Technical Insight
WiMax connectivity gives SiGe amplifiers the edge
Price erosion and the emergence of new high-throughput standards are two of the big issues facing chip manufacturers in the wireless LAN market, says SiGe Semiconductor's Andrew Parolin.
Compound Semiconductor: What products does SiGe Semiconductor supply to the wireless LAN [WLAN] space?
Andrew Parolin: We sell power amplifiers [PAs], switches and front-end modules. Our customers can create their own designs using our PAs and switches on the board. We ve also gone one step further and have designed complete front-end RF modules - everything from the uplink of the radio to the antenna - on a single device. Broadcom, which is the market leader, has taken our 2521A PA and that s a big win for us. Broadcom already uses our 2520 and 2529 PAs in its reference designs, and the company has now moved to our higher-integration module.
CS: The WLAN market boomed in 2003, how does it look this year?
AP: We estimate our total global available market to be around 65-70 million units in 2004, although there are some rumors that this may dip a little. For the 802.11g protocol, which is probably the dominant standard, we are the market leader. We have about a 48-50% share in 802.11g. The 802.11b market is dwindling and becoming commoditized, so we are focusing our efforts on 802.11g.
CS: How do you explain the relative success of SiGe Semiconductor PAs in WLAN compared with handsets?
AP: There has never really been a SiGe versus GaAs debate in WLAN. The integration advantage for SiGe has been a tremendous seller. Most of the GaAs guys require external transistors and regulators to provide a 2.8 V reference voltage and to turn the PA on and off. We ve actually integrated that into the SiGe die, which saves about 20 cents on the bill of materials (BOM).
We ve also been able to integrate the biasing with a robust power detector. I can get much more on the die and that reduces the BOM considerably [compared with GaAs]. SiGe has always met industry ruggedness requirements, it just comes down to performance and cost.
CS: Anadigics has said that the 802.11a/g dual-band protocol is more suited to GaAs than SiGe. What s your response?
AP: I disagree. I think there s a place for SiGe in the dual-band application. We have SiGe 5 GHz PAs available and we have combined the benefits of SiGe with those of GaAs for the 5 GHz band. Therefore we can provide a cost-effective dual-band solution that the GaAs guys still cannot.
Our dual-band solution is completely digitally controlled, so for the 5 and 2.4 GHz bands you can connect our PA directly to a CMOS baseband with no interface or glue logic. The GaAs guys are still going to have that integration problem so they will need some transistors and a reference voltage, and that s all extra cost. We ve already seen substantial price erosion in 802.11b and we now see it happening in 802.11g. It s going to happen in dual-band as well, and we ll be able to follow that curve down [better than GaAs].
The big figures of merit for WLAN are error vector magnitude [EVM] and output power. Our PAs combine high integration with a low EVM versus output power. The PA that we ve just launched [2528L] has less than 3% EVM at a 23 dBm output. That s important. It allows our customers to provide 20 dBm at the antenna, and for regulatory reasons this is where everyone wants to be. Getting above that figure is not really required: you would have to get into some sophisticated power control and no-one really wants to do that. Independent tests showed a throughput improvement of 5-6 Mbit/s and 20-30 m extra range.
CS: How do these figures compare with the best GaAs solutions?
AP: They are competitive. There is one GaAs PA that is [marginally] better, but the problem is that there is no integration, so you have to add another 40 cents to the overall cost to get the same level of functionality. So while with GaAs there might be a little bit more there in performance, you have to pay a price. That performance margin cannot be used, so at some point there will be diminishing returns. The 40 cent price advantage is huge: now the BOM for a whole card is approaching five to six dollars, so it s 5-10% of that cost in the consumer market.
CS: Has the 2528L been a success?
AP: It shares the same package as our previous 2525 device, so all of our customers can fit this higher-power device in the same socket. With three or four component changes you have the high-power design, and you don t have to change the layout or the board. Our customers love this because they can have a high-power/low-power design without having different boards.
To date we have design wins with four different customers. We gave out some early samples and people went crazy over it - it all goes back to higher performance and integration for lower BOM.
Revenue from this new PA will cannibalize some of our older designs, including the 2529 and the 2525. In the first four months or so it will probably take away about 20-30% of revenue from those existing products. By mid-2005 we would expect the old designs to be out completely, and we expect most of our existing customers to switch over.
CS: What opportunities are there for SiGe in the emergence of WiMax connectivity [IEEE 802.16]?
AP: WiMax was set up with two primary applications in mind. One is point-to-point backhaul for Wi-Fi: so if you have an access point on a lamp post and you want to connect to the main hub, then you would use 802.16. The second is point-to-multipoint, where you have one big transmitter to provide broadband services to homes. So it s very much like the local multipoint distribution services we had six or seven years ago, but now it s standardized and pushing forward. Those are two big markets.
Having said that, there is a new activity that is also going to support mobile applications and give WiMax support on laptops (see box). For this, you need both back-haul and point-to-multipoint links.
There are lots of synergies with 802.11g and a, but for the higher throughput the noise control has to be better. WiMax mainly operates in the 2.5-3.5 GHz spectrum, with support in the 5.8 GHz band as well. The frequencies are similar and the output power is a little higher, but essentially WiMax is very similar to Wi-Fi. Some people are using our existing PAs because of that overlap.
CS: What trends are you seeing in WLAN?
AP: An emerging theme in Wi-Fi is the new 802.11n protocol. The IEEE s intention is to reach 100 Mbit/s throughput through the 2.4-5 GHz spectrum. Right now the [theoretical] 54 Mbit/s throughput is actually more like 25 Mbit/s pure data. The "n" workgroup is trying to reach an actual 100 Mbit/s throughput to enable entertainment, video and Internet through the same pipe. That s the next big thing and we re working on it.
