[For a copy of this article in PDF format, which displays figures and equations,
Requires Adobe Acrobat Reader,
This year, more than 430 million cellular phones, 30 million notebook PCs and 12 million personal digital assistants (PDAs) are expected to be shipped. Each of these products will have two things in common: They will all be battery powered, and they will all have some kind of noise-sensitive radio circuitry (or add-on).
Cellular phones have always used sensitive RF receivers, but the receivers' addition into PDAs and handheld computers is a more recent phenomenon, driven, in part, by the proliferation of Bluetooth chipsets. At the same time, these products are replacing linear regulators with switching power supplies to increase battery life. The combination of noise generators (switching power supplies) and noise-sensitive circuitry creates the potential for interference.
The traditional solution to interference was to keep noise-generating circuits away from noise-sensitive circuitry. However, in modern handheld products, everything is so tightly packaged that this is no longer possible. Resorting to shielding isn't practical for cost or size reasons. Traditional switching power supplies concentrate noise energy into narrow-band harmonics. However, if one of those harmonics happens to coincide with a sensitive frequency — a receiver's intermediate frequency (IF) passband, for instance — interference will likely result.
An Existing Approach
One technique, which has been used with success, has been to dither the system clock of the dc-dc converter. This approach, with its resulting spread-spectrum operation, allows the switching frequency to be modulated by a pseudo-random number (PRN) sequence to eliminate narrowband harmonics. In essence, the noise is “spread” across the frequency range instead of being concentrated in distinct harmonics. Because the peak amplitude of the spread-spectrum noise is much lower, interference is significantly reduced. Although this method has been successfully implemented discretely in the past, process improvements now allow spread spectrum to be included in newer dc-dc converters, achieving significant space savings.
To understand how spread-spectrum operation is implemented on chip, consider the block diagram in Fig. 1. The dual-phase switched capacitor charge pump of is used to step down the input voltage to a regulated output voltage. Regulation is achieved by sensing the output voltage through an external resistor divider and modulating the charge pump output current based on the error signal. A two-phase nonoverlapping clock activates the two charge pumps. The two charge pumps work in parallel but out of phase from each other.
On the first phase of the clock, current is transferred from VIN through the external flying capacitor 1 to VOUT via the switches of charge pump 1. Not only is current being delivered to VOUT on the first phase, but the flying capacitor is also being charged.
On the second phase of the clock, flying capacitor 1 is connected from VOUT to ground, transferring the charge stored during the first phase of the clock to VOUT via the switches of charge pump 1. Charge pump 2 operates in the same manner, but with the phases of the clock reversed. This dual-phase architecture achieves extremely low output and input noise by providing constant charge transfer from VIN to VOUT. Using this method of switching, only half of the output current is delivered from VIN, thus achieving a 50% increase in efficiency over a conventional low dropout (LDO) regulator.
Spread-spectrum operation is achieved by modulating the oscillator frequency on a cycle-by-cycle basis in response to a changing PRN. Modulating the operating frequency only a few percent significantly reduces peak and harmonic noise.
Conventional Regulator vs. Spread Spectrum
Switching regulators can be particularly troublesome where electromagnetic interference (EMI) is concerned. Switching regulators operate on a cycle-by-cycle basis to transfer power to the output. In most cases, the frequency of operation is either fixed or is a constant based on the output load. This method of conversion creates large components of noise at the frequency of operation (fundamental) and multiples of the operating frequency (harmonics). Fig. 2a shows a conventional step-down switching regulator. Figs. 2b and 2c are the input and output noise spectrums for the step-down converter in Fig. 2a with VIN= 3.6V, VOUT= 1.5V and IOUT= 500mA.
An example of an IC with spread spectrum implemented on-chip is LTC3251, a 500-mA high-efficiency, low-noise, inductorless step-down dc-dc converter. The spread spectrum oscillator of this chip is designed to produce a clock pulse whose period is random on a cycle-by-cycle basis, but fixed between 1MHz and 1.6MHz. This has the benefit of spreading the switching noise over a wide range of frequencies. A typical application circuit for the converter is illustrated in Fig. 3a. Figs. 3b and 3c are the input and output noise spectrums for the switching regulator of Fig. 3a with VIN= 3.6V, VOUT= 1.5V and IOUT= 500mA. As a result of spread-spectrum operation, there is a significant reduction in peak output noise (>20dBm) with only one-half of the output capacitance. Input harmonics are virtually eliminated with only one-tenth the input capacitance. Spread-spectrum operation is used exclusively in “continuous” mode and for output currents greater than about 50mA in burst mode.
To improve efficiency at low output currents, a burst mode function is included in this switching regulator. An output current sense is used to detect when the required output current drops below an internally set threshold (50mA, typically). When this occurs, the part shuts down the internal oscillator and goes into a low current operating state. The LTC3251 will remain in the low current operating state until the output voltage has dropped enough to require another burst of current. When the output current exceeds 50mA, the LTC3251 will operate in continuous mode. Unlike traditional charge pumps where the burst mode current depends on many factors (i.e., supply, switch strength and capacitor selection), the LTC3251's burst current is set by the burst threshold and hysteresis. This means that VOUT ripple voltage in burst mode operation will be fixed and is typically 15mV with a 10-μF output capacitor. In this case, the operating current is only 35μA.
To further optimize the supply current for low-output current requirements, the LTC3251 includes a super burst mode operation, which is similar to burst mode operation, but much of the internal circuitry and switch is shut down to further reduce the supply current to only 8μA. In shutdown mode, all circuitry is turned off, and the IC draws only leakage current from the VIN supply. The MD0 and MD1 pins are CMOS inputs with threshold voltages of 0.8V to allow regulator control with low-voltage logic levels. The IC is shutdown when a logic low is applied to both mode pins.
To meet the current demands for high functionality, wireless portability and reasonable battery life in handheld applications, designers are forced to use high-efficiency switching power supplies together with noise sensitive RF circuitry. Traditional switching power supply architectures will improve battery life, but may necessitate costly and space consuming filtering or shielding. However, power supply ICs with “built-in” spread-spectrum switching will alleviate harmonic interference with sensitive RF circuitry without such measures. Furthermore, their small size, relatively high current outputs and low noise make them ideally suited for space-constrained battery-powered applications.
For more information on this article, CIRCLE 332 on Reader Service Card