Checking the water power supply board. Diagnostics of a computer power supply. Signs of a faulty power supply

The health of any living organism depends on how and what it eats. The same can be said about a computer - if the power supply is working well and correctly, electronic devices function “like a clock.” And vice versa: if the feeder malfunctions, working on a PC turns into torture or becomes completely impossible.

Problems with a computer power supply manifest themselves in different ways - from a lack of response to an attempt to turn it on to occasional “glitches”. Let's talk about what symptoms indicate a failure of the computer's power supply and how to check its functionality and serviceability without exposing yourself to danger.

Complete failure and malfunction of the power supply most often occur due to:

Voltage surges in the electrical network.
Low quality of the PSU itself.
Inconsistencies between power supply capabilities and load consumption (computer devices).

The consequences of a malfunction of the power supply, especially in combination with low quality manufacturing, can be not only breakdowns of the PC electronics, but also electric shock to the user.

How computer power supply problems manifest themselves

Symptoms of a malfunctioning feeder are very varied. Among them:

The PC does not turn on when you press the power button or turns on after pressing it multiple times.
Squeaking, crackling, clicking, smoke, burning smell from the power supply.
The mains fuse on the distribution board blows when the computer is turned on.
Discharges of static electricity from the case and connectors of the system unit.
Spontaneous shutdowns and restarts of the PC at any time, but more often under high loads.
Brakes and freezing (until reboot).
Memory errors, BSoD (blue screens of death).
Loss of devices from the system (drives, keyboards, mice, other peripheral equipment).
Stopping the fans.
Overheating of devices due to ineffective operation or stopping of fans.

Operating principle of the power supply

To figure out whether the power supply is working or not, you need to understand the basic principles of its operation. In a simplified way, its function can be described as follows: converting the input AC voltage of a household electrical network into a DC output of several levels: 12 V, 5 V 5 V SB (standby voltage), 3.3 V and -12 V.

The following devices receive power from a 12-volt source:

drives connected via SATA interface;
optical drives;
cooling system fans;
processors;
video cards.

The 12 V line wires are yellow.

Powered from 5 V and 3.3 V:

sound, network controller and the bulk of the motherboard microcircuits;
RAM;
expansion boards;
peripheral devices connected to USB ports.

According to the ATX standard, the 5 V line is indicated by red wires, 5 V SB by purple, and 3.3 V by orange.

The computer startup circuit on the motherboard receives power from a 5 V SB (standby) source. The -12 V source is designed to power COM ports, which today can only be found on very old motherboards and specialized devices (for example, cash registers).

The above voltages are produced by all ATX standard power supplies, regardless of power. The only differences are in the level of currents on each line: the more powerful the feeder, the more current it delivers to consumer devices.

Information about the currents and voltages of individual lines can be obtained from the power supply passport, which is pasted in the form of a label on one of the sides of the device. However, nominal indicators almost always differ from real ones. This does not mean anything bad: fluctuations in values within 5% are considered normal. Such minor deviations do not affect the operation of computer devices.

Among other things, a working power supply produces a Power Good or Power OK signal, which notifies the motherboard that it is working as it should and the board can start other devices. Normally, this signal has a level of 3-5.5 V and rises only when all supply voltages have reached the specified values. If the power supply does not produce Power Good, the computer will not start. If it produces too early, which is also not good, the device may turn on and turn off immediately, freeze during boot, or throw a critical error - blue screen of death.

The Power Good signal is transmitted to the motherboard via the gray wire.

ATX main power supply connector pins

We figured out the color coding of wires 12 V, 5 V, 5 V SB, 3.3 V and 3-5.5 V Power Good. The remaining contacts have the following voltages:

White:-5 V. Left for compatibility with older devices.
Blue:-12 V.
Black: 0 V. Common wire or ground.
Green: 3-5 V. Power On. Closing this contact to ground is equivalent to pressing the power button on the computer case. Starts the power supply. At the moment of pressing, the voltage at the button contacts should drop to 0 V.

The same voltages are present on other connectors that terminate the power supply cables. That is, in the yellow wire projection there should always be 12 V, in the red wire projection - 5 V, in the orange wire projection - 3.3 V, etc.

How to Test a Power Supply Using a Multimeter

The compliance of all voltages that the feeder produces with the specified levels and the preservation of their values under any load (if they do not exceed the capabilities of the power supply) indicate that the device is operational and, most likely, in good working order. And to determine them, you will need a multimeter - an inexpensive compact device that can be purchased at almost any electrical goods store.

Multimeters (testers), of course, are different. Among them there are expensive high-precision models with a lot of additional functions, but for our purposes a simple one is enough. To check the power supply, we don’t need measurements down to thousandths of a volt; tenths and sometimes hundredths are enough.

Conditions for taking measurements

Measurements of voltages at the power supply outputs should be made under conditions where failure occurs. If the problem appears in the first seconds and minutes of PC operation, the device readings should be taken immediately after turning on. If you are working intensively, to obtain reliable results, the computer should be loaded, for example, with a heavy game or a program designed for this (for example, the OCCT utility, Power Supply test).

To track changes in supply voltages during PC operation, measurements are best taken continuously over several minutes or tens of minutes. If for some reason this is difficult, you can take one-time measurements at certain time intervals.

The result of a single measurement during a floating fault is often not an indicator, since in the case of unstable operation of the feeder, the voltage values (or one of them) can constantly change.

The procedure for taking measurements

Turn on the computer and bring it into the state where the problem occurs.
Switch the multimeter to DC voltage measurement mode (the icon on the instrument panel is surrounded by a yellow frame). Set the upper scale limit to 20 V.
Connect the black probe to any metal pad on the motherboard where the voltage is 0 V (for example, near a mounting hole), or to a pin in the connector that the black wire goes to.
Place the red probe in the measurement area (in the connector opposite the corresponding wire). The number that you see on the tester display is the voltage indicator in Volts.

How to check the functionality of the feeder if the computer does not turn on

One of the common reasons why the computer does not respond to pressing the power button is precisely the malfunction of the power supply. To confirm or refute this version, all we need is a metal clip or tweezers, with which we can simulate pressing a button. Remember, a little earlier we found out that for this you need to short-circuit the green and black wires on the 24-pin connector of the power supply unit, which is connected to the motherboard? Just before that it needs to be disconnected from it.

Connect a certain load—an energy consumer—to the power supply, which is disconnected from the motherboard and computer devices. For example, an unused optical drive or light bulb. Please be aware that if the power supply is faulty, the connected device may be damaged. Therefore, use what you don't mind.
Plug in the power supply.
Use a paper clip to connect the 2 pins opposite the green and black wires. If the feeder shows signs of life - it starts the fan inside and turns on the connected load, then it is operational. However, performance does not mean serviceability, that is, this diagnostic method only allows you to differentiate a working device from a completely non-working one.

What diagnostic methods for computer power supplies still exist?

