I’m writing this blog a few days earlier this month, because I am headed off on my summer vacation on Saturday. It strikes me that the “9-5” crowd gets a bit of a distorted picture of summer and vacation. Unless you are a teacher, a student or with proper respect to my European friends, European, you get a week or two that defines the “summer”. Everything prior to your summer vacation is “leading up to” and everything that follows is “getting back in the swing from”. We associate summer with warmth and the weather in the Boston area has heated up a bit. We just registered the first 90-degree day of the year, but I have realized that I spend about 3 minutes a day outside of climate-controlled air conditioning, so I don’t really have an appreciation for what “summer heat” means.  However, for the next two weeks, I will be in South Carolina where the heat index (a combination of temperature and humidity) was 114 degrees earlier in the week! So, while I am excited about spending time with my favorite people, being wet, sand (not in the context of silicon ICs) and hot weather, my thoughts have turned to GaN. I think I need a vacation!

I just released two insights on particular aspects of the RF GaN market that I find interesting. I reported on some of the GaN developments announced at the IMS2015 Conference in a blog that I wrote at the end of May. The major GaN players were all touting their new GaN in plastic product developments and that piqued my interest. Back in my product days, I witnessed the conversion of GaAs devices from ceramic packages into plastic for high volume applications as I worked with customers and our engineering teams to identify and solve issues that plastic created with MMIC and package design. Plastic is a poor thermal conductor, so the initial thermal path was limited to conduction through the leads and the package lead frame. While this is sufficient for low power devices, the explosion of GaAs handset PAs forced package designers to develop other means for removing heat. The answer was exposed paddles and flanges, where the active semiconductor device sits on a metal carrier that served as the thermal and electrical ground.

The new GaN devices are an impressive blend of encapsulated, overmolded, flange and paddle configurations that show a lot of modeling and innovation. The largest commercial RF application for GaN is in base station PAs, so this got me curious about the performance sacrifices involved with putting GaN into plastic. The first challenge is the actual transmit power. The infrastructure market relies on macrocells that may have average transmit powers between 20W and 100W, with “small cells” encompassing power ranges from several milliwatts to about 10W average. The 4G and CDMA-based wireless standards use a linear modulation scheme that requires somewhere around 7 dB of power back off to achieve the spectral requirements of the particular standard. This results in a peak-to-average factor of five, so a 20W average output is actually closer to 100W of peak power.

With this in mind, I looked at 28 plastic packaged GaN devices targeting base station applications from SEDI, MACOM, Qorvo and Cree in Plastic Packaged GaN RF Power Devices for Base Station Applications. In the last several months, MACOM has made a big splash with their thoughts on the future of GaN-on-silicon, versus the more widely deployed GaN-on-SiC technology. Of the 28 parts I looked at, seven were GaN-on-silicon from MACOM. I won’t give away all the interesting conclusions, but the two juiciest ones are that the MACOM GaN-on-silicon plastic parts reach power levels competitive with the GaN-on-SiC plastic devices and it appears that plastic parts are not yet suitable for the highest power macrocell applications.

As I was doing this research, I got interested in the frequency distribution of revenue in the entire GaN market. When we think of GaN, it’s common to think about the high frequency and the high power attributes of GaN technology. These features are allowing GaN to capture share in applications that can leverage those performance advantages, but that is far from the entire story. Commercial applications are growing as operators come to understand the operating cost advantages of the same performance at lower power consumption that GaN provides. Defense applications make use of a wider range of GaN advantages and devices are used in a more diverse set of applications. This leads to an interesting segmentation of revenue, with the commercial revenue concentrated in lower frequency ranges and defense revenue spread over a much broader frequency range. I presented these results, along with some conclusions in Frequency Segmentation of GaN RF Device Revenue. I segmented frequency along the same lines as I’ve done with previous GaAs analyses, for consistency, so the ranges are <4 GHz, 4-10 GHz, 10-20 GHz and 20+ GHz.

As I mentioned earlier, we tend to think of GaN as a high power, high frequency, wide bandwidth technology and it is, but perhaps it is a bit surprising that slightly more than 2/3 of the total GaN RF revenue occurs below 4 GHz. This is because the two largest RF GaN commercial applications, base stations and CATV/broadband, along with military tactical radios reside in this frequency range. Defense applications dominate the remaining frequency ranges and the shift toward higher frequencies in commercial and defense applications will reduce the share of RF GaN revenue below 4 GHz going forward. I’ve sliced the revenue data in several different ways and there are some interesting conclusions, so if you are a client, please take a look and let me know if you have any questions. If you aren’t a client, you can also contact me and I’d be happy to discuss the benefits of a service subscription!

I hope everyone enjoys the rest of the summer. Think of me as I navigate loved ones through airports and if you are in South Carolina and see a guy on the beach thinking about GaN, it’s probably me, so come say hello!

- Eric