Disclaimer: I did not make this guide, this is made by TECHPOWERUP.
A Detailed Look Into Power Supplies
Introduction
The abbreviation PSU stands for Power Supply Unit and in the whole article we assume that it is also a Switching Mode Power Supply (SMPS) since in PCs, to the best of our knowledge, only SMPS units are used. A PSU is the heart and a vital part of a system, since it's the one that feeds power to the other components (CPU, VGA, HDD, etc.) If it stops supplying power, for any reason, then nothing else will run since there will be no power to the system. In case of damage to the PSU, there is a possibility that other components may be damaged, too. This is a fact that unfortunately many users ignore, otherwise they would first buy a decent PSU for their systems and then the rest of the components. On the contrary the opposite happens in many cases, users usually acquire all other components and they leave the PSU purchase for last, with the leftover money. If you belong in the category of users that we described above, we are pretty sure that after reading this article you will change tactics. However this article is not addressed only to users that do not know the important role that a PSU plays, but will give valuable information to experienced users, too.
In the beginning of the article, the fundamentals of Switching Power Conversion will be explained and a brief description of the various stages that compose a PSU will be made. In addition, we will make a quick reference to some switching regulator topologies used currently. A few pages dedicated to PSU protections are following and next we take a look at ATX, EPS and 80 PLUS specifications. Finally, in Appendix A we make a quick reference to the most significant electronic components currently used not only in PSU manufacturing but also in every modern electronics device. So you will learn the basic concepts of inductors, capacitors, resistors, transistors and diodes, in order to get familiar with them and understand better some concepts discussed in this article.
As you can imagine it's going to be an interesting and highly informative journey to the PSU world, so join us!
SPC and the Various Stages of a Power Supply
All PSUs that power every PC system today utilize Switching Power Conversion (SPC). The principle of SPC is quite simple, energy is drawn from the power grid, then it's chopped up with a high frequency in smaller energy packets and finally it's transferred with the help of components like capacitors and inductors. In the end all energy packets are merged and after some rectification processes energy is flowing smoothly from the output. So in PSUs we have as input 100–230V AC wall power (AC voltage and frequency differs from region to region) and several outputs that supply regulated DC (Direct Current), which are of course, always the same regardless the country/region. An interesting fact is that as the switching frequency increases the size of the energy packets is getting smaller; thereby the size of components (inductors and capacitors) that store and transfer those packets is also reduced. Finally, a PSU that utilizes SPC is called Switching Mode Power Supply or SMPS.
The two major advantages of an SMPS compared to a Linear Power Supply that uses a totally different design are drastically reduced size and weight and higher efficiency that can easily exceed 90%. On the other hand, the most significant drawbacks of an SMPS are its complexity and the production of Electromagnetic / Radio Frequency Interference (EMI/RFI) that makes necessary the use of an EMI filter (some of you may also know it as transient filtering stage because its role is twofold) and RFI shielding.

