I just got back from the CS ManTech Conference  in Indian Wells, CA, where I had the pleasure of giving a presentation (Is 5G the Next RF Compound Semiconductor Industry Driver?), moderating a 5G panel and of course, attending a very interesting conference.  Before I dive into some of the very intriguing things from the conference, I have to say I was not prepared for the weather in Indian Wells. As the weather in New England hovered in the 60s, with rain, I saw the temperature hit 114 degrees (nearly 46 degrees Celsius for my European friends) at the conference! The other unusual phenomenon was the heat index. My daughter moved to South Florida about a year ago and she is fond of pointing  out how the “feels like” temperature for her is roughly 10-12 degrees higher than the the actual temperature once the humidity moves in. In my case, it was the opposite. The 114 day had a heat index of 102 and that’s what they mean by a “dry heat”…hot nonetheless…but a dry heat…like an oven!!

The heat outside was matched by the technical energy flowing inside the convention center. As the name implies, the conference leans toward the manufacturing aspects of compound semiconductor technologies. Based on the presentatons, it’s very clear that there is significant effort involved with solving the design and manufacturing challenges of compound semiconductor technologies like SiC, GaN, SiGe and InP, along with the heterogenous integration of these technologies onto silicon for a variety of market applications. The CS ManTech link at the top of this blog gives you access to some of the conference presentations and I suggest that you take a look at the interesting work that was presented.

One of the market applications these developments addressed is, not suprisingly, 5G.  In the plenary session, Steven Kovacic of Skyworks  presented the Skyworks vision for front end modules in a 5G network. His paper contained the network graphic shown below and in the spirit of “a picture is worth a thousand words”, this diagram captures my thoughts on the early implementations of the 5G network succinctly.

Source:  “Technology Initiatives for 5G Radio Front-end elements”, Kovacic CS ManTech presentation

There is a tremendous amount of effort going into and experimental results coming out of millimeter wave network developments. I’ll loosely lump 28 GHz into “millimeter wave” because this band seems to be at the epicenter of high frequency activity and these activities are very interesting to the compound semiconductor industry. The diagram hints at a network architecture to address some of the challenges facing 5G deployment.

High on this list of challenges is cost and coverage. The compound semiconductor industry is very excited by the densification and performance requirements of millimeter wave networks, but the cost is ENORMOUS. I saw an estimate that a standalone 28 GHz network for the US would cost $300 billion! Deploying networks in other countries could easily drive operator capex to $1 trillion or more.

What seems more likely is a lower frequency network that utilizes as much of the existing 4G infrastructure as possible. This is what 3GPP refers to as the “non-standalone” (NSA) option.  US service provider T-Mobile captured the largest amount of spectrum in the recently concluded 600 MHz auction and they are aggressively touting a nationwide, mobile 5G network that adds to their existing 4G network. This presents some interesting possibilities.

I haven’t done this in a while, but it’s time to use my cousin April to illustrate an interesting possibility. If you don’t recall, April lives on a lake, deep in the Adirondack Mountains. She doesn’t get much wireless service, but she gets wired broadband at very competitive speeds, although it tends to be spotty and aggravating. I’ve used April as an icon for the difficulties of ubiquitous wireless broadband coverage and questioned how the vision of 5G will reach everyone.

So how does April avoid missing the benefits of 5G? Well, the propagation characteristics of 600 MHz present some very interesting possibilities. Despite being deep in the woods, she’s only 20 or 30 miles from towns with 4G cellular coverage. Perhaps the business model for a 600 MHz wireless network focused on two roads in and out works better than a 2 GHz network, because of the bigger footprint. This is intriguing, but T-Mobile has estimated it will cost them about $25 billion for nationwide mobile coverage. The devil is always in the details and one of the challenges with the 600 MHz business plan is a relative lack of bandwidth, but that’s a topic for another blog. While this may not be in the cards, someday I may be able to have a virtual reality session with April to see her beautiful lake on a low frequency network.

Strictly from a cost standpoint, the most likely 5G network architecture looks to be one that provides coverage at <6 GHz frequencies, augmented with millimeter wave elements to increase capacity. If you’ve stuck with me this long, you are probably wondering about the title of this blog. One of the clear trends that I’m seeing is a tremendous interest in InP technology for some of these higher frequency applications. Many of the leading compound semiconductor manufacturers are very excited about the frequency, performance and physical characteristics of the technology. The challenge is much of the manufacturing is done on smaller (3” or below) diameter wafers and the supply chain is not as robust as other compound semiconductor technologies. Manufacturers seem willing to investigate these challenges and I heard more than once “InP is now where GaAs was 20 years ago”.

It will be interesting to monitor how these developments unfold, so keep checking back!

-Eric