Interview: Canon engineers talk Canon EOS R development

jolyonralph · Sep 17, 2018

There is one huge advantage in the crop mode on the EOS R compared to the 5D Mark IV in that you can fit adapted EF-S lenses onto it. I wrote a whole lump about this that I won't repeat here on my blog (see link below) in my article about what on earth is going to happen to EF-M next...

www.everyothershot.com

CarlMillerPhoto · Sep 17, 2018

I guess this is what you get when photo-centric websites ask about video specs. The C700FF sensor is indeed full frame. Specifically, it's 38 x 20 and shoots in a 1.89 aspect ratio.

Canon is just reserving FF 4k readout for higher priced models. Why is that so hard for them to admit and consumers to understand?

NorskHest · Sep 17, 2018

No one is going to use 100% of a sensor for focusing, this sounds like a gimmick and that's probably because it is, sure more may be better but if you are trying to focus on the edge of what you are trying to capture is the image really worth capturing or is the image worth seeing? I'm all for advancement but this is cheap advertising also my c200 does a full sensor read out, the c700 does a full sensor read out. We all like canon but this interview is like listening to the apple keynote, just getting bludgeon with bs.

vjlex · Sep 17, 2018

I wonder if there's any technical reason why they wouldn't update codec options for 5D IV.

CarlMillerPhoto said:
Canon is just reserving FF 4k readout for higher priced models.

I do very much believe this. I have a hard time believing they lack any technical know-how regarding full-frame 4K readout.

sdz · Sep 17, 2018

shunsai said:
I wonder if there's any technical reason why they wouldn't update codec options for 5D IV.

I do very much believe this. I have a hard time believing they lack any technical know-how regarding full-frame 4K readout.

It's possible that Canon must maneuver around Sony patents, fabrication process issues, etc. Still, Canon has patents in hand that point to a new sensor platform.

melgross · Sep 17, 2018

This is all interesting. Thom Hogan states that the crop leads to BETTER quality in video, not worse, because full sensor use results in interpolation with resulting smear and artifacts. The negative is the loss of angle, requiring a wider lens.

vjlex · Sep 17, 2018

sdz said:
It's possible that Canon must maneuver around Sony patents, fabrication process issues, etc. Still, Canon has patents in hand that point to a new sensor platform.

Yeah, I wonder about the patent maneuvering from time-to-time. It's hard to know what is being withheld for technological, financial, or legal reasons.

ken · Sep 17, 2018

CarlMillerPhoto said:
I guess this is what you get when photo-centric websites ask about video specs. The C700FF sensor is indeed full frame. Specifically, it's 38 x 20 and shoots in a 1.89 aspect ratio.

Canon is just reserving FF 4k readout for higher priced models. Why is that so hard for them to admit and consumers to understand?

I think it is more than just segmentation. I firmly believe it's technical.

In the a7riii Sony is able to apply interpolation to the entire sensor readout of 7952 x 5304 to produce a video frame of 3,840 x 2,160, and do it fast enough to keep high data stream rates. Canon simply can't do that yet. They crop out of the middle of the high-resolution photocentric sensor to get the precise resolution needed for 4K rather than attempting to downsample in real time. Why? Because it's a CPU intensive process and they can't maintain the data stream rate required for video.

I'm fairly certain that Sony's big breakthrough was parallelizing this process on the sensor itself. It's the most logical place to do it, as so much of the sensor handling is inherently parallelized to the pixel. Think of how nVidia does similar processing with vertex shaders and pixel shaders on the GPU. One or two clock cycles and you have an interpolated frame output, vs iterating over each "effective" pixel on an external processor that has to iterate over every pixel and perform "nearest neighbors" interpolation. Canon's sensors don't do that. Yet.

There are other ways to approach the problem, but the most obvious place to solve this is on the sensor itself.

This is just an educated guess, but I bet a more programmable sensor is in the future for Canon.

dak723 · Sep 17, 2018

melgross said:
This is all interesting. Thom Hogan states that the crop leads to BETTER quality in video, not worse, because full sensor use results in interpolation with resulting smear and artifacts. The negative is the loss of angle, requiring a wider lens.

Yes, I have read other articles that also point this out.

