p.14 #1 · FF Mirrorless, what's it going to take Sony, Canon, Nikon
technic wrote:
What I really meant is that DSLR technology is running up against practical limits
I don't see any evidence to that. On the contrary, just three years ago there was a huge jump in DSLR AF performance (D5). Additionally, newer lenses have quieter, faster and more precise motors that reduce the shot to shot variability of focus.
and has little room for improvement while A9 technology (stacked sensor, lots of on-sensor processing, more possibilities for computational imaging etc.) has lots of room to grow.
Certainly there is a lot of room for growth in demand for processing power and battery consumption, if the latter wasn't yet satisfying. Personally I think pattern recognition can be a useful feature but not really necessary for what I do. One problem with it is when the pattern is not recognized, then the user may not have the skills to work in the situation without the technological assistance. When one is used to simpler approaches such as specifying the area to be focused on manually, one is not as dependent on the specific camera's capabilities.
I don't think Canikon will have these stacked sensors and the other hardware that needs to come with it in the near future, because for now Sony is the only one producing it and you need significant volume to make it profitable. I don't see Sony offering their crown jewels to Nikon but who knows
If the manufacturing volume is the problem, why limit the sensor to the A9 only (which has a relatively small market)? If Sony eventually want to make a profit on this technology, they should try to put it into as many cameras as possible. That they don't put it into the A7 series cameras suggests that the cost of manufacture is too high or yield is too low.
I doubt Sony would decline an offer to sell Nikon (or Canon) as many sensors as they ask (for a price), but the issue could be that Nikon and Canon want to do their own thing regarding sensors and not just use an old Sony sensor. Nikon have their own goals in terms of how they want to optimize the sensor optics to the camera lenses, and provide ISO 64; Canon have their dual-pixel AF technology which differentiates their sensors from others'. I don't think either company is interested in using off-the-shelf components as these would provide no competitive edge. Even if some see them at a disadvantage, they still continue to do their own research and development because in the long term this is necessary to avoid falling into a dependence on a single supplier (which could become a monopoly, and arbitrarily raise prices and force the competition out of business). No, they will continue on their own sensor design work, to some extent at least. As to when they might be able to compete with the A9 directly, difficult to say, but my guess is by the time of the 2024 Olympics both Nikon and Canon probably have achieved this. In the meanwhile they will do what they can to compete using DSLRs. I suspect they will do just fine.
p.14 #2 · FF Mirrorless, what's it going to take Sony, Canon, Nikon
ilkka_nissila wrote:
I don't see any evidence to that. On the contrary, just three years ago there was a huge jump in DSLR AF performance (D5). Additionally, newer lenses have quieter, faster and more precise motors that reduce the shot to shot variability of focus.
I'm talking about AF for action, not e.g. better precision for landscape shots. For sure there is a limit to DSLR framerate, and that also severely limits improvements in AF/tracking performance that depend on the image sensor seeing the scene all the time. Theoretically this could be worked around by having an extremely sophisticated PDAF sensor but I don't see that happening. What the A9 does is not feasible on a DSLR.
ilkka_nissila wrote:
Certainly there is a lot of room for growth in demand for processing power and battery consumption, if the latter wasn't yet satisfying. Personally I think pattern recognition can be a useful feature but not really necessary for what I do. One problem with it is when the pattern is not recognized, then the user may not have the skills to work in the situation without the technological assistance. When one is used to simpler approaches such as specifying the area to be focused on manually, one is not as dependent on the specific camera's capabilities.
Cameras like A9 are running up against some practical limitations too, but those are much less fundamental than the framerate of a DSLR. Obviously, when you can use MF it usually means that you don't need fast AF and tracking at all ...
ilkka_nissila wrote:
If the manufacturing volume is the problem, why limit the sensor to the A9 only (which has a relatively small market)? If Sony eventually want to make a profit on this technology, they should try to put it into as many cameras as possible. That they don't put it into the A7 series cameras suggests that the cost of manufacture is too high or yield is too low.
