Pages: 1 ... 11 12 [13] 14   Go Down

Author Topic: Nikon in difficulty?  (Read 94289 times)

davidgp

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 758
    • davidgp fotografia
Re: Nikon in difficulty?
« Reply #240 on: March 19, 2017, 06:00:52 am »

http://www.sonyalpharumors.com/sony-interview-ir-best-sony-sensors-will-remain-house-not-outsourced-nikon/

What was I saying about Sony withholding their top-tier sensors from competitors?


Yes, I remembered this conversation when I read the Image Resource interview yesterday. I didn't read the whole topic again, but I think I have to recognize that some of my informations were wrong.

If Nikon does not enter in mirrorless or video business (that maybe it is an error for them, but that it is another story), I don't see this as critical. For sure they will lose a bit of DR to whatever Sony releases next, probably less fps, and lower resolution. AF on chip not needed for Nikon unless they want to enter on video seriously...

What I'm seeing more it is that maybe Nikon needs to pay for custom designs of chips for them... Not sure if Sony is willing to do so... They did it with Phase ONE first, and later other medium format players, but they are not competing there... They are also making two variants for m43 20 megapixels chip, one for Olympus with phase AF on chip and other for Panasonic without it... (Although this last part is just don when they add the microlenses layer over the chip).

I think the main error here is for nearly everybody but Canon to go to Sony for sensors... This allowed Sony to develop faster their sensor technology (well, mobile phone market played more important role here)... Now probably they have a high dependency on Sony for their own taste...


davidgp

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 758
    • davidgp fotografia
Re: Nikon in difficulty?
« Reply #241 on: March 19, 2017, 06:10:29 am »


Buying from someone else might be an option, but who? Certainly not Sony or Canon, or the companies bought up by them.

They are other CMOS sensor players one the market... Canon it is not even one of the most relevant ones:
http://www.yole.fr/iso_album/illus_cis_playersranking_yole_june2016.jpg

Most of them are doing sensor not for cameras but for mobile phones (much bigger market than cameras and from where Sony is bringing most of its technology to their cameras: BSI, DRAM buffer...), also, nowadays everybody is trying to sell their sensors to car manufactures...

So, they are other players willing to get money for developing sensors... Samsung had a good one some years ago...

shadowblade

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2839
Re: Nikon in difficulty?
« Reply #242 on: March 19, 2017, 07:04:07 am »

They are other CMOS sensor players one the market... Canon it is not even one of the most relevant ones:
http://www.yole.fr/iso_album/illus_cis_playersranking_yole_june2016.jpg

Most of them are doing sensor not for cameras but for mobile phones (much bigger market than cameras and from where Sony is bringing most of its technology to their cameras: BSI, DRAM buffer...), also, nowadays everybody is trying to sell their sensors to car manufactures...

So, they are other players willing to get money for developing sensors... Samsung had a good one some years ago...

Making a phone sensor is quite different from a camera sensor. It mostly comes down to size - the camera sensor is much larger. A fab plant capable of making a camera sensor can make a phone sensor (provided it can make the circuitry small enough for the required specifications - newer technology is often introduced in phones/small sensors first, because it's cheaper than introducing it in a large-sensor plant) but a fab plant built for making small sensors can't possibly make a larger sensor. Fab plants making large sensors are few and far between.
Logged

davidgp

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 758
    • davidgp fotografia
Re: Nikon in difficulty?
« Reply #243 on: March 19, 2017, 07:42:03 am »

Making a phone sensor is quite different from a camera sensor. It mostly comes down to size - the camera sensor is much larger. A fab plant capable of making a camera sensor can make a phone sensor (provided it can make the circuitry small enough for the required specifications - newer technology is often introduced in phones/small sensors first, because it's cheaper than introducing it in a large-sensor plant) but a fab plant built for making small sensors can't possibly make a larger sensor. Fab plants making large sensors are few and far between.

No, it is the same sensor plant. This is CMOS technology after all, it is the same base technology to build microchips, like processors and so on. So, basically this works like this:

  • They have this big rods of pure Si - Silicon that are cut into thing wafers, typical sizes now a days are 300 mm in diameter. Purity of the Si is very important, as we will see later.
  • Over each one of these buffers, in really clean rooms, they start to apply several chemical and physical process. This is basically growing the different transistors, circuits, and sensor parts. In each one of this wafers they build several sensor chips. If they are little ones for smartphones, they can make more, if they are bigger ones for medium format they can make less.
  • After everything is finish, they wafer are cut into the sensors and testing starts.

