Has Canon Closed the Gap, or Is It Hobbled by Poor Performing Sensors?

Has Canon Closed the Gap, or Is It Hobbled by Poor Performing Sensors?

Canon is the most popular camera manufacturer in the world, but it has had a reputation in the past for using sensors inside its cameras that have lagged behind the best of the competition. Why is this and what does the future hold?

Digital image sensors are an expensive business, but with annual sales of smartphones at around 1.5B units, that's an awful lot of sensors, particularly when you factor in multi-camera models. The burgeoning image sensor business doesn't stop there, with industrial vision systems (particularly robotics and automotive, but also medical and science applications) adding to global demand that is continually rising. In fact, demand is such that earlier this year, Samsung announced that they were converting an existing DRAM manufacturing line over to camera sensor production. This isn't unusual (Samsung did the same thing back in 2018), as around 80% of the process and equipment overlaps; however, what gives an impression of scale of the capital and ongoing production costs is that this will still see a write-down of some $815 million to complete it!

Sensor Manufacturing

In terms of the global image sensor market, Samsung holds around 18% to Sony's 49%, with OmniVision trailing at 9%. In a market that's vastly expensive to enter and — at least for smartphone cameras — has low margins, the scale of production is all-important. Sony has an ability to invest in R&D to an extent that others are unable to. Of course, a single large manufacturer is never a recipe for a competitive sector, and it's for this reason that many camera manufacturers try to spread their risk between different sensor suppliers, although exactly which sensors are used in which cameras can only be surmised after a tear-down, and even then, it might be difficult to ascertain. Nikon has long used Sony sensors, but also produced its own LBCAST sensors and over the years continued manufacturing, although it appears they no longer have a fab facility. They now either source from other manufacturers or have their own designs produced. As a result, they are also quite promiscuous and not averse to using a range of suppliers (Toshiba, Aptina, and the recently rumored Tower Semiconductor).

This brings us to Canon, who is a major manufacturer with a particular focus upon industrial applications, although they make a point of stating that they produce the sensors in their EOS cameras. Holding somewhere around 5% of the market, Canon is an important — although small — player, which is something you don't often say about their business! Sony has dominated the sensor market both in terms of volume and top-end camera sensors for the best part of a decade, with Canon trailing in the image quality stakes. This is clearly something that Canon is aiming to redress: while they show no interest in entering the smartphone sector (which is where the volume lies), they have increased their production capacity and in 2016, announced they would sell to third parties. This began in 2018 (via Phase1); however, these are industrial sensors, which reiterates their manufacturing focus: horizontal expansion into related imaging markets.

EOS-1D X Mark III Image Quality

In terms of image quality, the gap has been narrowing with each iteration of sensor although DXOMark's recent review of the EOS-1D X Mark III again raises the issue of image quality. DXOMark score camera sensors on the basis of testing color, noise, and ISO sensitivity, with the overall score an average of the three. Garnering a score of 83, we could describe DXOMark's assessment of the image quality as "lackluster":

it’s not quite at the cutting edge in our metrics for sensor performance, but there’s far more to it than that.

The nearly two-year-old Nikon D850 and more recent Sony a9 II outperform in these stakes, scoring 100 and 99 respectively. The D850 isn't a direct competitor, although the a9 II is. Perhaps more pertinently, the low-light quality of the EOS-1D X III is only broadly on a par with the Nikon D5, which was released in 2016. Is the 1D X Mark III a miss for Canon in what should have been a superlative Olympic year?

It's still too early to get a consensus on how the camera performs; however, both the DPReview studio test and Photons to Photos dynamic range test paint a different story. DPReview clearly show a significant jump in performance over the EOS-1D X II, whilst ISO dynamic range performance at Photons to Photos shows a stop improvement at lower ISOs. More importantly, it edges ahead of the a9 II and offers significantly better low ISO performance than the Nikon D6, which trades this for very high ISO performance. It's also important to note that DXOMark does not publish their methods, which are frequently called into question.

