Understanding power supplies – Super Nade
In this article we will take a brief look at SMPS
terminology and a few basic design paradigms used in modern
PC power supply (Power supply Unit, PSU) designs. In part one
of this five part series, we look at the basics of the input features,
up to the inrush current limiting stage. The purpose of this article
is to demystify industry terminology and construction.
The PC Power supply is one of the most important com-
ponents in a well built PC. Consequently, this component
has been accorded a rather important status by computer
enthusiasts. In this day of power hungry parts such as video
cards and multi-core processors, it is all the more important
to have a good quality PSU.
Before making a purchase, it
would well serve an enthusiast to understand how a typical
PSU works and what are the things one should look for in
a good quality unit. In this five part series, we intend to
look at the features implemented in modern PSU’s. In this
article, we look at the input side of the PSU.
In Part Two, we
will look at Soft-Start, start up Over/Under voltage protection,
overcurrent protection. In part three, we will look at output
current limiting, output filtering/isolation/noise and snubber
networks. In part four, we will look at PFC and practical
identification and characteristics of the critical components in
use. In part five we will completely take apart and analyze a
PSU.
This is going to be a fun project so please hang in there
till we get to part five!
Quite often one can find words such as “robust unit”, “build
quality” and “well populated” being used to describe the
modern computer power supply. As computer enthusiasts have
come to realize, these terms very much reflect the critical
nature of the PC Power supply in the context of powering
an overclocked rig. This section will discuss, in brief, the
various input requirements specified by international standards
and their importance in modern PC Power supply design.
A. Input Transient Voltage Protection and Electromagnetic
Compatibility
A transient is a short lived electrical phenomena whose time
scale depends on the context.
In case of PSU’s, transients
can be either voltage or current spikes. IEEE Standard
587-1980 classifies various stress locations and investigates
this phenomena in some detail. Frequencies above 30 MHz
tend to radiate directly from the generating circuits, while
those below 30 MHz are usually conducted by the AC line
and other connections.
The IEEE standard recommends a 6
kV peak amplitude for a damped sinusoidal oscillation as
shown in the figure. PSU’s must filter out these high voltage
transients.
The ideal transient suppression device would be an open
circuit at normal voltages and would conduct without delay at a
certain voltage beyond spec and clamp the voltage at some
specified level. It should also be able to handle unlimited
surges or spikes and would not break down.
However, no
such ideal suppression device exists.
PSU manufacturers
try to approach an ideal setting by using a combination of
various suppression devices. We will look at a few of them
in brief:
Metal-Oxide Varistors (MOV’s): MOV’s are voltage de-
pendent resistors. At voltages below a certain level (called
turnover voltage), they have very high resistance and almost
act as an open circuit.
When the voltage ac cross it increases
beyond this turnover voltage, the resistance rapidly decreases so it can allow the surge current to be taken away from the
rest of the circuitry. Their clamping action as required by the
IEEE standard is very poor for high current transients. These
devices age quickly and such stress related degradation cannot
be easily measured or detected by visual inspection. They are
usually used in conjunction with other suppression devices.
Transient Protection Diodes (TPD’s): These are usually silicon based
diodes which have very good voltage clamping action. They
are expensive and are seldom used. When these diodes fail,
they fail as a short circuit, which usually means the PSU’s
fuse will blow.
Transient Suppression Capacitors: IEEE-587 identifies two
types of transients: Transients can either occur between live
and neutral lines or between live/neutral and ground.
Suppression capacitors are designated as Type-X and Type-Y
respectively. These transients are also called symmetric and
asymmetric transients. While X capacitors can be of any value,
Y capacitors are kept small and are usually limited to 4700
pF. Typical X capacitor values are 0.1 to 1.0 µF. In the
US, UL standards are most widely employed and capacitors
should adhere to these standards (refer to the Appendix for more
information).
Transient suppression Inductors: To suppress EMI, a typical
filter will include common mode inductors, differential mode
inductors and X and Y capacitors. The Y capacitors and the
common mode inductors contribute to the attenuation of the
common mode noise.