Here is where physical size of the sensor plays a big role. So, if they Si wafer has an impurity in any of the parts of the sensor, that sensor is not going to work. If any of the transistors, sensors, circuits does not work, that sensor needs to be discarded.

So, if you are making like 10000 sensors per wafer (invented number... I didn't do the math of the area of a mobile phone chip vs the area of the wafer), you are getting a high ratio of good sensors, even if you have some impurities or errors in the manufacturing, you still getting high number of good sensors per wafer that you can sell.

If you are making like 100 sensors per wafer (for example, full frame), you are probably getting worst ratio between good vs bad sensors. This is one of the main reasons bigger chips are more expensive. Apart that you are getting less good chips, the rest of the process to build the chips over the wafer is the same, so you have to divide the total costs between all the sensors coming out of the wafer, if you are getting more sensors, less they will have to pay.

Trying new technologies, like for example BSI, it is better to do this first in smaller ones. Because the technology is new and needs a tuning in the manufacturing process, the firsts wafers will have a very low yield, manufactures try to build first smaller sensors since it is easy for then to still get a good number of sensors that work (or the most simply ones, for example, memory chips are typically the ones used by Intel when then try a new physical/chemical process, they are the simplest chips they build, so easy to see how to fine tune the process). When that process is fine tune, you can take the risk of building bigger sensors with it and still get a good yield performance, so you are not risking too much money in the process.

So, big or small, all sensors are made in the same buildings. Take a look at the earthquakes that affected the manufacturing plants of Sony, all sensors lines were affected. Probably smartphone ones were the first one to recover because of what I said before, after rebuilding everything, having to calibrate again all machinery, doing it with the sensor size that has the highest probability of getting highest rate of good sensors make more sense. After those lines were starting to get good yield rates, Sony started to getting back to building bigger sensors.

Here: http://blog.teledynedalsa.com/2012/02/the-art-of-making-large-image-sensors-or-quilting-with-silicon/ and here: http://www.siliconimaging.com/cmos_fundamentals.htm you have images of wafers before being cut. Here you have a nice schema about how many full-frame sensors, APS-C sensors and 1" sensors are built per wafer: http://photo.stackexchange.com/questions/53826/why-does-increasing-sensor-size-necessarily-lead-to-lower-silicon-wafer-utilizat

Regards,

David

shadowblade

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2839
Re: Nikon in difficulty?
« Reply #244 on: March 19, 2017, 08:02:47 am »

No, it is the same sensor plant. This is CMOS technology after all, it is the same base technology to build microchips, like processors and so on. So, basically this works like this:

  • They have this big rods of pure Si - Silicon that are cut into thing wafers, typical sizes now a days are 300 mm in diameter. Purity of the Si is very important, as we will see later.
  • Over each one of these buffers, in really clean rooms, they start to apply several chemical and physical process. This is basically growing the different transistors, circuits, and sensor parts. In each one of this wafers they build several sensor chips. If they are little ones for smartphones, they can make more, if they are bigger ones for medium format they can make less.
  • After everything is finish, they wafer are cut into the sensors and testing starts.

Here is where physical size of the sensor plays a big role. So, if they Si wafer has an impurity in any of the parts of the sensor, that sensor is not going to work. If any of the transistors, sensors, circuits does not work, that sensor needs to be discarded.

So, if you are making like 10000 sensors per wafer (invented number... I didn't do the math of the area of a mobile phone chip vs the area of the wafer), you are getting a high ratio of good sensors, even if you have some impurities or errors in the manufacturing, you still getting high number of good sensors per wafer that you can sell.

If you are making like 100 sensors per wafer (for example, full frame), you are probably getting worst ratio between good vs bad sensors. This is one of the main reasons bigger chips are more expensive. Apart that you are getting less good chips, the rest of the process to build the chips over the wafer is the same, so you have to divide the total costs between all the sensors coming out of the wafer, if you are getting more sensors, less they will have to pay.

Trying new technologies, like for example BSI, it is better to do this first in smaller ones. Because the technology is new and needs a tuning in the manufacturing process, the firsts wafers will have a very low yield, manufactures try to build first smaller sensors since it is easy for then to still get a good number of sensors that work (or the most simply ones, for example, memory chips are typically the ones used by Intel when then try a new physical/chemical process, they are the simplest chips they build, so easy to see how to fine tune the process). When that process is fine tune, you can take the risk of building bigger sensors with it and still get a good yield performance, so you are not risking too much money in the process.

