Introduction

Modern digital communication technology requires a low level link known as the PHY or physical layer, to allow transmission and reception between connected devices. In most cases the integrated transceiver cannot produce enough power to realize the full potential of the specification, typically +20dBm.

This is due to the design of the transceiver which contains millions of CMOS gates to perform the DSP and MAC functions. These gates both take a lot of room, and dissipate a lot of power. For a fully integrated radio fabricated of CMOS and included on the single transceiver die, the output power seldom exceeds 0dBm.

Another approach is to include an additional die in the transceiver package to perform the radio PA (Power Amplifier) functions. This type of design can produce a few more dB of power output, but takes additional space horizontally, or results in a thicker package when the dies are stacked.

The additional power dissipation also can limit the performance. High power circuits can also leak into sensitive one causing undesired interference, especially with digital modulation schemes where phase and amplitude of the modulating signals are critical.

These limitations to a fully integrated high power radio have resulted in the need for external amplifiers for systems such as 802.11 WiFi. Initially these amps were added as discrete parts even down to the transistors configured into complex RF layout exercises.

The industry has moved to higher levels of integration to save on space and cost, and to reduce the development time. A number of companies have developed devices that incorporate the necessary parts into one external device known as an FEM (Front End Module).

These devices have now found extensive use in technology platforms such as 802.11, high powered Bluetooth, Zigbee, and other digital communication systems. These devices essentially are integrated versions of a discrete front end RF design, composed of a PA (power amplifier), a TX/RX (transmit/receive) switch, and various optional components such as an LNA (low noise amplifier) for the receive path, baluns, filters, diplexers, power detectors, and matching components.

Figure 1 – Conventional Discrete Transceiver Design

Figure 2 – FEM Transceiver Design

Conventional FEMs

The FEMs in use on typical WiFi platforms and other present day communication technology have similar configurations and designs. They are typically fabricated using multiple chips, discrete components, filters, and other devices wire bonded or connected through traces and packaged into a single device.

The IC technology used in FEMs varies by manufacturer, but generally falls into one of several types depending on the function performed. The following is a list of various IC technologies and the typical functions of the devices made from them.

GaAs MESFET

The Gallium Arsenide MESFET was the first of the GaAs semiconductors to be demonstrated. It is a great substrate for high frequencies as the bulk GaAs substrate has high resistivity.

This allows for the design of high quality passive components and low leakage switches with high isolation between ports. The MESFETs also have wide voltage swings and high breakdown voltage allowing it to be used in high output power applications.

Figure 3 – MESFET Horizontal Structure

Most GaAs MESFETs require a dual supply, making the designs more complex. The dual supply is typically provided through a charge pump integrated into the die, both increasing the size and also reducing the power efficiency. This can also result in spurious emissions in critical noise sensitive circuits and is especially undesirable in mobile applications.

The thermal resistance of GaAs material is 3 times greater than the equivalent silicon material, making heat dissipation a problem. This can be partially overcome by the use of substrate via-hole technology, which can also increase the RF performance by eliminating the parasitic and cross coupling resulting from wire bonding configurations.

MESFETs have a pure horizontal structure, so higher frequency applications required the use of high resolution optical lithography. Noise and linearity performance are very dependant of the surface state of the material, so mechanical and chemical processing of the chips must be tightly controlled.

It is relatively cheap to produce GaAs MESFETs, but they will always cost considerably more than silicon due to the higher cost of the starting materials, and are considerably more fragile than silicon.

GaAs HEMT & PHEMT

The GaAs high electron mobility transistor was invented in the 1990. This technology can operate at very high frequencies above the x-band with noise figures well under 1dB. It can be utilized in a power amplifier with very good efficiency, and can be configured as an RF switch. This technology also generally requires a dual supply, which complicates the circuit design as above. It is more expensive to produce then the GaAs MESFET due to the addition of the very thin layer of material with the high electron mobility.

The variation to the HEMT technology includes the PHEMT or pseudomorphic devices. There is also a class of PHEMT called E-PHEMT for enhanced-mode HEMT. These devices can operate with only one power supply which simplifies the design.

The transistors are also not as efficient requiring more surface area to achieve the same result raising the cost even further. The surface effects are not as pronounced as with MESFET, but still must be well controlled.

