Canon R6 mk2

[Thomas Smet]

Had a few minutes to play with my new R5. Very impressed. I still like the R6 and when shooting 4k video they are more or less the same. The R5 has a tad finer noise due to the 8k vs 5.4k down sampled to 4k.

Yeah Canon raw sucks. At least in terms of performance. Whats nice however is it's a real raw format and not a partially debarred raw format like Braw. FCP not only has no raw controls when using the plugin but I found it to be very unstable and forget about using Neatvideo on top of it. It comes to a halt and Neatvideo rarely updates the frames.

I am getting used to the Canon raw development application however. Its ugly but its not a bad solution t process the raw and batch convert to another format that works great in FCP or Resolve like ProRes444 which is still 12bit. Besides losing the ability to adjust the raw setting later its really a solid visual way to work.

I have heard some YouTubers claim the clog2 raw sucks and is not worth the hassle but I completely disagree with them and now have proof. Yes shooting for middle gray the clog2 has a ton of noise and looks really ugly compared to internal clog3. Shooting two stops over however compensates for that and remaps the stops so you move that noise floor two stops down. It really helps and the raw clog2 looks completely noise free. I used my light meter at ISO 800 then adjusted the camera and then bumped it to ISO 3200 to add two stops to the correct exposure. This not only shoots two stops over but the R5 uses dual native ISO unofficially which means ISO 3200 is a good clean way to shoot 2 stops over. I'm applying zero noise reduction and it looks great. I'm adjusting the ISO in the Canon development app back down to ISO 800 to compensate then exporting as ProRes444 8k. Takes a long time to export but once done I have a pretty insane looking 8k clog2 video to work with. I lose some highlight detail in this process but if noise is a major concern this is a valid way to work.

Thing with Canon is you cannot using clog as a one size fits all log format. Much like the Pocket cameras you need to map the stops below or above middle gray demanding on if shadow or highlight detail is more important. ISO 800 on Canon is like ISO 800 on the Pocket cameras. Great highlight detail but horrible shadow noise. Can't cannot use ISO to map the stops so we shoot over or under to basically do the same thing. Since ISO 800 is already the starting point with Clog3 and clog2 you really want to shoot over as much as possible or as base ISO. If you really want highlight detail you shoot base ISO. If you want shadow detail and there are no higher highlight values then you shoot one or two stops over. Two being the best at almost perfectly cleaning up clog. Thats like shooting ISO 200 on the Pocket cameras. Less highlight detail but amazing shadows.

Neatvideo is also an option and so for it's doing a really good job on the 8k and cleaning up the shadows when shooting at base ISO.

This weekend I will have more time to try out 5k cropped external raw on the Ninja and play around wit hthe internal raw more.

I'm saying all of this because yes I think the external raw on the R6 mk2 can have some benefit and might be worth it. In some ways I wish I went that direction instead. 8k is a bit much like I feared. Plus externally now I can only do 5k vs the R6 mk2 that can do 6k to the Ninja V. I'm not about to upgrade to he Ninja V+ either. The R6 mk2 can also do that FF vs now for me the R5 can only do that cropped. I do think the R5 has a better sensor however and the larger size helps make the noise even finer which can help.

All my tests so far are viewed on my new 32" Asus Pro art display in HDR PQ. I'm using a Decklink card in a eGPU enclosure to get a true UHD output to the Asus. Thats a rather large display sitting up close to not see any noise so far. I'm letting FCP down scale the 8k project to UHD to output to the Decklink.

Hope this helps somebody. There is an incorrect ,myth out there that Canon log and sensors suck. It's just that many don't know how to use them properly. Yeah a 15+ stop sensor that doesn't need to shoot over would be nice but its not a must.

 

[NorBro]

I'm one that does think the C-Log and C-Log3 in the Canon mirrorless systems is garbage with the amount of DR the cameras are producing (particularly the highlights clip way too soon compared to other companies).

The R5's RAW workflow was excruciating for me. CRD's exporting is more than real time (or was in July 2020).

After the R5, I rented a R3 and I was again horrified by the amount of noise in the RAW signal (like I called Canon to see if I can chat with someone about it), but if I switched the camera to H.265, it completely disappeared and the video was extremely nice just like on the R5 (meaning I guess their noise reduction is pretty good).

___

Then I got a C200B with Cinema RAW Light and it's a night-and-day experience. Better RAW, better RAW quality. Terrible AF tho.

 

[DLD]

Back to pixel shift -- there's a longstanding theory that only Panasonic gets to have a strong implementation of it and everyone else gets the cripple hammer

A7RV has the interpolation (or, as Sony calls it "motion compensation") that A7RIV did not have. Gordon Laing thinks that, while some motion is interpolated, some images - or really, parts of images - are just semi-randomly picked. In other words, the camera just picks a random frame - let's assume 8th or 9th out of 16 - and makes that a reference for the rest where it's not static. Without motion compensation, effected portions of the stitched together image looked awful, just a jumbled digital mess.

But A7RV is a recent unit while S1R is three-going-on-four-year-old model. Panasonic has just released a 24 MPX full frame update. I suppose S1R update/upgrade is not far away either. L-mount in general isn't a huge success - S1H sold well early on - but Leica must want a new camera as well. Sony has to play ball and these "do not do" limits - 8-bit recording, crappy AF, obsolete media and codecs, pixel shift, et cetera, et cetera - that have been imposed on all manufacturers are dropping one by one.

 

[Peter C.]

Had a few minutes to play with my new R5. Very impressed. I still like the R6 and when shooting 4k video they are more or less the same. The R5 has a tad finer noise due to the 8k vs 5.4k down sampled to 4k…

I guess I’m a very different shooter. When I get a camera I like to know the highest quality settings but very quickly I settle into the quickest easiest work flow. When I’m in front of a subject I want to spend most of my time and energy on the subject not the camera trying to eke out every possible pixel. I wouldn’t want to have to over expose by 2 stops and going through the process you describe sounds tedious. Not my cup of tea. Out of curiosity what sort of work are you doing with this camera?