Checking the power supply with a multimeter and a paper clip is enough to identify its malfunction in about 70-80% of cases. If you do not plan to repair it in the future, then you can limit yourself to this. In professional diagnostics of power supplies, not only these, but also other methods are used to localize the defect. Including:

Checking the output voltage ripple using an oscilloscope. This is a rather expensive device, so it is unlikely that anyone will decide to buy it for a one-time job.
Disassembly, inspection, checking voltages and resistances of printed circuit board elements for compliance with standards. It is dangerous to do this without special training, since power supplies accumulate household voltage in some parts. Accidentally touching any live part may result in electric shock.
Current measurement. This is done using an ammeter built into the tester, which is connected to the break in the line being tested. To create a gap, board elements are usually desoldered.
Testing on stands with specially selected equipment in various operating modes.

In short, there are quite a few methods for diagnosing power supplies, but not all of them are applicable or advisable at home. Except for research purposes, if, of course, the owner is interested in this.

Nowadays, many devices are powered by external power supplies - adapters. When the device has stopped showing signs of life, you first need to determine which part is defective, in the device itself, or the power supply is faulty.
First of all, an external examination. You should be interested in traces of a fall, a broken cord...

After an external inspection of the device being repaired, the first thing to do is check the power supply and what it outputs. It doesn't matter whether it's a built-in power supply or an adapter. It is not enough to simply measure the supply voltage at the power supply output. Needs a small load A. Without load it may show 5 volts, under light load it will be 2 volts.

An incandescent lamp at a suitable voltage does a good job of acting as a load.. The voltage is usually written on the adapters. For example, let's take the power adapter from the router. 5.2 volts 1 amp. We connect a 6.3 volt 0.3 ampere light bulb and measure the voltage. A light bulb is enough for a quick check. Lights up - the power supply is working. It is rare for the voltage to be very different from the norm.

A lamp with a higher current may prevent the power supply from starting, so a low-current load is sufficient. I have a set of different lamps hanging on the wall for testing.

1 and 2 for testing computer power supplies, with more power and less power, respectively.
3 . Small lamps 3.5 volts, 6.3 volts for checking power adapters.
4 . A 12-volt automotive lamp for testing relatively powerful 12-volt power supplies.
5 . 220 volt lamp for testing television power supplies.
6 . There are two garlands of lamps missing from the photo. Two of 6.3 volts, for testing 12 volt power supplies, and 3 of 6.3 for testing laptop power adapters with a voltage of 19 volts.

If you have a device, it is better to check the voltage under load.

If the light does not light, it is better to first check the device with a known good power supply, if one is available. Because power adapters are usually made non-separable, and to repair it you will have to pick it apart. You can't call it dismantling.
An additional sign of a malfunctioning power supply can be a whistle from the power supply unit or the powered device itself, which usually indicates dry electrolytic capacitors. Tightly closed enclosures contribute to this.

The power supplies inside the devices are checked using the same method. In old TVs, a 220 volt lamp is soldered instead of a line scan, and by the glow you can judge its performance. Partly, the load lamp is connected due to the fact that some power supplies (built-in) can produce significantly higher voltage without load than required.

— in the life of every radio amateur, sooner or later there comes a time when he has to start mastering minor equipment repairs. This could be desktop computer speakers, a tablet, a mobile phone and some other gadgets. I won’t be mistaken if I say that almost every radio amateur has tried to repair his computer. Some people succeeded, but others still took it to the service center.

Diagnosing PC power supply faults

In this article, we will walk you through the basics of self-diagnosis of PC power supply faults.

Let's assume that we got our hands on a power supply unit (PSU) from a computer. Now you need to find out how check the computer power supply— first we need to make sure whether it is working? By the way, you need to take into account that the standby voltage of +5 Volts is present immediately after connecting the network cable to the power supply.

If it is not there, then it would be a good idea to test the power cord for integrity with a multimeter in audio testing mode. Also, don’t forget to ring the button and fuse. If everything is OK with the power cord, then we turn on the PC power supply to the network and start it without the motherboard by closing two contacts: PS-ON and COM. PS-ON is abbreviated from English. — Power Supply On — literally means “turn on the power supply.” COM is short for English. Сommon - general. A green wire goes to the PS-ON contact, and the “common” one, also known as minus, is a black wire.

Modern power supplies have a 24 Pin connector. On older ones - 20 Pin.

The easiest way to close these two contacts is with a straightened paper clip

Although theoretically any metal object or wire will do for this purpose. You can even use the same tweezers.

Method for checking the power supply

How to check a computer's power supply? If the power supply is working, it should turn on immediately, the fan will begin to rotate and voltage will appear on all connectors of the power supply.

If our computer is malfunctioning, then it would be useful to check on its connectors that the voltage on its contacts corresponds. And in general, when the computer is buggy and a blue screen often appears, it would be a good idea to check the voltage in the system itself by downloading a small PC diagnostic program. I recommend the AIDA program. In it you can immediately see whether the voltage in the system is normal, whether the power supply is to blame, or whether the motherboard is “mandating”, or even something else.

Here is a screenshot from the AIDA program on my PC. As we can see, all voltages are normal:

If there is any decent voltage deviation, then it is no longer normal. By the way, when buying a used computer, ALWAYS download this program to it and fully check all voltages and other system parameters. Tested by bitter experience:-(.

If, however, the voltage value is very different at the power supply connector itself, then you should try to repair the unit, but for this you need to know how to check computer power supply. If you are generally very bad with computer equipment and repairs, then in the absence of experience it is better to replace it. There are often cases when a faulty power supply, when it fails, “drags” part of the computer with it. Most often, this causes the motherboard to fail. How can this be avoided and how to check the computer power supply?

You can never save on a power supply and you should always have a small power reserve. It is advisable not to buy cheap NONAME power supplies.

What to do if you have little knowledge of brands and models of power supplies, but your mother won’t give you money for a new, high-quality one))? It is advisable that it has a 12 cm fan, not 8 cm.

Power supply with 12 cm fan

Such fans provide better cooling of the radio components of the power supply. You also need to remember one more rule: a good power supply cannot be light. If the power supply is light, it means that it uses small-section radiators and such a power supply will overheat during operation at rated loads. What happens when it overheats? When overheated, some radioelements, especially semiconductors and capacitors, change their values and the entire circuit as a whole does not work correctly, which, of course, will affect the operation of the power supply.

Also, do not forget to clean your power supply from dust at least once a year and take good care of how to check computer power supply. Dust acts as a “blanket” for radioelements, under which they can function incorrectly or even “die” from overheating.

The most common failure of a power supply is power semiconductors and capacitors. If there is a smell of burnt silicon, then you need to look at what burned out from the diodes or transistors. Faulty capacitors are identified by visual inspection. Opened, swollen, with leaking electrolyte - this is the first sign that they urgently need to be changed.

When replacing, it is necessary to take into account that the power supplies contain capacitors with low equivalent series resistance (ESR). So in this case, you should get an ESR meter and choose capacitors with the lowest ESR possible. Here is a small plate of resistances for capacitors of various capacities and voltages:

Here it is necessary to select capacitors in such a way that the resistance value is no more than indicated in the table.

When replacing capacitors, two more parameters are also important: capacitance and their operating voltage. They are indicated on the capacitor body:

What if the store has capacitors of the required rating, but designed for a higher operating voltage? They can also be installed in circuits during repairs, but it must be taken into account that capacitors designed for higher operating voltages usually have larger dimensions.

