January 15, 1998
Postregulation technique efficiently
supplies multiple output voltages
Gedaly Levin, Cherry Semiconductor Corp
Various methods for regulating multiple-output power supplies exist. An
emerging technique--secondary-side postregulation, or SSPR--provides advantages for
high-frequency, high-power-density dc/dc converters.
Many electronic devices require two or more isolated and tightly regulated
voltages. For example, microprocessors require a precisely controlled supply voltage of
3.3V or lower, as well as the traditional 5V supply. Furthermore, electronic products
continue to decrease in size, demanding higher power densities and higher efficiency. The
combination of these trends presents a difficult challenge for design engineers. Choosing
the best approach to generate these multiple voltages requires an understanding of the
strengths and weaknesses of available techniques and of how each technique affects an
application. The main selection criteria are cost, power-conversion efficiency, and ease
of manufacture.
The problem of tight regulation for voltages of multiple-output power
supplies is not new. Some of the most popular methods include linear regulators, coupled
inductors, post dc/dc converters, magnetic amplifiers (mag-amps), and secondary-side
postregulators (SSPRs) with a semiconductor device as a switch. You can describe mag-amps
and SSPRs as postregulators with a programmable delay switch. This switch can be a
semiconductor device, a saturable reactor, or any other device with switch
characteristics. Each of these methods has positive and negative aspects.
Linear regulators
 Linear regulators are the simplest, most straightforward, and most
popular approach for low- and medium-current applications (Figure
1). These regulators are easy to design in, relatively inexpensive, and widely
available. The major disadvantage of linear regulators is relatively poor efficiency. The
only way to optimize the efficiency of linear regulators is to provide a tightly regulated
voltage at the input of the regulator and to satisfy the minimum dropout voltage
conditions at all line and load conditions.
Coupled inductors
 Coupled inductors are suitable for regulating secondary voltages for
which the tolerance can be ±5% to ±8% (Figure 2).
Tightly coupled inductor windings provide better regulation. However, tight coupling
creates the problem of dynamic interactions between the coupled outputs. Because all
outputs use the same inductor, little flexibility exists for component placement, and
laying out the pc board sometimes becomes difficult. For any protective or monitoring
features, additional circuits are required.
Second-stage dc/dc converters
 A second-stage dc/dc converter is another popular method for
postregulation. All you have to do is to connect another dc/dc regulator to the existing
output (Figure 3). However, several potential problems
lie behind the apparent simplicity of this technique. First, the output that serves as an
input to the second dc/dc converter must carry the combined power of both outputs. Second,
an additional ripple current resulting from dc/dc-converter switching action affects the
output of the main output, which requires beefing up the output filter with additional
capacitors. Third, the second dc/dc converter generates additional ripple and noise and in
many cases must synchronize with the main converter, requiring additional circuitry.
Fourth, this approach is fairly expensive because it requires an additional set of power
switches, inductors, capacitors, and PWM controllers. Last but not least, this technique
lowers the conversion efficiency because the voltage must be processed twice during the
ac-to-dc-to-ac-to-dc conversion.
 The function of a magnetic amplifier has nothing to do with magnetic
amplification. You could better describe the device as a "saturable reactor."
The main component of this regulator is a saturable reactor that acts as a magnetic
switch, exhibiting high-impedance characteristics during the blocking period (switch is
off) and low-impedance characteristics when in saturation (switch is on) (Figure 4).
The two basic types of mag-amps are set and reset. Popular as early as the
1930s and 1940s, the mag-amp approach became popular again in the early 1980s and is
probably the most common method for medium- and high-power postregulation. Core materials
that exhibit square BH loop characteristics are mainly used for the saturable reactor. In
the case of tape-wound cores, cores with very thin tape minimize losses as switching
frequencies increase. Unfortunately, 1-mil and thinner tape-wound cores are expensive.
With the introduction of so-called "amorphous cores" by Allied Signal
(Morristown, NJ) in the mid-1980s, it became possible to use mag-amps at higher
frequencies. Still, at 300 kHz and higher, amorphous-core losses are high, and designers
of high-frequency, high-density dc/dc converters must seek alternative solutions.
The main drawbacks of mag-amps include poor regulation at light or no
load. Because the mag-amp must block the pulse during overcurrent or shutdown conditions,
the core must be large, increasing both its cost and its losses (high core losses at high
frequencies greater than 300 kHz). Finally, except for toroids from a few companies, much
manual labor is necessary to wind and assemble a toroid, which also increases the cost.
Secondary-side postregulators
An SSPR uses a semiconductor device as a switch. This technique has not
been widely publicized, although it has been used in industry for many years. Not until
recently, when the requirements for tightly regulated multiple-output dc-dc converters at
high frequency (400 kHz) and high-power-density environments became part of everyday
design, have the benefits of SSPR been realized.