 

[Thomas Smet]

I guess I’m a very different shooter. When I get a camera I like to know the highest quality settings but very quickly I settle into the quickest easiest work flow. When I’m in front of a subject I want to spend most of my time and energy on the subject not the camera trying to eke out every possible pixel. I wouldn’t want to have to over expose by 2 stops and going through the process you describe sounds tedious. Not my cup of tea. Out of curiosity what sort of work are you doing with this camera?

Oh I wouldn't do that all the time. Plus it's actually super easy to shoot over or under. You can get to a point where you don't even think about it. Plus I create corrective luts to load on the Ninja V so it looks correct on there.

I shoot a wide range of stuff from live vide streaming to VFX. I typically save shooting raw for the VFX type stuff. I have zero problem with normal 10bit video if done well. I find the R6 and R5 base exposure Clog3 to be perfectly fine. A tad grainy but thats also not always a bad thing.

I'm actually shooting a lot of HDR PQ on Canon cameras these days. Doesn't have as much range as clog3 but it's also super clean and still gives me up to 530 nits HDR. I then have my own process to convert that to rec709 if I need to.

As I am also used to working with raw I'm also not 100% opposed to using noise reception in post. Noise reduction is fairly common with raw video. Either way video has some level of noise reduction. Either the crude form done in camera or the better form done in post.

You will find a lot of Canon reviews that suggest shooting over. Thats been a fairly common suggestion for Canon cameras for years. Although they tend to just say shoot ETTR and not why and not at a specific amount. ETTR, shooting over, +2 stops are all the same thing. Saying two stops over is just shooting ETTR by a specific amount. An amount that has to optimal benefit to the camera. Beyond that tends to be a waste and eats in to the highlight details and can start to desaturate or clip important parts of the image. Most cameras are ok shooting 3 stops over but it doesn't gain a lot really. Some even just prefer a single stop over.

It can be as easy as using the EV meter on a camera. 0 EV is middle gray. When the exposure hits +2 thats two stops over. That is an average however but some cameras can use a spot meter instead for video so it looks at a specific point. The EV meter can be off because of the scene average so you can't trust it 100%. Another tool is zebras which can be trusted more. If we set for skin and add two stops then we know any skin in the scene with zebras should be exposure two stops over. You can use middle gray if you have a gray card and then use it the same way.

For example if C-log3 uses 34.3% instead of setting the zebras to 35% you can set them two stops higher around 55%. That way if you expose on the gray card for 55% zebras then you know you are always two stops over. False color is another great tool. Just aim for skin, gray card or a white surface as a 90% reflectance and make sure the color it hits is one thats about two stops higher than middle gray.

I personally use a light meter. I measure the light hitting the subject using ISO 800. Once exposure is set I then bump the ISO to 3200 or the f stop down two stops to open up the lens if possible. I tend to not mess with the lens or camera exposure very much and instead more likely to adjust my light output by two stops. For exposure adjusting the lighting is always the best option first. Followed by aperture when we have no control over the lighting.

You are right however it is a bit more effort. There are plenty of times I just wing it as well and have to trust the EV meter and zebras or my good ole eyes. Panasonic EV meters always seemed more accurate to me. Canon tend to darken everything to protect the highlights. I end up accidents shooting darker a lot more on Canon than I ever did on Panasonic. Canon also cannot use the spot meter mode when shooting video which is super lame. On Panasonic I would use spot meter mode and use the joystick to move the spot meter point around the screen to check exposure. Like a crude poor mans false color seeing one area at a time. For my more scripted work like a documentary interview or some of the VFX work I do I always use my Ninja V, a gray card, light meter and adjustable lights so its pretty easy to shoot however I need to. It literally takes seconds to adjust the exposure under or over as needed.

If I'm shooting my daughter, more family based stuff or some kind of live event then yes I just shoot normal exposure as best I can with HDR PQ or C-log3 and deal with it.

One other thing I really like about the first R6 and R5 is they still include C-log. It may have a stop less in the highlights but it's also a much cleaner log format. For more run&gun stuff it's actually a solid log format to use. I know it's no longer as popular as C-log3 but it's really clean and still provides a lot more range than rec709. No need to mess around with shooting over but the highlights are not as good. I recent months I have gone back to being more of a fan of the old C-log. I think C-log3 is better if you have the time to use it well but C-log is more robust and easier to deal with out of the box with less messing around.

 

[Thomas Smet]

I'm one that does think the C-Log and C-Log3 in the Canon mirrorless systems is garbage with the amount of DR the cameras are producing (particularly the highlights clip way too soon compared to other companies).

The R5's RAW workflow was excruciating for me. CRD's exporting is more than real time (or was in July 2020).

After the R5, I rented a R3 and I was again horrified by the amount of noise in the RAW signal (like I called Canon to see if I can chat with someone about it), but if I switched the camera to H.265, it completely disappeared and the video was extremely nice just like on the R5 (meaning I guess their noise reduction is pretty good).

___

Then I got a C200B with Cinema RAW Light and it's a night-and-day experience. Better RAW, better RAW quality. Terrible AF tho.

Yeah Canon DSLR sensors are not very good out of the box handling raw. You really have to use noise reduction. To be fair the same is kind of true on the P4k and P6k as well if using anything thats not ISO 100,200,400 or 1250. Panasonic and Sony handle DSLR external raw much better. The signal to noise ratio on the R6 and R5 are not great with how Canon handles log. For photos using 14bit processing I find Canon to be great but they limit what the sensors can pump out for video way too much.