If our power supply starts up, then we measure the voltage at its output connector or connectors with a multimeter. In most cases, when measuring the voltage of ATX power supplies, it is sufficient to select a DCV limit of 20 volts.

There are two diagnostic methods:

— taking measurements “hot” with the device turned on

— carrying out measurements in a de-energized device

What can we measure and how are these measurements carried out? We are interested in measuring the voltage at specified points of the power supply, measuring the resistance between certain points, sound testing for the absence or presence of a short circuit, and also measuring the current strength. Let's take a closer look.

Voltage measurement.

If you are repairing a device and have a schematic diagram for it, it will often indicate what voltage should be at the test points on the diagram. Of course, you are not limited to just these test points and can measure the potential difference or voltage at any point in the power supply or any other device being repaired. But to do this, you must be able to read diagrams and be able to analyze them. You can read more about how to measure voltage with a multimeter in this article.

Resistance measurement.

Every part of the circuit has some kind of resistance. If, when measuring resistance, there is one on the multimeter screen, this means that in our case the resistance is higher than the resistance measurement limit chosen by us. Let me give you an example: for example, we measure the resistance of a part of a circuit consisting conventionally of a resistor of a value known to us and a choke. As we know, a choke is, roughly speaking, just a piece of wire with a small resistance, and we know the value of the resistor. On the multimeter screen we see a resistance slightly greater than the value of our resistor. Having analyzed the circuit, we come to the conclusion that these radio components are working and good contact is ensured with them on the board. Although at first, if you lack experience, it is advisable to call all the details separately. You also need to take into account that parallel connected radio components influence each other when measuring resistance. Remember the parallel connection of resistors and you will understand everything. You can read more about resistance measurement here.

Sound verification.

If a sound signal is heard, this means that the resistance between the probes, and accordingly the section of the circuit connected to its ends, is early zero, or close to it. With its help, we can verify the presence or absence of a short circuit on the board. You can also detect whether there is a contact on the circuit or not, for example, in the event of a broken track or a broken connection, or a similar malfunction.

Measuring current flow in a circuit

When measuring the current in a circuit, intervention in the board design is required, for example, by soldering one of the terminals of the radio component. Because, as we remember, our ammeter is connected to an open circuit. How to measure current in a circuit can be read in this article.

Using these four measurement methods with just one multimeter, you can diagnose a very large number of faults in the circuits of almost any electronic device.

As they say, there are two main faults in electrics: there is contact where there should not be one, and there is no contact where there should be one. What does this saying mean in practice? For example, when any radio component burns out, we get a short circuit, which is an emergency for our circuit. For example, this could be a breakdown of the transistor. In circuits, a break can also occur, in which current in our circuit cannot flow. For example, a break in a track or contacts through which current flows. It could also be a broken wire or the like. In this case, our resistance becomes, relatively speaking, infinity.

Of course, there is a third option: changing the parameters of the radio component. For example, as is the case with the same electrolytic capacitor, or burning of the switch contacts, and as a result, a strong increase in their resistance. Knowing these three failure options and being able to analyze circuits and printed circuit boards, you will learn how to easily repair your electronic devices. You can read more about the repair of radio-electronic devices in the article “Basics of Repair.”

You, like most personal computer users, have probably already encountered various problems associated with the failure of any vital configuration components. The PC power supply directly relates to such details, which tends to break if the level of care on the part of the owner is insufficient.

In this article, we will look at all currently relevant methods for testing PC power supplies for functionality. Moreover, we will also partially touch on a similar problem encountered by laptop users.

As we said above, the computer’s power supply, regardless of other components of the assembly, is an important part. As a result, a breakdown of this component can lead to complete failure of the entire system unit, which makes diagnostics significantly more difficult.

If your PC does not turn on, it may not be the power supply that is to blame - remember this!

The whole difficulty of diagnosing this kind of components lies in the fact that the lack of power in a PC can be caused not only by the power supply, but also by other components. This is especially true for the central processor, the failure of which manifests itself in a huge variety of consequences.

Be that as it may, diagnosing problems in the operation of a power supply device is much easier than in case of malfunctions of other elements. This conclusion is due to the fact that the component in question is the only possible source of energy in the computer.

Method 1: Check the power supply

If at any time during the operation of your PC you find it inoperative, you need to immediately check the availability of electricity. Make sure that the network is fully functional and meets the requirements of the power supply.

Sometimes power outages may occur, but in this case the consequences are limited to the PC turning off on its own.

It would not be superfluous to double-check the cord connecting the power supply to the network for visible damage. The best test method would be to try connecting the power cord you are using to another fully working PC.

If you are using a laptop, the steps to eliminate power problems are completely similar to those described above. The only difference here is that if there is a problem with the cable of a laptop computer, replacing it will cost an order of magnitude more than if there are problems with a full-fledged PC.

It is important to carefully inspect and test the power source, be it an outlet or a surge protector. All subsequent sections of the article will be aimed specifically at the power supply, so it is extremely important to solve any problems with electrical power in advance.

Method 2: Using a jumper

This method is ideal for initial testing of the power supply to determine its performance. However, it is worth making a reservation in advance that if you have never interfered with the operation of electrical appliances before and do not fully understand the principle of operation of a PC, the best solution would be to contact technical specialists.

If any complications occur, you can put your life and the condition of your PD in serious danger!

The whole point of this section of the article is to use a hand-made jumper to subsequently close the contacts of the power supply. It is important to note that the method is widely popular among users and this, in turn, can greatly help if any inconsistencies with the instructions arise.

Before proceeding directly to the description of the method, you will need to disassemble the computer in advance.

You can learn a little more about turning off the power supply from the dedicated article.

Having dealt with the introduction, you can proceed to diagnostics by using the jumper. And right away it should be noted that, in fact, this method was already described by us earlier, since it was created primarily to be able to start a power supply without using a motherboard.

Having familiarized yourself with the PSU startup method we have given, after supplying electricity, you should pay attention to the fan. If the main cooler of the device shows no signs of life, you can safely conclude that it is inoperable.

It is best to replace a broken power supply or send it to a service center for repair.

If after startup the cooler works properly and the power supply unit itself makes characteristic sounds, we can say with a high degree of probability that the device is in working condition. However, even under such circumstances, the verification guarantee is far from ideal and therefore we recommend a more in-depth analysis.

Method 3: Using a Multimeter

As can be seen directly from the name of the method, the method involves using a special engineering device "Multimeter". First of all, you will need to acquire such a meter, and also learn the basics of its use.

Typically, among experienced users, a multimeter is referred to as a tester.

Refer to the previous method after completing all testing instructions. After this, having made sure that it is working and maintaining open access to the main power supply cable, you can proceed to active actions.

First you need to find out what specific type of cable is used in your computer. There are two types of them:

20-pin;
24-pin.

You can make the calculation by reading the technical specifications of the power supply or by manually counting the number of pins of the main connector.

Depending on the type of wire, the recommended actions vary slightly.

Prepare a small but fairly reliable wire, which will then be needed to close certain contacts.

If you are using a 20-pin power supply connector, you should connect pins 14 and 15 to each other using a cable.

