Canon announces development of the EOS R5 full-frame mirrorless camera

[slclick]

Yennefer? Is that a code name for the EF 50mm f/1.4L IS USM?

I saw that he is a fan of Witcher.

 

[Michael Clark]

Delete button works the same way and gives you some additional space on the memory card.

At the possible expense of write speed as the card becomes fragmented (due to varying file size of each image depending on scene contents).

 

[davidhfe]

Since you mentioned the confusion about technology yourself, you sound like you know your thing. But I can't imagine the 8K 30p being anything other than RAW. What leads you to believe the opposite?

When the data comes off the sensor, it is RAW data. To change it to something else, a computationally expensive compression step would have to follow, that adds additional hardware requirements and produces heat. Just dumping the RAW to the memory card is the easier option. That's why Magic Lantern has been able to offer that for years (Yes, the new bodies are not supported).

If Canon had a way to deal with these issues, why is the 5.5K on the 1DX III only available in RAW?

OK the caveat here is that I am not an electrical engineer I just spend too much time on the internet.

My (limited) understanding is that you typically have different read modes on the sensor. You can trade speed for bit depth. So for photos, you can get 14-bit RAW at about 20fps in one read mode, or if you want speed you can drop that down to 10 or 12 bit and get a faster read. Even then, you'll hit another limit and then need to find a way to read fewer pixels: that's when crops, pixel bins and line skip modes kick in.

In human brain terms, multiplying 500x800 is easier than 543x822. You can trade speed for precision.

The 1DX3 has half the pixels to deal with, so a full RAW read is easier to do—there's just less data (and heat!)

All of the above is to just get the signal processed from the photo sites to the processor. Again, I am not an EE so I don't know what parts of the process are handled on chip vs on the DSP, but when you're talking about compression being computationally expensive, that's a different problem. The conflating thing is that canon has been behind in both areas:

- They've had sensor read issues, which have shown up in crop modes and fairly severe rolling shutter
- They've had DIGIC issues, giving us compression schemes like MJPEG

Quick edit: DIGIC X seems to have solved the latter, as it can move around a gigapixel per second and full depth. The sensor read issue remains to be seen. The 1DX3 sensor is a clear improvement (5.5K raw!) but does have other limits (still has rolling shutter, DPAF modes are limited, crops, etc). There are plenty of reasons why the R5's sensor might be more technically advanced than the 1DX3, though. We're just gonna have to sit tight on that.

Regarding ML, canon engineers (like most engineers) are pretty conservative. They want to ensure the modes they offer will work for the lifespan of the camera. ML is willing to squeeze every bit of headroom the engineers put in to get better performance.

 

[Michael Clark]

I never delete photos in-camera except for very obvious failures and the ones that have already been transferred to PC. Rate button in turn is very useful on 5DIV. I mark beginnings of sequences for stacking/blending and also some potential keepers.

I stopped doing it a long time ago after accidentally deleting some images I didn't realize were not "protected" and then using the card and overwriting some of them. Luckily, I was able to recover the ones that did not get overwritten, which was most of them.

For me there's just too much risk of accidentally deleted something unintentionally. I never delete anything until it's been backed up first.

 

[Jack Douglas]

No apologies needed. I am an EE but retired for quite a few years and my head spins reading some (not all) of this. It's amazing what is now being done.

Jack

 

[PureClassA]

You need the basic resolution of the video from the sensor to hit a specific video res. So it's the other way around from the way you have it.

For 8K in a 16:9 ratio captured on a 3:2 sensor, you need a 39-45 MP sensor (there are a few different 8K formats apparently) so you'll use just about the whole damn sensor to get that video resolution. Most people are assuming that Canon will more or less do the minimum necessary to reach 8K, we're means the R5 will have a 39-45 MP sensor and 8K will have little to no crop at all.

For 4K, they clearly don't need 39-45 MP on the sensor. So I'd assume that Canon has to choose to go with a pretty healthy crop (which is less heavy computationally to do, but it annoyingly changes your lens' FOV) or avoid a crop and do a lot of data handling to bin/sample/process the entire frame. Either, in theory, are possible, but having your FOV change when you drop from 8K to 4K would be kind of annoying, right?

(Video nerds: do I get a passing grade there?)

- A

Yup. Pretty much. Given the capability of the Digic X based on stated 1DX3 specs, we should still get up to 4K30 on the R5 in FF readout (obviously way oversampled and then compressed down to 4K) or at the very worst some minimal cropping. I think they could surprise us with 4k60. At 4K120 you're pretty much guaranteed a crop factor. That could easily be 1:1 which would put the crop at well north of 2X. 8K of course would need the entire 16:9 area of the sensor to operate anyway, so no crop.

The Sony A7R3 is roughly the same resolution and gets 4K up to 30fps. The Yet to be released Sony A7R4 is going up to 63MP and still can't break the 30fps barrier in 4K. That's why this R5 is so stunning. It's rumored to be able to slam a LOT more data through.

Then there's the Panny S1H. It can do FF internal and external 4K up to 60fps. But that isn't a fair comparison as the sensor is only 24MP, far more similar to the 1DX3 and this rumored R6. So a FF readout on those cameras is obviously far less taxing than the R5 or A7R3 or A7R4.

