Politically correct power

One of the intended consequences of universalpower-supply designs— those that accommodate inputs from 100V ac or less to 240V ac or more—is that the powermanagement circuits become source-independent and the designer need consider only matters specific to the intended load. Such a power-management approach unlocks considerable economies of scale with significant reductions in the number of supply models required to serve a global customer base. Similar advantages accrue to supply-chain, manufacturing, and inventorymanagement operations.



The unintended consequence of this approach is that the resultant supplies must meet the most stringent standards that the collective market imposes. Power standards have been evolving over the past several decades, and, importantly, the number of regulatory bodies that perceive a value to such norms has recently been growing. The rationale is simple: Left unchecked, the inefficient use of electric power leads to either rationing or mandatory major-infrastructure investment. Low power factors are pernicious detractors from grid efficiencies and have been the focus of mounting global concern. The European Union has been ahead of the power-factor-standards curve, but more recently the CCC (China Compulsory Certificate), the US EPA’s Energy Star program, and Japan’s JIC-C-61000-3-2 indicate that PF (power-factor) performance is evolving into a global mandate. Furthermore, high minimum power factors are no longer limited to highpower applications. Newer norms and standards apply to electronic products dissipating as little as 75W or 26W for lighting devices.


Making waves
To see how recent trends in the electronics industry have elevated concerns about PF, first examine the definitions of real and apparent power. A circuit’s real power dissipation is defined as the average of the instantaneous voltage-current product over a line cycle:



The apparent power dissipation, on the other hand, is simply the product of the rms voltage and current:



The PF, typically expressed in percent, is simply the ratio of the two:



If the voltage and current waveforms are both sinusoidal and in-phase, the real and apparent terms are equal, resulting in a unity PF. This relationship exists when the load is entirely resistiveessentially the case with noninductive heating elements and tungsten-filament lighting, for example.

Reactive loads impose a phase shift between their current and the applied voltage, but, because they represent linear impedances, the current waveform remains sinusoidal. The resulting PF is:



where Θ the phase shift, derives directly from the impedance:



Here, XL and RL are the reactive and resistive components of the net load impedance, respectively.

International Rectifier Chief Executive Officer Alex Lidow notes, “Half of the electricity that we use goes into motion. About 85% of [that half], goes into ac induction motors that are electromechanically or triacactivated” (Reference 1). This common arrangement forces the grid to supply current in excess of that required to drive the load and to do so in a way that is challenging to meter, because, as Equation 1 suggests, the net current comprises in-phase and quadrature components (Figure 1). The apparent power delivered to the system decomposes into a real component that does the work of the application and draws IP amperes and a quadrature component that draws the motor’s reactive or magnetizing current, IR.



In industrial installations, the utility-power provider may meter the two terms separately using a standard kilowatt-hour meter for the real component and a kilovarhour meter for the quadrature component. Billing, in this case, would take the rms value of the two readings over the billing period. But most utility-power customers have only one meter—a situation that points out a fundamental inequity in retail power distribution that works decidedly against the utility: Providers sell electric energy in energy units, sensibly enough. But the industry measures the powerdistribution infrastructure’s capacity in units of current. Given that the ac voltage is both a constant amplitude and the reference phase in Equation 3, that quantity cancels, and the PF becomes the ratio of real to apparent current, integrated cycle by cycle over a billing period, as the utility sees it. In this view, the PF gauges the current capacity fraction for which the utility can charge and, ominously, the reciprocal of the PF becomes an infrastructure-overbuild requirement. This relationship between load performance and both
operating and capital costs raises PF to the level of regulatory topic.




Harmonic divergence
As it happens, the largest and most common linear reactive loads tend to be inductive. These loads include several types of motors, vapor lamps, fluorescent ballasts, inductive heaters, and welders. All impose a phase lag (Figure 1). The customary PFC (power-factorcorrection) method has been the addition of capacitors and corresponding fusing and switching gear (Figure 2). If that were all there is to power-factor management, you’d be done. Unfortunately, virtually the entire electronics industry presents nonlinear loads to the grid and, as a result, substantially complicates matters by adding harmonics to the basic current waveform. Take, for example, a simple fullwave rectifier and capacitive filter with either a resistive or a current-sink load (Figure 3). The filter supplies the load current during most of the line cycle, recharging during the interval beginning when the absolute value of the input exceeds the capacitor voltage and ending at the input peak (Figure 4). The periodic gulp of recharge current has a fundamental at twice the line frequency because of the fullwave rectification. More important, the current waveform is rich with odd harmonics, the number and amplitude of which increase with decreasing recharging interval. An extension of Equation 3 accounts for the harmonic content of the current waveform in the PF calculation:



where IH is the harmonic component of the current waveform. In the limit, as the recharge interval approaches a Dirac impulse, (t), the harmonic series extends infinitely. With this in mind, consider the growing popularity and ever-increasing operating frequencies of switchmode power supplies, which bring the specter of long sequences of large-amplitude spurs in the current spectrum. This relationship





between switching behavior and power factor is the driver behind much of the growing regulatory concern. Indeed, the leading published norms on the topic, the EU’s EN61000-3-2 and EN60555, and their equivalents in international standards, IEC1000-3-2 and IEC555, limit the spectral content of input current in mainsconnected equipment from the fundamental to the 40th harmonic (Reference 2).



The historical approach to such problems in signal-processing applications is to add a lowpass filter (Figure 5). But, as the power level at which power-factor norms apply decreases, as it has over the last several years to less than 100W, the linear filter becomes increasingly uneconomic and unwieldy.



Power-IC designers have approached the problem using some of the same circuit-topology techniques that helped exacerbate it in the first place: the ability to quickly and efficiently switch power and to accurately measure the result in a small IC. Also key in active PFC implementations is the ability to perform high-quality analog-signal processing in a switching environment with minimum crosstalk between switching devices and linear signal-processing circuits. Among the key specifications of various PFC designs are net powerconditioning efficiency, THD, EMI, power density, and BOM cost (Reference 3).

As is often the case in electronic design, the task of solving a problem is much simpler if you have access to a signal in the shape of the desired result (Reference 4). In the case of power conditioning for PFC, the simplification hangs on the observation that the incoming voltage waveform all but hands the desired current waveform to the system. One PFC circuit—the CRM (critical-conduction mode)— typifies the signal-processing tasks (Figure 6). An error amplifier compares the output—dc plus ripple—with an on-chip dc reference. A multiplier scales a sample of the unfiltered, rectified input voltage by the lowpassfiltered output from the error amplifier, in effect providing a level of output-voltage regulation while producing the current reference waveform. A control circuit manages the shunt switch, ramping the inductor current to follow the reference waveform (Figure 7). The control cycle starts with the shunt switch on. A sense resistor detects and scales the shunt current, which a comparator measures against the multiplier’s output. When the shunt current reaches the level that the reference waveform requires, the comparator turns off the MOSFET. A secondary winding on the series inductor forms a current transformer and provides a signal to a detector, which turns on the MOSFET when the current ramps down to zero.



Other PFC structures implement different control algorithms, but replicating the voltage waveform with the current drawn from the power line remains the primary objective. CCM (continuous-current mode) PFCs maintain an average current equal to the ac reference signal. This approach reduces the output ripple and fixes the ripple frequency, which simplifies downstream filtering but does so with an increase in control-loop complexity.



The CRM PFC is currently common for applications that dissipate as much as 100W. CCM PFCs are common choices for applications with dissipations greater than 200W. In the 100 to 200W region, which coincides with a number of large consumer applications, power-supply designers must choose a topology based upon the trade-offs that are most appropriate for their overall system goals. When investigating the available PFC methods for your application, be aware that some literature refers to the CRM method as “transitional mode” and CCM as ACM (“average-current mode”). Most power-IC manufacturers that provide PFCs offer both types— often integrated with a variety of associated functions.

For example, the L6561 from STMicroelectronics is a CRM PFC IC that accommodates inputs of 86 to 265V ac. The IC develops 400V on its output and operates with efficiencies and power factors of 92.8≤≤97.3, and 0.89≤PF≤0.999, respectively, over the input-voltage range. With the addition of three passives and a diode, you can reduce the 6561’s distortion from its native 3.7%≤THD≤13.7% to 2.9%≤THD≤8.1%, also input-voltage dependent.



The 30-cent (100,000) PFC IC draws a maximum 90µA at start-up and a maximum 5.5mA operating at 70kHz. Grounding the ZCD (zerocurrent-detection) pin disables the part and typically reduces the quiescent current to 1.4mA. When you release the ZCD pin, the internal start-up timer restarts the circuit. In addition to the basic PFC function, the 6561 provides overcurrent protection and resistor-programmable overvoltage protection. STMicroelectronics provides the L6561 in either SO-8 or DIP-8 packages.