The inductors become high impedances
to the high frequency noise and either reflect or absorb the
noise, while the capacitors become low impedance paths to
ground and redirect the noise away from the main line. To be
effective, the common mode inductor must provide the proper
impedance over the switching frequency range.
Common mode inductors are wound with two windings of
equal numbers of turns. The windings are placed on the core
so that the line currents in each winding create fluxes that are
equal in magnitude but opposite in phase. These two fluxes
cancel each other, leaving the core in an unbiased state.
The differential mode inductor has only one winding requiring the core to support the entire line current without
saturating. Herein lies the great difference between common
mode and differential mode inductors. To prevent saturation,
the differential mode inductor must be made with a core that
has a low effective permeability (gapped ferrites or powder
cores). The common mode inductor, however, can use a high
permeability material and obtain a very high inductance on a
relatively small core.
Faraday Screens: Faraday Screens are copper screens used
to eliminate parasitic capacitive coupling between components
and RF coupling. These are seldom used in PSU’s because of cost and because the case provides adequate shielding.
Hence, we will not discuss this any further.
{mospagebreak}
B. Input Fuse Selection
The fuse (fusible wire link) is one of the oldest available
protection paradigms. In the most elementary case, it consists of
a special wire which melts when excess current flows through
it, thereby protecting the device it was designed for. There are
several parameters one must look out for while replacing a
fuse.
Current Rating: The rated current must exceed the maxi-
mum DC RMS (root mean square) current demanded by the
circuit. The tolerance depends on the requirements.
Voltage Rating: This is not necessarily related to the supply
voltage but generally is a measure of its ability to extinguish
the arc that is generated when the fuse blows. Failure to select
a proper voltage rating will mean a lot of transient energy is
let through by arcing and may be a fire hazard.
Slow-Blow Fuse: These are low cost, low-melting point
alloys and the fuse element is very thick. They are used in
devices where the inrush current is large upon start up, like
motors and alternators.
Standard-Blow Fuse: These are usually made of copper in clear glass
enclosures and can handle short-term high current transients;
selected for short circuit protection in most cases.
Very Fast Acting Fuses: These are used to protect semi-
conductor components from spikes and they blow very very
quickly. The response time is in the order of micro-seconds.
SCR (Silicon controlled Rectifier) Crowbar Protection: This
scheme is as shown below. It is a bit expensive to implement
due to the number of components being used. These days, one
can find integrated elements which do the job.
This is typified by the use of a large electrolytic reservoir
capacitor. Either a single or multiple storage capacitors may be
used. Manufacturers like Zippy and Seasonic take the single
capacitor approach by picking a low ESR 105 C rated Hitachi
for the primary energy bank.
A simple criterion for selecting
the input capacitor would be
C = 1.5 µF/W
where C is the
effective capacitance of the entire input capacitor bank. There
are other important considerations one must look at such as:
RMS Ripple Current Rating: This is actually tied in with the
amount of energy lost as heat and depends on the ESR rating of
the capacitor. The larger the ESR (Effective Series Resistance),
the more energy is dissipated as heat.
Capacitors with larger
ESR are prone to heat related failure, as in most cases ESR
dramatically increases with rising ambient temperature. If you
every wondered why your PSU blew its caps out, this is the
reason.
Ripple Voltage: This requirement defines the minimum
capacitor value during short timescales. The output from the
rectifier assembly needs to be be smoothed out and this is
accomplished by a capacitor.
Undersized caps are frequently
used when economy overrides design prudence and this man-
ifests as undesirable voltage ripple. Quite often, the following
stage is a voltage regulator based on DC-DC conversion. This
is dependent on having a clean DC input in the first place.
If the input has excessive ripple, the PWM assembly will be
under unnecessary stress.
Holdup Time: It is the minimum time for which the supply
will maintain the output voltage within design limits when the
input supply is removed or falls below the input regulation
limits. Holdup time is a very important parameter and plays
a big role in the choice of the input capacitor and is given as:
where Ei is the energy used during holdup time and the
voltages are supply and failure voltages respectively.
{mospagebreak}
In AC/DC power converters above a few watts, a large
inrush current flows when the input capacitors are suddenly
charged during the initial application of power. If unrestricted,
this current can easily exceed 50 A at the peak of the AC
cycle.
