Will Silicon LDMOS High Power RF Devices Drive the Rollout of 5G?

Now that it is a reality, 5G is starting to build out. Most systems currently being installed are at 5G New Radio (NR) mid-band (1 GHz to 6 GHz). This is more or less a worldwide phenomenon; 4G/Long-Term Evolution (LTE) infrastructure is or will be slowly replaced, although some countries are asking that the older technology infrastructure be replaced before 5G is installed.

Radio Frequency (RF) power semiconductor devices are playing an integral part of this build-out, and there are two competing technologies vying for sockets in 5G base stations: the first is Gallium Nitride (GaN), and the second is Silicon (Si) Laterally Diffused Metal Oxide Semiconductor (LDMOS). The former holds much promise and has the required performance for the new systems, especially at higher frequencies. GaN is relatively new in a volume sense, although its development took about 15 years. Si LDMOS, on the other hand, has been the mainstay in mobile wireless base stations for 25 years.

When the 5G base station topology was being formed, assigning frequencies was a difficult problem. There are a number of allocated frequencies; at least initially, most were in the above-mentioned 1–6 GHz slot. There was some conjecture at first that those frequencies above 3.5 GHz would be the jumping-off point.

This has turned out not to be the case as large service providers and Original Equipment Manufacturers (OEMs) actually took a much more pragmatic and conservative course. Si LDMOS can be used at or below about 3.8 GHz, so at these lower frequencies the two competing technologies, GaN and Si LDMOS, are viable.

Si LDMOS is less expensive than GaN and always will be in some sense, even though the cost difference is closing. Also, up until now, there has not been a high-volume use of GaN in any commercial application. At this point let’s discuss what “high-volume” really means for wireless infrastructure. At a minimum, volumes would need to be in the tens of millions in terms of devices shipped. No GaN application except for the one case mentioned below comes close to meeting that level. Lower shipment rates just won’t cut it in reaching economies of scale and efficient manufacturing.

The volume 5G applications that are now shipping and being built out are at 2.6 GHz and 3.5 GHz. As mentioned above, these frequencies can use Si LDMOS. Aside from the frequency considerations, Si LDMOS was chosen by the large OEMs for the following reasons:

• In Si LDMOS’s 25-year history, about one billion Si LDMOS RF power devices have been shipped.
• Manufacturing large quantities of these parts holds no surprises.
• Present production lines are already configured for the volume required.
• Si LDMOS is the most inexpensive technology.
• Performance and reliability are well known and documented.

ABI Research has found out that Si LDMOS is currently shipping into 5G base stations at the above-mentioned frequencies in volumes of tens of millions. GaN is being used only by Huawei for high volume 5G base station use, with shipments in the millions.

The conventional wisdom at 5G’s outset was that frequencies greater than 3.5 GHz would be the starting point and that GaN would be the technology of choice. Neither of these has happened.

So where does this leave us?

We know the following:

• The volume 5G rollout will be at 5G NR mid-band with initial frequencies at 2.6 GHz and 3.5 GHz.
• Tens of millions of Si LDMOS devices have been shipped for wireless infrastructure at these two frequencies.
• GaN is only be being used by Huawei in any large quantity sense.

Moving forward we can look for this trend to continue. Si LDMOS will be the driver for this new technology, perhaps garnering 70% of the total parts shipped. Evaluation of GaN for 5G will continue with test systems and low-scale deployment (in thousands of base stations), but really, high-volume shipments, with the exception of Huawei, are just not in the cards yet. Eventually GaN devices will capture respectable share.

Rollout at other frequencies is likely to be with Si LDMOS as long as it is at 3.5 GHz or less. Expect that the larger OEMs, driven by the service providers, will take a conservative and proven base station rollout.

No discussion of 5G will be complete without some mention of 5G NR high-band (> 24 GHz). These frequencies, also known as Millimeter Wave (mmWave), are a difficult and troublesome area. While this foresight is not the proper place to discuss issues of propagation and system operability, a look at the basic rollout of high band is appropriate. Although there have been continuing trials, test systems, and evaluations, many of the propagation and penetration issues still exist.

Strictly from the hardware production standpoint, the market is not there yet. At best, deployments are in the hundreds and perhaps optimistically thousands of base stations, but it is certainly not at the millions that large OEMs need to be viable. The same volume guidelines that apply to 5G NR mid-band also apply to high band.

There is a further complication: pricing of subsystems and their associated costs can have a disconnect rate of up to 10 times. This has led to some mmWave integrators (an important third player in this hardware class) to reconsider investing in mmWave hardware. Nothing is easy at these frequencies, and because of operational issues, there is little visibility to volume shipments. Investment in mmWave devices and subsystems are real expenses, and the perhaps dubious promise that production is “around the corner” is not a pleasant prognosis for management. Look for more of the involved companies to reconsider their mmWave investment.

The 5G large-scale deployment has started. There has been a nearly seamless transition from 4G/LTE to 5G NR mid-band using base station RF that uses Si LDMOS devices and peripheral components. Although GaN will capture respectable share, Si LDMOS will probably drive deployment.