in the graph below) – still, a very good result!

Fig. 4: Waveform plot of the Xyla 21 stepchart for the Canon C300 Mark III at 75fps, ISO 800, CLOG2 / C.Gamut, 4K DCI. The dual gain architecture is no longer active – 11.6 stops are calculated.

Rolling Shutter of the Canon C300 Mark III

As mentioned previously, we are using a strobe light at 300 Hz to calculate rolling shutter of CMOS sensors.

First interesting result – at 25 fps and up to 60 fps 15.6 ms rolling shutter are measured, see figure 5 below:

Fig. 5: at 25 frames per second up to 60 frames per second the rolling shutter in 4K DCI mode is 15.6 ms

This is a good but not outstanding value for the rolling shutter – other Super35 cameras like the FUJIFILM X-T3 are on the same level, the aforementioned BMPCC6K has 19.8ms.

Now, this camera also offers up to 120 frames per second in 4096×2160 – now how does this work? 15.6ms rolling shutter would not support higher frame rates than 60 fps.

Obviously, by eliminating the dual gain read out at 75fps and higher, the rolling shutter drops to 7.7ms, see figure 6:

Fig. 6: at 75fps up to 120 frames per second the rolling shutter in 4K DCI mode is 7.7 ms, as the dual gain readout is no longer active at these frame rates

Latitude (underexposure) Test of the Canon C300 Mark III at ISO800

As mentioned earlier, “latitude” is the capability of a camera to retain details and colors when over- or underexposed and pushed back to zero exposure.

In our Lab scene, the zero exposure level represents a base exposure of F4 at ISO800, shutter angle 360° for 25fps DCI 4K. The exposure level is checked by the waveform reading of my colleague Nino’s face which shows a maximum of 60% Luma value, the white sheet of paper comes in at 65%.

Then, exposure is dropped successively to 180°, 90°, 45°, 22.5° and finally 11.25° shutter angle – 5 stops underexposed. The resulting images are pushed back to the zero exposure level in postproduction using DaVinci Resolve Studio.

The good dynamic range results are giving already a strong indication that this test will go well with the C300 Mark III – spoiler alert!

Let’s have a look first at the zero exposure level:

Fig. 7: zero / base exposure level (graded) for the studio scene with the C300 Mark III at 25fps, 360° shutter angle F4

Let’s jump right ahead and have a look at the 3 stops underexposed image, figure 8 below:

Fig. 8: 3 stops underexposed and pushed back – the C300 Mark III at 25fps, 45° shutter angle F4

Not much to see in figure 8 above, right? The C300 Mark III behaves very well at this exposure level. The image becomes a bit grayish, and some red speckles of noise start to appear (look at the shadow area below Nino’s cheek). Those speckles can be easily removed by just a bit of temporal noise reduction in post.

Fig. 9: 4 stops underexposed and pushed back – the C300 Mark III at ISO800, 25fps, 22.5° shutter angle F4

A similar observation holds true for 4 stops of underexposure pushed back – figure 9: the noise in the shadows increases but can be removed quite effectively by just a bit temporal noise reduction. The noise is finely dispersed, like film grain, which looks really good (and is easily removed by noise reduction).

Fig. 10: 5 stops underexposed and pushed back – the C300 Mark III at 25fps, 11.25° shutter angle F4

Finally, at 5 stops underexposure, the noise becomes a bit nasty. You can remove this noise again quite effectively in post by adding temporal noise reduction (still no spatial NR needed), but the shadow areas nevertheless have those strong reddish noise speckles, which become posterized areas of red after noise reduction.

But I would say you could still get away with this image, as it does not show those fixed horizontal/vertical lines like all other cameras tested so far –  which are very distracting in a moving image and cannot be removed easily in post.

Here is an example of using 3 frames temporal noise reduction in DaVinci Resolve, 15 temporal threshold (no spatial NR), see figure 11 below:

Fig. 11: 5 stops underexposed and pushed back – the C300 Mark III at ISO800, 25fps, 11.25° shutter angle F4 using noise reduction in DaVinci Resolve (3 frames temporal, 15 threshold, no spatial)

Summary

This year Canon surprises us all with their innovative new cameras – the C500 Mark II full-frame camera from late last year already set the benchmark in our lab for a full-frame sensor, and the Super 35 sensor sized C300 Mark III is almost on the same level – a superb lab result, leaving behind all other cameras in this sensor class – with the exception of the ARRI Alexa of course.

Hence, overall a third place behind the C500 Mark II and the ARRI Alexa. Well done, Canon!

What do you think about the new Canon C300 Mark III, do you think it will gain its firm place in the production world after the world opens up for business again? Let us know in the comments below. 

