How a Switchmode Power Supply Works

[LongRunner]

Discuss and Comment on "How a Switchmode Power Supply Works" by c_hegge...
http://hardwareinsights.com/how-a-sw ... ply-works/

[duckula]

Very informative article.

[Galib]

Hi,
I am Galib from India . The Post by C_HEGGE is really excellent . I am highly appreciated with this Post . This Post is really very helpful for those who works with ATX PSU .

I have some Computer Power Supply - Which all are Made by CHINA and look like as "NSCom" design and they have the Same Problem . The Problem of those Power supply is that - All The Parts of PSU is OK but it does not Start . I am getting 5VSB and 5V PS-ON on all PSU . I have checked Feedback Transformer, Primary Transistor 2SC13007, Main Transformer, Output Rectifier Diodes, Output Capacitor and everything in Output Section - all the parts are OK . But still the PSU does not start . Why ?? Even I have changed the PWM IC TL494 and Comparator/Power Good IC LM339 with new one but still the Power supply does not start .

I am totally confuse . I spend a lot of time to repair them but all the time failure is the result . I do not understand why these power supply are not starting What's the problem in there as they are not running ??

Guys can you help me to figure out the problem and repair this power supply .... Please give me a favour ........

[c_hegge]

Does the fan twitch when you try to power it on? If so, it sounds like one of the secondary rectifiers is shorted.

[Galib]

Does the fan twitch when you try to power it on?

Thanks Hegge for your Reply . I could remember, Few years ago I got a power Supply which twitch the Fan once and shut down . I failed to repair that because I was new at PSU repairing and was not so much experience at that time . Currently I have more than 20 Power Supply at non-working condition . None of them twitch the fan for a single moment . I have changed their Output Rectifier Diode and even all the Capacitor but it does not listen to me in any how .

Then still where is the problem ?? Of course there is a problem which I can not figure out . Hegge have you got such kind of power supply which have all parts virtually OK but does not run ?? What's Problem you find in those power supply ??

By the way - How do you test the Output Rectifier Diode ?? I use a analogue Multimeter . I keep the Red Probe to Rectifier's Middle Pin and Black Probe to other two side Pin . In that condition - I can see that 12V Rectifier shows almost 30 ohms . But 5V and 3.3V Rectifier are almost sort - they show only 5 to 7 ohms .

And againg in reversely, Black probe in the Middle pin and Red Probe to other two side pin and this time every Rectifier Diode show high ohms . For 12V Rectifier Meter does not move and for 5V Rectifier meter moves to 300K and for 3.3V Rectifier meter moves to 200K .

And Also When Two probe to Side Pin ( not middle Pin ) and Meter in X10K Positoin - For 12V Rectifier Meter does not move and for 5V Rectifier meter moves to 300K and for 3.3V Rectifier meter moves to 200K .

Is my these Diode testing Procedure is correct ?? If I wrong then what is the correct procedure ??

[shovenose]

This might seem stupid but have you checked the fuse?

[Galib]

This might seem stupid but have you checked the fuse?

Thanks Shovenose for your reply . Hahahaha ....... These Power supply really made me frogy angel .... hahahaha . I said it before - all the Power supply are showing 5VSB, So Checking of fuse is totally out of question .

In the Control Section the TL494 IC's Output Pin 8 and 11 is also giving 2V and the Feedback Transformer's 3 Pin at controller side is also showing 2V, Which we generally found in a good and workable Power Supply at Controller unit . But still it is not switching - Why ........... ???

[c_hegge]

Could we have a picture of the PSU's internals?

Another possibility is bad start capacitors. They are usually located near the switching transistors and are usually very small with no vent stamp.

[LongRunner]

It's been over two years and while it's a decent start, this article has plenty of room for improvement. I've made some example texts myself (placed in the unattributed "quote" boxes) which you can use if you want. You may want to take some pictures of a more modern (and maybe higher-quality) unit than those shown (e.g. one with active PFC, double-transistor forward, an IC-based +5VSB and maybe high-quality capacitors) for examples of differing designs.

Taking on each section...

Introduction
I know an in-depth look at linear PSUs would be a waste of space, but could we at least provide a brief description of them??? Something like this:

The role of a power supply is convert the mains voltage into the lower DC voltages required by connected equipment. There are two types of PSUs out there - linear and switching/switch-mode.

Linear PSUs (which came first) use a large iron-core transformer, connected directly to the mains, to change the voltage and provide electrical isolation. Its output is still AC, however, so to change it into DC, diodes (which allow current to flow only in one direction, turning the AC into pulsing DC) along with hefty filtering capacitors (which smooth out the voltage pulsations, known as ripple, by storing and releasing energy) are used. If the load is very sensitive to voltage variations and ripple, a regulator (which drops the ingoing DC to the constant voltage required by the load) is added. The big (literally) disadvantage of linear PSUs is that the low frequency (50 or 60Hz) requires a massive transformer and capacitors, and especially small regulated-output units tend to be very inefficient. For these reasons, it is not practical to use them in PCs.

Switching supplies work very differently, though they do share some of the same basic parts.