When the power supply is equipped with a 24-pin connector, you need to close pins 16 and 17, also using a previously prepared piece of wire.

Having completed everything exactly according to the instructions, connect the power supply to the mains.

At the same time, make sure that by the time you connect the power supply to the network, nothing intersects with the wire, or rather its uninsulated ends.

Don't forget to use hand protection!

As in the earlier method, after power supply is supplied, the power supply may not start, which directly indicates a malfunction. If the cooler does work, you can proceed to more detailed diagnostics by using a tester.

All values given are rounded figures, as minor differences may still occur due to certain circumstances.

After completing our instructions, make sure that the data obtained corresponds to the voltage level standard. If you notice significant differences, the power supply can be considered partially faulty.

The voltage level supplied to the motherboard is independent of the PSU model.

Since the power supply itself is a rather complex component of a personal computer, it is best to contact specialists for repairs. This is especially true for users who are new to the operation of electrical devices.

In addition to the above, a multimeter may well come in handy when checking a laptop’s network adapter. And although breakdowns of this type of power supply are rare, you can still find problems, in particular when operating the laptop in rather harsh conditions.

The laptop model does not affect the level of supplied electricity at all.

If these indicators are missing, you need to carefully examine the network cable again, as we said in the first method. If there are no visible defects, only complete replacement of the adapter can help.

Method 4: Using a Power Supply Tester

In this case, for analysis you will need a special device designed for testing the power supply. Thanks to such a device, you can connect the pins of PC components and get the results.

The cost of such a tester, as a rule, is somewhat lower than that of a full-fledged multimeter.

Please note that the device itself may differ significantly from the one shown by us. And although power supply testers come in different models that differ in appearance, the principle of operation is always the same.

Read the specifications of the meter you are using to avoid difficulties.

Connect the corresponding wire from the power supply to the 24-pin connector on the case.

Depending on your personal preferences, connect other contacts to special connectors on the case.

It is recommended to use a Molex connector.

It is also advisable to add voltage from the hard drive using the SATA II interface.

Use the power button of the measuring device to take performance indicators of the power supply.

You may need to press the button briefly.

The final results will be presented to you on the device screen.

There are only three main indicators:

+5V – from 4.75 to 5.25 V;
+12V – from 11.4 to 12.6 V;
+3.3V – from 3.14 to 3.47 V.

If your final measurements are lower or higher than normal, as stated earlier, the power supply requires immediate repair or replacement.

Method 5: Using system tools

Including cases where the power supply is still in working order and allows you to start the PC without any difficulties, you can diagnose faults using system tools. Please note that checking is mandatory only when there are obvious problems in the computer’s behavior, for example, spontaneous turning on or off.

The article we bring to your attention describes the methodology we use for testing power supplies - until now, individual parts of this description have been scattered across various articles with tests of power supplies, which is not very convenient for those who want to quickly familiarize themselves with the methodology based on its current state.

This material is updated as the methodology develops and improves, so some of the methods reflected in it may not be used in our old articles with power supply tests - this only means that the method was developed after the publication of the corresponding article. You will find a list of changes made to the article at the end.

The article can be quite clearly divided into three parts: in the first, we will briefly list the block parameters we check and the conditions for these checks, and also explain the technical meaning of these parameters. In Part 2, we will mention a number of terms often used by block manufacturers for marketing purposes and explain them. The third part will be of interest to those who want to familiarize themselves in more detail with the technical features of the construction and operation of our stand for testing power supplies.

The guiding and guiding document for us in developing the methodology described below was the standard , the latest version of which can be found at FormFactors.org. At the moment, it is included as an integral part of a more general document called Power Supply Design Guide for Desktop Platform Form Factors, which describes blocks not only of ATX, but also of other formats (CFX, TFX, SFX, and so on). Although PSDG is not formally a mandatory standard for all power supply manufacturers, we a priori believe that unless otherwise explicitly stated for a computer power supply (that is, it is a unit that is in regular retail sale and intended for general use , and not any specific computer model from a particular manufacturer), it must comply with PSDG requirements.

You can view the test results for specific power supply models in our catalog: " Catalog of tested power supplies".

Visual inspection of the power supply

Of course, the first stage of testing is a visual inspection of the block. In addition to aesthetic pleasure (or, conversely, disappointment), it also gives us a number of quite interesting indicators of the quality of the product.

First, of course, is the quality of the case. Metal thickness, rigidity, assembly features (for example, the body can be made of thin steel, but fastened with seven or eight bolts instead of the usual four), the quality of the block's painting...

Secondly, the quality of internal installation. All power supplies passing through our laboratory are necessarily opened, examined inside and photographed. We do not focus on small details and do not list all the parts found in the block along with their denominations - this, of course, would give the articles a scientific appearance, but in practice in most cases it is completely meaningless. However, if a block is made according to some generally relatively non-standard scheme, we try to describe it in general terms, as well as explain the reasons why the block designers could choose such a scheme. And, of course, if we notice any serious flaws in the quality of workmanship - for example, sloppy soldering - we will definitely mention them.

Thirdly, the passport parameters of the block. In the case of, let's say, inexpensive products, it is often possible to draw some conclusions about the quality based on them - for example, if the total power of the unit indicated on the label turns out to be clearly greater than the sum of the products of the currents and voltages indicated there.

Also, of course, we list the cables and connectors available on the unit and indicate their length. We write the latter as a sum in which the first number is equal to the distance from the power supply to the first connector, the second number is equal to the distance between the first and second connectors, and so on. For the cable shown in the figure above, the entry will look like this: “removable cable with three power connectors for SATA hard drives, length 60+15+15 cm.”

Full power operation

The most intuitive and therefore most popular characteristic among users is the full power of the power supply. The unit label indicates the so-called long-term power, that is, the power with which the unit can operate indefinitely. Sometimes the peak power is indicated next to it - as a rule, the unit can operate with it for no more than a minute. Some not very conscientious manufacturers indicate either only peak power, or long-term power, but only at room temperature - accordingly, when working inside a real computer, where the air temperature is higher than room temperature, the permissible power of such a power supply is lower. According to recommendations ATX 12V Power Supply Design Guide, a fundamental document on the operation of computer power supplies, the unit must operate with the load power indicated on it at an air temperature of up to 50 ° C - and some manufacturers explicitly mention this temperature to avoid discrepancies.

In our tests, however, the operation of the unit at full power is tested under mild conditions - at room temperature, about 22...25 °C. The unit operates with the maximum permissible load for at least half an hour, if during this time no incidents occur with it, the test is considered successfully passed.

At the moment, our installation allows us to fully load units with a power of up to 1350 W.

Cross-load characteristics

Despite the fact that a computer power supply is a source of several different voltages at the same time, the main ones being +12 V, +5 V, +3.3 V, in most models there is a common stabilizer for the first two voltages. In his work, he focuses on the arithmetic mean between two controlled voltages - this scheme is called “group stabilization”.

Both the disadvantages and advantages of this design are obvious: on the one hand, cost reduction, on the other, the dependence of voltages on each other. Let’s say, if we increase the load on the +12 V bus, the corresponding voltage sags and the unit’s stabilizer tries to “pull” it to the previous level - but, since it simultaneously stabilizes +5 V, they increase both voltage. The stabilizer considers the situation corrected when the average deviation of both voltages from the nominal is zero - but in this situation this means that the +12 V voltage will be slightly lower than the nominal, and +5 V will be slightly higher; if we raise the first, then the second will immediately increase, if we lower the second, the first will also decrease.

