How to efficiently power LED luminaires

How to efficiently power LED luminaires
Workarounds needed to fulfil the requirements of electromagnetic compatibility can
mar the reliability and efficiency of power circuits in LED luminaires.
By Arnoldas Bagdonas
Field Application Engineer
Future Electronics
Compromises and workarounds undertaken in an effort to meet the requirements of electromagnetic compatibility
(EMC) have had the potential in the past to markedly harm the reliability and efficiency, as well as increase the total
production cost, of the power circuit in an LED luminaire.
A typical example of a power circuit for an LED luminaire using conventional power components is shown in figure
1. Excessive radiated emissions will normally lead the design team to shield the entire housing. In practice,
however, this increases the parasitic capacity between the (now larger) conductive area – that is, the chassis plus its
shield – and the reference ground of anundery EMC measurement equipment. Common-mode conducted
interference then becomes a large enough phenomenon to require attention. While a low-cost EMI filter will
eliminate this problem, this author has seen designs in which even these counter-measures are not sufficient, since
higher-frequency emissions radiated by the mains cable persist, and must be blocked by a more expensive filter
with higher attenuation.
The root cause of the problem in LED luminaires is the power supply’s high-speed switching circuits, which create
wide spectrum current and/or voltage ripples. Shielding and filtering might mitigate the emission problem, but do
not eradicate it. A better solution would be to avoid generating high emissions at particular frequencies in the first
place – and this is now possible through the use of new power components that use soft switching to minimise
ripple currents, or to spread the noise energy over a wide frequency band.
Figure 1: a typical AC/DC LED driver design. The H-field is the result of winding leakage, the primary loop area and the
secondary loop area. The E-field is the result of high dV/dt on conductive surfaces and of high-frequency ripple in
Figure 2 shows that there are broadly five architectures used today, each suited to different power outputs. Each of
these LED driver topologies enables the designer to comply with the strict requirements of today’s EMC regulations.
While figure 2 indicates the power range in which each topology is most commonly used, it should be noted that
each can be adapted for use in a higher or lower power output range.
EE Times-India |
Copyright © 2012 eMedia Asia Ltd.
Page 1 of 4
Figure 2: The power range in which each topology is most commonly used.
The Power Factor Controller (PFC) is the most common block in modern AC/DC LED drivers. A boost converter is
inserted between the bridge rectifier and the main input capacitors. This regulator can operate in three modes. In
Discontinuous-Conduction Mode (DCM), the energy stored in the inductor (L) during the conduction interval of the
switch is equal only to the energy required by the load for one switching cycle (figure 3). The energy in the inductor
drops to zero before the end of each switching cycle, resulting in a period of no energy flow, or discontinuous
operation. In Transition Mode (TM) – also called Boundary Conduction Mode (BCM) or Critical conduction Mode
(CRM), the converter operates at the boundary between DCM and Continuous Conduction Mode (CCM), reducing
the idle time of DCM to close to zero.
Figure 3: Peak and average current in the inductor (IL) in a) discontinuous
conduction mode b) transition conduction mode and c) continuous
conduction mode.
In CCM, the inductor has continuous current during the operation of the converter. The extra energy stored in the
inductor during the conduction time of the switch is equal to the energy discharged into the output during the nonconductive time of the switch; at the end of the discharge interval, residual energy remains in the inductor.
During the next conduction interval of the switch, energy builds from that residual level to that required by the load
for the next switching cycle. CCM has a lower peak-to-average current ratio; thus it has lower ripple currents, lower
coil conduction and core losses, and lower electromagnetic emission levels. The drawbacks are that it requires a
very fast boost diode (otherwise diode recovery current starts to dominate, resulting in increased power losses and
additional electromagnetic emissions). Unfortunately, it also requires hard MOSFET switching, and this results in
high switching losses, and these are the main source of electromagnetic emissions in a CCM system. The biggest
advantage of TM or DCM operation is the absence of reverse recovery in the boost diode, which means that the
circuit can use a low-cost diode with a low forward voltage. On the other hand, the cost of filters to block the
EE Times-India |
Copyright © 2012 eMedia Asia Ltd.
Page 2 of 4
electromagnetic emissions generated at the high peak currents might be excessive.
New components in the second DC/DC stage of the LED driver (figure 4) also offer new ways to reduce
electromagnetic emissions. They often contain ‘resonant’ LC networks, of which the voltage and current waveforms
are sinusoids. The turn-on or turn-off transitions of semiconductor devices can occur at zero crossings of the tank
voltage or current waveforms, thereby reducing or eliminating some of the switching loss. This means that resonant
converters can operate at higher switching frequencies than comparable PWM converters, leading to smaller
inductor and capacitor values and costs. In addition, zero-voltage switching reduces converter-generated EMI as it
has no current or voltage ripples during switch commutation.
Figure 4: New components in the second DC/DC stage of the LED driver also offer new ways to reduce electromagnetic
Figure 5: operation of a quasi-resonant, zero-voltage switching circuit.
Resonant converters do, though, have several disadvantages. Performance can be optimised at a single operating
point, but not across a wide range of input-voltage and load-power variations. Also, the quasi-sinusoidal waveforms
found in a resonant converter exhibit higher peak values than their equivalent rectangular waveforms. In addition,
current can circulate through the tank elements even when the load is disconnected, leading to poor efficiency at
light load. A similar switching technology is today often employed in low-power LED drivers in which a quasiresonant or valley-switching topology is implemented. To start, current IQ1 ramps up until the desired energy level
is charged in to coil L (figure 5). Then switch Q1 is turned off. When the switching transient is complete and the coil
current equals zero, the drain voltage starts to oscillate around the input voltage level VDC. The amplitude equals
V0. Circuitry connected to the Q1 drain pin senses when the voltage on the drain of the switch has reached its
lowest value. The next cycle is then started. The effect of this topology is to reduce capacitive switching losses and
EE Times-India |
Copyright © 2012 eMedia Asia Ltd.
Page 3 of 4
electromagnetic emissions. On the other hand, a quasi-resonant converter has the same disadvantages as a resonant
converter, as described above.
EMC standards for conducted emissions generally set peak energy limits within the frequency band from 150kHz to
30MHz. Although carefully selected fixed-frequency modulation can be effective in spreading harmonic content, the
disadvantage is that in some cases it might not provide sufficient attenuation of the fundamental. New research,
however, suggests that modulating at a fixed frequency is not as effective in reducing the peak energy in the
fundamental as modulating the carrier with a complex, random, or pseudo-random waveform. Perhaps the most
important lesson to learn from experience is the importance of testing prototypes for EMC, before going to preproduction. About the author
Arnoldas Bagdonas is Field Application Engineer at Future Electronics (Lithuania).
EE Times-India |
Copyright © 2012 eMedia Asia Ltd.
Page 4 of 4