Canon to start using 0.18um (180nm) process for FF?

[jrista]

Chipworks recently released an article analysing the CMOS Image Sensor (CIS) processes from a variety of manufacturers, including Nikon, Sony, and Canon. Historically, Canon has used a 0.5 micron (500nm) process for all of their FF sensors since the original 1Ds. In the Canon analysis, they noted that Canon has a 0.18 micron (180 nanometer) fabrication process (possibly what they used for the 120mp APS-H?) that they may begin using for future FF sensors:

[quote author=Chipworks]Canon does have a 0.18 µm generation CIS wafer fab process, featuring a specialized Cu back end of line (BEOL) including light pipes (shown below). It is possible to speculate that Canon may be preparing to refresh its FF CIS line to supply devices for a new FF camera system.[/quote]

A move from their 0.5um process to a 0.18um process for FF CIS manufacture would be a fairly significant move for Canon. The accompanying image figure also seems to indicate a double microlens layer...one above the CFA and one below...which could lead to higher Q.E. The article also mentions the use of "Light Pipes", a term I had not heard before. According to a few papers I've read, lightpipes in CMOS sensor design make use of high refractive index materials and a reflective wall in the optical stack the to improve transmission of light from the color filter/microlens to the photodiode, which exists at the end of a narrow tube where all the readout wiring exists (in a frontside-illuminated design). Seems like a lightpipe is an alternative to using a backside-illuminated design that aims to improve Q.E while avoiding some of the complexities and issues with BSI designs. Additionally, the use of copper interconnects should improve efficiency, allowing lower power usage, and hopefully leading to a lower level of electronic noise (the great bane of Canon these days.)

Seems Canon is most definitely not out of the CMOS Image Sensor design game yet. They seem to have some new tricks up their sleeves, and hopefully they will see the light of day in their next FF camera. Ah, competition is good!

 

[Neutral]

Interesting article.

Mostly Interesting is that Canon for the last 10 years starting with the first 1D was using the same 500nm technology process even for the latest 1Dx !!!
And only now Canon is planning to move to 180nm technology process.
This was really amazing for me from technology evolution prospective in general

Just to think of it - Intel for it's latest IvyBridge is using 22 nm technology process and in near future will be moving to 14nm for second generation of Haswell and then to 10nm process for Skylake
http://en.wikipedia.org/wiki/Ivy_Bridge_(microarchitecture)
Plus to this new 3D (Tri-gate) transistors technology that drastically reduce currents leakage and reduce power consumption (http://en.wikipedia.org/wiki/Tri-gate_transistor#Tri-gate_transistors). This in general significantly reduce noise produced by them .

More to this ARM year back demonstrated real 3D circuitry technology in their 1mm cube microchip where only circuit layer is grown above another layer forming multilayer chip (3D instead of flat 2D).
All that combined together is promising amazing things now and in the new future.

And just compare current level of Intel microchips technology (22nm) to the same of Canon (500nm)
Intel introduced 600nm process for Pentium P54C in October 1994 and 350nm process in June 1995.
http://en.wikipedia.org/wiki/Pentium
Though it is not very fair to compare digital circuits with analog ones but still this shows very huge gap in microchips technology process itself - 20 years of difference.
And explanation for that is simple - technology process development is so costly and requires so much investments in R&D and technology equipment that is is possible itself only for few really big players on the market like Intel or AMD or Samsung and other such companies that focus on development and production and manufacturing of microchips for the rest of the industry.
Smaller companies just do not have enough resources to keep pace in microchip technology evolution and if they do so then eventually technology gap would increase and in order to survive they would need to use technology end equipment licensed (leased) for the major players on this market.

To me the best way is not to reinvent the wheel but try to use whatever is already available and to combine all the best in the top level end-user product (similar what Apple is doing including purchasing small companies that invented and patented something really useful but do not have enough resources for further quick development)