GaAs HBT

The heterojunction bipolar transistor is a relatively new commercial development. The transistor is a vertical structure and the optical process does not have to be as precise as with the FETs mentioned above, resulting in a lower cost to manufacture and higher yields. These transistors also operate on a single power supply simplifying circuit and IC design.

The InGaP (indium gallium phosphide) process was added to the basic HBT to provide stability and extended temperature range as well as intrinsically higher linearity for RF circuits. The bulk substrate is the same as MESFETs so passive circuits generally are high quality. GaAs HBT has a fairly high turn on in the 1.2 volt range, which can limit applications that require complex circuits or low voltage mobile designs.

The typical integrated designs include power amps and LNAs, but the process has not yielded good RF switches. Reverse isolation is not as good as some of the other compound IC technologies. Heat dissipation can be a problem but can be reduced by the via-hole technology mentioned for MESFETs.

Composite GaAs

Several companies are working on composite GaAs devices, which would include InGaP HBT and PHEMT on the same die. The desire is to use the best of both technologies to allow design of an RF front end device on a single die that includes the amplifiers and switches. This has proven so far to be too expensive and the yields have been less than satisfactory.

Figure 4 – HBT Vertical Structure

All GaAs based technologies have the limitation of always being a subsystem, because the difference in the fabrication techniques will never allow them to be directly integrated into base band transceivers. GaAs also suffers from limitations in complexity that reduces the features, capability, and interface flexibility available in pure silicon technology.

Silicon (CMOS, BiCMOS)

By far the cheapest to manufacture, silicon devices are produce in the foundry by an optical process, and the yields generally are very good with very low cost starting material.

Silicon devices are also by far the most rugged of any of the RF devices produced today making the die packaging simple, compact, and strong. Silicon also has excellent thermal dissipation properties improving heat transfer. These qualities make silicon the first choice for portable and mobile electronic devices.

Pure CMOS can be used to fabricate RF devices, but they generally have less desirable performance in noise figures and output power.

However, RF CMOS can be fabricated on sub 100nm that can achieve noise figures well below 1dB, but these fine line geometries are too expensive for commercial production of RF parts. SOI (Silicon on Insulator) technologies also have promising RF characteristics, allowing the creation of high performance RF circuits including RF switches, but at the expense of higher manufacturing costs and low output power.

The far better alternative is BiCMOS which has reasonably good noise figures and usable output power up through 70 GHz. Typical Fmax and Ft for SiGe (Silicon Germanium) is well above 150GHz today. The turn-on voltage of silicon is typically 0.7 volts which allows it to be very flexible for implementing multi-circuit complex designs.

One of the main disadvantages of Silicon is the poor electrical insulating characteristics of the substrate making it a poor medium for micro strip. This characteristic can be overcome somewhat through the use of up to 6 metal layers that are available in BiCMOS fabrication, two to three times as many as are typically available in GaAs.

These metal layers can create shielding allowing the performance of the passive components on the die to approach the performance of the passives fabricated on the lower loss GaAs substrates. These metal layers also allow very complex circuits to be designed and implemented, including the possibility of a full function transceiver composed of the RF front end, the ADC and DACs, LO, and LDOs.

In addition to the complexity of the BiCMOS, full CMOS logic can be fabricated and included on the same die enabling ease of interface and a full range of digital functionality making possible the dream of RF plug and play. The only piece missing is that traditionally there has not been good techniques for implementing a transmit/receive antenna switch using BiCMOS, which has limited the implementation of a single die RF front end module using silicon.

FEM Design

In order to take advantage of the GaAs HBT process with high yields, the designers of the FEMs are forced to use additional dies of different technology in order to provide the all important transmit receive switch. The typical configuration might involve one PA die and one LNA die both made from InGaP HBT, and a switch made with the GaAs PHEMT process.

Additionally, a bias and switch control circuit is required which will often mean an additional die of some type. All these ICs are then mounted in a package and connected through bond wires or some other technique. Matching and filtering components are often included in this package for the I/O ports.

This technology is proven and reliable, and is shipping in quantity worldwide, but the use of multiple higher cost dies and complex packaging results in extra costs passed on to the product manufacturers.

