Minimize GSM buzz noise in mobile phones

Today's handset design requirements
Compact design's used with today's handsets results in highly sensitive electronic components placed to very near to one another due to market requirements. These electronic components must not only be extremely small, but also must not intrude into the operation or corrupt adjacent devices. In addition, today's handset designs must operate without flaw despite the extremely small signal strengths transmitted from the base station which, in turn, result in the handset transmitting at its maximum power level.

With the expansion of handset marketing to more people worldwide, and given the expectation of faster design times along with the greater number of features required to expand market share, the handset designer is often forced to develop PCB layouts with placement of large components that dictates trace routing within extreme schedule constraints. This means that traditional routing and crosstalk reviews may be limited or even skipped, as long as the prototype handset functions. Often the prototype needs to only demonstrate basic functionality, and may not include full RF (RF Transmitter) power-output testing. It is often this last stage of evolution where problems regarding 'GSM Buzz' are identified.

User problem description
How often have you been in your company's conference room and placed your handset close to a speakerphone, and then later heard an intrusive, repeated humming sound? Sometimes this occurs even when the handset is not being used, but turned on. Move the phone a greater distance from the speakerphone and the problem is reduced. Another similar situation can be observed when placing a GSM handset near an AM or even FM radio. This objectionable repeated humming sound is often referred to by handset design engineers as 'GSM Buzz'.

This problem can occur completely within a handset by itself, especially with the RF power amplifier (PA) chip is at its maximum output level.

Technical explanations of the problem
Today's handsets contain efficient batteries that must provide on-demand power to the RF PA. The PA delivers large amounts of RF energy within either the 850/950 or 1800/1900 MHz GSM bands. The +33 dBm GSM power is delivered within these bands at 217 Hz-modulated bursting intervals. Figure 1 illustrates the burst content within the audio range typical in a GSM phone.

Figure 1: Spectral content of GSM burst
(Click on image to enlarge)

How the burst of RF energy can create audio buzz
The 217-Hz bursting results in large power excursions from the handset battery. These excursions are typically 1 to 2 amperes, depending on the efficiency of the power amplifier. Because of these large current transients, slight amounts of resistance and inductance in the battery and its leads result in 217-Hz voltage transients throughout the entire cell phone.

Figure 2 illustrates a typical voltage transient resulting from the 217-Hz RF power transients at an audio chip. Note that both the power-supply rail and the ground reference have shifted. In this case, this suggests that there is impedance in each of the nodes. The 217-Hz burst, in effect a square-wave envelope, may also contain harmonics that exist well into the audio band.

Figure 2: Typical 217-Hz power transients
(Click on image to enlarge)

This parasitic resistance and inductance can be identified in numerous places within a handset. Often, these are found in within the battery itself; within PCB (Printed Circuit Board) interlayer vias, traces resistance and even connectors. Figure 3 illustrates typical places this resistance and inductance can be found.

Figure 3: Common high-resistance and high-inductance locations in a handset
(Click on image to enlarge)

Solutions to the problem
1) Bypass and decoupling capacitors
The first and most logical solution is to implement both bypass and decoupling capacitors local to both the power amplifier chip, and then also to the audio components themselves which are supporting the audio paths. It is good practice to first focus on the PA and then retest, after these components have been added. Because of the large currents, often the size of the decoupling capacitor adjacent to the RF power chip must be very large.

Due to size and cost limitations, the size of this capacitor must be reduced, thus resulting in greater then normal voltage excursions. Bypass capacitors (0.01 μF typical) should also be used as the edge rates of the power transients can easily exceed the bandwidth of a decoupling capacitor. These capacitors must be placed close to the PA chip.

A large decoupling capacitor may be placed in parallel with the battery to reduce the transients from the internal resistance of the battery, and may even eliminate a problem in a handset. However, the drops associated with the PCB and flex circuit remain, and the possibility for GSM buzz remains.

2) Path to the power amplifier chip
The PA must have large low-inductance and low-resistance ground and power paths. These paths need to be consistent from the battery through every node. This includes connectors, flex circuits, printed circuit boards and solder pads. The routing must be as direct as possible with minimum excursions. Ideally, both power and ground are designed as planes through to the PA chip. However, this is often not the case due to space limitations. It is also important to not run a power path to the PA chip directly over audio paths. Modulated power rails can easily couple into audio circuits and can be amplified.

3) Vias
Poor power and grounding through vias remain the largest contributor to difficult GSM buzz problems. If at all possible, route traces through PCBs, flex circuits and connects to minimize via utilization. If vias must be used, it is good practice to double up or use quadruple vias, all in parallel, and ensure that both power and ground paths contain adequate protection. Numerous calculators are available on line to approximate resistance and inductance as part of the design

4) Isolating high-resistance points
Common resistance-measurement techniques usually do not have accurate resolution to identify high-resistance nodes in a handset assembly. More-expensive multimeters offer a 4-wire measurement mode to precisely measure resistance. This well-established technique uses two meter leads to source current ,and two leads to sense voltage drop, thus effectively leaving only the resistance of interest in the measurement while removing the resistance of the meter leads from the measurement.

5) Voltage drops
Basic Ohm's law can be used to approximate the voltage drop across a suspect path of a handset. Knowing that the transient currents can be 1-2 amps, a resistance of only 50 mΩ can result in a voltage drop of 100 mV in each path (both ground and power). It is important to recognize, that sometimes the drop may be in a power path, but not a ground path. The converse may be true.

Summary
Excessive GSM RF noise can be greatly reduced with proper handset design, if basic good-design principles are followed. These include:

1. Use adequate bypass and decoupling capacitors at the PA chip.
2. Ensure direct routing of power and ground from the battery to the PA chip. This should also include minimal crosstalk of power to sensitive analog audio signals.
3. Ensure low-resistance and low-inductance paths from battery to PA chip. Ground and power planes are most desired, but difficult to achieve. Exceeding 50 mΩ can cause problems.
4. Minimize ground and power vias to the PA chip and also to key audio circuits.
5. If budget for cost of material allows, use bypass and decoupling capacitors on audio devices such as amplifiers and filters.

About the author
John Whalen is currently Senior Manager of Business Development focusing on Reference Designs worldwide within Fairchild Semiconductor. John previously held positions in Engineering and Product Development Management, Regulatory Affairs, at Biotronik Pacemaker Defibrillator Group. Early in his career, John was a power-supply hybrid designer and test engineer at GE Aircraft Engines.

Editor's Note: If you are interested in the low-glam but essential topic of decoupling and bypassing capacitors, check out these in-depth articles which appeared at Planet Analog (they contain links to their preceding parts, as well, of course):

• "Choosing and Using Bypass Capacitors (Part 3 of 3)"
• "A bypass-capacitor dialogue peels back the layers, Part 3: Continuing the discussion on layout considerations"
• "More about decoupling and bypass capacitors, power-supply rails, models, and their interactions (Part 5.6)"