If I would be the project development manager in Canon responsible for this imaging area then I would consider establishing partnership with Intel (or AMD or other major players in the microchip technology) to get access to the latest technology processes. Having such technology at hand gives huge possibilities for the new designs for image sensors itself that would be still done within the Company.
Having high density of active elements on the image chip provided by current 3D 22nm process technology from Intel it could be feasible and possible to design new chip with extremely low lower power consumption and current leaks that in turn would could help to reduce electronic noise.
The other thing that would become feasible is to implement phase AF detection for every pixel on the image sensor - that would be biggest revolution in the camera technology ever - just think of the possibilities that this would give to the camera users. Camera could be mirrorless with AF performance of the existing top level PRO DSLRs. And you could be use any variable group of pixels on the sensor to start AF and track subject in AI Servo mode across the full frame.
And if that done on high resolution high ISO performance sensor (e.g. 80 mpx) you could do perfect shots which would be almost impossible to do before - e.g. object tracking without camera move - e.g. shooting acrobatics - you can put in frame the whole performance area , focus on object and camera will be tracking it across all the frame without need to move the camera itself. The same is very useful e.g. for shooting bailey performance.
Then you can do required size crops from the final images and get perfect subject close up.
And if that would be combined also with Foveon type sensor itself that would another step in this revolution
Plus to this each pixel could have it's own ADC with on chip pixel response uniformity calibration processor - this could totally eliminate low light pattern noise . Similar to what is done currently in astronomy for telescope with multi element mirrors arrays where each channel in real time compensated for atmosphere turbulence and light propagations variances.
There so many exiting possibilities - there would be not enough space to count all of them
And this is what could be done now or in near future with current technology level.
Just combine together all the best technology pieces already available around and get the best product ever
I am talking now just about current technology level possibilities -not about the cost - would be high now and this is different subject. But for building prototype - is something that could be considered
So all above is more to the vision of the future and how to make this future to come faster)))

 

[gmrza]

Seems Canon is most definitely not out of the CMOS Image Sensor design game yet. They seem to have some new tricks up their sleeves, and hopefully they will see the light of day in their next FF camera. Ah, competition is good!

My suspicion, which remains speculation as I have no proof, is that Canon has managed to get the very last bit it can out of its 500nm process. - What has been achieved with the 1DX seems pretty amazing considering the maturity of the process.

What I suspect Canon has been trying to do is to commercially leverage the 500nm process to the fullest extent. The current crop of FF bodies has probably been released at the last possible moment when Canon can still get away with using the current process. That gives Canon 3 to 4 years (or thereabouts) until the 1DX and 5DIII have to be replaced (unless market conditions change). - That gives Canon quite a leverage in terms of sweating an asset.

There are two ways we might see a new 180nm sensor first - either in an APS-C format, where the production yields are higher, and as a result the cost per sensor is more palatable on a new fab, or in a high end high resolution FF body where a high cost per sensor can be absorbed through the premium price of the product.

I, for one, am hoping that the next year will show us Canon's hand when Canon's next APS-C sensor gets unveiled. - There seems to be an increasingly strong body of opinion that a new APS-C sensor will reach the market next year - possibly in the successor to the 7D.

 

[marekjoz]

Seems Canon is most definitely not out of the CMOS Image Sensor design game yet. They seem to have some new tricks up their sleeves, and hopefully they will see the light of day in their next FF camera. Ah, competition is good!

My suspicion, which remains speculation as I have no proof, is that Canon has managed to get the very last bit it can out of its 500nm process. - What has been achieved with the 1DX seems pretty amazing considering the maturity of the process.

What I suspect Canon has been trying to do is to commercially leverage the 500nm process to the fullest extent. The current crop of FF bodies has probably been released at the last possible moment when Canon can still get away with using the current process. That gives Canon 3 to 4 years (or thereabouts) until the 1DX and 5DIII have to be replaced (unless market conditions change). - That gives Canon quite a leverage in terms of sweating an asset.
(...)

I also think, that they don't bring to market some technologies they have already developed. Someone could say that this is some naive point of view, but if they can still make money on old technology, then why waste this possibility killing it with quite new devices with superior specification? Everyone - I think - will agree, that they are doing incremental steps forward, rather than revolutionary. Those who bought either 5d3 or 1dx will still spend their money for new toys, if they will be better.
It's just a pity, that the life is too short to see the photos made with full DR, 200MP and 24 bits And if they achieve this moment, what would they work on next? Assuming that they have a technology allowing today to make a sensor having 14 stops DR, 100MP and 16bits at 25600 ISO... would it be smart from their point of view to put it into the next DSLR? I am sure, that the answer is NO and if I would be responsible for that in Canon I would simply not allow it to happen, no matter how much as photographer I would enjoy such a tool.

 

[Woody]

Actually, Paul Pope, a highly reliable source in DPReview, has indicated on many occasions that Canon has the next level sensor technology; Canon is not as far behind as what many think. However, he believes the bean counters in Canon are not willing to move their latest technology into manufacturing phase because of low(er) yield.