So, big or small, all sensors are made in the same buildings. Take a look at the earthquakes that affected the manufacturing plants of Sony, all sensors lines were affected. Probably smartphone ones were the first one to recover because of what I said before, after rebuilding everything, having to calibrate again all machinery, doing it with the sensor size that has the highest probability of getting highest rate of good sensors make more sense. After those lines were starting to get good yield rates, Sony started to getting back to building bigger sensors.

Here: http://blog.teledynedalsa.com/2012/02/the-art-of-making-large-image-sensors-or-quilting-with-silicon/ and here: http://www.siliconimaging.com/cmos_fundamentals.htm you have images of wafers before being cut. Here you have a nice schema about how many full-frame sensors, APS-C sensors and 1" sensors are built per wafer: http://photo.stackexchange.com/questions/53826/why-does-increasing-sensor-size-necessarily-lead-to-lower-silicon-wafer-utilizat

Regards,

David

Maybe for the big, multi-billion-dollar plants. The little ones, owned by small companies making specialised sensors for non-camera applications? No way they can deal with a 36x24mm sensor.
Logged

davidgp

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 758
    • davidgp fotografia
Re: Nikon in difficulty?
« Reply #245 on: March 19, 2017, 08:40:32 am »

Maybe for the big, multi-billion-dollar plants. The little ones, owned by small companies making specialised sensors for non-camera applications? No way they can deal with a 36x24mm sensor.

Not 100% sure here, since for work motives, I'm following more microprocessors things that image sensors. But typically, very few companies have CMOS manufacturing plants: Intel, GlobalFoundries, TSMC, Samsung, Sony, etc... Intel with very few exceptions is doing chips for their own. But for example, GlobalFoundries (they are building AMD chips among others), TSMC (Apple and image sensors chips), Samsung (Apple, Qualcomm, memory, image sensor, own designs... ) are building chips for other companies. What I want to say, you don't need to have a fab foundry to do a big chip, you can hire a third party and specially TSMC and Samsung have top end technology and they let other use it... so, if you get a good design, if you have money, other will build it. Take for example Red or Arry video cameras, they are building big CMOS chips, full-frame area equivalents with no problem. CMOSIS is building sensors for Leica, they don't have a foundry (I think) and they hire other company to build the chips.

Probably, right now Sony is the leader of sensor technology. Look at the mobile phone market, even Samsung is using Sony sensors on their phones even though they have their own sensor designs in house (Samsung is strange case, they also use Qualcomn ARM chips in their phones even they have a similar performance ARM chip in-house). That does not mean that in few years other player could build a better design, specially an small one. There are lot of companies doing research nowadays in image sensor technology, not for photography camera applications, that market stopped growing, but for mobile phones (now high end models are mounting three chips, two in the back (iPhone 7 Plus, Huawei P9, ...)) and specially the automative and manufacturing industries, now they are automating everything, they need image sensors to process the information... video consoles with those sensor bars also include camera sensors... Panasonic already announced they are coming back to building image sensors...

scyth

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 584
Re: Nikon in difficulty?
« Reply #246 on: March 19, 2017, 10:07:10 am »

They are also making two variants for m43 20 megapixels chip, one for Olympus with phase AF on chip and other for Panasonic without it... (Although this last part is just don when they add the microlenses layer over the chip).
it is the same chip - the difference is only in CFA layer, which is on top of the chip (unless you are Canon and have PDAF implemented _in_ sensor with "dual" sensels)... microlenses go on top of CFA and present always in modern designs... unless some people wash off that + CFA to make monochrome sensor DIY way
Logged

armand

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 5540
    • Photos
Re: Nikon in difficulty?
« Reply #247 on: March 19, 2017, 10:09:09 am »

The Samsung APS-C sensor from few years ago is still competitive. There were rumors that didn't go through for some alliance between Nikon and Samsung. Getting access to that sensor or even better, a newer version, and to production facilities might improve Nikon's chance in a mirrorless market.

scyth

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 584
Re: Nikon in difficulty?
« Reply #248 on: March 19, 2017, 10:13:47 am »

Pentax is also using the 36MP sensor.

