It seems an advancement in technology inside the EOS R1’s image sensor is being overlooked when comparing it to the modern DPAF stacked sensors in EOS R5 Mark II and EOS R3. We have had a lot of “isn’t the autofocus the same?” types of comments whenever the EOS R1 and EOS R5 Mark II are brought up in a comparison conversation, which is going to be an inevitable comparison. That’s fine, we don’t think Canon explained it all that well when the two cameras were announced.
Don’t feel too bad, we haven’t seen any reviewers (no, we haven’t seen them all) address this difference either when comparing the EOS R1 and EOS R5 Mark II/EOS R3.
This is a groundbreaking sensor design from Canon. We’ll briefly run down how autofocus sensors have evolved to lead us to where we are today with the EOS R5 Mark II and EOS R1.
DSLR Autofocus
Back in the day when DSLRs were roaming the earth, we didn’t have some of the problems we have today, because our auto-focus sensors were separate from the image sensor and looked something like this;
This is one of the basic sensors (the 7D Mark II sensor), and you’ll notice that there are horizontal and vertically orientated lines. These lines would then group up into both orientations and form what we call a “cross-type” auto-focus point.
There was also one other focus point, arranged in an “X” orientation that allowed for even greater precision, and also allowed for what is called “major defocus” autofocus. Major defocus isn’t something that we are talking about today, but just a basic description is that on very fast lenses especially super telephoto lenses, if the lens is focused far away from the target, the entire target is blurred so much that phase difference can’t determine how far away the subject is from the camera. Major defocus focus points in DLSRs helped with that, by being able to determine focus based on what was a blurry mess for other auto-focus points.
Dual Pixel Sensors
With the advent of the EOS 70D, Canon announced dual-pixel sensors. This allowed the sensor to directly auto-focus with the speed and precision of a DSLR auto-focus sensor.
The sensor works by taking the pixel and splitting it in half (thus dual pixel). Canon can then read the sensor and measure the phase difference between each half of the pixel to determine if the feature is in focus and if not, how far away it is.
This was groundbreaking for Canon and set the stage for the Canon mirrorless cameras we have today.
Dual Pixel Sensors Have an Issue
However, we have a problem with dual-pixel sensors: all the pixels are orientated in the same orientation.
This gives a Canon mirrorless camera an autofocus “blind spot”. The camera can focus on either horizontal or vertical features when the camera orientation is parallel to the feature. For instance, your R5, R5 Mark II will struggle to lock onto horizontal features, when the camera is in landscape orientation. This is because on all of Canon’s DPAF sensors up to the R1, all the DPAF pixels were orientated in the same direction.
But this is no longer the case when the R1 sensor.
Canon EOS R1 Dual Pixel Sensor
With the R1 sensor, Canon has changed the orientation of one of the two green pixels, allowing edge detection to now occur on any orientation of the edge that you are attempting to focus on.
This allows the R1 to focus reliably on any feature regardless of its orientation. I am sure you can see the benefit of this, as the last thing you want in fast-paced sports photography is to have the sensor simply not recognize a feature.
Are there any downsides?
I don’t think the sensor is that much more complicated to design for Canon, but it does need a different sensor design than Canon has done since dual-pixel sensors were announced back in 2013. There is a bunch of “stuff” that exists around each photodiode pair, and that will have to be adjusted on a sensor that supports cross-type. While that shouldn’t be an insurmountable problem – Canon spends a huge amount of sensor research and development on improving the subtle aspects of its sensor designs to extract the maximum performance from them.
Any change to the underlying structure of the sensor has to be supported by Canon’s patent library. Even on a stacked sensor where there are separate “substrates” sandwiched together to create the sensor, the engineers have to plan the sensor to be as efficient as possible for the wiring paths and components that absolutely need to exist on the photodiode substrate.
Even after the sensor is created, then there’s the math. The camera still has to calculate the phase difference between the pairs of the photodiodes on the dual-pixel auto focus sensor, and this determines the distance the feature is away from the sensor.
I’m unsure about the maths and if that gets considerably more difficult when Canon is calculating out the phase difference and doing all those predictive distance calculations. But my thought is that it does complicate the calculations, and that is probably one of the reasons we haven’t seen it until now.
The Future
I hope most Canon sensors will be slowly changed to include the cross-type green pixel.
This will improve the focus of all cameras in Canon’s lineup. Right now, we know this isn’t the case with the R5 Mark II (and also didn’t exist on the R3) so we can assume that Canon right now is making this a 1 series feature. But we can hope that as with other technologies that Canon has introduced in the 1 series, it moves downward through Canon’s other professional and prosumer camera systems.
Will Canon improve the sensor? Certainly. Canon had other designs to combat this issue, including the Quad Pixel AF sensor design. Instead of just one subpixel having a different orientation, the entire subpixel can read the phase difference in horizontal or vertical directions, allowing for more reliable autofocus. But I suspect these sensors also come with performance penalties, so it may be a while until we see them.
Canon EOS Autofocus Technology
|
When you purchase through links on our site, we may earn an affiliate commission. Here's how it works. |
The R7 shows this quite well: more advanced AF algorithms, much worse AF performance.
From the readout speed alone I would expect a noticeable improvement.
That's definitely a rabbit hole to go down when both of these cameras are being used in situations where these differences may show up. I think it'll be with acquisition/tracking in messy scenes.
Or it'll be that time when tracking a BIF and the 1 in 20 that isn't perfect was the one you wanted to be.
We all need something else to read, write and converse about between the product cycles.
And just to state again, I feel the R3 was also kind of dismissed and downplayed and ignored by many that went for the R5 and R5 II. A store rep even recommended me the R6 II over the R3 because of the specs saying it’s «basically the same camera». I left. The R3 is the perfect compromise for me and has performed way beyond any other Canon body I’ve owned or tried..
They could. They started with contrast detection and then they moved to phase detection on the DPAF equipped EOS cameras.
That was the first implementation of DPAF, DPAF II took a big leap and it only continues to get better.
A major innovation! Well done!
Sadly, the "internet experts" will still be focused on "only" 24 MP...disregarding real technological advances (unless they came from soni).
I am wondering if the smaller phase difference of smaller pxiels may also need more processing or cause problems.
Just a thought.:unsure:
With the r1 vs r5ii, the processing appears to be identical. The big difference is the actual AF sensors (cross type vs non). I’m definitely curious to see real world experiences of people who use both.
Also worth noting the price difference between the 1 and 5 series is smaller than it has ever been before