Of course, block developers make some efforts to mitigate this problem - the easiest way to evaluate their effectiveness is with the help of the so-called cross-load characteristics graphs (abbreviated CLO).

Example of a KNH schedule

The horizontal axis of the graph shows the load on the +12 V bus of the unit under test (if it has several lines with this voltage, the total load on them), and the vertical axis shows the total load on the +5 V and +3.3 V buses. Accordingly, each a point on the graph corresponds to a certain block load balance between these buses. For greater clarity, we not only depict on the KNH graphs the zone in which the output loads of the unit do not exceed permissible limits, but also indicate their deviations from the nominal in different colors - from green (deviation less than 1%) to red (deviation from 4 to 5 %). A deviation of more than 5% is considered unacceptable.

Let's say, in the above graph we see that the voltage of +12 V (it was built specifically for this) of the tested unit is kept well, a significant part of the graph is filled with green - and only with a strong imbalance of loads towards the +5 V and +3 buses, 3V it goes red.

In addition, on the left, bottom and right of the graph is limited by the minimum and maximum permissible load of the block - but the uneven upper edge is due to stresses exceeding the 5 percent limit. According to the standard, the power supply can no longer be used for its intended purpose in this load range.

Area of typical loads on the KNH graph

Of course, it is also of great importance in which area of the graph the voltage deviates more from the nominal value. In the picture above, the area of power consumption that is typical for modern computers is shaded - all of their most powerful components (video cards, processors...) are now powered by the +12 V bus, so the load on it can be very large. But on the +5 V and +3.3 V buses, in fact, only hard drives and motherboard components remain, so their consumption very rarely exceeds several tens of watts even in computers that are very powerful by modern standards.

If you compare the above graphs of the two blocks, you can clearly see that the first of them turns red in an area that is insignificant for modern computers, but the second, alas, is the opposite. Therefore, although in general both blocks showed similar results over the entire load range, in practice the first will be preferable.

Since during the test we monitor all three main buses of the power supply - +12 V, +5 V and +3.3 V - then the power supplies in the articles are presented in the form of an animated three-frame image, each frame of which corresponds to the voltage deviation on one of the mentioned tires

Recently, power supplies with independent stabilization of output voltages have also become increasingly widespread, in which the classic circuit is supplemented with additional stabilizers according to the so-called saturable core circuit. Such blocks demonstrate a significantly lower correlation between output voltages - as a rule, the KNH graphs for them are replete with green color.

Fan speed and temperature rise

The efficiency of the unit's cooling system can be considered from two perspectives - from the point of view of noise and from the point of view of heating. Obviously, achieving good performance on both of these points is very problematic: good cooling can be achieved by installing a more powerful fan, but then we will lose in noise - and vice versa.

To evaluate the cooling efficiency of the block, we step by step change its load from 50 W to the maximum permissible, at each stage giving the block 20...30 minutes to warm up - during this time its temperature reaches a constant level. After warming up, using a Velleman DTO2234 optical tachometer, the rotation speed of the unit’s fan is measured, and using a Fluke 54 II two-channel digital thermometer, the temperature difference between the cold air entering the unit and the heated air leaving it is measured.
Of course, ideally both numbers should be minimal. If both the temperature and the fan speed are high, this tells us that the cooling system is poorly designed.

Of course, all modern units have adjustable fan speed - however, in practice, the initial speed can vary greatly (that is, the speed at minimum load; it is very important, since it determines the noise of the unit at moments when the computer is not loaded with anything - and therefore the fans video cards and processor rotate at minimum speed), as well as a graph of speed versus load. For example, in power supplies of the lower price category, a single thermistor is often used to regulate the fan speed without any additional circuits - in this case, the speed can change by only 10...15%, which is difficult to even call adjustment.

Many power supply manufacturers specify either the noise level in decibels or the fan speed in rpm. Both are often accompanied by a clever marketing ploy - noise and speed are measured at a temperature of 18 °C. The resulting figure is usually very beautiful (for example, a noise level of 16 dBA), but does not carry any meaning - in a real computer the air temperature will be 10...15 °C higher. Another trick we came across was to indicate for a unit with two different types of fans the characteristics of only the slower one.

Output voltage ripple

The principle of operation of a switching power supply - and all computer units are switching - is based on the operation of a step-down power transformer at a frequency significantly higher than the frequency of the alternating current in the supply network, which makes it possible to reduce the dimensions of this transformer many times.

The alternating mains voltage (with a frequency of 50 or 60 Hz, depending on the country) at the input of the unit is rectified and smoothed, after which it is supplied to a transistor switch, which converts the direct voltage back into alternating voltage, but with a frequency three orders of magnitude higher - from 60 to 120 kHz, depending on the power supply model. This voltage is supplied to a high-frequency transformer, which lowers it to the values we need (12 V, 5 V...), after which it is straightened and smoothed again. Ideally, the output voltage of the unit should be strictly constant - but in reality, of course, it is impossible to completely smooth out the alternating high-frequency current. Standard requires that the range (distance from minimum to maximum) of the residual ripple of the output voltages of power supplies at maximum load does not exceed 50 mV for the +5 V and +3.3 V buses and 120 mV for the +12 V bus.

When testing the unit, we take oscillograms of its main output voltages at maximum load using a Velleman PCSU1000 dual-channel oscilloscope and present them in the form of a general graph:

The top line on it corresponds to the +5 V bus, the middle line – +12 V, the bottom – +3.3 V. In the picture above, for convenience, the maximum permissible ripple values are clearly shown on the right: as you can see, in this power supply the +12 V bus fits it’s easy to fit into them, the +5 V bus is difficult, and the +3.3 V bus doesn’t fit at all. High narrow peaks on the oscillogram of the last voltage tell us that the unit cannot cope with filtering the highest frequency noise - as a rule, this is a consequence of the use of insufficiently good electrolytic capacitors, the efficiency of which decreases significantly with increasing frequency.

In practice, if the power supply ripple range exceeds the permissible limits, it can negatively affect the stability of the computer and also cause interference with sound cards and similar equipment.

Efficiency

If above we considered only the output parameters of the power supply, then when measuring efficiency, its input parameters are already taken into account - what percentage of the power received from the supply network the unit converts into the power it supplies to the load. The difference, of course, goes to useless heating of the block itself.

The current version of the ATX12V 2.2 standard imposes a limit on the efficiency of the unit from below: a minimum of 72% at rated load, 70% at maximum and 65% at light load. In addition, there are the figures recommended by the standard (80% efficiency at rated load), as well as the voluntary certification program “80+Plus”, according to which the power supply must have an efficiency of at least 80% at any load from 20% to the maximum permissible. The same requirements as 80+Plus are contained in the new Energy Star certification program version 4.0.

In practice, the efficiency of the power supply depends on the network voltage: the higher it is, the better the efficiency; the difference in efficiency between 110 V and 220 V networks is about 2%. In addition, the difference in efficiency between different units of the same model due to the variation in component parameters can also be 1...2%.