I doubt Sony would decline an offer to sell Nikon (or Canon) as many sensors as they ask (for a price), but the issue could be that Nikon and Canon want to do their own thing regarding sensors and not just use an old Sony sensor. Nikon have their own goals in terms of how they want to optimize the sensor optics to the camera lenses, and provide ISO 64; Canon have their dual-pixel AF technology which differentiates their sensors from others'. I don't think either company is interested in using off-the-shelf components as these would provide no competitive edge. Even if some see them at a disadvantage, they still continue to do their own research and development because in the long term this is necessary to avoid falling into a dependence on a single supplier (which could become a monopoly, and arbitrarily raise prices and force the competition out of business). No, they will continue on their own sensor design work, to some extent at least. As to when they might be able to compete with the A9 directly, difficult to say, but my guess is by the time of the 2024 Olympics both Nikon and Canon probably have achieved this. In the meanwhile they will do what they can to compete using DSLRs. I suspect they will do just fine. ...Show more →
So like I said, Sony is the only one who can offer these stacked sensors and they don't offer them to others for now. Maybe Sony is using these sensors at a loss just because they want to show what they can do, increasing production might be a really bad idea then as long as there is no jump in production technology. Probably because they are too expensive indeed (no mini-A9 also suggests there is a cost/yield problem with this technology).
Agree that by 2024 both Canon and Nikon might have this technology available in their new ILC's but for some of us that's a long wait
p.14 #3 · FF Mirrorless, what's it going to take Sony, Canon, Nikon
AcuteShadows wrote:
Isn't Sony using machines made by Nikon to produce sensors that are installed in Nikon cameras according to Nikons specifications? Or at least machines that are similar to those made by Nikon? And it would be perfectly possible to have two sensor in a DSLR camera - one that is exclusively focused on image quality, and another that is primarily geared towards autofocus, but catches enough data for contrast AF, face recognition, exposure fine-tuning etc.
First of all, I really doubt that Sony is using only Nikon lithography equipment for their sensors; as to "similar", there are really only three important players in this market now: Dutch ASML (about 85% of the market), Nikon (about 10%) and Canon (5%). Take those numbers with a grain of salt because they depend on who you ask and how you frame the question but you get the idea.
Yes, a second image sensor in DSLR (in the mirror box, so it isn't blocked by the mirror) for just very fast AF/tracking that can compete with a Sony A9 may be theoretically possible. But I don't see it happening because it is much easier and cheaper to integrate it in one sensor-based system. The cost for developing such "next generation PDAF" systems must be spread over a strongly declining number of DSLRs - not going to happen IMHO.
p.14 #4 · FF Mirrorless, what's it going to take Sony, Canon, Nikon
technic wrote:
I'm talking about AF for action, not e.g. better precision for landscape shots. For sure there is a limit to DSLR framerate, and that also severely limits improvements in AF/tracking performance that depend on the image sensor seeing the scene all the time. Theoretically this could be worked around by having an extremely sophisticated PDAF sensor but I don't see that happening. What the A9 does is not feasible on a DSLR.
Cameras like A9 are running up against some practical limitations too, but those are much less fundamental than the framerate of a DSLR. Obviously, when you can use MF it usually means that you don't need fast AF and tracking at all ...
So like I said, Sony is the only one who can offer these stacked sensors and they don't offer them to others for now. Maybe Sony is using these sensors at a loss just because they want to show what they can do, increasing production might be a really bad idea then as long as there is no jump in production technology. Probably because they are too expensive indeed (no mini-A9 also suggests there is a cost/yield problem with this technology).
Agree that by 2024 both Canon and Nikon might have this technology available in their new ILC's but for some of us that's a long wait ...Show more →
AFAIK, the Fuji XT-3 uses sony's stacked sensor as well.
May 03, 2019 at 10:16 AM
Steve Spencer Online • • • • • • • Upload & Sell: On
p.14 #5 · FF Mirrorless, what's it going to take Sony, Canon, Nikon
AcuteShadows wrote:
Isn't Sony using machines made by Nikon to produce sensors that are installed in Nikon cameras according to Nikons specifications? Or at least machines that are similar to those made by Nikon? And it would be perfectly possible to have two sensor in a DSLR camera - one that is exclusively focused on image quality, and another that is primarily geared towards autofocus, but catches enough data for contrast AF, face recognition, exposure fine-tuning etc.
DSLR's are of course already using a second sensor in the mirror box for AF, but keep in mind that any sensor you put in that mirror box is going to have important physical limitations. The two obvious and very important ones are: 1) it won't be able to capture light from anything but about the central 1/3 of the image, so it can't focus on any area outside of that. You could theoretically make the mirror box bigger, so you wouldn't have this limitation but that would require a new mount with a longer registration distance. 2) Such a sensor will always only be getting a fraction of the light coming through the lens and if you try to divert more light to this sensor you will be diverting it away from the viewfinder making it darker.
