Never miss a story from Breakfast Bytes. Subscribe for in-depth analysis and articles.
When cell-phones first became a consumer product, a VP of Nokia drew me an upside-down triangle, with radio at the top. chips in the middle, and the little point at the bottom being software. When cell-phones first became a consumer product, getting the radios to work was a huge challenge, the baseband chips (maybe 5 of them at first, gradually going down to one as Moore's Law ticked inexorably) were challenging but understood, and there was minimal software. "Snake" was the only game. The earliest phones couldn't even handle text messages since they were only used under the hood for the network to send configuration data to the handsets. In fact, in those days, the handsets were called terminals before the companies realized that no consumer had a clue what a "terminal" was. The VP told me that hiring RF engineers was a huge challenge and that Nokia even had a spreadsheet that listed pretty much all the experienced RF designers in the world, a few hundred.
The VP then drew another triangle, this time the other way up. Radio was a solved problem since each design pretty much used the radio from the handset before. But software was becoming a big problem. This was in the era before true smartphones, before software really exploded, but when email on your phone was starting.
Well, with 5G, radio is once again a huge challenge. There are new mm-wave bands, there is the need to aggregate multiple channels using different radios, phased-array antennas, and more. RF design is back. I wouldn't say it is the old triangle, since software and application processor design are still both big tasks. The triangle is now a broad rectangle. It is also getting much more multi-technology, with modules sometimes containing many (or even all) of CMOS, SoI, CaAs, GaN, SMD, laminate, package) and as many as 20 ICs in the highest end modules.
The three big areas for RF today are:
If you want to design any sort of aggressive RF design, then one of your first ports of call is likely to be National Instruments. But the complexity of modern RF design, especially the verification, means that a disjointed flow using Cadence and National Instruments tools together is not longer good enough. So the two companies have worked together to create an integrated product, Virtuoso RF (or VRF for short). The integration is at the level of source code, not just a few menu items that allow one product to launch another.
The first challenge was that the OA database didn't support all the weird (by IC design standards) geometries required for RF design: circles, arcs, any-angle routes, bond-wires, and more. These simply don't occur in traditional IC design, which OA was originally architected for.
Cadence MTS (multi-technology simulation) was extended to layout and a co-design environment was created.
For EM, Sigrity 3D-EM and National Instruments' AXIEM are part of the same integration, with results back-annotated onto the extracted view.
Cadence recently hosted an internal training with our own engineers and engineers from National Instruments. Since the two design tools are joined at the source code, it requires a good level of expertise in both systems to create an effective RF design and do all the verification. At short notice, I was invited over to see a little of the presentations (it was an all-day event) and to talk to a few of the participants.
Autopilots on planes are often colloquially named "George". The reason seems to be disputed but it seems to date to the second world war, where all the planes were nominally owned by the King, who was George VI. Other theories are available.
The first computers that I ever used, the ICT (later ICL) 1900 series, had an operating system called George. Since one of my tasks at the high school where I went was to run all the students' jobs at our local RAF base, where we had scrounged some mainframe computer time, I actually sat and the control terminal and ran everything. Of course, this sounds very archaic, in these days of smartphones and 100,000 server cloud datacenters.
All this is a very roundabout segue to the fact that the VP in charge of these products on the National Instruments side is George Zafiropoulos. In the early 2000s I worked with him at Cadence. He had been the VP marketing at Quickturn, which Cadence acquired. That product line eventually morphed into the current Cadence emulation product line, Palladium. George was there at the internal training, so it was fun to meet up with him for the first time in about a decade. In between Cadence and National Instruments, he held various senior positions at Synopsys, not surprisingly mostly in verification.
Sign up for Sunday Brunch, the weekly Breakfast Bytes email.