In the present business environment, the drive for lowest component cost and maximum device efficiency have been the two deciding factors in winning most RF power-device sockets at the larger cellular infrastructure OEMs. The reduction in both operating expenses and system costs are the main overriding factors behind these requirements, but in some instances, there are other application-specific cost drivers that make the efficiency requirement take precedence over cost.

Some of the most stringent application conditions can be found in the requirements for tower-top basestation-amplifier systems. In these designs, the complete basestation is located not at the bottom of the antenna tower in an environmentally controlled housing (radio shack) but at the top of the tower in a small, weather-exposed, sealed enclosure.

Reliability
The tower-top exposure forces a unique set of operating requirements on the RF amplifier, some that are apparent but also some that do not readily come to mind. One of the more obvious requirements is one of ultra-high reliability. Given that it is relatively difficult, expensive and slightly dangerous for service personal to climb to the top of a tower to replace a failed unit, the tower-top application must be completely bullet proof with a very long projected operational life. As an ancillary requirement, the design must use convection cooling, as no conventional-fan designs have a suitably long mean time between failures (MTBF) to be considered for this ultra reliable, relatively remote site.

Efficiency
Another relatively obvious requirement is the need for very high DC-to-RF conversion efficiency. Because of the lack of a controlled, air-conditioned enclosure and even the lack of forced-air cooling, RF power amplifiers in tower-top applications require the highest efficiency of all basestation infrastructure PA designs.

Conversion efficiency controls the physical size of the amplifier due to the convection-cooling requirements of the tower-top application. In order to sink of a lot of dissipated heat, you need a lot of surface area to transfer large amounts of energy to the outside air.

Tower Loading
In the complete set of tower-top application requirements, there is one additional overriding factor that almost no one thinks about. This is the wind-loading conditions on the tower structure itself. Sure, you electrical engineers thought you had considered all options when comparing GaN-based amplifier to those of LDMOS, but you need to let the mechanical engineers into the room and let them have their say, too.

When you start throwing a relatively large, heavy enclosure with a lot of surface area at the top of a very long lever, the integral strength of the tower structure needs to be significantly increased. The structure has to support both the vertical load of the mass of the amplifier and the horizontal stresses created by the wind -ail properties of the enclosure.

Under some extreme weather conditions, such as maybe a Category 3 hurricane with wind speeds up to 130 mile per hour (210 km/hr), this tower structure has to be very strong, which easily translates to very expensive. In the tower-top application, the physical size—and to a lesser degree, the weight̵of the enclosure drives the structural cost of the tower and therefore drives the overall system cost. This one overriding system cost requirement for minimal physical size and maximum conversion efficiency can tip the scale towards the highest efficiency designs.

LDMOS devices have a proven reliability track record, are highly linear in back-off conditions, and are very easily DPD-corrected (digital predistortion). In comparison, GaN devices are not as linear, correctable, reliable or affordable as the LDMOS products.

In the past, GaN did have a slight efficiency advantage over previous generations of LDMOS devices when tested in conventional symmetrical Doherty designs. That efficiency advantage has somewhat disappeared with the latest 8^th generation of LDMOS products now being designed in asymmetrical Doherty applications and is down to the two to four percent advantage.

Given the comparable efficiencies of both the LDMOS and GaN technologies, one has to question if GaN devices are worth the increase in cost. For a standard three-sector, 40W average basestation, this cost differential is on the order of $900 for GaN vs $120 for a LDMOS solution. Oh yes, the operators want the maximum efficiency possible, but no one is will to pay that amount of price differential for no real added benefit.

Is every basestation going tower top? No, but a significant portion of them are. In the future, what percentage will be tower top? I don't know, but it will be an interesting topic of conversation whenever RF amplifier designers get together.

Do you have an opinion you want to share with me? Come by the Freescale booth #322 at the CTIA convention (March 23-25, Las Vegas) and I would love to hear your opinion on this matter.♦

Leonard Pelletier is lead RF applications engineer at Freescale Semiconductor and is a regular Guest contributor on RF topics for RF DesignLine. He can be reached for questions and suggestions for future topics by either commenting in the box below or by contacting him directly at leonard.pelletier@freescale.com. You can follow him on Twitter at www.twitter.com/RFLeonard.