CEO interview: The joy of RFSOI with MixComm’s Mike Noonen
Noonen is a semiconductor veteran who has started to specialize in guiding startups. He has worked for many companies included spells in senior management with National Semiconductor, NXP Semiconductors, and Globalfoundries. Most recently he was with Silego Technology Inc. prior to its acquisition by Dialog Semiconductor in 2017 (see Dialog buys configurable mixed-signal IC pioneer).
I started the interview by asking Noonen if the Covid-19 pandemic had been a problem for startup MixComm. It had caused the cancellation of Mobile World Congress and turned the launch of MixComm’s first product – the ‘Summit 2629’ 28GHz beamforming front-end IC for 5G cellular communications – in to a slightly lower key affair than it might have been (see Startup launches 5G front-end beamforming IC).
Lightning-quick to see the sales angle Noonen replied: “One thing we have learned from the pandemic is that communications infrastructure needs to be upgraded. 5G is exceedingly important.” And that’s what MixComm is betting on with its application of circuit design to a radio frequency silicon-on-insulator (RFSOI) manufacturing process.
Noonen pointed out that the company, although founded in 2017, is building on a legacy. “The company was co-founded by Dr Harish Krishnaswamy, a professor at Columbia University, and Frank Lane, previously a vice president of technology with Qualcomm. But the company leverages almost a decade of RF design advances made at the CoSMIC laboratory that was led by Professor Krishnaswamy.
Noonen observed that prior to that, the US Defense research agency DARPA had funded research into the RFSOI manufacturing process, which had been brought to realization by IBM, before the technology was passed to Globalfoundries in 2014.
With a 45nm RFSOI process to design to and exploit, MixComm has got to first product revenue with an equity investment of less than $10 million, Noonen said. “This is the best time to be a semiconductor entrepreneur if you can do it frugally. In the digital domain, it can cost $100 million to find out you’ve built the wrong chip,” commented Noonen.
On the assumption that MixComm has built something approximating the right chip, where does the company’s commercial advantage come from?
“It comes from that deep understanding of RFSOI acquired over many years and knowledge of the circuit architectures it supports,” said Noonen.
RFSOI is a specialized planar process that can be used to make numerous RF chips such as high frequency switches and antenna tuners for smartphones. It is akin to fully-depleted silicon-on-insulator (FDSOI) with similar back-biasing options. Not only does it allow performance in silicon close to compound semiconductor devices but it supports integration in a similar manner to CMOS and on large diameter wafers.
As cellular communications move from 3G to 4G and on to 5G more and more carrier frequencies are being used. Simply adding additional gallium-arsenide piece parts is an expensive solution, Noonen said.
“RFSOI provides not only a tremendous Fmax but also things like [topologically] stacked power amplifiers. Allowing designers to stack multiple power amplifiers provides a breakthrough in output power. We are approaching III-V and wide-band-gap device performance with CMOS efficiency and integration opportunities,” said Noonen.
He added that one of the pain points in 5G roll out is in the size of antenna arrays. When built with conventional technology they can be large, costly and unaesthetic. When there are numerous smaller cells, making 5G antennas unobtrusive is a key consideration. “Increasing the amplification from 60dBm to 70dBm can change the cost of coverage from $1.5 million per square kilometer to $500,000,” Noonen said.
Next: New integration
RFSOI not only allows performance across the range of voltage and temperature but also allows as yet untried integration looking back into the signal chain and looking forward at the antenna. “Front-end plus antenna co-design is underappreciated,” Noonen added.
So, what are the limitations of RFSOI or is it generally applicable across RF circuits and frequencies? Noonen said: “It is generally applicable but we are focused on where novel architectures meet millimeter-wave frequencies. That could be cellular communications, automotive radar, Ka-band [26.5-40GHz] satellite communications. MixComm has a 5G infrastructure focus for now.”
Noonen gives the example of beamforming, which is addressed by MixComm’s first chip. “RFSOI allows beamforming to be looked at holistically, including very large beam tables. Beamforming has not previously been the domain of RF engineers.”
Noonen emphasised the support MixComm gets from working with his former employer Globalfoundries, who can make RFSOI circuits on 300mm-diameter wafers, bringing cost advantages. These wafers are run at the former IBM East Fishkill wafer fab – now owned by On Semiconductor – under a legacy manufacturing agreement (see On Semi buys ex-IBM fab from Globalfoundries). However, there is a plan for a transition to Globalfoundries’ Malta wafer fab in upstate New York, Noonen added.
What about competition? “There are some companies focused on silicon-germanium but that tends to be a few dB short. There are others working in bulk CMOS. The integration possible is good for smartphones but it is inadequate for infrastructure. Movandi is using bulk CMOS, but at a disadvantage,” Noonen asserted.
And what about the application of RFSOI to sub-6GHz frequencies and Wi-Fi? “At the moment we need to do millimeter-wave correctly. There’s more than enough to do there; 5G, remote sensing applications and automotive radar,” said Noonen. He added that millimeter-wave benefits for infrastructure will include decreased latency, which could open up new use cases and applications for the technology. At the simplest level, reduced latency would mean less use of the radio and better battery life for smartphones.
It helps to have Qualcomm evangelizing to have millimeter-wave be used as widely as possible, Noonen added. The technology is also applicable in broadband to the home in the form of fixed-access radio. In terms of next steps for the company MixComm has a 39GHz solution coming and has formed a partnership with Ireland’s Taoglas to build a smart antenna array.
Next: Startup explosion
More generally we asked Noonen whether he agreed that there has been a dramatic increase in startup formation around the world in recent years.
“I agree and I am somewhat gratified. When we founded Silicon Catalyst in 2014 we were concerned about what would happen if the number of hardware-based startups dried up. At that time Sand Hill Road was treating silicon like it was plutonium. We wanted to foster an extended, global Silicon Valley. More-than-moore is where it is at. There are so many opportunities that don’t need aggressive lithography.”
Silicon Catalyst did not single-handedly turn the investment climate around or nurture all the semiconductor startups that are now coming out of stealth, but it did seem to catch and help foster the mood in the second half of the last decade.
But success comes not only from the re-ignition of a vision within the financial community. That’s a vital component but not the only one. “MixComm did not happen overnight. DARPA focused on fundamental research over many years. IBM, Qualcomm and Globalfoundries have leveraged that research and that creates the opportunity for MixComm,” said Noonen.
Related links and articles:
News articles:
Startup launches 5G front-end beamforming IC
Dialog buys configurable mixed-signal IC pioneer
GloFo 45nm RFSOI process ready for 5G
Toshiba improves SOI process for RF
Tower, Soitec, Maxscend ramp RFSOI
Silego signs Noonen to drive sales, marketing
On Semi buys ex-IBM fab from Globalfoundries
Sibling startup offers RF for 5G
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