The figure above shows the block diagram of an SMPS, there are seven main components that turn AC wall power into several DC voltages used by the components in your computer.
*EMI/Transient Filter: Suppress incoming and out coming EMI/RFI and protects from voltage spikes
*Bridge Rectifier: Rectifies the AC power stream to DC
*APFC: Controls the current supplied to the PSU so that the current waveform is proportional to the mains boltage waveform
*Main Switches: Chop the DC signal to very small energy packets, with high frequency
*Transformer: Isolates primary from secondary side and converts (steps down) the voltage
*Output Rectifiers & Filters: Generate the DC outputs and filter them
*Protection Circuits: Shut down the PSU when something goes wrong
*PWM Controller: Adjusts the duty cycle of the main switches, in order to keep steady output voltage under all loads
*Isolator: Isolates the voltage feedback that comes from the DC outputs and heads to the PWM controller
The part of the SMPS before the power transformer is called “primary” side and the part after it “secondary” side.
In the following pages we will analyze all these individual stages that compose a power supply, in more detail.
EMI/Transient Filtering Stage
The problem with PSUs is that their switching transistors produce EMI/RFI that could seriously affect other electronic devices in the house. Also we must protect the PSU from incoming noise and voltage spikes coming out of the power grid, so the role of this stage is twofold and serves as protection in both directions.
Noise can be classified into two types, according to the conduction mode: Common Mode Noise (CMN) and Differential Mode Noise (DMN).
1. CMN is electrical interference with reference to the ground or common wire. It consists of high frequency spikes and comes from faulty wires or from EMI/RFI of nearby devices. Common mode choke coils along with Y capacitors are used to suppress CMN.
2. DMN represents the noise that is measured between two lines with respect to a common reference point, excluding common-mode noise. To suppress DMN, X capacitors are placed across the lines.
The EMI/Transient filter in PSUs is always placed before the bridge rectifier, because in this position it also suppresses the noise coming from the bridge's diodes (yes, even those produce noise, especially at the moment they are turning off). The necessary parts for a proper EMI/Transient filter are two Y and two X capacitors, two coils, an MOV (Metal Oxide Varistor) and a fuse. Very briefly an MOV is a voltage-dependent resistor that protects the PSU/system from voltage spikes coming from the power grid.
However, especially in low-end PSUs, manufacturers omit some components in order to save money. Usually the first to be left out is the MOV. If your PSU does not have an MOV in the EMI/Transient filter then you should always operate it along with a surge suppressor or a UPS, otherwise a spike could damage permanently not only the PSU but your system too.
Bridge Rectifier

The AC power stream, after it passes the EMI/Transient filter is rectified by one or more bridge rectifiers. So AC is converted to DC with increased voltage (e.g. if we have 230V input then the DC output of the bridge rectifier will be v2 × 230= 325.27V DC). Afterwards the DC signal is fed to the APFC stage.
Active Power Factor Correction (APFC)
Before we talk about the Power Factor Correction stage let's take a quick look at the concept of power factor. Power factor is defined as the ratio of real power to apparent power (kW/kVA) and power is the product of voltage and current (P = V × I).
There are two basic types of loads, resistive (the load consists only of resistors) and reactive (the load consists of inductors, capacitors or both).
In a system with only linear load both current and voltage curves are sinusoidal (the sine wave or sinusoid is a mathematical function that describes a smooth repetitive oscillation @
http://en.wikipedia.org/wiki/Sine_wave ). If the load is purely resistive then both current and voltage reverse their polarity at the same time (the phase angle between voltage and current is 0 degrees), so at every instant the product of voltage and current is positive, meaning that the "direction" of energy flow does not reverse, so only real power is transferred to the load.

In case that the load is purely reactive, there is a time shift (maximum theoretically 90 degrees, typical 45 degrees) between voltage and current so the product of these two for half of each cycle is positive and for the other half is negative (when voltage is at its peak, either positive or negative, current is zero and vice-versa). Thus, on average as much energy flows to the load as flows back to the source (power grid). If we analyze a whole cycle then we will see that there is no net energy flow and only reactive energy flows since there is no net transfer of energy towards the load.

However both scenarios above are only theoretical, as in real life all loads/circuits present resistance, inductance and capacitance at the same time, so both real and reactive power will flow to them. Apparent power is the vector sum of real and reactive power or the product of the root-mean-square of voltage and current. As we already mentioned above, power factor is the ratio between real and apparent power. Ideally we want the power factor to be close to 1. Here we must note that residential consumers pay only for the real power (Watts) they use and not for apparent power. On the contrary business consumers (e.g. factories) pay for apparent power usage, too.
Although we residential consumers do not pay for apparent power, in order to minimize apparent power usage the EU standard EN61000-3-2 states that all switched-mode power supplies with output power of more than 75 W must include passive PFC, at least. In addition 80 PLUS power supply certification requires a power factor of 0.9 or more. Some years ago many PSU manufacturers used Passive PFC (PPFC) in their products. PPFC uses a filter that passes current only at line frequency, 50 or 60 Hz, so the harmonic current is reduced and the non-linear load is transformed to a linear one. Then with the usage of capacitors or inductors the power factor can be brought close to unity. The downfalls of PPFC are the smaller attained power factors compared to APFC and the need for a voltage doubler, for the PSU to be compatible with 115/230V (the proper voltage is manually selected via a switch near the AC input). On the other hand passive PFC has higher efficiency compared to active PFC!

