Switch technology delivers longer cellular talk times

Agilent Technologies, which is following this route, believes it can grasp the winning formula now that it has acquired the Korean power-amplifier (PA) specialist Wavics (Compound Semiconductor March p9). According to Agilent mobile and wireless industry analyst Joe Madden, the combination of the firm s existing filter technology, which discriminates between call signals and far more powerful electromagnetic radiation, its E-PHEMT technology, and Wavics advanced circuit design will propel Agilent up the RFIC vendor rankings. Currently placed ninth in RF and microwave, the California-based chip maker has its sights set on the top spot.

"Before the [Wavics] acquisition, Agilent s biggest strength was in duplexers and filters fabricated in our Fort Collins fab," revealed Madden. "We were shipping about 2 million GSM amplifiers per month to some large European OEMs [original equipment manufacturers], and we had introduced an EDGE amplifier. However, we had not got very far in terms of CDMA amplifiers, so our product portfolio for the entire RF space for mobile handsets was limited. We had a hole in the portfolio." With the addition of Wavics, a firm with customers including Samsung and LG and a specific strength in CDMA amplifiers, Agilent has now plugged that gap.

Incorporating Wavics HBT technology into Agilent s product line is not a trivial matter, because Agilent has always used an E-PHEMT process that was developed at its own labs. But although the fundamental transistor technology is quite different, Madden doesn t see this as a problem, because Wavics technology is not specific to any particular type of amplifier. "[Wavics technology is] agnostic in terms of the transistor or process used," said Madden. He adds that Wavics success has resulted primarily from its circuit design: "Wavics has this great circuit technique and Agilent has an inherently efficient transistor process for amplifiers. When we put those two together we ll have the most efficient power amplifiers on the market by far."

Wavics amplifiers use patented "CoolPAM" technology, so-called because the circuit techniques deployed allow amplifiers to run cooler and more efficiently, thereby increasing handset talk time. According to Madden, this technology is particularly well-suited to efficient power amplification for handsets operating in the low-power regime, and integrates well with the high-power output provided by Agilent s E-PHEMTs.

If you re out in the countryside and you want to use your cell phone to call several miles back to your base-station, all of the amplifier stages in the phone would be used. "We would be sending the signal through the two output stages to get the maximum output power, and in that case the E-PHEMT technology is quite strong," said Madden, who claims that Agilent s amplifiers are 5% more efficient than those based on HBTs. But that subtle difference isn t enough to make cell-phone OEMs chose an Agilent product, says Madden, because handsets don t normally operate at maximum power. "Most of the time people s phones are backed off by about two orders of magnitude in power. What that means is that phone designers care much more about low-power operation."

Wavics circuit design offers high efficiency at low operating powers. "The CoolPAM technique is a way of bypassing those output stages so that for low-power operation in class AB mode, we can have up to 15% better efficiency than traditional HBT amplifiers," said Madden. Although other companies are also switching-out their output stages, Madden says that the key to Wavics technology is the absence of physical switches on the MMICs. "Essentially [Wavics technology] characterizes the impedance of the output stage in both the on and off state," he said, explaining that this is achieved by inserting a matching circuit. Switching off the output transistors presents a different impedance to the output of the driver stage, which in turn causes the output signal to take a different route. "We are able to take advantage of that to reduce the insertion loss within the amplifier and improve the efficiency," said Madden, who reveals that this process also reduces manufacturing costs by using 15% smaller die.

Prior to Agilent s acquisition, Wavics was producing reliable devices in large volumes and was on the verge of becoming a top-tier supplier to companies such as LG and Samsung. According to Madden, however, these big customers were nervous about using Wavics exclusively, because the company, which was fabless and working with foundries, had an unproven supply chain.

Offering an interesting perspective on winning contracts with LG and Samsung, Madden claims that their barrier to entry is lower than with other OEMs: "[LG and Samsung] are open-minded when it comes to trying new things. If something doesn t work, they move on and try something else very quickly. That s their mode of operating in Korea, and that s why Wavics had the opportunity to get in."

Madden explains that Agilent will continue with the foundry model that Wavics established for HBTs. "[The outside fabs] have done a good job for Wavics. That business has been successful, and we see no reason to change."

"The strength of Agilent is that we have armies of people that know how to fabricate devices and we can specify to a much greater level of detail what we want these foundries to do. That s part of the synergy," he said. According to Madden, Agilent s efforts to date have focused on developing discrete components: "We ve taken the individual duplexer and E-PHEMT power amplifiers and simply put them together on the same board."

Agilent s approach in 2003 was to mount a filter and amplifier on the same substrate, but its next-generation modules will be influenced by Wavics design principles. "Rather than using a 50 Ω interface between the filter and amplifier we can design for optimal impedance for the filter you choose to use," said Madden. "That will allow us to shrink the filter and improve the insertion loss between components by eliminating several steps of matching and fine-tuning. And you ll see another step of improved efficiency for talk time."

Madden now expects Agilent to enter into a period of increased integration: "We ll have these multiband or multifunction filters and put them together with multiband amplifiers. For example, you might have a CDMA front-end module that has both CDMA frequency bands, 800 and 1900 MHz, with the two different duplexers, all in one sort of super module."

The natural extension of this, he says, is a phone with one chip from Qualcomm that handles the RF transceiver and baseband components, one chip from Agilent that handles the non-CMOS components, and an antenna. The improvements that are expected from Agilent s increased integration are illustrated in figure 1. Next year the company plans to extend the integration to include Wavics circuit technique. The result should be an improvement in talk time of up to 90 minutes.

Madden believes that Wavics circuit design is already giving Agilent an edge, by offering talk times at power outputs of 10 dBm that are 25-30 minutes longer than those of its competitors. He believes that this advantage is significant: "Customers like Samsung, Nokia and Siemens are very much focused on the talk time. From what I hear, half an hour is a significant improvement. Ten to 15 minutes might not be noticed by a consumer but when we get to half an hour, that starts to be a differentiator for the OEM."

As Agilent continues to turn around the fortunes of its semiconductor products group, the RF sector offers the greatest room for improvement. If successful, the integration of E-PHEMT and CoolPAM technology should see it move up the RFIC ranks quickly.

"Agilent delivers E-PHEMT PAs with low-voltage operation" Compound Semiconductor May 2004 p23.