Andrew Parolin: We sell power amplifiers [PAs], switches and front-end modules. Our customers can create their own designs using our PAs and switches on the board. We ve also gone one step further and have designed complete front-end RF modules - everything from the uplink of the radio to the antenna - on a single device. Broadcom, which is the market leader, has taken our 2521A PA and that s a big win for us. Broadcom already uses our 2520 and 2529 PAs in its reference designs, and the company has now moved to our higher-integration module.
CS: The WLAN market boomed in 2003, how does it look this year?
AP: We estimate our total global available market to be around 65-70 million units in 2004, although there are some rumors that this may dip a little. For the 802.11g protocol, which is probably the dominant standard, we are the market leader. We have about a 48-50% share in 802.11g. The 802.11b market is dwindling and becoming commoditized, so we are focusing our efforts on 802.11g.
CS: How do you explain the relative success of SiGe Semiconductor PAs in WLAN compared with handsets?
AP: There has never really been a SiGe versus GaAs debate in WLAN. The integration advantage for SiGe has been a tremendous seller. Most of the GaAs guys require external transistors and regulators to provide a 2.8 V reference voltage and to turn the PA on and off. We ve actually integrated that into the SiGe die, which saves about 20 cents on the bill of materials (BOM).
We ve also been able to integrate the biasing with a robust power detector. I can get much more on the die and that reduces the BOM considerably [compared with GaAs]. SiGe has always met industry ruggedness requirements, it just comes down to performance and cost.
CS: Anadigics has said that the 802.11a/g dual-band protocol is more suited to GaAs than SiGe. What s your response?
AP: I disagree. I think there s a place for SiGe in the dual-band application. We have SiGe 5 GHz PAs available and we have combined the benefits of SiGe with those of GaAs for the 5 GHz band. Therefore we can provide a cost-effective dual-band solution that the GaAs guys still cannot.
Our dual-band solution is completely digitally controlled, so for the 5 and 2.4 GHz bands you can connect our PA directly to a CMOS baseband with no interface or glue logic. The GaAs guys are still going to have that integration problem so they will need some transistors and a reference voltage, and that s all extra cost. We ve already seen substantial price erosion in 802.11b and we now see it happening in 802.11g. It s going to happen in dual-band as well, and we ll be able to follow that curve down [better than GaAs].
The big figures of merit for WLAN are error vector magnitude [EVM] and output power. Our PAs combine high integration with a low EVM versus output power. The PA that we ve just launched [2528L] has less than 3% EVM at a 23 dBm output. That s important. It allows our customers to provide 20 dBm at the antenna, and for regulatory reasons this is where everyone wants to be. Getting above that figure is not really required: you would have to get into some sophisticated power control and no-one really wants to do that. Independent tests showed a throughput improvement of 5-6 Mbit/s and 20-30 m extra range.
CS: How do these figures compare with the best GaAs solutions?
AP: They are competitive. There is one GaAs PA that is [marginally] better, but the problem is that there is no integration, so you have to add another 40 cents to the overall cost to get the same level of functionality. So while with GaAs there might be a little bit more there in performance, you have to pay a price. That performance margin cannot be used, so at some point there will be diminishing returns. The 40 cent price advantage is huge: now the BOM for a whole card is approaching five to six dollars, so it s 5-10% of that cost in the consumer market.
CS: Has the 2528L been a success?
AP: It shares the same package as our previous 2525 device, so all of our customers can fit this higher-power device in the same socket. With three or four component changes you have the high-power design, and you don t have to change the layout or the board. Our customers love this because they can have a high-power/low-power design without having different boards.
To date we have design wins with four different customers. We gave out some early samples and people went crazy over it - it all goes back to higher performance and integration for lower BOM.
Revenue from this new PA will cannibalize some of our older designs, including the 2529 and the 2525. In the first four months or so it will probably take away about 20-30% of revenue from those existing products. By mid-2005 we would expect the old designs to be out completely, and we expect most of our existing customers to switch over.
CS: What opportunities are there for SiGe in the emergence of WiMax connectivity [IEEE 802.16]?
AP: WiMax was set up with two primary applications in mind. One is point-to-point backhaul for Wi-Fi: so if you have an access point on a lamp post and you want to connect to the main hub, then you would use 802.16. The second is point-to-multipoint, where you have one big transmitter to provide broadband services to homes. So it s very much like the local multipoint distribution services we had six or seven years ago, but now it s standardized and pushing forward. Those are two big markets.
Having said that, there is a new activity that is also going to support mobile applications and give WiMax support on laptops (see box). For this, you need both back-haul and point-to-multipoint links.
There are lots of synergies with 802.11g and a, but for the higher throughput the noise control has to be better. WiMax mainly operates in the 2.5-3.5 GHz spectrum, with support in the 5.8 GHz band as well. The frequencies are similar and the output power is a little higher, but essentially WiMax is very similar to Wi-Fi. Some people are using our existing PAs because of that overlap.
CS: What trends are you seeing in WLAN?
AP: An emerging theme in Wi-Fi is the new 802.11n protocol. The IEEE s intention is to reach 100 Mbit/s throughput through the 2.4-5 GHz spectrum. Right now the [theoretical] 54 Mbit/s throughput is actually more like 25 Mbit/s pure data. The "n" workgroup is trying to reach an actual 100 Mbit/s throughput to enable entertainment, video and Internet through the same pipe. That s the next big thing and we re working on it.