On Semiconductor’s NCP1601 dual-mode correction controller suits midpower applications, such as light ballasts, television monitors, and ac adapters. The IC can operate in fixed-frequency DCM (discontinuous-conduction mode); CRM, which is an inherently variable frequency process; or in a combination of the two. The architecture allows you to set the switching frequency in DCM and provides a synchronization capability. Overvoltage- and undervoltage-protection thresholds are, respectively, 107% and 8% of the nominal output. The controller also provides resistor-programmable overcurrent protection and hysteretic thermal protection.

A nominal 100W application circuit using the 1601 measures efficiencies of 93% with an 85V-ac input, rising to 96% at 265V. Over the same input-voltage range, PF and THD vary from 0.995 and 8.3% to 0.901 and 38.9%. The 1601’s start-up, operating, and shutdown currents are 40µA, 5mA, and 50µA, respectively. On provides the 1601 in SO-8 packaging for 59 cents (1000) and in DIP-8 packaging for 72 cents (1000).



International Rectifier has eliminated the analog multiplier and input sensing with a proprietary integrator-based chip that reduces the external-parts count. The integrator operates over a clock cycle, which leads to fast recovery from disturbances from either load or line. The IR1150 avoids input sensing by taking advantage of the PWM’s duty-cycle dependency on the line voltage. The control loop uses this derived reference waveform to set the average current. The chip does exhibit somewhat greater distortion near the power cycle’s zero crossings and at light load but nonetheless maintains conformance to EN61000-3-2. Unfortunately, hard numbers for power factor, THD, and operating efficiency were unavailable at press time but should become available as the PFC moves from its preliminary documentation toward formal release.

The 1150 provides undervoltage, overvoltage, peakcurrent, and open-loop protection. The PFC draws a maximum of 22 mA with a 1nF load and 200µA in sleep mode. The IC is available in an SO-8 package in consumer and industrial grades that differ by their operating-temperature range. The $1.05 (10,000) consumer-grade device operates at 0 to 70°C; the $1.38 (10,000) industrial-grade version operates from -25 to +85°C. The PFC IC’s operating frequency is resistor-programmable from 50 to 200kHz, allowing smaller magnetics than devices operating at lower frequencies.

A few semiconductor manufacturers have combined the PFC and regulator in one IC. TI’s UCC2851x family complies with IEC1000-3-2 using seriesconnected PFC and PWM stages. The PFC stage manages the leadingedge modulation while the PWM stage modulates the trailing edge. The PFC uses an ACM control loop. Of the family’s eight models, the first four operate the PWM at the PFC clock rate, and the last four operate the PWM at twice the PFC rate.

As is the case with many PFCs, the UCC2851x family provides undervoltage lockout, here with selectable hysteresis, overvoltage protection, and a peak-current limiter. The controllers operate at a nominal 200kHz and draw 6mA. TI rates the $1.80 (1000) controllers for -40 to +105°C operation and offers them in SO-20 and DIP-20 packages.

Another example of a PFC-plusregulator IC is the iW2202 from iWatt, which the company presented at this year’s APEC (Applied Power Electronics Conference). The IC manages the PFC loop in the digital domain and provides conversion efficiencies in excess of 88.3% at 90W operating on 90 to 264Vac supplies. An iW2202-based 19.5V ac/dc adapter reference design provides less-than-5% THD and greater-than-0.98 power factors across the input range. Output ripple at 4.62A is on the order of 1V.

The $1.29 (1000) controller consumes less than 300mW in standby mode. The regulator requires neither secondary feedback nor external loop compensation, further reducing the external-parts count. On-chip features include overvoltage, overcurrent, and overtemperature protection. The iW2202 is available in an SO-14 package.

Recent introductions of PFCs have been occurring with increasing frequency—a trend that is likely to continue and accelerate if regulations continue to expand into new geographic areas or make new demands on the system/ power-source interface. It is likely, for example, that system-level sleep-mode dissipation will be the focus of increasing regulation. If such is the case, it is also likely that power frontend ICs will be responsible for implementing the new regulations.

Author information
You can reach Technical
Editor Joshua Israelsohn at
jisraelsohn@edn.com.

References
1. Lidow, Alex, PhD, “Variable speed motion: a key to energy-savings,” EOEM Design Expo, March 16, 2005, www.eoemdesignexpo.com.
2. Bourgeois, JM, “Circuits for power-factor correction with regards to mains filtering,” AN510/0894, STMicroelectronics, 1999.
3. “Power factor correction handbook,” On Semiconductor, August 2004.
4. Black, Harold, “Stabilized Feedback Amplifiers,” Bell System