Patent: Canon Back Side Illuminated (BSI) DPAF image sensor

[Mt Spokane Photography]

It wasn’t all that long ago that we were being told that this is only useful for small sensors. What’s changed?

Video has changed it (faster readout speed due to a better ground plane, as has the ability to put more complex circuitry on the sensor. Multi layer sensors as in the Sony A9 are the next step as far as faster sensor readout, but very expensive. Photosites on sensors are about as good as they can get as far as low light, some sort of tech breakthru would be needed to have a big boost, so small gains from BSI are now the best low cost way to improve. We see lots of claims of new tech, but most of them are really just fishing expeditions looking for more R&D money.

 

[dak723]

Yes, sensors are about as good as they can get, and there has been very little improvement for years now.

But "gear-heads" need to believe that technology can be improved endlessly. And they will continually flock to those companies that give them "new" - even if "new" is no better than old. Thus Sony is wildly popular while Canon is continually bashed. Even though the photos are essentially indistinguishable for most photographers (not counting differences in color, which of course have nothing to do with being "new" and thus doesn't count).

 

[3kramd5]

Video has changed it (faster readout speed due to a better ground plane, as has the ability to put more complex circuitry on the sensor. Multi layer sensors as in the Sony A9 are the next step as far as faster sensor readout, but very expensive. Photosites on sensors are about as good as they can get as far as low light, some sort of tech breakthru would be needed to have a big boost, so small gains from BSI are now the best low cost way to improve. We see lots of claims of new tech, but most of them are really just fishing expeditions looking for more R&D money.

Can you elaborate on how a better ground plane improves readout speed? Also, ground plane in the semiconductor or in the circuit card?

 

[jd7]

Yes, sensors are about as good as they can get, and there has been very little improvement for years now.

But "gear-heads" need to believe that technology can be improved endlessly. And they will continually flock to those companies that give them "new" - even if "new" is no better than old. Thus Sony is wildly popular while Canon is continually bashed. Even though the photos are essentially indistinguishable for most photographers (not counting differences in color, which of course have nothing to do with being "new" and thus doesn't count).

Sony is wildly popular and Canon is bashed in some (many?) online forums ... but it's the other way around in the market. My guess is the market generally agrees that new isn't necessarily better here, even if the gear head segment struggles with that.

 

[juststeve]

This patent actually makes for fascinating reading. Yes, I lead a boring life.

But the majority of the patent is not so much about back side illuminated sensors, although there is BSI stuff there, it is about introducing impurities like boron and arsenic to the light sensitive silicon layer(s) in different concentrations and depths to create signal amplification and isolation. The patent even talks about sensitivities changing to varying wave lengths at varying depths of the light sensitive layer. That sounds a bit Foveon-like.

An important clue is the n and n+ layers of the top photo sensitive layer. Apparently, the impurities to the silicon are injected and then layered here as the n and n+ layers.

Other parts of the patent mention things like on sensor memory and on sensor noise reduction. Sounds like Canon may be getting ready to give Sony's A9 sensor a run for the money, like by the 2020 Olympics, perhaps.

The caveat here is that I am neither an engineer or a patent lawyer, so I could be full of poo.

 

[Mt Spokane Photography]

This patent actually makes for fascinating reading. Yes, I lead a boring life.

But the majority of the patent is not so much about back side illuminated sensors, although there is BSI stuff there, it is about introducing impurities like boron and arsenic to the light sensitive silicon layer(s) in different concentrations and depths to create signal amplification and isolation. The patent even talks about sensitivities changing to varying wave lengths at varying depths of the light sensitive layer. That sounds a bit Foveon-like.

An important clue is the n and n+ layers of the top photo sensitive layer. Apparently, the impurities to the silicon are injected and then layered here as the n and n+ layers.

Other parts of the patent mention things like on sensor memory and on sensor noise reduction. Sounds like Canon may be getting ready to give Sony's A9 sensor a run for the money, like by the 2020 Olympics, perhaps.

The caveat here is that I am neither an engineer or a patent lawyer, so I could be full of poo.

At last, someone else who has actually read the patent. It is pretty difficult to wade thru, I had to print it out and mark the various claims versus the numbers on the drawings.

BTW, all CMOS sensors have noise reduction built in to the sensor, even Canon's first CMOS sensor has it.Its nothing like a foveon Sensor, Canon has patents for those as well. The issue with front side wiring is that there is no room for more complex noise reduction and amplification, BSI allows that, and layered sensors are better yet.

Its about improving the pixels for dual or quad pixel sensors by tweaking the manufacturing process.

 

[MrToes]

ABOUT FREAKING TIME CANON!

 

[drob]

I can't pretend to know about the pluses and minuses of BSI sensors...I do own a 3-4 year old Samsung NX50 with a BSI 28MP sensor...so Canon hasn't looked into this when Samsung was doing this years ago? I find it hard to believe. I'll give them my NX50 so they can tear it apart and backward engineer newer sensors. I'm still a little bitter about Samsung exiting the camera market...especially after their "Ditch the DSLR" campaign.