These are important limitation of the way DSLRs work and mirrorless cameras do not have these limitations. Of course mirrorless cameras have other limitation DSLRs do not, but some of those limitations (like lag between light hitting the sensor and being shown on the EVF) are getting to be less of a limitation as technology improves. In fact, the lag until light is shown on the EVF in some cases has gotten so short it isn't perceptible.
p.14 #6 · FF Mirrorless, what's it going to take Sony, Canon, Nikon
That's correct, but just because it's a Sony sensor doesn't mean it can hold its own against the A7 gen 3 sensors.
May 03, 2019 at 10:24 AM
Steve Spencer Online • • • • • • • Upload & Sell: On
p.14 #7 · FF Mirrorless, what's it going to take Sony, Canon, Nikon
Charlie N wrote:
AFAIK, the Fuji XT-3 uses sony's stacked sensor as well.
No, that camera appears to have a version of Sony's IMX 571 sensor which is back side illuminated but not stacked. See this page from Sony's sensor division that describes their sensors:
Sony's sensor division only lists smaller sensors (1" and smaller) that are stacked. Of course we know the A9 has a stacked sensor, and I would be very surprised if the rumored 31mp APS-C sensor that is supposed to have a global shutter and may well show up in the A6700 is not a stacked BSI sensor, but Sony doesn't list these larger stacked sensors yet.
p.14 #8 · FF Mirrorless, what's it going to take Sony, Canon, Nikon
Steve Spencer wrote:
No, that camera appears to have a version of Sony's IMX 571 sensor which is back side illuminated but not stacked. See this page from Sony's sensor division that describes their sensors:
Sony's sensor division only lists smaller sensors (1" and smaller) that are stacked. Of course we know the A9 has a stacked sensor, and I would be very surprised if the rumored 31mp APS-C sensor that is supposed to have a global shutter and may well show up in the A6700 is not a stacked BSI sensor, but Sony doesn't list these larger stacked sensors yet.
Dang it, I got caught up in fake news
I could have sworn all those rumor sites were saying it had a stacked sensor. 30 fps on a non stacked sensor, it must have fast readout.
May 03, 2019 at 10:53 AM
Steve Spencer Online • • • • • • • Upload & Sell: On
p.14 #9 · FF Mirrorless, what's it going to take Sony, Canon, Nikon
Charlie N wrote:
Dang it, I got caught up in fake news
I could have sworn all those rumor sites were saying it had a stacked sensor. 30 fps on a non stacked sensor, it must have fast readout.
If you look at the specifications closely it only has 30 fps with a 1.25 crop, so you aren't capturing all the pixels. The sensor seems to have a 16 fps limit with all the pixels, but Fuji seems to have a tweak that allows 20 fps uncompressed RAW and uncropped. I doubt the readout is that fast. I would guess it is 1/30 of a second, but it might be as fast as 1/60 (which would be faster than any non stacked sensor). We will know when someone tests it. Still this is a great sensor, it just doesn't have the benefits of a stacked sensor, but in the A9 that stacked sensor pays for speed with a bit of reduced performance at low ISO. I doubt in time all stacked sensors will have that limitation, but the proof will be in the pudding.
p.14 #10 · FF Mirrorless, what's it going to take Sony, Canon, Nikon
with respect to Nikon's sensors - article fyi, admittedly it was sponsored but the other recent interviews etc with Nikon and Sigma by Dave Etchells hold water
p.14 #11 · FF Mirrorless, what's it going to take Sony, Canon, Nikon
chambeshi wrote:
with respect to Nikon's sensors - article fyi, admittedly it was sponsored but the other recent interviews etc with Nikon and Sigma by Dave Etchells hold water
That is a great article! So Sony has nothing to do with these Nikon sensors then, except making them. And so if Nikon didn't choose to go with BSI or Stacked sensor tech, it's because they don't want to and not because Sony is holding out on them. Personally, I haven't really cared whether Nikon sourced Sony sensor or actually designed them from ground up. But this article should be a great source to cut short any of the ridiculous debate whether it's important Nikon went with Sony or Towerjazz....
p.14 #12 · FF Mirrorless, what's it going to take Sony, Canon, Nikon
hiepphotog wrote:
That is a great article! So Sony has nothing to do with these Nikon sensors then, except making them. And so if Nikon didn't choose to go with BSI or Stacked sensor tech, it's because they don't want to and not because Sony is holding out on them. Personally, I haven't really cared whether Nikon sourced Sony sensor or actually designed them from ground up. But this article should be a great source to cut short any of the ridiculous debate whether it's important Nikon went with Sony or Towerjazz....