Ricoh is using it to simply save money because with their situation they have to watch the bottom line + their intended audience was simply dying for FF dSLR so 36mp or 42mp did not matter (36mp sensor is still quite good if implemented properly by a buyer as all Nikon cameras with variations of it show... certainly not on order of magnitude worse that 42mp)

And given that Sony people themselves said it...

again IR interview was withdrawn... so did Sony say it or IR screwed the translation/interpretation (from/of Japanese) is unknown

Logged

davidgp

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 758
    • davidgp fotografia
Re: Nikon in difficulty?
« Reply #249 on: March 19, 2017, 10:54:57 am »

again IR interview was withdrawn... so did Sony say it or IR screwed the translation/interpretation (from/of Japanese) is unknown

Apart that the interview is down now, for me the most interest part was when they were talking about what Sony is doing in software, like Eye AF autofocus. They were talking about image recognition to improve the follow focus performance of their system. I think it is in things like that were mirrorless manufactures need to improve, it is the main advantage that a mirrorless system has in front of an reflex one. The reflex one can not use the image information... that could become interesting in the following years, when the image processors inside the camera improve in performance.

Just thinking aloud...

razrblck

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 482
  • Chill
    • Instagram
Re: Nikon in difficulty?
« Reply #250 on: March 19, 2017, 12:28:43 pm »

Apart that the interview is down now, for me the most interest part was when they were talking about what Sony is doing in software, like Eye AF autofocus. They were talking about image recognition to improve the follow focus performance of their system. I think it is in things like that were mirrorless manufactures need to improve, it is the main advantage that a mirrorless system has in front of an reflex one. The reflex one can not use the image information... that could become interesting in the following years, when the image processors inside the camera improve in performance.

Just thinking aloud...

They need to move quickly. AF/Metering sensors in modern DSLR bodies are already capable of recognizing faces and tracking them. As soon as they get even more pixels, they will be able to recognize eyes too.
Logged
Instagram (updated often)

shadowblade

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2839
Re: Nikon in difficulty?
« Reply #251 on: March 19, 2017, 04:20:17 pm »

Apart that the interview is down now, for me the most interest part was when they were talking about what Sony is doing in software, like Eye AF autofocus. They were talking about image recognition to improve the follow focus performance of their system. I think it is in things like that were mirrorless manufactures need to improve, it is the main advantage that a mirrorless system has in front of an reflex one. The reflex one can not use the image information... that could become interesting in the following years, when the image processors inside the camera improve in performance.

Just thinking aloud...

What they said about AI-driven, image recognition-based AF is what I've been saying here for years - eye focus and face recognition is just the start of it. Improved processing and greater resolution potentially allow for such things as, 'track the eyes of the cheetah and shoot as it leaps on its prey' or 'track the striker (in soccer) and shoot as the ball is kicked'. Couple it with electronically-controlled lenses and tripod heads and you could even say, 'get a close-up of the face, thn zoom out for a whole-body action shot while the ball is being kicked, with the head in the upper right quadrant'. Or, even, 'track this animal while I set up and shoot a landscape on the other camera'.
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
Re: Nikon in difficulty? Outsourcing 36x24 sensor fab is probably doable
« Reply #252 on: March 20, 2017, 07:21:26 pm »

Maybe for the big, multi-billion-dollar plants. The little ones, owned by small companies making specialised sensors for non-camera applications? No way they can deal with a 36x24mm sensor.
I think there is no big problem outsourcing the fab of a 36x26mm sensor (except the familiar significant cost penalty compared to APS-C and smaller sensors.)

Firstly, consider how Leica gets a custom 45x30mm CMOS sensor for the S. It is designed by CMOSIS, which is fabless, so the actual manufacturing is outsourced, probably to TowerJazz.  (See this article at businesswire for some details of the connection.)
So Nikon's larger volume order for a smaller sensor seems doable.

Secondly, the dominant maker of chip fab equipment, ASML, now offers the on-wafer stitching (as needed to produce sensors large than the 33x26mm field size) as a standard feature on some of it step-and-scan sensor fab equipment. So, although it is probably still significantly slower and more expensive to make these big sensors compared to ones smaller than about 33x26mm, it is probably doable on machines already owned by outsourcing fabs like TowerJazz.
Logged

BernardLanguillier

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 13983
    • http://www.flickr.com/photos/bernardlanguillier/sets/
Re: Nikon in difficulty? Outsourcing 36x24 sensor fab is probably doable
« Reply #253 on: March 20, 2017, 10:05:38 pm »

Firstly, consider how Leica gets a custom 45x30mm CMOS sensor for the S. It is designed by CMOSIS, which is fabless, so the actual manufacturing is outsourced, probably to TowerJazz.  (See this article at businesswire for some details of the connection.)