This high inrush current severely stresses the converter’s
fuse, input rectifiers and power switch. It can significantly
reduce the reliability and life expectancy of the modules.
Most PSU’s use some form of bridge rectification to convert
the AC from the mains to DC (with input capacitors).
This is
usually in the form of a diode arrangement or an integrated
bridge rectifier assembly. If the line input is switched to
this type of rectifier-capacitor assembly, very large current is
drawn by all components in a short timescale. Hence, there
is a need for “inrush control” and is usually some form of a
series limiting resistive device.
A few techniques used for this purpose are:
Using Resistors: For very small power supplies, a few watts
at most, adding a resistor in series with the line is a simple
and practical solution to limit the inrush current. The large
resistance required to limit peak inrush current causes too
great a loss in efficiency to be used in higher wattage power
supplies. It is sufficient to use resistors immediately after
the surge suppression filter section and between the bridge
rectifier/capacitor stages.
Using NTC Resistors/Thermistors in series with line:
Many power supply manufacturers use a negative temperature
coefficient (NTC) resistor in series with the line. An NTC
resistor offers tens of ohms of resistance when cool, dropping
to less than one ohm as its temperature increases.
If the
power supply is cool when turned on, the NTC provides good
inrush current limiting. Its effect on efficiency is reduced as
the power supply warms up. However, this approach is not
effective over large temperature extremes. A power supply
used outdoors in the northern winter may never warm up
enough for the NTC resistance to drop. Conversely, a supply in
the hot summer sun will be very warm even with the power off,
so that the warm NTC resistor will fail to provide adequate
inrush current protection on start up.
An NTC resistor can
also be problematic when a user turns the system off and then
immediately switches it back on again. The capacitor voltage
may drop, but the NTC resistor will not cool quickly enough
to provide inrush current protection. That is why it is bad to
switch on and off the PSU rapidly.
Using an Active Limiting device: For a high powered PC
power-supply (anything available today, really), it is more
energy efficient to have the limiting device shorted out of
the circuit, unless needed. This will reduce operational losses
when the unit is in steady state.
A TRIAC is usually the
active device of choice in such an implementation. The SCR
or TRIAC approach limits the inrush current by progressively
varying the phase of the AC line voltage at which the
SCR/TRIAC is switched on during start up. The instantaneous
line voltage at which the SCR is activated is incrementally
higher at each subsequent cycle ensuring that the difference
between the line voltage and the output bulk capacitor voltage
is small enough to result in negligible inrush current.
This
kind of circuit is considered to be non-power dissipating.
However, the triggering circuit itself consumes power and
the SCR has to be fed with a continuous pulse to prevent
it switching off due to line glitches The power needed for this
circuit is therefore not insignificant.
These methods are “one shot” which means that (unless the input power is cycled on/off
or additional, complex circuitry is added) they cannot limit
inrush due to subsequent disturbances on the power supply
line after initial power on.
{mospagebreak}
We have seen in brief, a few important characteristics and
requirements of modern PSU’s. We will be looking at more
characteristics in the next part of this article.
To be continued…
Auf Wiedersehen!
IV. REFERENCES
1) SMPS Handbook by Keith Billings
2) Tech notes, Bear Power supplies
3) Meeting Military Requirements for EMI & Transient
Voltage Spike Suppression, xp-military.com
4) Common Mode Inductors for EMI Filters Require Care-
ful Attention to Core Material Selection by Robert West,
Magnetics, Division of Spang & Co., Butler, Pennsyl-
vania
5) Capacitors for RFI Supression of the AC line: Basic
Facts,300 Tri-State International, Su. 375 Lincolnshire,
IL 60069 847 948 9511 Copyright c 1996 Evox-Rifa,
Inc
6) Ripple Current Capabilities 2004 KEMET Technical
Update by John Prymak Applications Manager, P.O. Box
5928 Greenville, SC 29606 Phone (864) 963-6300 Fax
(864) 963-66521 WNW.kemet.com.