 Early SSPRs required several ICs, several discrete components, and
considerable ingenuity on the part of the design engineer. In 1996, Cherry Semiconductor
introduced the first inexpensive SSPR-controller IC, and two other manufacturers now also
produce ICs for SSPRs (Unitrode (Merrimack, NH) and Linfinity (Garden Grove, CA)). These
first-generation SSPR-controller ICs offer combinations of modulation and synchronization
techniques. None of these approaches provides an ideal or universal solution, and it
remains to be seen if one method will prevail. However, the high efficiency, low cost, and
simple implementation of the SSPR technique are rapidly making it the most popular choice
for medium-power applications (Figure 5).
 You can regulate the voltage on the secondary side of the transformer
either by controlling the volt-second product across the output inductor for buck-derived
topologies or by controlling the amount of energy for boost- and flyback-derived
topologies (Figure 6). In either case, the power-control
device is in series with the appropriate winding and performs either a delayed turn-on
function (leading-edge modulation) or a delayed turn-off function (trailing-edge
modulation). Both types have many similarities. The transfer functions for both modes are
identical except for the negative sign in front of the transfer function for the
leading-edge modulation, because the greater the required duty cycle, the earlier the
power switch must turn on.
Both methods offer virtually lossless turn-on. With trailing-edge
modulation, the power switch turns on before the arrival of the pulse. With leading-edge
modulation, the turn-on of the power switch usually happens after the arrival of the
pulse. At turn-on, the current in the winding is in steady state. The main transformer's
leakage inductance determines the time necessary to change this current. As a result of
fast-turn-on characteristics of the power switch (usually an n-channel FET), the switch
turns on before any measurable change of current in the winding can take place. Thus,
turn-on is nearly lossless. Leading-edge modulation offers lossless turn-off as well.
 Leading-edge modulation is compatible with any PWM topology and any
control method. Trailing-edge modulation creates the current waveform on the primary side
with a negative step. Thus, this method is incompatible with peak-current-mode control,
which is by far the most popular mode of operation (Figure 7).
Synchronize to main controller
Another important consideration with SSPRs is frequency synchronization
with the main controller. SSPR controllers may have their own oscillators, but an error
amplifier and a ramp generator to create the intersection that defines a duty cycle are
essential.
The easiest way to accomplish synchronization is by sensing either the
leading or the trailing edge of the pulse on the secondary side of the transformer. The
advantage of synchronization to the leading edge is that this edge coincides with the
start of the switching cycle of the main converter. Secondary effects from this method of
synchronization are nonexistent. However, the delay resulting from IC propagation and
sensing results in the loss of some available volt-seconds of the pulse. High-frequency
dc/dc converters, especially those that derive both 5 and 3.3V outputs from one secondary
winding, have tight volt-second budgets. In this case, trailing-edge synchronization is
preferable because the entire pulse is available for use.
Note that the main output dictates the trailing edge of the pulse. Thus,
the duty cycle varies with changes in the input voltage and load. This variation means
that the start of the SSPR ramp voltage also varies, requiring the error amplifier of the
SSPR to adapt to these changes to keep voltage in regulation. The most critical situation
happens when you subject the main output to step loads. Interaction between major and
minor SSPR loops may create transient conditions between the outputs.
All multiple-output postregulation methods use some common components to
process the pulse. Depending on the method and number of power-conversion stages,
different components facilitate regulation of the output. SSPRs that use an n-channel FET
as the power switch require only a few low-loss components. The losses of
leading-edge-modulation SSPRs are basically the sum of FET conduction losses and losses
associated with powering the IC controller and driving the FET. As a result, conversion
efficiency greater than 97% is possible.
Depending on the modulation and synchronization techniques, SSPR
controllers can operate in current- or voltage-control mode. Controllers that use
voltage-mode control can easily incorporate a feed-forward feature to improve line
regulation. Either method lets you achieve load/line regulation in fractions of percentage
points. In this tight regulation, SSPR has a clear advantage over mag-amps, which have
degrading regulation characteristics at light or no-load conditions.
 With each voltage- or current-mode-control method, the main and the
secondary loops interact. Thus, designers need to carefully compensate for stability. With
the peak current-mode control, you should set the crossover frequency of the control loop
one decade lower than the main loop's crossover frequency (Figure
8).
One of the most useful features of an SSPR with an n-channel FET as a
power switch is that it easily implements shutdown and overcurrent protection. An SSPR
controller can incorporate either pulse-by-pulse or average overcurrent protection. In
mag-amps, the saturable reactor must be large enough to block the entire pulse, making it
larger than regulation requires. Hence, the conversion efficiency is lower, and the cost
of the converter is higher.
You can connect an SSPR power switch several ways using both single- and
dual-ended topologies. To directly drive the n-channel FET, place a diode between the
winding and the FET because n-channel FETs have a parasitic body diode. If you need to use
the single-package diode, you need to place the FET in series with the secondary winding
and drive the FET through the gate-drive transformer, which adds cost and complexity to
the power supply.
|