I was under the impression that Leica sensors were manufactured by ST Micro-electronics?

Cheers,
Bernard


BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
Re: Nikon in difficulty? Outsourcing 36x24 sensor fab is probably doable
« Reply #254 on: March 20, 2017, 10:25:48 pm »

I was under the impression that Leica sensors were manufactured by ST Micro-electronics?

Cheers,
Bernard
Could be; I only know that Tower is one fab used by CMOSIS, and was the one used by the predecessor of CMOSIS to fab the 36x24mm sensors for the Kodak DSLRs. Either way, there are "fabs  for hire" out here that Nikon could use, once they have a decent sensor design.

P. S. You are right about the M, so likely for the S also: http://image-sensors-world.blogspot.com/2012/09/cmosis-sensor-in-leica-m-rangefinder.html?m=1
« Last Edit: March 20, 2017, 10:45:09 pm by BJL »
Logged

BernardLanguillier

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 13983
    • http://www.flickr.com/photos/bernardlanguillier/sets/
Re: Nikon in difficulty?
« Reply #255 on: June 17, 2017, 12:58:19 am »

I know that equipment quality is mostly irrelevant in this debate, but for what it's worth, Lloyd calls the Nikon 70-200 f2.8 E FL the only zoom he would be willing to use.

https://diglloyd.com/blog/2017/20170611_2240-Nikon70_200f2_8E-aseries-70mm-TailingsPiles.html

This matches my experience.

Cheers,
Bernard

shadowblade

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2839
Re: Nikon in difficulty?
« Reply #256 on: June 17, 2017, 02:59:38 am »

I know that equipment quality is mostly irrelevant in this debate, but for what it's worth, Lloyd calls the Nikon 70-200 f2.8 E FL the only zoom he would be willing to use.

https://diglloyd.com/blog/2017/20170611_2240-Nikon70_200f2_8E-aseries-70mm-TailingsPiles.html

This matches my experience.

Cheers,
Bernard

Then he probably hasn't tried the Canon 200-400 or 11-24. The latter is sharper than most of the primes within its focal length range.
Logged

BernardLanguillier

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 13983
    • http://www.flickr.com/photos/bernardlanguillier/sets/
Re: Nikon in difficulty?
« Reply #257 on: June 17, 2017, 07:22:32 am »

Then he probably hasn't tried the Canon 200-400 or 11-24. The latter is sharper than most of the primes within its focal length range.

The 200-400 is probably even better (haven't used it), but what I have seen from the 11-24 tells me it is the best wide zoom available, and a very good wide lens for sure, but not in the same category.

The 70-200 Nikon renders like an Otus.

Cheers,
Bernard

shadowblade

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2839
Re: Nikon in difficulty?
« Reply #258 on: June 17, 2017, 08:01:00 am »

The 200-400 is probably even better (haven't used it), but what I have seen from the 11-24 tells me it is the best wide zoom available, and a very good wide lens for sure, but not in the same category.

The 70-200 Nikon renders like an Otus.

Cheers,
Bernard

The point is, the 11-24 is sharper than any other lens - prime or zoom - in the 18mm-or-wider focal length category. A crown it took from another zoom - the Nikon 14-24.

Also, he clearly doesn't shoot landscapes either. 'Foot zooming' isn't an option there, and performance at f/5.6 or narrower, quality of sunstars, etc. tend to matter a lot more than things like bokeh or wide-open performance. It takes a lot of primes to replace a single zoom, and still leaves a lot of gaps where you're left cropping.
Logged

JKoerner007

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 262
  • "A picture's worth a thousand words."
    • John Koerner Photography
Re: Nikon in difficulty?
« Reply #259 on: June 17, 2017, 08:29:55 am »

The point is, the 11-24 is sharper than any other lens - prime or zoom - in the 18mm-or-wider focal length category. A crown it took from another zoom - the Nikon 14-24.

You're comparing a Canon lens that was launched in 2015 with a Nikkor that was launched in 2007 ::)

For 8 years, the Nikkor 14-24 was the standard to which all other zooms aspired ... and I am quite sure that Nikon will be updating this to an FL ED version in the very near future.

So Canon will once again take a back seat, as every FL ED Nikkor lens that has ever come out has set a benchmark in its class.

This will likely include a 200-400, which hasn't been refreshed since 2010.
Pages: 1 ... 11 12 [13] 14   Go Up