You know the price of this devices? I think this is quite attractive for Sony. They push into other even smaller niches as well, like UV. They lower the hurdle for integration of a SWIR sensor. It is funny to watch how one after another of the industrial camera vendors comes with its IMX990 camera, just like with the polarizer sensor (I wonder how many they actually sell of these...). But in SWIR i think Sony doesnt just take significant parts of the existing market but also opens new ones, since they also lower the price point. I think Sony doesnt just spend money on SWIR, they might get a ROI quite soon.
The price is not secondary ;-) if you require more than vga, the pre imx990 cost for camera+optics+light+accessories is what? 70k? So you have to charge 150-200k which very little will pay. Now you can build the same machine feature for <20k and sell for 50. And suddenly some problems jump to a solution in swir. Once again ;-) I think many people here overestimate pixel performance and underestimate the rest of the story.
What I mean is that with this CIS-like approach, the price will be soon CIS-like too, so it's secondary since it will drop naturally. If you like at the overall complexity, it's not more complicated than a BSI CIS. In a BSI CIS, you have to Cu-Cu bonding, backside thinning and also backside passivation. If you look at the modern CIS, there are much more stuffs such as DTI, pyramid silicon surface, etc. From this optic, InGaAs is much simpler if the pixel performance is NOT overlooked :).
ah - sorry for the misunderstanding. But I still do not agree in this point - it is still a ingaas FPA hybridized on top of a Cmos ROIC. Plus I dont see a high volume market for this. I think SWIR based on InGaAs will remain a niche. The niche will grow for sure with IMX990 but it will remain a niche. Its a big step but still a factor 1000(?) in cost would be required to make it suitable for mass market. I guess if there would be a real mass market for swir resp. a CIS like scaling then it might be CQD based, like http://image-sensors-world.blogspot.com/2021/10/st-to-start-quantum-dot-image-sensor.html, but what should the application be? What would create a benefit large enough to integrate this into most smartphones? or at least most cars? It will not be a plastic sorting or semiconductor active alignment or inspection problem ;-) I dont know about the complexity and cost of ingaas FPA layer fabrication itself. Around the next step - hybridization - there is a lot of progress recently. Die to wafer cu-cu hybrid bonding is picking up strongly, AMD soon brings first processors with multiple cu-cu die to wafer bonds, in a few years this packaging process will be in high volume. There will be solutions soon for low um contact pitch with placement and the preparation steps required. But still, more costly than a deposition of a cqd layer I think. But lets see what the future brings, it is for sure very interesting to follow.
i think it is not the high techishness of applications or pixel performance, the key here is applicability. The design in hurdle and also the price point in the context of multiplicative margins. If you can integrate on top of a CMOS type ROIC because it is basically the same interface than whats around in very similar flavor plus you are able to create the full assembly / machine option and in the end put your margin instead of multiplying margins of multiple steps of vendors before, then the assembly gets economically attractive for end users, also for this type of applications. If the option is to design in a module or full camera you have one margin multiplication more on top of a expensive base component. Why do you think that all the lucids, allieds etc. come with swir products now? even mipi modules are around with the IMX990 and also integration of bare sensor into assemblies is around. It is possible because it is surprisingly easy to integrate the sensor if you have a pregius knowhow or firmware infrastructure. The pixel performance is good, at least good enough for many applications. And due to this, the lower tech, "known since decades" types of applications get adressable.
Nice.
ReplyDeletePuzzled though why Sony spends all that money to go after the market of differentiating cooking oil from water. Fascinating.
You know the price of this devices? I think this is quite attractive for Sony. They push into other even smaller niches as well, like UV. They lower the hurdle for integration of a SWIR sensor. It is funny to watch how one after another of the industrial camera vendors comes with its IMX990 camera, just like with the polarizer sensor (I wonder how many they actually sell of these...). But in SWIR i think Sony doesnt just take significant parts of the existing market but also opens new ones, since they also lower the price point. I think Sony doesnt just spend money on SWIR, they might get a ROI quite soon.
DeleteThe price of the products is secondary and it will be adjusted by the corresponding market practice. What is important is that it introduces CMOS practice into this chasse-gardée sector and triggers a revolution. The CIS compatible interface in this family of sensors will be deadly attractive for small camera companies to enter this market. I heard that there is ultra-low noise version under development.
DeleteThe price is not secondary ;-) if you require more than vga, the pre imx990 cost for camera+optics+light+accessories is what? 70k? So you have to charge 150-200k which very little will pay. Now you can build the same machine feature for <20k and sell for 50. And suddenly some problems jump to a solution in swir. Once again ;-) I think many people here overestimate pixel performance and underestimate the rest of the story.
DeleteWhat I mean is that with this CIS-like approach, the price will be soon CIS-like too, so it's secondary since it will drop naturally. If you like at the overall complexity, it's not more complicated than a BSI CIS. In a BSI CIS, you have to Cu-Cu bonding, backside thinning and also backside passivation. If you look at the modern CIS, there are much more stuffs such as DTI, pyramid silicon surface, etc. From this optic, InGaAs is much simpler if the pixel performance is NOT overlooked :).
Deleteah - sorry for the misunderstanding. But I still do not agree in this point - it is still a ingaas FPA hybridized on top of a Cmos ROIC. Plus I dont see a high volume market for this. I think SWIR based on InGaAs will remain a niche. The niche will grow for sure with IMX990 but it will remain a niche. Its a big step but still a factor 1000(?) in cost would be required to make it suitable for mass market. I guess if there would be a real mass market for swir resp. a CIS like scaling then it might be CQD based, like http://image-sensors-world.blogspot.com/2021/10/st-to-start-quantum-dot-image-sensor.html, but what should the application be? What would create a benefit large enough to integrate this into most smartphones? or at least most cars? It will not be a plastic sorting or semiconductor active alignment or inspection problem ;-) I dont know about the complexity and cost of ingaas FPA layer fabrication itself. Around the next step - hybridization - there is a lot of progress recently. Die to wafer cu-cu hybrid bonding is picking up strongly, AMD soon brings first processors with multiple cu-cu die to wafer bonds, in a few years this packaging process will be in high volume. There will be solutions soon for low um contact pitch with placement and the preparation steps required. But still, more costly than a deposition of a cqd layer I think. But lets see what the future brings, it is for sure very interesting to follow.
DeleteAll these are known for decades - it looks very naive
ReplyDeleteWell known but utile people's well-being. You prefer laser gun, laser bomb...
Deletei think it is not the high techishness of applications or pixel performance, the key here is applicability. The design in hurdle and also the price point in the context of multiplicative margins. If you can integrate on top of a CMOS type ROIC because it is basically the same interface than whats around in very similar flavor plus you are able to create the full assembly / machine option and in the end put your margin instead of multiplying margins of multiple steps of vendors before, then the assembly gets economically attractive for end users, also for this type of applications. If the option is to design in a module or full camera you have one margin multiplication more on top of a expensive base component. Why do you think that all the lucids, allieds etc. come with swir products now? even mipi modules are around with the IMX990 and also integration of bare sensor into assemblies is around. It is possible because it is surprisingly easy to integrate the sensor if you have a pregius knowhow or firmware infrastructure. The pixel performance is good, at least good enough for many applications. And due to this, the lower tech, "known since decades" types of applications get adressable.
Delete