 

[nightbreath]

Actually, Paul Pope, a highly reliable source in DPReview, has indicated on many occasions that Canon has the next level sensor technology; Canon is not as far behind as what many think. However, he believes the bean counters in Canon are not willing to move their latest technology into manufacturing phase because of low(er) yield.

They might re-consider this taking into account decrease in profit. Even if they don't gain much in camera sales, they would rather hold the first position than loosing their current place (I assume )

 

[dolina]

Well and good that Canon will be migrating to 0.18um but will it show up in the next generation 1 and 5 series? I chose to skip this generation as I do not shoot as much as last year.\\

BTW folks the single digit bodies tend to have a 3 year product cycle. Double digit 2 years and three or more digits 12 months.

 

[jrista]

For those who think Canon needs to build a new fab. The article clearly indicates that Canon already has 0.18um fabrication process. That means they wouldn't have to build some new multi-billion dollar wafer fab...simply that they would need to start using the one they already for commercial products. My guess is they developed it several years ago, as they would have required something smaller than 500nm to produce a 120mp APS-H sensor, which had 2um pixels.

 

[x-vision]

There are two ways we might see a new 180nm sensor first - either in an APS-C format, where the production yields are higher, and as a result the cost per sensor is more palatable on a new fab, or in a high end high resolution FF body where a high cost per sensor can be absorbed through the premium price of the product.

My thoughts exactly.

FYI, the 180nm sensor mentioned in the Chipworks article is the S100/S110 sensor.
This sensor has been on the market for about a year now (the S100 was announced in September 2011).

So, Canon't 180nm sensor technology is in production already; it's not some R&D project.
The question is, when are they going to scale this process to DSLR sensors.

Indeed, because of yields, it makes the most sense to first implement it in a premium APS-C model (7DII?).
But Canon is not known for doing things that make much sense (6D anyone :). So, who knows.

OTOH, after the disappointing sales numbers from last quarter, Canon is surely under pressure to show improved sales.
This bodes very well for bringing the (pending) 180nm DSLR sensors to market ASAP.

 

[Mt Spokane Photography]

Switching technologies is expensive, and the camera industry is in a tailspin. Canon has released a recent patent which makes a rear illuminated sensor practical for FF and APS-C sensors (It is not used now because there is no benefit).
Unless Canon has already committed and is installing the equipment, don't look for any billion dollar facility updates soon.

 

[Stu_bert]

For those who think Canon needs to build a new fab. The article clearly indicates that Canon already has 0.18um fabrication process. That means they wouldn't have to build some new multi-billion dollar wafer fab...simply that they would need to start using the one they already for commercial products. My guess is they developed it several years ago, as they would have required something smaller than 500nm to produce a 120mp APS-H sensor, which had 2um pixels.

Yes it did, but sufficient for large volume? Unless they predict minimal usage of the new tech, over a 3 year period they would replace most of their sensor tech with it and thus would need the capacity. And that's a big investment.

As another poster stated, it is indeed about the bean counters, they keep the company operating within a risk profile the board / investors are happy with. And with a downturn in profits, I doubt they are to invest majorly anywhere.

I concur with others, bringing out a high end camera, with high MP, is a good signal to your competitors and the market that you have the tech, plus it does not require such high volumes and as you have more headroom to absorb the costs.

I'd vote against the APS-C version unless technicalities prevent Canon from releasing a FF sensor first. Just my tuppence.

 

[gmrza]

Actually, Paul Pope, a highly reliable source in DPReview, has indicated on many occasions that Canon has the next level sensor technology; Canon is not as far behind as what many think. However, he believes the bean counters in Canon are not willing to move their latest technology into manufacturing phase because of low(er) yield.

As has been pointed out - that makes perfect sense. The 1DX demonstrates exactly how much Canon has been able to pull out of the current process. Launching the 1DX, 5DIII and 6D using the current 500nm process probably gives Canon 3 years (maybe 4) to get a 180nm process to a point where it will deliver sufficient margins for full frame sensors. I suspect that the economics of the new process will start working for smaller sensors first - i.e. APS-C.

I haven't been able to dig out any information about what process Canon is currently using on smaller sensors for instance in the G15 - that would be interesting to know as well.

 

[verysimplejason]

I'm just hoping they try it with 6D. If they do, 6D will be selling like pancakes should performance improve greatly. They've done it with 5D2, now is the time to do it with another one.

 

[gmrza]

I'm just hoping they try it with 6D. If they do, 6D will be selling like pancakes should performance improve greatly. They've done it with 5D2, now is the time to do it with another one.