As usual, all the internet nobodies that cry and whine when crop video is mentioned are just Canon bashers and can be ignored.

dak723 · Sep 17, 2018

shunsai said:
Yeah, I wonder about the patent maneuvering from time-to-time. It's hard to know what is being withheld for technological, financial, or legal reasons.

While no doubt some of the reasons that certain features are withheld or downgraded are product segmentation, only ignorant reviewers and forum dwellers automatically believe it is all because Canon is BAD and wants to screw the consumer. It has been brought up numerous times that perhaps Canon's dual pixel architecture is one of the problems in processing speed. So their can indeed be technical issues that they have not yet overcome. Patents play a big part. Just cause Sony does it, doesn't mean that Canon can do it. Not sure why people can't get that. Overheating issues also get scorned by the internet crowd. Perhaps that is one of the issues why their is no Canon IBIS yet. Maybe Sony can do it because they also have the worst weather sealing in the business. Maybe better weather sealing doesn't allow for heat dissipation? Maybe there is a trade-off there (just guessing). If so, what would you rather have, top notch weather sealing with no IBIS, or IBIS with awful weather sealing?

Roy Hunte · Sep 17, 2018

dak723 said:
While no doubt some of the reasons that certain features are withheld or downgraded are product segmentation, only ignorant reviewers and forum dwellers automatically believe it is all because Canon is BAD and wants to screw the consumer. It has been brought up numerous times that perhaps Canon's dual pixel architecture is one of the problems in processing speed. So their can indeed be technical issues that they have not yet overcome. Patents play a big part. Just cause Sony does it, doesn't mean that Canon can do it. Not sure why people can't get that. Overheating issues also get scorned by the internet crowd. Perhaps that is one of the issues why their is no Canon IBIS yet. Maybe Sony can do it because they also have the worst weather sealing in the business. Maybe better weather sealing doesn't allow for heat dissipation? Maybe there is a trade-off there (just guessing). If so, what would you rather have, top notch weather sealing with no IBIS, or IBIS with awful weather sealing?

On the topic of heat dissipation, I borrowed my cousin's A7R II to do a 4K shoot, and I wasn't impressed with how hot the body got. Also it ate battery life like crazy. So maybe there is something to what Canon is saying. I wonder how the Z6/Z7 fares temperature wise in full frame 4K.

Etienne · Sep 17, 2018

melgross said:
This is all interesting. Thom Hogan states that the crop leads to BETTER quality in video, not worse, because full sensor use results in interpolation with resulting smear and artifacts. The negative is the loss of angle, requiring a wider lens.

Sony A7sII reads every pixel in FF and downsizes ... much better than crop

tmroper · Sep 17, 2018

ken said:
I think it is more than just segmentation. I firmly believe it's technical.

In the a7riii Sony is able to apply interpolation to the entire sensor readout of 7952 x 5304 to produce a video frame of 3,840 x 2,160, and do it fast enough to keep high data stream rates. Canon simply can't do that yet. They crop out of the middle of the high-resolution photocentric sensor to get the precise resolution needed for 4K rather than attempting to downsample in real time. Why? Because it's a CPU intensive process and they can't maintain the data stream rate required for video.

I'm fairly certain that Sony's big breakthrough was parallelizing this process on the sensor itself. It's the most logical place to do it, as so much of the sensor handling is inherently parallelized to the pixel. Think of how nVidia does similar processing with vertex shaders and pixel shaders on the GPU. One or two clock cycles and you have an interpolated frame output, vs iterating over each "effective" pixel on an external processor that has to iterate over every pixel and perform "nearest neighbors" interpolation. Canon's sensors don't do that. Yet.

There are other ways to approach the problem, but the most obvious place to solve this is on the sensor itself.

This is just an educated guess, but I bet a more programmable sensor is in the future for Canon.

Doesn't a dedicated piece of silicon process raw information it gets from the sensors? And if so, maybe Sony is better able to design that chip (dedicated registers, etc) to process video, and get patents on what they design. Canon doesn't have the silicon flexibilty that Sony apparently has (being such experts in the whole silicon process).

fullstop · Sep 17, 2018

As I always said: "EF lenses are legacy in relation to Canon R. They will perform *within their mount- and communications-related limits*.

Some folks here did not (want to) believe it. So, here's one prof. Other EF limitations will surface in due course.

