You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
Does anyone have an idea if we will have a combination MEM/optical/semi all on the same chip or whether it will be separate chips in the same package. I feel it will be separate chips in the same package because most mems will use a much larger scale than optical and semis. I feel the combination on these three technologies in some fashion are going to be a very large market in the future, especially in numerous IOT areas and functions, also in control systems. Thoughts and comments solicited. Also will their be a single chip with multiple node sizes as a solution?
Sounds reasonable to me Arthur. I could see a pre-amp on the MEMS chip and interface logic and optical muxing etc on the optical chip, but no compelling advantage and a lot of disadvantages to trying to squeeze onto one piece of silicon.
Arthur, I do not think optical can come on the same chip, but there are efforts to standardize MEMS processes. In future, we may see MEMS coming together with Semi, but there needs to be a compelling reason to drive that. I do not see that reason as of now.
I do think most of the MEMS currently are already done on top of a standard CMOS technology. Currently I think one of the main optical R&D subjects is to make it silicon compatible; I don't think it is in volume production yet.
I've been hearing about optical backplane technology for 20 years now, and not a lot of progress has been made. Most optical installations use a SFP transceiver module and over-the-top cabling. Unless backplane integration gets better, the motivation to integrate optical at a system level is low - it's just not a cost driver. There is a niche in supercomputing where this may happen, where cost is no object.
I've been hearing about optical backplane technology for 20 years now, and not a lot of progress has been made. Most optical installations use a SFP transceiver module and over-the-top cabling. Unless backplane integration gets better, the motivation to integrate optical at a system level is low - it's just not a cost driver. There is a niche in supercomputing where this may happen, where cost is no object.
The recent push again to get electrical/optical interface closer to the chip is to reduce power consumption of electrical data transfer; some are even dreaming of making CPU-DRAM interface optical.