I think your conclusion about stacked sensor tech is wrong because Sony is the one producing the sensors (for now, I'm not aware of any other party that can produce such large stacked image sensors), NOT Nikon. Designing and producing is not the same thing. This makes a big difference when the yields are too low for commercial production or the sensor is simply too expensive to sell to third parties - for Nikon or Canon it really isn't an option at the moment. Remember what the Sony manager said recently about the cost of these sensors and the problem of e.g. offering an APS-C model with this technology.
Probably at some point the yields or cost will improve and then Sony could offer the technology to others but in the current situation that sounds unlikely (just imagine what an A9 type camera from Nikon would cost now, I don't think they would sell significant qty so it doesn't make sense...).
p.14 #13 · FF Mirrorless, what's it going to take Sony, Canon, Nikon
You draw way to much from that article.
Nikon will go with the more advanced architectures as they work to catch up. Nikon is resource restricted compared to the World's largest image sensor company. There is still a lot of patent, cross licensing, technology sharing (or not sharing), cost efficiencies going on.
As the author states, letting someone in to see highly proprietary work like this was only hesitantly done to combat the perception in the industry that Nikon was just using second tier Sony technology. So, the whole thing was structured for this purpose. The truth likely is somewhere in the middle. But, it must be acknowledged that Nikon and Sony both offer cameras with the best IQ in the industry. That is what is really important to a camera buyer.
p.14 #14 · FF Mirrorless, what's it going to take Sony, Canon, Nikon
technic wrote:
I think your conclusion about stacked sensor tech is wrong because Sony is the one producing the sensors (for now, I'm not aware of any other party that can produce such large stacked image sensors), NOT Nikon. Designing and producing is not the same thing. This makes a big difference when the yields are too low for commercial production or the sensor is simply too expensive to sell to third parties - for Nikon or Canon it really isn't an option at the moment. Remember what the Sony manager said recently about the cost of these sensors and the problem of e.g. offering an APS-C model with this technology.
Probably at some point the yields or cost will improve and then Sony could offer the technology to others but in the current situation that sounds unlikely (just imagine what an A9 type camera from Nikon would cost now, I don't think they would sell significant qty so it doesn't make sense...)....Show more →
I'm not sure why you think the whole stacked sensor is not commercially viable since the A9 has it and there are certainly enough A9 out there to go around. And I don't think too expensive to sell to third parties would be a good reason either. If anything, if someone is willing to pay, it's in their interest to sell to help with the R&D anyway. It's whether Nikon chooses to do so or not.
Now, I can buy the argument that Nikon engineers didn't have the know-how to incorporate stacked sensor into the rest of the system that is due to Sony IP protection or other reasons.
p.14 #15 · FF Mirrorless, what's it going to take Sony, Canon, Nikon
1bwana1 wrote:
You draw way to much from that article.
Nikon will go with the more advanced architectures as they work to catch up. Nikon is resource restricted compared to the World's largest image sensor company. There is still a lot of patent, cross licensing, technology sharing (or not sharing), cost efficiencies going on.
As the author states, letting someone in to see highly proprietary work like this was only hesitantly done to combat the perception in the industry that Nikon was just using second tier Sony technology. So, the whole thing was structured for this purpose. The truth likely is somewhere in the middle. But, it must be acknowledged that Nikon and Sony both offer cameras with the best IQ in the industry. That is what is really important to a camera buyer....Show more →
Ha, if anything, it shows exactly that. Nikon is using second tier Sony tech whether they're picking it from a drop-down menu or coming up with their owns. After all, whatever they come up with, Sony has to be able to do it. It's not like Nikon can just ask Sony to make a dual pixel sensor.
p.14 #16 · FF Mirrorless, what's it going to take Sony, Canon, Nikon
Steve Spencer wrote:
DSLR's are of course already using a second sensor in the mirror box for AF, but keep in mind that any sensor you put in that mirror box is going to have important physical limitations. The two obvious and very important ones are: 1) it won't be able to capture light from anything but about the central 1/3 of the image, so it can't focus on any area outside of that. You could theoretically make the mirror box bigger, so you wouldn't have this limitation but that would require a new mount with a longer registration distance. 2) Such a sensor will always only be getting a fraction of the light coming through the lens and if you try to divert more light to this sensor you will be diverting it away from the viewfinder making it darker.