QFN Packaging

One of the simplest and lowest cost packages is the QFN (Quad Flat No leads). This surface mount package is built on a lead frame where the die or dies can be mounted and bond wires attached to connect to the pads and pins. Often with RF devices, the bottom center of the QFN package will include a large pad called a slug to help with grounding and heat dissipation. After the ICs and passives are bonded to the frame, the QFN is finished by molding the body with a thermal setting plastic and optionally a lid is attached if an air cavity is desired. A QFN package and side view is shown in Figure 5 and Figure 6.

Figure 5 - QFN Package Bottom View

Figure 6 - QFN Cut Away Side View

LTCC Packaging

The low temperature co-fired ceramic packaging is a high performance RF package, but is generally less compact and much more expensive. It works well for multi-die configurations as the inter trace impedances can be tightly controlled and RF matching and filtering components can be integrated into the substrate.

It is also strong and can tolerate extreme temperatures. The design of LTCC packages generally includes a number of layers in the substrate which facilitates complex interconnections for MCMs (Multi Chip Modules). Some sample LTCC packages are show in Figure 7 and a design example is show in Figure 8.

Figure 7 – LTCC packages

Figure 8 – LTCC design example

Flip Chip Packaging

Flip chip packaging is a way to connect the die pads directly to the packaging lead frame without using bond wires. Pads on the die are “bumped” with a conductive material, and the chip is placed upside down on the lead frame and bonded to the leads.

This technique has the advantages of improving RF performance while simplifying the packaging effort, and can help reduce the thickness of the finished parts which is very important for mobile platforms.

Flip chip techniques are generally used with the QFN package, but can be also be used with LTCC packages. When used with a QFN, only a single die can be place within the lead frame. A QFN flip chip contrasted to a QFN with wire bonds is shown in Figure 9.

Figure 9 – QFN Wire Bonded and Flip Chip

RFaxis’ RFeICs

RFaxis has developed a new design process that allows the fabrication of all the components needed for a front end module on a single die, including a proprietary method for switching the RF signal received from the antenna, also known as the RF switch.

We are currently using this technique to fabricate single die front end devices in BiCMOS, though the technology can be applied to other IC fabrication methods as well. This new innovation is called RFeIC for RF Front end Integrated Circuit.

This breakthrough allows the missing piece, the RF switch, to be included in the single die fully integrated front end device (RFeIC) for use with modern digital communications systems.

The innovation of the RFeIC has opened up a new era in the design of radio system front ends. RFaxis has designed multiple configurations of these devices to support highly integrated solutions for WLAN 802.11, Bluetooth, Zigbee, MIMO, WAVE, and other systems that require two way communications between RF connected platforms.

The RFaxis designs include a high efficiency and high linearity PA for the transmit channel, a high power LNA for the receive channel, a transmit/receive switch, matching and harmonic filters, a transmit power detector, and control circuitry.

These parts are packaged in a 3mm X 3mm QFN, and will be available in a flip chip version. Dual band devices will soon be available supporting application operating both at 2.4GHz and 5GHz. The RFeIC is also an excellent device to use as a building block for a MIMO channel.

Operating each MIMO channel through a separate single channel RFeIC has the advantage of less complexity and more flexibility in the layout, and additional isolation between channels and a shorter antenna run between the antenna and the TX/RX input of the front end device.

Basic Building Blocks

RFaxis has developed a number of “Basic Building Blocks” that allows configuration of the needed components for digital radio front end design. These block are variations of a power amplifier, a low noise amplifier, and the transmit/receive RF switch with appropriate matching and filtering. The following block diagrams in Figure 10 include the basic function and technology supported.

Figure 10 – Basic Building Blocks

Advantages

In addition to the simplicity of a single die in a radio front end RFeIC, our design has a number of other advantages. As all the front-end functions (PA, LNA, Switch, Matching, Bias, and Logic) are included on a single die, the size is compact allowing for a very small finished package, with the flip chip technique further reducing cost and increasing performance.

Small die size also translates to low cost as well, and the BiCMOS process is a main stream manufacturing technique that lends itself well to high volume, high reliability and high yield. As this design includes an LNA, performance will be substantially improved over designs without the receive channel amplifier. (See also the System Noise Figure White Paper)

The following are some charts that show system improvements that can be realized while using the Rfaxis RFeIC as the system RF front end. In Figure 11, two different design examples show the system sensitivity improvement verses LNA current. It is worth noting that sensitivity and range are often easily substituted for data bandwidth and transmit duty cycle which can result in substantial system power savings.

