One transistor for the future of mmWave?

We’ve heard recently from several sources that millimeter wave radios, once the exclusive realm of defense and satellite use, are now finding homes in applications such as automotive radar and 5G networks. Therein lies a significant opportunity for digital design: moving frequency conversion and filtering from the analog domain into the digital domain.

It’s akin to the breakthrough in mobile baseband technology at the origin of GSM where much of the signal processing moved into a digital chip, typified early on by TI OMAP architecture with an ARM core and a DSP core. This greatly reduced component costs, handset size, and power consumption. Analog front ends (AFEs) still did much of the heavy lifting, pulling the higher frequency carrier signals off the antenna and downconverting them into baseband frequencies.

For 4G LTE and Wi-Fi systems, teams have been working hard on CMOS implementations of advanced AFEs, with broad success down to 28nm. An LTE front end is complex: there are up to 18 LTE bands plus 2G/3G bands, some at 2.5 GHz and higher; the peak to average power ratio (PAPR) is increasing and is now over 8dB; and carrier aggregation adds channels that must be correlated with better sensitivity, higher linearity, and more isolation.

Millimeter wave ups the stakes with even higher frequencies – and transistor f[SUB]T[/SUB] and f[SUB]MAX[/SUB] need to be at least 3 to 5 times the application frequency. SiGe has entered the mix with outstanding RF performance at some process cost, but investigation is underway on both FinFET and FD-SOI processes for compatibility with digital processes.

Research is working on direct sampling data converters for millimeter wave, moving the beamforming for a phased array antenna into digital domain and greatly simplifying the analog subsystem. Jariet Technologies is one company looking at digital millimeter wave, and their website indicates they are working on a 14nm FinFET implementation:

image courtesy Jariet Technologies

GLOBALFOUNDRIES thinks they have a better idea. 22FDX offers a more flexible layout for optimizing analog performance with both lower BEOL loss and lower R[SUB]g[/SUB], reducing noise and increasing f[SUB]MAX[/SUB]. Initial investigations show FD-SOI transistor applicability across the range of power amplifiers (PAs), antenna switches, antenna tuners, and low-noise amplifiers (LNAs). GF also claims an ADC and DAC in 22FDX would consume less power compared to 14nm FinFET.

You may have heard some automotive radars are working at 77 GHz, and that is an important point for this chart presented by Jamie Schaeffer at the recent Linley Group Mobile Conference. Long story short, due to its higher R[SUB]g[/SUB], according to Schaeffer FinFET meets some RF requirements but runs out of gas as frequencies increase:

Considering GF is also a foundry for FinFET and SiGe, Schaeffer is trying to position 22FDX as the digital RF integration solution, and SiGE as the high performance RF solution. The trend is unmistakable, however: digital integration ultimately wins, especially as volumes increase. We haven’t even talked about the other advantages of 22FDX, particularly leakage power and the body-bias configurability.

As GF continues to improve FD-SOI, these advantages in mmWave applications only get stronger. Phased-array technology drives a change in front end technology, making the case for integration in more applications. This should rekindle the “integrated baseband” discussion, right? 2017 should bring news on who has picked up GF 22FDX for RF applications, and we’ll have a better idea of how research translates to economic success then.

Schaeffer’s presentation from the Linley conference isn’t available in the clear, but there is a presentation by Anirban Bandyopadhyay of GF at Tech Shanghai 2016 with more details from investigations in 45nm PD-SOI paving the way for the 22FDX story:

RF Front-End Module Roadmap and Handset Architecture Challenges in Future mmWave Technology