Future Sensor Strategy

All of which makes the upcoming release of the R5 and R6 immensely important for Canon, particularly given the headline specifications that have been teased so far. Make no mistake, this is an era-defining year for Canon as it transitions its camera business from DSLR to mirrorless. Nikon beat it to market by some considerable distance with the release of the Z 6 and Z 7, although it's as much about the lens lineup as it is about the camera bodies themselves. The lead it has given Sony, both in terms of the time advantage for developing a system and relinquishing market share in the ILC sector, is substantial. It is no longer number one in its home market, a position it has held for a long time. Manufacturers may bemoan the cataclysmic drop in unit sales, but the truth is that this isn't about cameras. It's about companies that produce cameras, and their manufacturing influence goes much broader. The pivot to mirrorless and implosion of camera sales is an opportunity for disrupting the sector, one that Sony has taken advantage of.

Could it be that 2020 will actually be a return to form for Canon that sees its engineering teams not only fill out its RF lens lineup, but produce genuinely world-class mirrorless camera bodies? On top of that, could we also see Canon's sensor fabrication finally leapfrog Sony in terms of image quality?

Mike Smith's picture

Mike Smith is a professional wedding and portrait photographer and writer based in London, UK.

Log in or register to post comments
72 Comments
Previous comments

I don't see ads on Fstoppers or elsewhere unless I specifically unblock them. This works for me and millions of others. Free by Chrome. Using something else, just search out an ad blocker that works for you.

https://chrome.google.com/webstore/detail/adblock-plus-free-ad-bloc/cfhd...

Thanks, will give it a try....

I don't care about sensor scores for the following reason.
I had to take a photograph of the unilluminated dark interior of a shed using an f/2.8 lens on a six year old canon 7d mk2 in order to actually see what was in there, as I couldn't actually see anything in it with my eyes. I forgot that I had disabled the flash, but it didn't matter. The resulting photograph still showed me what was in there.

You are not using that word correctly. lol.

hobnail [ hob-neyl ]
noun
a large-headed nail for protecting the soles of heavy boots and shoes.
a small allover pattern consisting of small tufts, as on fabrics, or of small studs, as on glass.

I am sure you meant "hobbled"

This is a purely academic debate as while differences may be seen on screen at high magnifications and at extreme abuse of files, precisely zero people will see differences in final results. Moreover, the PP skills employed will have a far greater effect on final results than any sensor. This is particularly true in a situation with a rotten file.

Here we go again with tech talk. I think that a great sensor can make the photographer's job easier. Nevertheless, photographic skill can make up for equipment limitations. If you walk around doing street photography in low light, you really want that great high performing sensor. If you are shooting landscapes on a tripod or portraits in the studio, you just give the camera more light by exposing to the right. I see many fine photographs on this site taken with Canon cameras. In fact, I see no evidence that pictures from Sony's are any better than those from Canons. In fact, one of my favorite photographs on F Stoppers is one where the photographer used a Canon 5D MK II, yes a 12 year old camera. Sensor technology is just one of many considerations for purchasing a camera. If you already have a camera and you know how to make great photographs, it is not that compelling to go out and buy another camera just for the sensor alone. Just my opinion...

As anyone who has used this camera knows, the score this camera received at DxO Mark says more about DxO's "secret" weighting formula for deriving a composite score that can't be peer reviewed than it does about the capabilities of this camera. Even using DxO's own measurements, the composite score makes absolutely no sense as an indicator of what one can expect from this camera.

Looking at photonstophotos, which is entirely transparent with regard to their methodology, a much more accurate indication of what this sensor is capable of can be seen.

Honestly, it looks like DXO used electronic shutter for the test which has a lower bit depth. Photons to photos has results for both settings independently, and the 1DXIII with mechanical shutter trounces a lot of other bodies. The only way DXO's results make sense to me is them using electronic shutter:

https://www.photonstophotos.net/Charts/PDR.htm#Canon%20EOS%201D%20X%20Ma...(ES),Sony%20ILCE-9M2

Isn't this a question better asked six months from now?

The author talks about making cameras and then jumps to the image sensor semiconductor chip as if that is only important part for making an image. Nothing could be farther from reality! Images are made by the camera in optimizing the entire combination of the capabilities provided by the many parts of the camera including software, focusing technologies, lenses, user interfaces and ergonomics all of which contribute to the making of fine images.