I think that is highly unlikely - the cost base of the 6D has to be significantly lower than the 5DIII. The $2000 to $2100 odd price-tag of the 6D is the high water mark for the price. The price of the 6D will, if anything, follow a similar trajectory to the 7D - possibly testing $1500. Canon will need every bit of margin it can get to sustain that kind of a price drop. That would just not be possible with a new manufacturing process.

 

[dr croubie]

I'm just hoping they try it with 6D. If they do, 6D will be selling like pancakes should performance improve greatly. They've done it with 5D2, now is the time to do it with another one.

I think that is highly unlikely - the cost base of the 6D has to be significantly lower than the 5DIII. The $2000 to $2100 odd price-tag of the 6D is the high water mark for the price. The price of the 6D will, if anything, follow a similar trajectory to the 7D - possibly testing $1500. Canon will need every bit of margin it can get to sustain that kind of a price drop. That would just not be possible with a new manufacturing process.

Yeah, I reckon we'll see a new process launched with a 7Dmk2 or 70D, I doubt the 6D is it (although it would go some way to explaining the price vs features of the 6D). Seriously, they can't use the same 18MP sensor in *yet another* aps-c camera. I'm curious as to whether anyone's x-rayed the 650D, maybe that's already on the new process? Either way, the 70D might use the same sensor as 650D if it's launched first, otherwise if the 7Dmk2 is first then it could lead with the new process.

(Also, maybe that explains the 7D v2 firmware? Obviously it was a stop-gap because the 7Dmk2 wasn't ready for market, could it be troubles with the new process? Or just maybe once they go to the new process, the 7Dmk2 will have so much freakingly better IQ than the 5D3 (or even close-to would be nice), that instead they held off on it until the 5D3 is a year old before the 7Dmk2 launch, so as to not cannibalise too much?)

 

[gmrza]

(Also, maybe that explains the 7D v2 firmware? Obviously it was a stop-gap because the 7Dmk2 wasn't ready for market, could it be troubles with the new process? Or just maybe once they go to the new process, the 7Dmk2 will have so much freakingly better IQ than the 5D3 (or even close-to would be nice), that instead they held off on it until the 5D3 is a year old before the 7Dmk2 launch, so as to not cannibalise too much?)

I suspect you are right - I would expect a 7DII to be announced around a year after the 5DIII was. - The 7D was launched roughly a year after the 5DII.

I don't expect the 7DII to have the same relationship to the 5DIII as the 7D did to the 5DII - a lot of people who could not afford a 1DIV bought a 5DII and 7D. Given the upgrades to the 5DIII AF (AF points, accuracy and f/8 coming) and frame rate, the 7DII will probably fulfil a different role.
For instance, wildlife photographers will probably start to opt for a 5DIII and 1.4x TC rather than a crop body.

That said, I hope a 7DII and enthusiast-level mirrorless will be platforms on which we might see a new manufacturing process being used.

 

[tron]

I don't expect the 7DII to have the same relationship to the 5DIII as the 7D did to the 5DII - a lot of people who could not afford a 1DIV bought a 5DII and 7D.

Now, if -and that's a big IF- 7DII uses the new technology sensors and an AF system that is between the 7D's and 5D3's (hopefully 5D3's) then I cannot think why it couldn't be the new wildlife camera. An 18Mpixel APS-C sensor with a quality close to 5D3's would kill... Of course this would cannibalize the 1DX and the new super telephotos sales

 

[verysimplejason]

I'm just hoping they try it with 6D. If they do, 6D will be selling like pancakes should performance improve greatly. They've done it with 5D2, now is the time to do it with another one.

I think that is highly unlikely - the cost base of the 6D has to be significantly lower than the 5DIII. The $2000 to $2100 odd price-tag of the 6D is the high water mark for the price. The price of the 6D will, if anything, follow a similar trajectory to the 7D - possibly testing $1500. Canon will need every bit of margin it can get to sustain that kind of a price drop. That would just not be possible with a new manufacturing process.

Yeah, I reckon we'll see a new process launched with a 7Dmk2 or 70D, I doubt the 6D is it (although it would go some way to explaining the price vs features of the 6D). Seriously, they can't use the same 18MP sensor in *yet another* aps-c camera. I'm curious as to whether anyone's x-rayed the 650D, maybe that's already on the new process? Either way, the 70D might use the same sensor as 650D if it's launched first, otherwise if the 7Dmk2 is first then it could lead with the new process.