55
Leave a reply

guest
1
1
- 2
Filter:
all
Sort by:
latest
 Frank Ostyn
Frank Ostyn
Member
June 17th, 2020

Thx for these tests, I which you would do more of these. However, I am intrigued by the discrepancy between what you measured and what Canon claimed. Especially as you tested a preproduction camera. How did Canon react on the test results?

Member
May 27th, 2020

This really seems to draw into question the imatest DR testing method given how much better the c300iii underexposure looks than the c500ii. It seems clear the c300iii has at least 1-2 stops more DR yet imatest says the opposite, which seems to leave more questions than answers. If c500ii is as good as imatest says and c300iii has even more DR that we see in the underexposure, then the c300iii could be quite amazing and even compare somewhat favorably to Alexa, but the big question is why aren’t your imatest testing procedures able to capture and quantify this?

Horatio
Horatio
Guest
April 28th, 2020

So, the fx9 test, coming?

Horatio
Horatio
Guest
May 10th, 2020
Reply to  Gunther Machu

When?

Horatio
Horatio
Guest
July 2nd, 2020
Reply to  Gunther Machu

Is it?

Sam
Sam
Guest
April 27th, 2020

I’m wondering if there’s any word on CMOS smearing or banding that has been such an issue with the C300mkii. Particularly noticeable with a dark element in the foreground with a bright background. Can usually be avoided of course, but for documentary work it’s been a nuisance. I don’t think the tests you ran here would get to the bottom of this, but curious to know if anyone has some thoughts. Thanks.

Michi
Michi
Guest
April 25th, 2020

Why is there no FX9 Lab Test available yet?

Steve-o
Steve-o
Guest
April 23rd, 2020

I was honestly hoping for better DR numbers, this only puts the DR at 0.1 higher than the Panasonic s1.

George
George
Guest
April 22nd, 2020

How can we achieve 16+ stops of dynamic range? Only in raw?

Ben J
Ben J
Member
April 22nd, 2020
Reply to  George

Only in marketing!

Will McGough
Will McGough
Guest
June 25th, 2020
Reply to  Gunther Machu

Gunther, one difference of the ≤30 Canon Raw Light vs XF-AVC is that it is 12-bit (whereas the XF-AVC stays at 10-bit regardless of the framerate). Do you think it might be necessary to shoot in 12-bit in order to properly encode the camera’s full dynamic range?

Anecdotally, I perceive a loss in DR when switching between 12bit 24p and 10bit 60p raw on the C200 – although, admittedly, I have not conducted scientific tests. (I could have adjusted exposure incorrectly when switch framerates and the comparison isn’t perfect since the raw compression ratio also changes from 3:1 to 5:1. Nevertheless, I did notice an appreciable difference when grading the footage side-by-side in Resolve when the mids were exposed similarly.)

From my reading of Canon’s CLog2 specs, changing the bit depth only increase the smoothness of the curve as opposed to expanding the actual IRE range represented. However, I have to wonder if the steep parts of the log curve (shadows and highlights) might suffer from descritization as similar but distinct IRE values have to be bucketed together when the bitrate is lower. (If so, I would imagine this might impact results from this style of testing.)

Member
April 22nd, 2020

Thanks for the test, very promising. Did you ever test the C200? I would love to know how it compares.

 XiaoYang Jin
Member
April 22nd, 2020

Look at the side by side underexposure test between c500 mark ii vs c300 mark iii
the new sensor clearly shows strong wins. I think the DR is highly related to how
we measure the noise level but I can see the improvement of this sensor in terms
of color representation. just curious why there is no any over-exposure test?

 Derek Doublin
Derek Doublin
Member
April 22nd, 2020

I’d love to see you guys do an Alexa Mini or an Amira so we can use that as a baseline to interpret and compare these other tests from.

Kyle Bainter
Kyle Bainter
Guest
April 22nd, 2020
Reply to  Derek Doublin

+1

 salty salt
salty salt
Member
April 23rd, 2020
Reply to  Derek Doublin

Yes this is needed to make this a professional source of info.

Ben J
Ben J
Member
April 23rd, 2020
Reply to  salty salt

It would be interesting, but I’m not sure it’s ‘needed’ in any particular way. The vast majority of professionals who operate cameras haven’t used an Arri camera and don’t need to – these people will be interested in how the C300/3 performs in comparison to similar cameras.

