Technical Insight
GaN: ready for cellular 3G?
3G base stations are finally appearing on the cellular infrastructure scene. But can high-power GaN devices grab a foothold in the market this year? Jon Newey investigates.
One of the regularly discussed future applications for wide-bandgap devices, particularly AlGaN/GaN HEMTs, is in base stations for 3G cellular networks. Wide-bandgap devices promise to make such installations more compact and energy-efficient. In 2003 GaN made some impressive strides towards the 3G goal, with both RF Micro Devices and Nitronex announcing that GaN-based power amplifiers (PAs) were sampling to wireless infrastructure customers, but will we see the first GaN PAs in a base station in 2004?
Recent technological progress has promised much, with several announcements of increasing power density and output powers from GaN amplifiers. Those involved have indicated that such developments were placing them nearer to the stringent requirements for wireless infrastructure PAs. These requirements demand an output above 150 W at 48 V operation – the standard operating voltage in base-station systems – as well as low adjacent channel power ratio.
With this in mind, device developers have been pushing the drain voltage as far above 50 V as possible in order to improve the efficiency and output power. In December 2003, Cree announced that it had reached 32 W/mm output power density and 55% power-added efficiency (PAE) at 4 GHz. Just a week earlier, Fujitsu had released details of its GaN PA operating at 2.1 GHz in a W-CDMA modulation scheme with 150 W output power and 54% PAE at a drain voltage of 63 V.
Reliability is a key issueHowever, despite the rush to higher powers hitting the headlines, the key area to address before GaN devices can be fielded in cellular base stations is reliability. "Reliability is a key issue in GaN, and improving it is our key focus," Cree s John Palmour told Compound Semiconductor. "Getting to 30 W/mm is of limited benefit if the device soon degrades. We ve made a lot of progress, but still have a way to go like everyone else."
Fujitsu admits that it is still some way from producing commercial devices, with the company currently researching mass-production techniques and building up its library of reliability data. Commercial products are estimated to be "one or two years" away.
Nitronex has also created interest with its GaN HEMT technology and applications for cellular infrastructure. It is currently sampling 10 W devices to wireless infrastructure and radar customers. The Raleigh, NC, company s technology differs from that of most other companies in that it is growing GaN/AlGaN HEMT device structures onto silicon substrates rather than SiC. SiC is generally favored because of its closer lattice match to GaN and its high thermal conductivity.
Paul Williams, Nitronex s European managing director, seems less worried than most about reliability, which is surprising given that at first glance silicon and GaN seem rather incompatible as a substrate–epilayer pair and one might expect the GaN quality to suffer.
"Reliability issues surround material that s grown onto SiC," said Williams. "With Nitronex s proprietary buffer layer scheme for growing GaN/AlGaN structures onto silicon, the material quality is such that reliability is vastly better." However, Williams concedes that the company wants, and expects, to see better figures for gain and efficiency from its devices before they are properly deployed. Although Nitronex is chasing design wins now for its 10 W devices, the lengthy design cycle means that actual implementation in base stations will not be seen in 2004, even if those design wins come soon.
Meanwhile, silicon LDMOS, the incumbent base-station PA technology, continues to raise the bar on its performance. LDMOS has the advantage of being a mature technology with proven reliability over many years of service. Motorola s next-generation HV6 high-power RF LDMOS transistor technology has been optimized for W-CDMA applications. Operating at 2.1 GHz, the devices show 29% drain efficiency at 20 W power output, and a PAE of 61% at 100 W.
While 2004 looks certain to bring more technological progress in the field of GaN microelectronics, it seems likely that we will have to wait until at least 2005 before commercial deployments in cellular infrastructure begin to appear.
Recent technological progress has promised much, with several announcements of increasing power density and output powers from GaN amplifiers. Those involved have indicated that such developments were placing them nearer to the stringent requirements for wireless infrastructure PAs. These requirements demand an output above 150 W at 48 V operation – the standard operating voltage in base-station systems – as well as low adjacent channel power ratio.
With this in mind, device developers have been pushing the drain voltage as far above 50 V as possible in order to improve the efficiency and output power. In December 2003, Cree announced that it had reached 32 W/mm output power density and 55% power-added efficiency (PAE) at 4 GHz. Just a week earlier, Fujitsu had released details of its GaN PA operating at 2.1 GHz in a W-CDMA modulation scheme with 150 W output power and 54% PAE at a drain voltage of 63 V.
Reliability is a key issueHowever, despite the rush to higher powers hitting the headlines, the key area to address before GaN devices can be fielded in cellular base stations is reliability. "Reliability is a key issue in GaN, and improving it is our key focus," Cree s John Palmour told Compound Semiconductor. "Getting to 30 W/mm is of limited benefit if the device soon degrades. We ve made a lot of progress, but still have a way to go like everyone else."
Fujitsu admits that it is still some way from producing commercial devices, with the company currently researching mass-production techniques and building up its library of reliability data. Commercial products are estimated to be "one or two years" away.
Nitronex has also created interest with its GaN HEMT technology and applications for cellular infrastructure. It is currently sampling 10 W devices to wireless infrastructure and radar customers. The Raleigh, NC, company s technology differs from that of most other companies in that it is growing GaN/AlGaN HEMT device structures onto silicon substrates rather than SiC. SiC is generally favored because of its closer lattice match to GaN and its high thermal conductivity.
Paul Williams, Nitronex s European managing director, seems less worried than most about reliability, which is surprising given that at first glance silicon and GaN seem rather incompatible as a substrate–epilayer pair and one might expect the GaN quality to suffer.
"Reliability issues surround material that s grown onto SiC," said Williams. "With Nitronex s proprietary buffer layer scheme for growing GaN/AlGaN structures onto silicon, the material quality is such that reliability is vastly better." However, Williams concedes that the company wants, and expects, to see better figures for gain and efficiency from its devices before they are properly deployed. Although Nitronex is chasing design wins now for its 10 W devices, the lengthy design cycle means that actual implementation in base stations will not be seen in 2004, even if those design wins come soon.
Meanwhile, silicon LDMOS, the incumbent base-station PA technology, continues to raise the bar on its performance. LDMOS has the advantage of being a mature technology with proven reliability over many years of service. Motorola s next-generation HV6 high-power RF LDMOS transistor technology has been optimized for W-CDMA applications. Operating at 2.1 GHz, the devices show 29% drain efficiency at 20 W power output, and a PAE of 61% at 100 W.
While 2004 looks certain to bring more technological progress in the field of GaN microelectronics, it seems likely that we will have to wait until at least 2005 before commercial deployments in cellular infrastructure begin to appear.