During our tests, we change the load on the unit in small steps from 50 W to the maximum possible and at each step, after a short warm-up, we measure the power consumed by the unit from the network - the ratio of the load power to the power consumed from the network gives us the efficiency. The result is a graph of efficiency depending on the load on the unit.

As a rule, the efficiency of switching power supplies increases rapidly as the load increases, reaches a maximum and then slowly decreases. This nonlinearity gives an interesting consequence: from the point of view of efficiency, as a rule, it is slightly more profitable to buy a unit whose rated power is adequate to the load power. If you take a block with a large power reserve, then a small load on it will fall into the area of the graph where the efficiency is not yet maximum (for example, a 200-watt load on the graph of a 730-watt block shown above).

Power factor

As you know, in an alternating current network two types of power can be considered: active and reactive. Reactive power occurs in two cases - either if the load current in phase does not coincide with the network voltage (that is, the load is inductive or capacitive in nature), or if the load is nonlinear. A computer power supply is a clear second case - if no additional measures are taken, it consumes current from the mains in short, high pulses that coincide with the maximum mains voltage.

Actually, the problem is that if the active power is entirely converted in the block into work (by which in this case we mean both the energy supplied by the block to the load and its own heating), then the reactive power is not actually consumed by it at all - it is completely returned back to the network. So to speak, it just walks back and forth between the power plant and the block. But it heats the wires connecting them no worse than the active power... Therefore, they try to get rid of reactive power as much as possible.

A circuit known as active PFC is the most effective means of suppressing reactive power. At its core, this is a pulse converter, which is designed so that its instantaneous current consumption is directly proportional to the instantaneous voltage in the network - in other words, it is specially made linear, and therefore consumes only active power. From the output of the A-PFC, the voltage is supplied to the pulse converter of the power supply, the same one that previously created a reactive load with its nonlinearity - but since it is now a constant voltage, the linearity of the second converter no longer plays a role; it is reliably separated from the power supply network and can no longer affect it.

To estimate the relative value of reactive power, a concept such as power factor is used - this is the ratio of active power to the sum of active and reactive powers (this sum is also often called total power). In a conventional power supply it is about 0.65, and in a power supply with A-PFC it is about 0.97...0.99, that is, the use of A-PFC reduces reactive power to almost zero.

Users and even reviewers often confuse power factor with efficiency - although both describe the efficiency of a power supply, this is a very serious mistake. The difference is that the power factor describes the efficiency of the power supply's use of the AC network - what percentage of the power passing through it the unit uses for its operation, and the efficiency is the efficiency of converting the power consumed from the network into the power supplied to the load. They are not connected with each other at all, because, as was written above, reactive power, which determines the value of the power factor, is simply not converted into anything in the unit, the concept of “conversion efficiency” cannot be associated with it, therefore, it has no effect on efficiency.

Generally speaking, A-PFC is beneficial not to the user, but to energy companies, since it reduces the load on the power system created by the computer's power supply by more than a third - and when there is a computer on every desktop, this translates into very noticeable numbers. At the same time, for the average home user there is practically no difference whether his power supply contains A-PFC or not, even from the point of view of paying for electricity - at least for now, household electricity meters only take into account active power. Still, manufacturers' claims about how A-PFC helps your computer are nothing more than ordinary marketing noise.

One of the side benefits of the A-PFC is that it can be easily designed to operate over the full voltage range from 90 to 260 V, thus making a universal power supply that works on any network without manual voltage switching. Moreover, if units with mains voltage switches can operate in two ranges - 90...130 V and 180...260 V, but cannot be run in the range from 130 to 180 V, then a unit with A-PFC covers all these tensions in their entirety. As a result, if for some reason you are forced to work in conditions of unstable power supply, which often drops below 180 V, then a unit with A-PFC will either allow you to do without a UPS altogether, or significantly increase the service life of its battery.

However, A-PFC itself does not yet guarantee operation in the full voltage range - it can only be designed for a range of 180...260 V. This is sometimes found in units intended for Europe, since the rejection of the full-range A-PFC allows slightly reduce its cost.

In addition to active PFCs, passive ones are also found in blocks. They represent the simplest method of power factor correction - they are just a large inductor connected in series with the power supply. Due to its inductance, it slightly smoothes out the current pulses consumed by the unit, thereby reducing the degree of nonlinearity. The effect of P-PFC is very small - the power factor increases from 0.65 to 0.7...0.75, but if the installation of A-PFC requires serious modification of the high-voltage circuits of the unit, then P-PFC can be added without the slightest difficulty into any existing power supply.

In our tests, we determine the power factor of the unit using the same scheme as efficiency - gradually increasing the load power from 50 W to the maximum permissible. The obtained data is presented on the same graph as the efficiency.

Working in tandem with a UPS

Unfortunately, the A-PFC described above has not only advantages, but also one drawback - some of its implementations cannot work normally with uninterruptible power supplies. At the moment the UPS switches to batteries, such A-PFCs abruptly increase their consumption, as a result of which the overload protection in the UPS is triggered and it simply turns off.

To assess the adequacy of the A-PFC implementation in each specific unit, we connect it to an APC SmartUPS SC 620VA UPS and check their operation in two modes - first when powered from the mains, and then when switching to batteries. In both cases, the load power on the unit gradually increases until the overload indicator on the UPS turns on.

If this power supply is compatible with a UPS, then the permissible load power on the unit when powered from the mains is usually 340...380 W, and when switching to batteries - a little less, about 320...340 W. Moreover, if at the time of switching to batteries the power was higher, the UPS turns on the overload indicator, but does not turn off.

If the unit has the above problem, then the maximum power at which the UPS agrees to work with it on batteries drops noticeably below 300 W, and if it is exceeded, the UPS turns off completely either right at the moment of switching to batteries, or after five to ten seconds . If you are planning to acquire a UPS, it is better not to buy such a unit.

Fortunately, recently there are fewer and fewer units that are incompatible with UPS. For example, if the blocks of the PLN/PFN series of the FSP Group had such problems, then in the next GLN/HLN series they were completely corrected.

If you already own a unit that is unable to work normally with a UPS, then there are two options (in addition to modifying the unit itself, which requires good knowledge of electronics) - change either the unit or the UPS. The first, as a rule, is cheaper, since a UPS will need to be purchased with at least a very large power reserve, or even an online type, which, to put it mildly, is not cheap and is not justified in any way at home.

Marketing noise

In addition to technical characteristics, which can and should be checked during tests, manufacturers often like to supply power supplies with a lot of beautiful inscriptions telling about the technologies used in them. At the same time, their meaning is sometimes distorted, sometimes trivial, sometimes these technologies generally relate only to the features of the internal circuitry of the block and do not affect its “external” parameters, but are used for reasons of manufacturability or cost. In other words, beautiful labels are often mere marketing noise, and white noise that does not contain any valuable information. Most of these statements do not make much sense to test experimentally, but below we will try to list the main and most common ones so that our readers can more clearly understand what they are dealing with. If you think that we have missed any of the characteristic points, do not hesitate to tell us about it, we will definitely add to the article.