 Derek Doublin
Derek Doublin
Member
April 25th, 2020
Reply to  Ben J

Yes, but as it stands the Alexa is the camera that sets the bar. It is the industry standard cinema camera. To date, no other camera has managed to achieve what they’ve achieved concerning dynamic range and highlight rolloff, which is saying a lot since their sensor is now 10 years old. Unlike most other camera companies out there, Arri is not known for exaggerating their specs. In fact, to the contrary, their specs are almost always dead on. So if Cinema5D would test the Alexa, it would help some of us understand how to interpret their own tests, or even highlight any margins of error.

Daniel
Daniel
Guest
April 21st, 2020

Would be amazing to see a 5 stop over test as well.

 Derek Doublin
Derek Doublin
Member
April 21st, 2020

Hmmmm. Can’t say I’m super happy about this test. Canon’s claiming 16+ stops, which I knew would be an exaggeration. But I was hoping we would at least get a usable 14 stops from the C300 III. This news seems to take the shine off some of the hype around this camera for me.

Reluker
Reluker
Guest
April 23rd, 2020
Reply to  Gunther Machu

Hi Gunther! Canon add to a model only +0.7/+0.5 Stops DR at every 5 years :)

 Derek Doublin
Derek Doublin
Member
May 7th, 2020
Reply to  Gunther Machu

When compared to the waveform test you performed for the C500 II, it does look like the C300 III might exhibit better highlight rolloff. Whereas the C500 II drops off sharply after 13 stops, in the C300 III it seems we can see some very small, slight hints of detail after that same point in the chart. Might this translate to better rolloff in the C300 III?

Thomas
Guest
April 21st, 2020

Hi Gunther,

did you do an ABB before you did the test?

Clayton
Clayton
Guest
May 11th, 2020
Reply to  Gunther Machu

Hi Gunther, as a c200 owner I definitely see a noise floor improvement when I black balance – so i try to always do so before a shoot. It’d be great to see a test with the c300 m3 black balanced first.
Also, are you going to include the Red cameras in your lab tests? Would be great to see the Gemini’s results! And of course the Alev 3 sensor as a benchmark.

Andrée
Guest
April 21st, 2020

I really like your standardised, @imatest based tests for DR. You should have an easy to read chart of all the cameras tested so far (under identical, unemotional conditions). Perhaps a small explanatory text above to clarify why your numbers might differ from other tests.
(https://t.co/h73TPXK8gP)

Kyle Bainter
Kyle Bainter
Guest
April 21st, 2020

Where is the extra 3 stops? In raw?

Kyle Bainter
Kyle Bainter
Guest
April 22nd, 2020
Reply to  Gunther Machu

Have you tested the c200 the same way? I couldn’t find the results..

Jonathan Constantinou
Jonathan Constantinou
Member
April 21st, 2020

Hi Gunther, thanks as always for the work you do. I am interested to know what C5D finally settled on the DR rating of the C300mk2? As a mk2 owner that relative figure is more important than the absolute DR – I want to know how much more is in the mk3! Thanks again, Jonathan

Elio Stamm
Guest
April 21st, 2020

Thanks. Interesting that in your lab tests the C500 II shows slightly more Dynamic Range than the C300 III, as Canon themselves mention 1 stop more for the C300 III (16 vs 15 stops if I recall correctly).

Robert Niessner
Robert Niessner
Guest
April 22nd, 2020
Reply to  Gunther Machu

Gunther, do you think it would be possible to include IMTEST results for samples which have been denoised in post? So one could estimate how well the shadow parts can be salvaged later.

Elio Stamm
Guest
April 22nd, 2020
Reply to  Gunther Machu

Thanks for this additional information. Very interesting indeed. Btw, looking back at your FS7 test I see a Dynamic Range of 12.1 for FS7 with a SNR of 2 (which is clearly lower than C500ii and C300iii), but 17.7 detected patches (which is higher).

Felix
Felix
Guest
April 21st, 2020

Comparing the 6-8ms rolling shutter of the C300ii the MKiii is disappointing…

Felix
Felix
Guest
April 22nd, 2020
Reply to  Gunther Machu

Would be great if Canon would give us the choice between less rolling shutter and DPAF and/or dual gain.

sebastian
sebastian
Guest
April 21st, 2020

Looks like the c300 MIII handles under exposure much better than the C500 MKII. No streaking when you reach 4 under. I’m impressed.

 Guillaume Bression
Member
April 21st, 2020
Reply to  sebastian

The C500 mark II use an old tech sensor (same as C700FF that use just a bigger version of the C300 mark II sensor) so it doesn’t surprised me. I would have even expected better results from this new generation of sensor. Anyway, it is a great camera that will not be the main limitation when most of us will try to tell a good story… even if I would love to have 4K 120p on my C700FF…

1
1
- 2
Filter:
all
Sort by:
latest

Take part in the CineD community experience