Figure 11 – System Sensitivity Improvement Using the RFeIC

In Figure 12 the data represents improvements in the range for two design examples above in comparison with Bluetooth class-2 devices, with the upper lines representing the increase in the range when the RFeIC is used on both ends of the link, while the lower lines representing the increase of the range when the RFeIC is present on one end of the link and the only sensitivity increase is due to LNA which improves the system noise figure.

Figure 12 – System Range Improvement Using the RFeIC

Performance

The typical performance of the RFX series can be found in Table 1.

Table 1 – RFeIC Typical Performance Characteristics

Applications

Figure 13 is a typical application circuit for Bluetooth and Zigbee based on the RFX2401.

Figure 13 – Application Circuit Using the RFX2401 RFeIC

Many Bluetooth and Zigbee transceivers have an internal TX/RX switch to support direct connection through a band pass filter to the antenna. Even though this will result in a simple circuit design, the performance will yield low power output and an undesirably high system noise figure. With the addition of the simple, low cost RFeIC, performance will be greatly increased.

Figure 14 is a typical application circuit for single-band 802.11 based on the RFX2402.

Figure 14 – Application Circuit Using the RFX2402 RFeIC

This application can be used for both 802.11g and 802.11n. The small size and versatility of the RFeIC make it an ideal building block for MIMO application requiring 2 to 4 channels of spatial diversity as shown in Figure 15. With the RFeICs located near the antennas, excellent channel isolation will be maintained.

Figure 15 – 3 X 3 MIMO Application Circuit Using the RFX2402 RFeIC

Figure 16 presents additional MIMO configurations. Using the RFaxis building blocks, these configurations are easily supported with a single die, compact, high performance device. The dual LNA supports the spatial diversity.

Figure 16 – Additional MIMO Application Circuits

Dual Band & 5GHz Applications

There has been a strong need for additional channels for communications beyond the 2.4GHz band. Many channels are available in the 5 GHz band as supported by 802.11a and n.

The growth of products supporting the 5GHz domain have been limited for a number of reasons mostly related to complexity and cost of supporting the additional band, with additional concerns due to smaller coverage area.

Rfaxis is developing basic building blocks presented in figure 16 above that operate in the 5GHz band in addition to the 2.4GHz parts.

The RFaxis dual band RFeICs offer simplicity and performance, enabling a low cost product that can support the 5GHz band with very little additional cost over single band configurations. Our highly integrated 802.11 building block RFeIC includes all the RF functions to allow direct connection between the transceiver and antenna as shown in Figure 17.

Extended frequency range of 4.9GHz through 6GHz, robustness, and high linearity allows use of the RFaxis building blocks for automotive applications such as 802.11p WAVE. The applications include traffic control, car telematics, electronic tolling with the technology located either on-board or in roadside units.

Figure 17 – 2.4GHz/5GHz Application Circuit

Mobile Applications

RFaxis has developed innovative solutions with high linearity, high sensitivity, and very low power for mobile platforms that both save cost and space. The RFX2405 is a single-die device composed of the Bluetooth and single band WiFi building blocks, and includes a single antenna connection.

With the use of Bluetooth/WiFi coexistence, these two competing technologies can operate seamlessly on the same platform, and both technologies can be in receiving mode through the same antenna at the same time.

The application circuit is detailed in Figure 18.

Figure 18 – Mobile Platform Bluetooth WiFi Application Circuit

Conclusion

The innovative designs of the RFeIC from RFaxis have accomplished a breakthrough allowing the lowest cost fabrication technique available for RF devices to be used to make front end systems for applications including 802.11 a/b/g/n, Bluetooth, Zigbee, WiMAX, WAVE, and other custom digital radio communication technologies.

By designing a method to fabricate a completely integrated matching and switching network for transmit/receive operation along with a high performance linear power amplifier and low-noise amplifier in a BiCMOS single die, the integration of the digital radio can now be taken to the next level resulting in a reduction in cost and size, and an increase in performance.

For more information visit www.RFaxis.com