Dual +12V output circuits

In the old, old days, power supplies had one bus for each of the output voltages - +5 V, +12 V, +3.3 V and a couple of negative voltages, and the maximum power of each bus did not exceed 150...200 W, and only in some particularly powerful server units the load on the five-volt bus could reach 50 A, that is, 250 W. However, over time, the situation changed - the total power consumed by computers kept growing, and its distribution between the buses shifted towards +12 V.

In the ATX12V 1.3 standard, the recommended +12 V bus current reached 18 A... and this is where the problems began. No, not with an increase in current, there were no particular problems with that, but with safety. The fact is that, according to the EN-60950 standard, the maximum power on connectors freely accessible to the user should not exceed 240 VA - it is believed that high powers in the event of short circuits or equipment failure can most likely lead to various unpleasant consequences, for example, fire. On a 12-volt bus, this power is achieved at a current of 20 A, while the output connectors of the power supply are obviously considered freely accessible to the user.

As a result, when it was necessary to further increase the permissible load current by +12 V, the developers of the ATX12V standard (that is, Intel) decided to divide this bus into several, with a current of 18 A each (the difference of 2 A was included as a small margin). Purely for safety reasons, there are absolutely no other reasons for this decision. The immediate consequence of this is that the power supply doesn't actually need to have more than one +12V rail at all - it just needs to trigger protection if it tries to load any of its 12V connectors with more than 18A of current. That's all. The simplest way to implement this is to install several shunts inside the power supply, each of which is connected to its own group of connectors. If the current through one of the shunts exceeds 18 A, the protection is triggered. As a result, on the one hand, the power on any of the connectors individually cannot exceed 18 A * 12 V = 216 VA, on the other hand, the total power removed from different connectors may be greater than this figure. And the wolves are fed, and the sheep are safe.

Therefore - in fact - power supplies with two, three or four +12 V rails are practically not found in nature. Simply because it’s not necessary - why put a bunch of additional parts inside the block, where it’s already quite cramped, when you can get by with a couple of shunts and a simple microcircuit that will control the voltage on them (and since we know the resistance of the shunts, then does the voltage immediately and unambiguously imply the magnitude of the current flowing through the shunt)?

However, the marketing departments of power supply manufacturers could not ignore such a gift - and now on the boxes of power supplies there are sayings about how two +12 V lines help increase power and stability. And if there are three lines...

But it’s okay if that’s all there is to it. The latest fashion trend is power supplies in which there is, as it were, a separation of lines, but it is as if not. Like this? It’s very simple: as soon as the current on one of the lines reaches the treasured 18 A, the overload protection... is turned off. As a result, on the one hand, the sacred inscription “Triple 12V Rails for unprecedented power and stability” does not disappear from the box, and on the other hand, you can add some nonsense next to it in the same font that, if necessary, all three lines merge into one. Nonsense - because, as stated above, they were never separated. It is generally absolutely impossible to comprehend the full depth of the “new technology” from a technical point of view: in fact, they are trying to present to us the absence of one technology as the presence of another.

Of the cases known to us so far, the companies Topower and Seasonic, as well as, respectively, brands that sell their units under their own brand, have been noted in the field of promoting “self-switching protection” to the masses.

Short circuit protection (SCP)

Block output short circuit protection. Mandatory according to the document ATX12V Power Supply Design Guide– which means it is present in all blocks that claim to comply with the standard. Even those where there is no "SCP" inscription on the box.

Overpower (overload) protection (OPP)

Protection against unit overload based on total power across all outputs. Is mandatory.

Overcurrent protection (OCP)

Protection against overload (but not yet short circuit) of any of the unit outputs individually. Present on many, but not all blocks - and not for all outputs. Not mandatory.

Overtemperature protection (OTP)

Protection against block overheating. It is not so common and is not mandatory.

Overvoltage protection (OVP)

Protection against exceeding output voltages. It is mandatory, but, in fact, it is designed in case of a serious malfunction of the unit - the protection is triggered only when any of the output voltages exceeds the nominal value by 20...25%. In other words, if your unit produces 13 V instead of 12 V, it is advisable to replace it as quickly as possible, but its protection does not have to work, because it is designed for more critical situations that threaten immediate failure of the equipment connected to the unit.

Undervoltage protection (UVP)

Protection against underestimation of output voltages. Of course, too low a voltage, unlike too high, does not lead to fatal consequences for the computer, but it can cause failures, say, in the operation of a hard drive. Again, the protection is triggered when the voltage drops by 20...25%.

Nylon sleeve

Soft braided nylon tubes in which the output wires of the power supply are tucked away - they make it a little easier to lay the wires inside the system unit, preventing them from getting tangled.

Unfortunately, many manufacturers have moved from the undoubtedly good idea of using nylon tubes to thick plastic tubes, often supplemented with shielding and a layer of paint that glows in ultraviolet light. Glowing paint is, of course, a matter of taste, but the power supply wires need shielding no more than a fish needs an umbrella. But thick tubes make the cables elastic and inflexible, which not only prevents them from being placed in the case, but simply poses a danger to the power connectors, which bear considerable force from the cables that resist bending.

This is often done supposedly for the sake of improving the cooling of the system unit - but, I assure you, packaging the power supply wires in tubes has very little effect on the air flow inside the case.

Dual core CPU support

In fact, nothing more than a beautiful label. Dual-core processors do not require any special support from the power supply.

SLI and CrossFire support

Another beautiful label, indicating the presence of a sufficient number of video card power connectors and the ability to produce power considered sufficient to power an SLI system. Nothing more.

Sometimes the block manufacturer receives some kind of corresponding certificate from the video card manufacturer, but this does not mean anything other than the aforementioned availability of connectors and high power - and often the latter significantly exceeds the needs of a typical SLI or CrossFire system. After all, the manufacturer needs to somehow justify to buyers the need to purchase a block of insanely high power, so why not do this by sticking the “SLI Certified” label only on it?..

Industrial class components

Once again a beautiful label! As a rule, industrial-grade components mean parts that operate in a wide temperature range - but honestly, why put a microcircuit in the power supply that can operate at temperatures from -45 °C if this unit still won’t be exposed to the cold? .

Sometimes industrial components mean capacitors designed to operate at temperatures up to 105 °C, but here, in general, everything is also banal: capacitors in the output circuits of the power supply, heating up on their own, and even located next to hot chokes, are always designed at 105 °C maximum temperature. Otherwise, their operating life turns out to be too short (of course, the temperature in the power supply is much lower than 105 °C, but the problem is that any An increase in temperature will reduce the life of capacitors - but the higher the maximum permissible operating temperature of a capacitor, the less the effect of heating on its life will be).

Input high-voltage capacitors operate practically at ambient temperature, so the use of slightly cheaper 85-degree capacitors does not affect the life of the power supply in any way.

Advanced double forward switching design

Luring the buyer with beautiful, but completely incomprehensible words is a favorite pastime of marketing departments.

In this case, we are talking about the topology of the power supply, that is, the general principle of constructing its circuit. There are quite a large number of different topologies - so, in addition to the actual two-transistor single-cycle forward converter, in computer units you can also find single-transistor single-cycle forward converters, as well as half-bridge push-pull forward converters. All these terms are of interest only to electronics specialists; for the average user, they essentially mean nothing.