Note that the primary-side bridge and capacitors are essentially the same as the bridge and capacitors in a linear PSU, though rated for mains-voltage operation.

Input filtering
The PSU itself wouldn't be affected much by RF from the grid, which self-generated RF would overwhelm anyway. The second role stated - keeping noise generated by the PSU from going back into the grid - is the important one.

There should be a description of how the NTC works, such as:

The NTC thermistor limits the initial charging current of the primary capacitors (mentioned below). It is essentially a resistor whose value drops when it heats up under load, reducing wasted power compared to using a conventional resistor with the same "cold" resistance. Some high-end PSUs use a relay (an electrically operated switch) to short it out when the main supply is active, providing further savings.

As far as I know, X capacitors more often use polypropylene dielectric than polyester. (X1, X2, Y1, and Y2 are just safety ratings and don't require any specific dielectric. Plastic film X caps and ceramic Y caps are just most common. Also, if the PSU manufacturer cheaps out and uses non-safety-rated caps, then they aren't actually X or Y class.) And there is (or at least should be) a high-value resistor across mains live and neutral to discharge the X capacitors after the unit is unplugged. The value of that resistor is a compromise between the discharge time constant and wasted power when plugged in, but there is a device called CAPZero that can work around that problem.

Also, there is usually an additional Y capacitor, not mentioned in the article, between one side of the high-voltage DC supply and earth. (There is no definite rule about whether it's on the positive or negative side.)

The bridge rectifier
I think this is due for quite a re-write...

This is very similar to the rectifier in a linear power supply, albeit rated for a much higher voltage. Its effect on the PSU's efficiency is minimal, but it does determine how much current the PSU can draw (too much and the bridge will overheat and short out). Usually, a single part with the required four diodes integrated into one four-leaded package (2 for the AC input and 1 each for DC positive and negative), saving space on the printed circuit board, is used. Some cheaper power supplies (like the NSCom unit pictured earlier) use four discrete diodes instead of a single-piece rectifier. While functionally the same, these generally have much lower ratings, 2A and 3A being common.

To get an idea of how much it can handle, take its nominal rating, multiply it by the mains voltage and multiply that by the unit's efficiency (typically around 75% for a low-end unit on 120V input), and for units without active PFC (see below), de-rate by 20%. [source: any number of datasheets] So for example, if the rectifier is rated at 4A, the mains is 120V, the unit has an efficiency of 75% and lacks PFC, 288W is what you can, theoretically, pull (if the other components can take it).

It may or may not have a heatsink attached; those designed to be mounted to a heatsink (such as GBU and GBJ series) usually require one to deliver their label rating.

The primary capacitors
Add a better explanation of how a voltage doubler works, complete with a diagram (preferably something that a layperson could understand i.e. drawings of actual parts with graphical "lines" joining them). For example:

Like the capacitors on the secondary of a linear PSU, these smooth the pulsed DC coming from the rectifier into a continuous DC for the primary switchers to work with. Power supplies with active PFC have a single 400-450V capacitor (or sometimes two in parallel, but this makes no functional difference), which holds the boosted output from the PFC circuit (see below). Power supplies without active PFC usually have two 200V capacitors arranged as a switchable voltage doubler - see the diagram.
[insert diagram here]
When set for 230V, it functions as a normal bridge (the resistors are to ensure the voltage is evenly split between the caps). When set for 115V, the switch connects one side of the mains to the center point of the two capacitors. Therefore, the two caps are alternately charged, one on the positive half-cycle of the mains and one on the negative half-cycle. The total voltage across both caps is about the same either way.

These capacitors are generally relatively unstressed, so even those from low quality manufacturers don't fail that often.

Include, in the diagram, the rectifier, voltage selector and both of the capacitors, along with the voltage equalisation resistors.

An example of a PFC section (well overdue):

Power factor correction

Not all AC loads are easy on the power grid. Pure resistances - incandescent lightbulbs and heating elements - no problem; those don't put any more stress on the wires and transformers than is necessary to do their job. But other loads aren't so graceful, heating wiring more than their actual power consumption would suggest. The discrepancy is known as power factor, a number equivalent to the actual power consumed divided by the apparent power. Large energy consumers are billed for power factor for exactly that reason. There's more to it than fits into the space here [yes, I tried] - if you're interested, this is a highly informative article about it.

Power supplies with no PFC generally have a PF of up to 0.65. Inserting a large inductor (usually with a film capacitor in parallel) before the primary capacitors (whether in the AC or DC side of the bridge is the designer's choice, but either way has advantages and disadvantages) forms passive PFC. While better than no PFC, the improvement is very limited due to size constraints. The PF is increased to maybe 0.75.

Active PFC circuits, together with correcting the power factor, boost the voltage going into the primary capacitor. Between the bridge and the primary capacitor, they have: A film capacitor (which filters the high-frequency pulses drawn by the converter), a large coil (which in conjunction with the MOSFET and diode, provides the actual voltage boosting), a MOSFET switcher (like those used to drive the transformers), and a high-speed power diode (which rectifies the high-frequency pulses from the coil). The MOSFET and diode require a heatsink, which may be shared with the primary switchers (see below) or may be separate. They are far more effective than passive PFC, easily achieving a power factor of over 0.95.