Active PFC is basically an AC/DC converter which controls the current supplied to the PSU via PWM (Pulse Width Modulation). At first the AC voltage is rectified by the bridge rectifier. Then the PWM triggers the APFC MOSFETs (usually two) which separates the intermediate DC voltage into constant pulse sequences. These pulses are smoothened by the smoothing capacitor and are fed to the main switches. Right before the smoothing capacitor we always find an inductor (coil) that has the ability to limit the sudden rise of current without dissipating energy, because it's a reactive component. This coil is necessary because all capacitors that are connected directly to a DC signal show uncontrolled inrush current, so the inductor limits this inrush. In addition sometimes there is also a thermistor used with active PFC to further limit the inrush current especially in the switch-on phase of the PSU, where the smoothing capacitor is fully discharged.
In active PFC two different types of control are mostly used, Discontinuous Conduction Mode (DCM), where the PFC MOSFETs are turned on only when the inductor current has reached zero and Continuous Conduction Mode (CCM) where the MOSFETs are turned on when inductor current is still above zero and therefore all reverse recovery energy is dissipated in the MOSFETs. In the APFC stage of PSUs the second mode (CCM) is mostly used since it is ideal for over 200W output power, because it offers the lowest peak to average current ratio for the converter throughput power. The main drawbacks of CCM are the losses and EMI generation associated with the turn-off of the boost diode (the reverse recovery currents of the diode causes significant power dissipation to the MOSFETs and increased EMI). Because of that we usually see an X capacitor after the bridge rectifier.
Main Switches – Transformer
The main switches operate in two modes only, ON (fully conduction) and OFF (fully non-conduction) and chop the DC signal, coming from the smoothing capacitor, into pulses whose amplitude is the magnitude of the input voltage and whose duty cycle is controlled by a switching regulator controller (Marty Brown – Power Supply Cookbook). Thus the DC signal is converted to an AC rectangular waveform that is fed to the transformer. The latter steps down the voltage which feeds the secondary rectifiers that generate all DC outputs (+12V, 5V, 3.3V, 5VSB, -12V). The transformer also plays the role of an isolator between the primary and the secondary side.

When the switches are ON there is zero voltage across them (theoretically) and when they are OFF we have zero current through them. So always the product V × I is zero. This means that there is no power loss on the switches. However this represents an ideal situation and in real life there are power losses since there is no transistor/MOSFET than can switch instantly. There is always a small period that a transistor (switch) is between ON and OFF state (also called linear region) and during this period the V × I product is not zero. Because of that all MOSFETs in a PSU (and not only) are cooled down by heatsinks and a fan (although there are some PSUs that use only passive cooling).
Output Rectifiers and Filters

The roles of output rectifiers and filters are, as their names state, to rectify and filter the high frequency switching waveform created by the switches (MOSFETs) and fed through the secondary of the main transformer(s). In this stage we meet two types of rectification designs, passive and synchronous. In the first Schottky Barrier Rectifiers (SBRs) are used and in the latter MOSFETs take the place of SBRs. In synchronous rectification efficiency is increased since we get rid of the forward voltage drops of SBRs. To make this crystal clear let's give an example. A typical SBR has 0.5V voltage drop so if we want to conduct 40A then we have 40 × 0.5 = 20W. Now if we use a MOSFET instead of an SBR and assuming that the MOSFET's RDS(on) is 3 mO then we have 40 × 40 × 0.003 = 4.8W. This results in 15.2W less power dissipation and 24% increase in efficiency.
Besides the two above designs sometimes a hybrid one, called semi-synchronous design, may be used. In semi-synchronous MOSFETs and SBRs are used to reduce cost and increase the efficiency above passive design's levels.
The generation of -12V is done with the usage of a conventional diode, since we do not demand much power from this rail (below 1A in most cases). For 5VSB a completely independent circuit with a transformer is usually used, since 5VSB are working continuously even when the PSU is OFF (in standby mode). For the generation/filtering of the main outputs (+12V, 5V and 3.3V) there are three types of regulation. Group regulation, independent regulation and DC-DC conversion. We will analyze each one of them in the below paragraphs.