We can feel very safe assuming FF reads at up to 4K30. We get fuzzy beyond that when we're talking 45MP.

This is why if you're looking for (real world) video, the 1DX3 and R6 with its lower MP count is going to be more favorable for doing those higher frame rates.

Of course the real question mark is the fact the R5 is still being reported at 20fps still, which presume are still at 14bit RAW and not some otherwise compressed format. If so, that reopens the doors to 4K60 at FF readouts in perhaps 8bit 420 and maybe externally at better bitdepths. If you look at the Panny and Sonys, you'll see better record mode available via external only which is not at all uncommon, including the EOS R

 

[Michael Clark]

So there is no physical remote port on the R5? I am not seeing the symbol for one on the port covers. I use cabled remotes a ton. Love them, hard to lose.

It looks like there's one on the front in the same spot that the 5D Mark IV has one.

 

[davo]

8K literally means ~8000 (either 8192 or 7680 depending on who you ask) horizontal pixels. If your sensor does not have at least that number of horizontal pixels, you cannot record 8K by definition. Simple as that.

That makes sense

 

[Michael Clark]

Indeed not seen on the video. Maybe wireless control?

They moved it off of the side to the front of the camera with the 5D MArk IV. There appears to be a cover in the same spot on the R5.

 

[PureClassA]

We have crop mirrorless Canon bodies now, and have had them for some time. It's called the EOS M series.

I have an EOS M. But they are not equipped with RF mounts obviously. Canon is going to probably over time slowly semi-replace the M with small compact crop body RF mounted rigs and a line of RF-S (like EF-S) glass.

 

[ahsanford]

I saw that he is a fan of Witcher.

Come on people. Drilling down on Twitter bios is a wee bit off topic, is it not?

(My friend makes that show. Watch that show and help my friend out, would ya?)

- A

 

[davidhfe]

The Sony A7R3 is roughly the same resolution and gets 4K up to 30fps. The Yet to be released Sony A7R4 is going up to 63MP and still can't break the 30fps barrier in 4K. That's why this R5 is so stunning. It's rumored to be able to slam a LOT more data through.

The A7R4 is very much available for purchase, isn't it? BH lists it as in stock.

 

[Michael Clark]

I am about 300 posts behind on this thread, but that article doesn’t seem super informed. Stuff like “Their processor division” (doesn’t canon source their DSPs?) or “the 5D4 processor is too slow” when rolling shutter, and crappy DR before that, indicate more of a sensor fab issue.

It was clear that Canon was lagging from chronic lack of investment in ILCs for a while, across multiple fronts. These improvements all took years of R&D to fix. Article makes it sound like in July of 2019 some VP said “oh hey let’s not make under spec’d camera anymore”

The R5 and the improvements that it has must have all been in the works for years.

I don't think it was lack of investment. I think it was the commitment to Dual Pixel CMOS AF and the increased processor load of twice as much data for the same resolution sensors.

 

[PureClassA]

The A7R4 is very much available for purchase, isn't it? BH lists it as in stock.

My mistake. I dont follow Sony.

 

[Viggo]

It took my a few episodes to find out that he also played Superman

 

[Michael Clark]

I have an EOS M. But they are not equipped with RF mounts obviously. Canon is going to probably over time slowly semi-replace the M with small compact crop body RF mounted rigs and a line of RF-S (like EF-S) glass.

Not as long as EOS M is the best selling mirrorless system on the planet. APS-C and smaller is still a much larger portion of the ILC market than FF is.

 

[slclick]

It took my a few episodes to find out that he also played Superman

There's a man in that show?

 

[sanj]

Sorry to pick on you a bit here, but I am blown away by how effective the youtube and camera site contingent has been at drilling in "anything canon offers for video will be a crop!" in to people's minds. We're at the point where people are talking about crop factors that are physically impossible. There are a lot of folks who don't understand the technical side of things (which is fine!) who have been done a real disservice by photos sites/etc.

To actually answer your question, my guess would be:
8k30 - Full width from a 45 mp sensor. You cannot "crop in" on a 45 mp sensor and get an 8k image*
4k30 - Full width, oversampled from 8k
4k60 - Either line skipping, pixel binning or (most likely) a 1.5x crop (about super 35, will work great with EFS lenses an adaptor!)
4k120 - 2.0 crop

What modes have what limits beyond that is anyone's guess:
- I am not expecting anything RAW
- Likely 10bit 4:2:2 C-Log with 2 H.265 compression options for most modes
- I expect DPAF to work in 8/4K 30fps. Maybe 60fps. Unlikely 120fps.

*there are a lot of shenanigans that COULD happen here. In camera up-scaling from a 6K image, etc. Canon would never EVER live it down, though. It would be a serious mistake to put out a development announcement with limited details, and have one of those details be BS.

Thank you. So for sure, this will be full-frame at 4k 30? Am happy!!!

 

[sanj]

A 1.6x crop on 8k would make it an 80+MPix sensor.

I do not understand the point.

 

[davidhfe]

Thank you. So for sure, this will be full-frame at 4k 30? Am happy!!!

I am sure that's my guess!

Only canon (and maybe CR Guy) know the truth