These are important limitation of the way DSLRs work and mirrorless cameras do not have these limitations. Of course mirrorless cameras have other limitation DSLRs do not, but some of those limitations (like lag between light hitting the sensor and being shown on the EVF) are getting to be less of a limitation as technology improves. In fact, the lag until light is shown on the EVF in some cases has gotten so short it isn't perceptible....Show more →
What I wanted to stress is that the AF sensor could be augmented, so that it can see and interpret the actual image (face detection etc.). If the mirror provides the full image to the optical viewfinder, I don't see a physical reason why it couldn't provide the full image to the augmented AF sensor. Maybe my idea about the physical setup of a DSLR and its limitations are mistaken, though.
May 03, 2019 at 06:16 PM
Steve Spencer Online • • • • • • • Upload & Sell: On
p.14 #17 · FF Mirrorless, what's it going to take Sony, Canon, Nikon
AcuteShadows wrote:
What I wanted to stress is that the AF sensor could be augmented, so that it can see and interpret the actual image (face detection etc.). If the mirror provides the full image to the optical viewfinder, I don't see a physical reason why it couldn't provide the full image to the augmented AF sensor. Maybe my idea about the physical setup of a DSLR and its limitations are mistaken, though.
It can't provide the full image to the AF sensor because the mirror box is too small, it will cut off the light path. Remember the light going to the AF sensor in a DSLR is passing through the mirror not reflected off it, so it has a hard limit of how big of an area it will illuminate. You could make the mirror box bigger but then you would need a new mount with a longer registration distance. That is the physical reason you can't have AF sensors (i.e., points) outside the central area in a DSLR and that wouldn't change if you put any other type of sensor in place of the AF sensor. Also quite obviously the light is split between the AF sensor and the viewfinder (it is a mirror that let's some light through and reflect some light) so either you have a dark viewfinder or a significantly reduced amount of light on the AF sensor. These are both hard physical limits on any AF sensor for a DSLR--less light and only central light. No getting around that unless you just flip the mirror up and in effect make it a mirrorless camera.
p.14 #18 · FF Mirrorless, what's it going to take Sony, Canon, Nikon
Steve Spencer wrote:
It can't provide the full image to the AF sensor because the mirror box is too small, it will cut off the light path. Remember the light going to the AF sensor in a DSLR is passing through the mirror not reflected off it, so it has a hard limit of how big of an area it will illuminate. You could make the mirror box bigger but then you would need a new mount with a longer registration distance. That is the physical reason you can't have AF sensors (i.e., points) outside the central area in a DSLR and that wouldn't change if you put any other type of sensor in place of the AF sensor. Also quite obviously the light is split between the AF sensor and the viewfinder (it is a mirror that let's some light through and reflect some light) so either you have a dark viewfinder or a significantly reduced amount of light on the AF sensor. These are both hard physical limits on any AF sensor for a DSLR--less light and only central light. No getting around that unless you just flip the mirror up and in effect make it a mirrorless camera....Show more →
At least, it should be possible to construct something useful with an APS-C sensor and a FF flange distance (or using a medium format mount for a full frame sensor). Even if you limit the AF area, face detection (or eye detection, for that matter) with OVF composing would be a useful addition to DSLR technology.
May 03, 2019 at 07:21 PM
Steve Spencer Online • • • • • • • Upload & Sell: On
p.14 #19 · FF Mirrorless, what's it going to take Sony, Canon, Nikon
At least, it should be possible to construct something useful with an APS-C sensor and a FF flange distance (or using a medium format mount for a full frame sensor). Even if you limit the AF area, face detection (or eye detection, for that matter) with OVF composing would be a useful addition to DSLR technology.
Why would you do that when you could use the full sensor and all the light with a mirrorless? You are not acknowledging the significant drawback of having only a fraction of the light with such an arrangement. That is going to hurt AF in low light especially. You have to reflect a lot of the light to the OVF, trying to do what a mirrorless can do with DSLR just doesn't make sense. It will never be able to do it as well. You will always be stuck with only central light and only a fraction of the light.
Oh, and DSLRs have tried to expand the zone for AF points for a number of years now. There really isn't room to cover a wider area.
p.14 #20 · FF Mirrorless, what's it going to take Sony, Canon, Nikon
AcuteShadows wrote:
What I wanted to stress is that the AF sensor could be augmented, so that it can see and interpret the actual image (face detection etc.). If the mirror provides the full image to the optical viewfinder, I don't see a physical reason why it couldn't provide the full image to the augmented AF sensor. Maybe my idea about the physical setup of a DSLR and its limitations are mistaken, though.
The days of a silver mirror reflecting 100% light to the OVF are very far behind us.
