In Post World-War II Blues, Al Stewart (who happened to go to the same high school as I did) sang: 1959 was a very strange time A bad year for Labour and a good year for wine Well, it's too early to say whether 2018 will be a good year for wine. But it is certainly a good year for Virtuoso. The Virtuoso 2018 is a fine vintage, ready to drink now, especially if you are facing new challenges: 5nm, automotive reliability, More-than-Moore integration of different die designed in different process technologies, statistical corners, variability, and so on. Even a little off the bleeding edge, say 16nm, there is still need for multiple coloring (red, white, and rosé!) and FinFET design. Since this is the main node for automotive, you may also have to worry about ISO 26262 and other reliability requirements. Even if you aren't on cutting-edge process nodes, and you don't have to concern yourself with FinFETs, you still want increased productivity and more automation. Like balance in a good wine, these things are desired in every vintage. New Challenges Every design is different, so any example I pick is unlikely to match what you are doing. But let's pick a real example. Do you know what this is? It's the lidar unit from the top of a Waymo self-driving car (although I took this photo back in pre-Alphabet days when it was all still Google). Lidar is the laser equivalent of radar, sending out laser pulses and then timing when the pulses return and building up a picture of what surrounds the car. The "li" on the beginning of lidar stands for light, an equivalent to the "ra" that stands for radio in Radar. This is a still photo, but even though this car was parked outside Peet's coffee with nobody in it (or maybe the car just wanted a coffee and drove itself to the café), the lidar was still spinning. Do you know how much it costs? I've seen numbers from $50K to $80K. But even if we go with $10K, it's clear that is a price point that even a $60K Mercedes can't support. Lidar has to get cheaper, and the way to make it cheaper is to use solid-state laser diodes, use MEMS mirrors to direct the light, use silicon photonic detectors, and substitute for high-precision optics with lots of digital signal processing. That has the potential to reduce the price to perhaps $10 a unit. A vehicle would probably need several to get adequate coverage since a unit would be more like a camera that doesn't have all-round coverage. That leads to a system as in the diagram above, with a laser-diode transmitter (at the top) built in III-V semiconductor, with MEMS mirrors to make the light beam scan, a super-sensitive SPAD detector (single photon avalanche diode), and signal processing to pull the system together. Light travels a foot in a nanosecond, and the return pulses is about 150ps wide, to give an idea of the processing accuracy required. Anyway, this is not a tutorial on lidar, the important thing is that there are several different technologies that need to be packaged together: Laser Diode (III-V) MEMS mirrors (Si) Photonics IC (SiPo) Analog sensor (non-leading edge CMOS) SoC (advanced node CMOS) What's needed is a system that can handle all of these technologies not just one at a time (they are unlikely to all be designed by the same group), but in the context of the packaging, and the other die. These go from advanced node (let's say, looking to the future, 5nm) to analog to MEMS to photonics. Virtuoso 2018 is just that system. What's New? Well, first, what I've been casually calling Virtuoso 2018 is officially ICADVM 18.1. At the layout level, there is full support for advanced nodes. This includes full support for multiple coloring, both the self-aligned (such as SADP) and not (LELE) types. Second-generation 7nm and all 5nm are expected to use EUV for some masks (contacts, vias, and cuts at a minimum), and that is supported, too. Layout in any advanced node has reached the stage that it is not really possible to design DRC-correct layout by hand, it requires advanced layout automation and, now, simulation-driven layout. This is all enabled by the tight integration of ADE, Virtuoso layout, and the Spectre simulator. Going up to the system level, any die can be opened in the context of the package and the rest of the system. There is full interoperability between Virutoso (layout), Allegro (PCB and packaging layout), and Sigrity (signal and power integrity). This supports mixed technologies for MEMS, photonics, power devices, and RF, as well as mainstream analog and digital silicon. I said above that multiple die will not be all designed by the same group, and that is often true even for different parts of the same chip. The Virtuoso layout suite supports concurrent design, with the capability to parallelize effort, do incremental editing, compare and merge changes, and generally allow multiple designers to be working in parallel. The focus initially is on chip-finishing, fixing design rule violations, and manual routing. All of these are activities where it is a big problem if only one designer can work at a time. The result: More productivity, better performance, higher quality results. Santé. Sign up for Sunday Brunch, the weekly Breakfast Bytes email![]()
↧