 

[Mr Majestyk]

Canon cannot be taken seriously in mirrorless without all new sensor tech (for them that is). The whole myriad of issues with the EOS R stem from the not-designed-for-mirrorless sensor from the 5D4 being used in the first place. Read speed is king in mirrorless, a new Sony class sensor would transform the EOS R series in AF, eye-AF, video, DR, VF lag etc. If they can deliver then I may just stand by them for a lot longer to come. I want to like the EOS R, they have some basics right, nice size, weight and ergonomics but this is almost Rebel class in most ways.

 

[Kit.]

Canon cannot be taken seriously in mirrorless without all new sensor tech (for them that is).

You cannot be taken seriously with the first sentence like this. Doesn't Canon dominate the mirrorless market with their current sensor tech?

 

[chik0240]

I just hope they can get me a 5D mk V with a sensor of better shadow noise which I once a while would prefer and that’s all I would want

 

[melgross]

Video has changed it (faster readout speed due to a better ground plane, as has the ability to put more complex circuitry on the sensor. Multi layer sensors as in the Sony A9 are the next step as far as faster sensor readout, but very expensive. Photosites on sensors are about as good as they can get as far as low light, some sort of tech breakthru would be needed to have a big boost, so small gains from BSI are now the best low cost way to improve. We see lots of claims of new tech, but most of them are really just fishing expeditions looking for more R&D money.

Well, yeah, it was mostly an observation. I know about the advances. I just find it interesting that just a few years ago “experts” were insisting that it was useless for serious photography, and yet, here it is!

 

[Mt Spokane Photography]

Well, yeah, it was mostly an observation. I know about the advances. I just find it interesting that just a few years ago “experts” were insisting that it was useless for serious photography, and yet, here it is!

What I heard (long before video in DSLR's) was that there were diminishing returns on the % increase in sensor space as sensors got larger. It was said that the first sensors to use it would be cell phones where there is a large advantage, with larger sensors well off in the future because its expensive and difficult to do with large sensors. That's still true, but now we have video which we did not have then, and we have users demanding more FPS, so Sony has introduced them. I think that Canon is a ways off because dual pixel sensors will be much more difficult to do. They may even go straight to multi layer bonded sensors.

 

[YuengLinger]

Even if this approach leads nowhere, at least Canon is covering its backside!

 

[3kramd5]

Even if this approach leads nowhere, at least Canon is covering its backside!

 

[blackcoffee17]

You cannot be taken seriously with the first sentence like this. Doesn't Canon dominate the mirrorless market with their current sensor tech?

Dominating the market with the M series is one thing, but they are behind in sensor tech. And i don't mind if the DR is 0.5 stops less than the competition but now they don't have a separate AF and metering sensor like in the DSLR's, everything has to be done by the main sensor.

The EOS R has very slow burst rate with AF-C, cropped and soft 4K video and this is likely because they have a bottleneck in sensor readout speed or processing. In AF-C the the camera sometimes drops to 1-2 FPS, that is not good enough.

 

[Kit.]

Dominating the market with the M series is one thing,

Dominating the market cannot be taken seriously? By whom?

but they are behind in sensor tech. And i don't mind if the DR is 0.5 stops less than the competition but now they don't have a separate AF and metering sensor like in the DSLR's, everything has to be done by the main sensor.

The EOS R has very slow burst rate with AF-C, cropped and soft 4K video and this is likely because they have a bottleneck in sensor readout speed or processing. In AF-C the the camera sometimes drops to 1-2 FPS, that is not good enough.

The EOS R has best in class low light autofocus performance and best in class video autofocus performance. All thanks to Canon's sensor tech.

Yes, one needs to pay for that by reading twice as much pixels per frame.

 

[Diko]

Yes, sensors are about as good as they can get, and there has been very little improvement for years now.

But "gear-heads" need to believe that technology can be improved endlessly. And they will continually flock to those companies that give them "new" - even if "new" is no better than old. Thus Sony is wildly popular while Canon is continually bashed. Even though the photos are essentially indistinguishable for most photographers (not counting differences in color, which of course have nothing to do with being "new" and thus doesn't count).

This is not quite the case.

Low light photos a few years ago prior to 5D4 was somewhat behind compared to the back then trendy SONYKON & SONY devices. Talking from personal experiences. It was not nice to see results from similiar lenses but different vendor bodies resulting mostly from lack of enough DR

 

[Diko]

...The whole myriad of issues with the EOS R stem from the not-designed-for-mirrorless sensor from the 5D4 being used in the first place...

Could you please elaborate on that part?

 

[CanonFanBoy]

Yes, sensors are about as good as they can get, and there has been very little improvement for years now.

But "gear-heads" need to believe that technology can be improved endlessly. And they will continually flock to those companies that give them "new" - even if "new" is no better than old. Thus Sony is wildly popular while Canon is continually bashed. Even though the photos are essentially indistinguishable for most photographers (not counting differences in color, which of course have nothing to do with being "new" and thus doesn't count).

Sony is only wildly popular on this website, and mostly brand new account holders.