Group regulation is usually used in low capacity PSUs and budget ones. A fast way to identify group regulation is by checking the number of coils used in the secondary side. If you find only two then group regulation is present. The bigger coil is used for 12V/5V and the smaller one for 3.3V. In this regulation type +12V and 5V are generated together and both of them feed back their output voltage error to the regulator controller. This means that if the load is unbalanced between the rails, then the regulator controller has a really hard time to properly implement regulation (e.g. if the load at +12V is high and at 5V is low then the voltage on the second rail will be raised because the regulator controller tries to raise the +12 rail voltage, but since the latter is connected to 5V then both of them are raised.) This is why most group regulated PSUs fail to keep their rails within +/-5% tolerance at Crossload tests. In group regulation the 3.3V rail is usually regulated by a mag-amp post-regulator from 5V or from 12V (scarcely).

Independent regulation is used in higher capacity and performance PSUs where cost comes in second place. In this type of regulation all main DC outputs have their independent regulation circuit and unbalanced loads do not cause problems to the rails' voltages. The +12V rail is regulated by the main regulator controller and 5V/3.3V by mag-amp post-regulators. You can easily indentify a PSU that uses independent regulation by the number of toroidal coils in its secondary side. If you find three of them (one for each rail) then the PSU uses independent regulation.

Now in many contemporary PSUs the generation of the minor rails is done with the use of buck -step down- converters (DC-DC converters or Voltage Regulation Modules – VRMs). In these PSUs 5V/3.3V are generated from the +12V directly. This has a positive impact on efficiency and in cross-regulation. Here we must note, PSUs that utilize DC-DC conversion are independent regulated too.

A last note before we move to the next stage, the toroidal chokes located after the rectifiers take part not only in the rectification but also in the filtering process, since they are used for ripple, voltage and current, reduction on the DC outputs. However in PSUs that utilize LLC resonant topologies usually there are no toroidal chokes in the secondary (for +12V generation) and if there is any then it is used only for filtering.
PWM Controller - Isolator

The main purpose of the PWM controller is to maintain a regulated output voltage and control the amount of energy being delivered to the load (system). The aforementioned are accomplished by adjusting the duty cycle of the main switches. The duty cycle can be adjusted from 0 to 100 percent but usually its range is smaller. With great approximation we could say that output voltage is the product of input voltage and duty cycle (Vout = Vin × duty cycle).
The PWM controller uses a voltage reference as the PSUs “ideal” reference to which the output voltage is constantly compared. In the PWM IC there is a voltage error amplifier that performs a high gain voltage comparison between the output voltage and the above-mentioned reference. According to this comparison an error voltage-to-pulse width converter sets the duty cycle in response to the level of the error voltage from the voltage error amplifier. Besides determining the duty cycle of the main switches the PWM controllers usually incorporate and other functionalities, as soft-start circuit which starts the PSU smoothly reducing large inrush currents, over-current amplifier that protects the PSU from overloading, undervoltage lockout that prevents the PSU starting when the voltage within the control IC is too small to drive the main switches etc.

In order for the voltage feedback, from the DC outputs, to reach the voltage error amplifier of the PWM IC an isolated feedback is needed. There are two methods of electrical isolation, optical (optoisolator) and magnetic (transformer). In modern PSUs optoisolators are commonly used. The voltage error amplifier is placed on the secondary side of the optoisolator.
Switching Regulator Topologies used in Contemporary PSUs
According to the peak current that will pass from the main switches, the desired efficiency levels, the maximum operating voltage across the switches the cost etc. a manufacturer has several switching regulator topologies at his disposal.
below you will find a table that compares several switching regulator topologies (Marty Brown: Power Supply Cookbook)
Lately, many PSUs use the LLC resonant topology. This topology, utilizing a resonant combination of inductors and capacitors, shapes the voltage and current waveforms in switching MOSFETs allowing for soft (zero voltage) switching which by its turn leads to RFI and EMI reduction and minimized switching losses, so we gain increased efficiency. Utilizing LLC resonant converters higher switching speeds along with efficiencies of 93 – 95% can be achieved.
Finally, here you will find a very informative pdf file that describes many commonly used topologies @
http://www.techpowerup.com/articles/...topologies.pdf