I don't think it's totally accurate to say that two-transistor self-oscillators are more failure prone - just that when they fail, it results in the catastrophic over-voltage. They are crude circuits, though, usually with no protections all, making them easy to destroy when overloaded or short-circuited. Here's how I'd write the section on switchers:

These supply pulsed current to the transformer at a much higher frequency than that of the grid (usually from 50kHz on up), which allows the transformer to be far smaller than that in a linear power supply. The duty cycle is adjusted on-the-fly to compensate for varying input voltage, changing loads, and the ripple remaining on the primary capacitors. They produce enough heat to require a heatsink to maintain a safe operating temperature. There are two types of switchers used - bipolar junction transistors and MOSFETs. MOSFETs are preferable as they switch much faster, providing higher efficiency, but are a bit more expensive. Usually, BJTs are used in the old half-bridge topology and MOSFETs in more modern designs such as forward and LLC resonant. When overstressed, as often happens when dodgy power supplies are run at their rating, they tend to fail explosively.

The standby supply may use either a two-transistor self-oscillator or a switching IC. The two transistors in question are the switcher (which may be either a BJT or MOSFET) and a smaller transistor nearby. Those are crude circuits that typically have no protections, and often contain an electrolytic capacitor which, when it fails, causes catastrophic overvoltage on the output. The Bestec ATX-250-12E was notorious for failing that way, with up to 18V being reported. Modern switching ICs, such as DM311, TinySwitch, TOPSwitch and VIPer22A, are much better protected from abuse of the supply and do not result in overvoltage even if the supporting capacitors do fail.

There is a recently released set of devices providing active PFC, LLC resonant, or two-switch forward + flyback standby supply, each on a single chip (although, presently, they are only good for smaller PC PSUs). It's your decision whether those are worth mentioning.

And yes, I do have more to say than this. But I'll save it for later when I'm ready.

[LongRunner]

Okay, part two...

The transformers

These change the voltage and provide electrical isolation between the primary and secondary of the PSU. They consist of coils of wire wound around a plastic former which is fitted with a ferrite core. Layers of tape provide safety insulation.

There is generally one large transformer, for the main supply, and one or two smaller transformers. One of the small transformers is used for the standby supply. The other is used in the old half-bridge topology to drive the switching transistors. There are also one or more optocouplers, which resemble integrated circuits but only have four pins, that transfer small signals across the isolation barrier. The two-transistor forward topology also has a tiny transformer (contained entirely on the primary side and not located near these) driving the high-side MOSFET.

The secondary rectifiers

These work in a similar way to the bridge rectifier on the primary side, only allowing current to flow in one direction. However, these are special types designed for high-speed operation. There are four common types of rectifier:
Standard rectifiers as used in linear PSUs and on the primary-side bridge. These drop about 1.0V when fully loaded. They are too slow for the high switching frequencies, however, and are never used on the secondary of even shoddy switching supplies.
Fast recovery rectifiers have a typical reverse recovery time of 150ns [at low voltage ratings]. Unfortunately, they drop about 1.3V when fully loaded, resulting in low efficiency.
Ultra-fast rectifiers have reverse recovery times as low as 35ns [again, at low voltage ratings]. They are often used interchangeably with the above type. At ratings of up to 200PRV, they drop about 1.0V when fully loaded, the same as standard rectifiers. [At 400PRV they only break even with fast-recovery rectifiers and above that they can drop up to 1.7V, but no sane PSU manufacturer would use parts with that rating on the secondary.]
Schottky rectifiers are the fastest of all and [at low voltage ratings] can drop <0.5V, resulting in greatly improved efficiency. They are the only type of rectifier used on the +3.3V and +5V rails.
There is also synchronous rectification, which uses MOSFETs switched on and off at timed intervals to avoid the problem of diode drop.

The main rails use 3-pinned components containing two diodes with a shared cathode terminal. These get hot and require a heatsink. The -12V output uses small cylindrical diodes (rarely if ever Schottky as efficiency is not the goal) usually rated for only 1~2A. The +5VSB output may use either a larger cylindrical diode (generally rated for 3~5A) or a power rectifier similar to those on the main rails. Some ultra-low-quality PSUs use two cylindrical diodes (usually fast recovery and rated for 3A each) in place of the +12V rectifier, crippling its current capacity to a theoretical 6A, which is inadequate for any modern system that draws most of its power from that rail.

Unlike the switchers, they don’t go out with a fireworks display when they fail. The rectifier will short internally, the output voltage for whichever rail it is on will briefly spike, damaging attached hardware and the PSU’s SCP (Short Circuit Protection) will step in and shut the PSU down.

The output filtering
This looks about right, except that most standby supplies do have both capacitors. Also, "pi" refers to the arrangement of the two capacitors and coil, so it isn't technically correct if you remove one capacitor.

The Feedback and Protection
This is a bit more difficult, but I think the important thing is that the old half-bridge topology has the PWM controller on the secondary, while other topologies have it on the primary side.