7) A new approach to PFC Inrush protection by Linus Liu,
Copyright c 2006 Astec International Limited
APPENDIX
Underwriters Laboratory (Trivia)
The United States actually does not have a national safety
agency. Underwriters Laboratories (UL) is a private corpo-
ration. Various federal and state laws require that electronic
products be listed with any Nationally Recognized Testing
Lab (NRTL) of which UL is the most widely known. NRTLs
typically test end products to UL standards (plus those of
other countries) and very few test primary components such
as capacitors.
UL was established over 100 years ago by a group of
insurance companies to promote product safety as a means
of reducing insurance claims. Accordingly UL’s focus has
generally been on traditional consumer products – this is the
case even today.
An RFI capacitor in an electronic ballast has less stringent
requirements. The same was true of RFI capacitors in switched
mode power supplies until UL adopted the IEC950 standard.
At the present time there are two UL standards related to
RFI capacitors: UL1414 and UL1283. UL1283 is actually a
standard for potted RFI filters.
The only meaningful reason to have UL1283 recognition for
a capacitor is to demonstrate the capacitor’s ability to survive
the tests that the filter must undergo. UL1283 recognition of a
capacitor is not required by any UL equipment standard. The
requirements are not very stringent, consisting primarily of a
dielectric withstand test.
UL1414 is specifically required for
television and radio receivers and certain telecommunications
equipment. The tests are quite stringent. A capacitor may
be rated for either 125 or 250 VAC (nominal) at 85ºC. The
requirements of UL1414 at 250 VAC are summarized:
- A 1500 VAC dielectric withstand test for 1 minute.
- Be subjected to 50 discharges from a cap charged to 10
kV through a 1000 W resistor, then pass a 1 kV dielectric
withstand test. - A 1008 H endurance test at +85ºC with an applied voltage
of 440 VAC 60 Hz. Once per hour, for 0.1 s, the voltage
is raised to 880 VAC. - A passive flammability test.
- An active flammability/expulsion test.
Material Choices for Surge Supression Inductors:
For the most part, ferrites are the material of choice for
common mode inductors and they are divided into two groups
viz., Nickel-Zinc and Manganese-Zinc.
Nickel Zinc materials
are characterized by low initial permeability’s (< 1000µ), but
they maintain their permeability's at very high frequencies ( >
100 MHz). Manganese Zinc materials, on the other hand, can
attain permeability’s in excess of 15,000 µ but may start to
“roll-off” at frequencies as low as 20 kHz.
Because of their low
initial permeability’s, nickel Zinc materials will not produce a
high impedance at low frequencies. They are most often used
when the majority of unwanted noise is greater than 10 or
20 MHz. Manganese Zinc materials, however, offer very high
permeabilties at low frequencies and are very well suited to
EMI suppression in the 10 kHz through 50 MHz range.
High permeability ferrites come in many different shapes
like Toroids, E cores, Pot Cores, RMs, EPs, etc; but for the
most part, common mode filters are wound on toroid’s. There
are two main reasons for using toroids:
First, toroids are
generally less expensive than the other shapes because they
are one piece, whereas other shapes require two halves. When
cores come in two halves, they must be flat ground on their
mating surfaces to make them smooth and to minimize the
air gap between them. Furthermore, high permeability cores
often require an additional lapping procedure to make them
even smoother (this produces a mirror-like finish).
Toroids require none of these extra manufacturing steps.
Second, toroids have the highest effective permeability of any
core shape. The two-piece construction of the other shapes
introduces an air gap between the halves, which lowers the
effective permeability of the set (typically by about 30%).
Lapping improves this but does not eliminate it. Because
toroids are made as one piece, they do not have an air gap
and do not suffer a reduction in effective permeability.
Toroids do have one disadvantage, their high winding cost.
Bobbins, which are available for the other shapes, can be
wound quickly and economically. Toroids require special wind-
ing machines or must be wound by hand, making the per-piece
winding cost higher. Fortunately, the number of turns on com-
mon mode inductors is usually quite low, so the winding costs
do not become too prohibitive.
For these reasons, toroids are
the geometry of choice in common mode inductors. There are
several other material characteristics and design considerations
one must take into account while selecting inductors, but that
would be beyond the scope of this article.
Be the first to comment