(Also, maybe that explains the 7D v2 firmware? Obviously it was a stop-gap because the 7Dmk2 wasn't ready for market, could it be troubles with the new process? Or just maybe once they go to the new process, the 7Dmk2 will have so much freakingly better IQ than the 5D3 (or even close-to would be nice), that instead they held off on it until the 5D3 is a year old before the 7Dmk2 launch, so as to not cannibalise too much?)

I don't think it's impossible though highly unlikely. With the recent Canon sales performance, anything and everything is possible. Sometimes the goal is to sell move even at the cost of little profit per unit. Also 6D will be competing with D600 which already have superior specs at least on paper. In order to compete, Canon might do something revolutionary in their lineup. I don't know what it is but certainly WIFI and GPS isn't it. 5D3 at least has better AF, better build quality and better in low-light performance than D800 (without downsizing). Of course, I might be wrong. You know how stubborn Canon is...

 

[LetTheRightLensIn]

Chipworks recently released an article analysing the CMOS Image Sensor (CIS) processes from a variety of manufacturers, including Nikon, Sony, and Canon. Historically, Canon has used a 0.5 micron (500nm) process for all of their FF sensors since the original 1Ds. In the Canon analysis, they noted that Canon has a 0.18 micron (180 nanometer) fabrication process (possibly what they used for the 120mp APS-H?) that they may begin using for future FF sensors:

[quote author=Chipworks]Canon does have a 0.18 µm generation CIS wafer fab process, featuring a specialized Cu back end of line (BEOL) including light pipes (shown below). It is possible to speculate that Canon may be preparing to refresh its FF CIS line to supply devices for a new FF camera system.

A move from their 0.5um process to a 0.18um process for FF CIS manufacture would be a fairly significant move for Canon. The accompanying image figure also seems to indicate a double microlens layer...one above the CFA and one below...which could lead to higher Q.E. The article also mentions the use of "Light Pipes", a term I had not heard before. According to a few papers I've read, lightpipes in CMOS sensor design make use of high refractive index materials and a reflective wall in the optical stack the to improve transmission of light from the color filter/microlens to the photodiode, which exists at the end of a narrow tube where all the readout wiring exists (in a frontside-illuminated design). Seems like a lightpipe is an alternative to using a backside-illuminated design that aims to improve Q.E while avoiding some of the complexities and issues with BSI designs. Additionally, the use of copper interconnects should improve efficiency, allowing lower power usage, and hopefully leading to a lower level of electronic noise (the great bane of Canon these days.)

Seems Canon is most definitely not out of the CMOS Image Sensor design game yet. They seem to have some new tricks up their sleeves, and hopefully they will see the light of day in their next FF camera. Ah, competition is good!
[/quote]

Hopefully. Hopefully they take advantage of it and can fit column ADC and stuff on their sensors now.
It sounds like there is some promise there, hopefully it will be delivered.

 

[DB]

A lot of people here are making assumptions about Canon not releasing new products for a variety of reasons, from pricing policy, to milking older technology, to marketing decisions, even to a reluctance to invest capital expenditure in the new machinery required for even thinner chips and sensors, but few people are considering the simple logistical facts: manufacturing enough quantity (inventory) prior to shipping.

Canon is a victim of it's own success, it has made and sold more than 50 million digital-SLR bodies since 2000 (starting with the D30, then D60 up to the 1DX and 6D bodies). In the first 8 years Canon sold 24 million DSLR's or about 3 million per year. Over the last 4 years they've sold almost as many again, now averaging 6.5 million DSLR bodies each year - that is nearly 542,000 per calendar month and more than 27,000 per working day! Assuming that Canon requires a minimum of a 3-month inventory of a new product prior to launch + another 2-3 months supply in-transit (got to ship those boxes half way around the world, time in bonded warehouses, customs etc.), then they have to have a half years Sales Supply in hand before launch. This is probably why the 6D is not being delivered till early-December (which means they could have been manufacturing it since May).

These logistical supply constraints add up to a considerable lead time. This delay in manufacturing enough new product to physically deliver to specialist camera stores goes a long way to explaining why many CR bloggers feel that Canon has slipped behind competitors in the recent product launch stakes - perhaps the decisions to go ahead with a specific feature-list were made 18 months prior to actual launch, or some figure like that. Whichever way you look at it, Canon has to produce hundreds of thousands if not millions of sensors and digic processors for their cameras, months even before they arrive in your country.

It is a physical impossibility for Canon to launch e.g. a 7D2 in Spring 2013, if they're not already making the parts for that particular body right now....in large quantities.