Since EF-mount lenses used via the adapter are so much further from the sensor surface, it seems likely that they wouldn’t be able to support as large an AF area as native RF-mount ones. I asked, and it turns out this is true. It will vary some based on the specific lens design, but some EF-mount lenses won’t have quite as wide AF coverage as the native ones. Native RF-mount coverage is 100% vertically and 88% horizontally, but some EF-mount lenses will only have 80% horizontal coverage

mppix · Sep 17, 2018

Full frame video is and remains problematic in any foto-cam. While good downsampling improves rather than deteriorates IQ, it is computationally expensive. Hence several photo-centric cams use line-skipping, which enhances iq issues, especially in sensors without low pass filter. To my knowledge, this is why videographers often prefer the crop image.

The fundamental issue is that a camera needs to read a 6k-8k image and process it for proper downsampling. This is still teritory of cams with dedicated sensor or 6k-8k video cams costing $30k+ (with active cooling)

6degrees · Sep 17, 2018

Wow, this is like a bombshell.
I think neither Sony nor Nikon want their customers to read this: https://www.imaging-resource.com/news/2018/09/15/canon-eos-r-qa-with-the-canon-engineers, especially Sony. The consequenc of this can last 30 years long.
- This may impact DSLR/EF lenses sales if customers read this. DSLR/EF look obsolete to me.
- It implies that RF lenses will be much better than EF lenses. It is really bad news for the EF lenses providers, which are still developing EF lenses and more to come.
- This seems to imply that Canon EF-to-RF/ND filter adpater + Zeiss Milvus 1.4/25 may not be a smart purchase. Better wait for RF wide angle lenses?
- If Canon can develop RF 100-200 F2 or RF 135-200 F2 with similar image quality to Canon EF 70-200mm f/2.8L IS III USM, and if Zeiss will develop RF native wide angle lenses with manual focus only, like 1.4/20 or 1.4/25, it will be perfect.
- Another issue, left untouched, is about adapting Leica M-lenses to Canon RF. Will that present similar issue that Sony E Mount has due to the sensor glass thickness?

Graphic.Artifacts · Sep 17, 2018

CarlMillerPhoto said:
I guess this is what you get when photo-centric websites ask about video specs. The C700FF sensor is indeed full frame. Specifically, it's 38 x 20 and shoots in a 1.89 aspect ratio.

Canon is just reserving FF 4k readout for higher priced models. Why is that so hard for them to admit and consumers to understand?

I don't think that's entirely accurate nor is the interviewer entirely incorrect. I beleive the C700FF reads 6K from the entire sensor. 4K is a crop to a size similar to Super 35 as stated in the interview. In each case Canon is reading the sensor at 1:1 and it's a much lower resolution sensor than the one in the Canon R.

If Canon has demonstrated that they can read an entire high resolution sensor and downsample to 4K in camera I haven't seen it. I think we can assume that it is a technical challenge they are working on vigorously.

tron · Sep 17, 2018

Regarding "(and in steps of just 1/8 EV, vs. the 1/3 EV of EF-series optics)" mentioned in the article. The writer is not familiar with the Technical Back E which worked with Canon EOS 620, 600(630 and RT.

With its help autobracking could be done and the step was selectable starting from 1/4 of a stop! YES 1/4 of a stop (Referred in the manual as 0.25) from the late 80s! So all EF lenses already supported that.

BurningPlatform · Sep 17, 2018

Graphic.Artifacts said:
I don't think that's entirely accurate nor is the interviewer entirely incorrect. I beleive the C700FF reads 6K from the entire sensor. 4K is a crop to a size similar to Super 35 as stated in the interview. In each case Canon is reading the sensor at 1:1 and it's a much lower resolution sensor than the one in the Canon R.

If Canon has demonstrated that they can read an entire high resolution sensor and downsample to 4K in camera I haven't seen it. I think we can assume that it is a technical challenge they are working on vigorously.

Actually C700 FF reads full sensor for 4k. This is from their marketing material:

"Capable of ProRes and XF-AVC internal recording up to 4K to CFast™ cards, the camera can also record 5.9K RAW up to 60 frames per second with the optional Codex CDX-36150 Recorder. The camera uses the full 5.9K of the sensor to create oversampled 4K, resulting in improved video quality. "