The article obsesses on dynamic range as the only thing that matters and many modern cameras are simply not limited in my opinion by their dynamic range while creating great images but the are limited by a combination of one or more of the above mentioned capabilities working together well. At least this is my opinion

I see this seems to be also the same opinion of many of the comments made above. Once the dynamic range is large enough like it is in many cameras today it ceases to be the most important limit to making good camera images.

Canon is the number #1 camera company in the world because they do a industry leading job a delivering the complete end to end system for making camera images that most photographers like me want at a price we can afford to pay. This is more important to many buyers than small improvements in any one of the many image making components.

DXO Mark testing procedures are truly flawed. I work for a large motion picture rental company. We test sensors on all cameras we use, its a very controlled process but its not the only process. We can accurately read the dynamic range and see the variance between sensors in the same type of camera including Sony cameras. We also separately test lenses and finally test lenses on cameras. Because of our size we test batches of cameras & lenses.
From a technical perspective Sony make sone of the best sensors (so do TowerJazz). However a 10year old sensor Arri have made for them is still the most widely used. Why is that? Technology is only part of the story artistic lighting, post processing, the integration of different types of lenses, filters etc all add or detract from what we eventually view.
Cameras & lenses are tools nothing more & nothing less its how you use them that counts.

Well you've said it all: "It's also important to note that DXOMark does not publish their methods, which are frequently called into question."
So let's forget about these b.s. tests, as we have no clue what they are actually testing.

I have (and had) some Sony bodies although I am primarily a Canon shooter. Yes they are great, but I have never managed to achieve the same finesse in post-processing with any Sony file compared to Canon. Is that taken into account in those tests? I am far from being the only one noticing this.

I was pretty surprised by DXO's review of the 1DXIII specifically because it was so far in contrast to that of photons to photos. Based on the results I am betting DXO did their review using only electronic shutter which has a lower bit depth. According to photons to photos, the 1DXIII out performs both the 1DXII and the a9II in dynamic range at base ISO when using mechanical shutter, while electronic shutter lands right about where DXO has it:

shorturl.at/afruO

The problem is that in my opinion from testing, the 1Dx Mark III still loses, by a very slight margin to the 2012 Nikon D4 in base ISO dynamic range. I tested the 1Dx Mark III next to my D5 and D4s, but many say the D4 has better dynamic range as the D4s was tuned more towards high ISO like the D5. So congratulations Canon your new $6500 camera has almost as good dynamic range as the 2012 Nikon D4!

Totally agree, but the AF in the D5, D6, and 1DX3 are way better than the D4.

I must be the only weirdo on this earth who's happy with his EOS 6D from 2014. Still rocking, still delivering.

One of my mates is also very happy with his 6D, and he's waiting for Black Friday to buy a... 6D Mark II. He's happy with his work, his clients are happy with his work, his prints look great - and that's only a problem to pixel peepers and those who are paid by or are trying to be paid by Sony to make articles and videos reminding everyone that Sony's alpha series has always had superior DR.

True, since I stopped with pixel peeping, I started enjoying photography as it is. My 6D delivers great results, professional grade ones, and I don't feel I need to change to anything new anytime soon.

Are we really sure is Canon that should close the gap?

While I am just an amateur, I don't feel at all limited by Canon's sensors. If it is easy enough for someone like me to avoid being limited by Canon's sensors, it should be child's play for a professional. When you have to push a file by more than a couple stops to see a meaningful difference, it doesn't matter anymore.

Shouldn't there be a correction on this article after DXOMark corrected their Canon EOS-1D X Mark III review? It's pretty comparable to Sony a9 II: https://www.dxomark.com/Cameras/Compare/Side-by-side/Sony-a9-II-versus-C...

The cool thing is that the R5 now has the best performing sensor in terms of dynamic range of any full frame camera (photons to photos.net) beating out the best from Sony. That's pretty amazing for a company with only 5% of the sensor share.

Additionally, the R5's sensor produces lower noise and better colours than the A7RIV at high ISOs.