The choice of a specific power supply topology is determined by many reasons - the range and price of transistors with the necessary characteristics (and they differ significantly depending on the topology), transformers, control microcircuits... For example, a single-transistor forward version is simple and cheap, but requires the use of a high-voltage transistor and high-voltage diodes at the output of the block, so it is used only in inexpensive low-power blocks (the cost of high-voltage diodes and high-power transistors is too high). The half-bridge push-pull version is a little more complicated, but the voltage on the transistors in it is half as much... In general, it is mainly a matter of the availability and cost of the necessary components. For example, we can confidently predict that sooner or later synchronous rectifiers will begin to be used in the secondary circuits of computer power supplies - there is nothing particularly new in this technology, it has been known for a long time, it’s just too expensive and the benefits it provides do not cover the costs.

Double transformer design

The use of two power transformers, which is found in high-power power supplies (usually from a kilowatt) - as in the previous paragraph, is a purely engineering solution, which in itself, in general, does not affect the characteristics of the unit in any noticeable way - simply in some cases it is more convenient to distribute the considerable power of modern units over two transformers. For example, if one full power transformer cannot be squeezed into the height dimensions of the unit. However, some manufacturers present a two-transformer topology as allowing them to achieve greater stability, reliability, and so on, which is not entirely true.

RoHS (Reduction of Hazardous Substances)

New EU directive restricting the use of a number of hazardous substances in electronic equipment from July 1, 2006. Lead, mercury, cadmium, hexavalent chromium and two bromide compounds were banned - for power supplies this means, first of all, a transition to lead-free solders. On the one hand, of course, we are all for the environment and against heavy metals - but, on the other hand, a sudden transition to the use of new materials can have very unpleasant consequences in the future. Thus, many are well aware of the story with Fujitsu MPG hard drives, in which the massive failure of Cirrus Logic controllers was caused by packaging them in cases made of the new “eco-friendly” compound from Sumitomo Bakelite: the components included in it contributed to the migration of copper and silver and the formation of jumpers between tracks inside the chip body, which led to almost guaranteed failure of the chip after a year or two of operation. The compound was discontinued, the participants in the story exchanged a bunch of lawsuits, and the owners of the data that died along with the hard drives could only watch what was happening.

Equipment used

Of course, the first priority when testing a power supply is to check its operation at various load powers, up to the maximum. For a long time, in various reviews, the authors used ordinary computers for this purpose, into which the unit under test was installed. This scheme had two main drawbacks: firstly, it is not possible to control the power consumed from the block in any flexible way, and secondly, it is difficult to adequately load blocks that have a large power reserve. The second problem has become especially pronounced in recent years, when power supply manufacturers began a real race for maximum power, as a result of which the capabilities of their products far exceeded the needs of a typical computer. Of course, we can say that since a computer does not require a power of more than 500 W, then there is little point in testing units at higher loads - on the other hand, since we generally began testing products with a higher rated power, it would be strange at least it is not possible to formally test their performance over the entire permissible load range.

To test power supplies in our laboratory, we use an adjustable load with software control. The system relies on a well-known property of insulated gate field-effect transistors (MOSFETs): they limit the current flow through the drain-source circuit depending on the gate voltage.

Shown above is the simplest circuit of a current stabilizer on a field-effect transistor: by connecting the circuit to a power supply with an output voltage of +V and rotating the knob of variable resistor R1, we change the voltage at the gate of transistor VT1, thereby changing the current I flowing through it - from zero to maximum ( determined by the characteristics of the transistor and/or the power supply being tested).

However, such a scheme is not very perfect: when the transistor heats up, its characteristics will “float”, which means that the current I will also change, although the control voltage at the gate will remain constant. To combat this problem, you need to add a second resistor R2 and an operational amplifier DA1 to the circuit:

When the transistor is on, current I flows through its drain-source circuit and resistor R2. The voltage at the latter is equal, according to Ohm's law, U=R2*I. From the resistor this voltage is supplied to the inverting input of the operational amplifier DA1; the non-inverting input of the same op-amp receives the control voltage U1 from the variable resistor R1. The properties of any operational amplifier are such that when turned on in this way, it tries to maintain the voltage at its inputs the same; it does this by changing its output voltage, which in our circuit goes to the gate of the field-effect transistor and, accordingly, regulates the current flowing through it.

Let’s say resistance R2 = 1 Ohm, and we set the voltage at resistor R1 to 1 V: then the op-amp will change its output voltage so that resistor R2 also drops 1 volt - accordingly, current I will be set equal to 1 V / 1 Ohm = 1 A. If we set R1 to a voltage of 2 V, the op-amp will respond by setting the current I = 2 A, and so on. If the current I and, accordingly, the voltage across resistor R2 change due to the heating of the transistor, the op-amp will immediately adjust its output voltage so as to return them back.

As you can see, we have received an excellent controlled load, which allows you to smoothly, by turning one knob, change the current in the range from zero to maximum, and once set, its value is automatically maintained for as long as desired, and at the same time it is also very compact. Such a scheme, of course, is an order of magnitude more convenient than a bulky set of low-resistance resistors connected in groups to the power supply being tested.

The maximum power dissipated by a transistor is determined by its thermal resistance, the maximum permissible temperature of the crystal and the temperature of the radiator on which it is installed. Our installation uses International Rectifier IRFP264N transistors (PDF, 168 kbytes) with a permissible crystal temperature of 175 °C and a crystal-to-heatsink thermal resistance of 0.63 °C/W, and the cooling system of the installation allows us to keep the temperature of the radiator under the transistor within 80 °C (yes, the fans required for this are quite noisy...). Thus, the maximum power dissipated by one transistor is (175-80)/0.63 = 150 W. To achieve the required power, parallel connection of several loads described above is used, the control signal to which is supplied from the same DAC; You can also use parallel connection of two transistors with one op-amp, in which case the maximum power dissipation increases by one and a half times compared to one transistor.

There is only one step left to a fully automated test bench: replace the variable resistor with a computer-controlled DAC - and we will be able to adjust the load programmatically. By connecting several such loads to a multi-channel DAC and immediately installing a multi-channel ADC that measures the output voltages of the unit under test in real time, we will get a full-fledged test system for testing computer power supplies over the entire range of permissible loads and any combinations of them:

The photo above shows our test system in its current form. On the top two blocks of radiators, cooled by powerful fans of standard size 120x120x38 mm, there are load transistors for 12-volt channels; a more modest radiator cools the load transistors of the +5 V and +3.3 V channels, and in the gray block, connected by a cable to the LPT port of the control computer, the above-mentioned DAC, ADC and related electronics are located. With dimensions of 290x270x200 mm, it allows you to test power supplies with a power of up to 1350 W (up to 1100 W on the +12 V bus and up to 250 W on the +5 V and +3.3 V buses).

To control the stand and automate some tests, a special program was written, a screenshot of which is presented above. It allows:

manually set the load on each of the four available channels:

first channel +12 V, from 0 to 44 A;
second channel +12 V, from 0 to 48 A;
channel +5 V, from 0 to 35 A;
channel +3.3 V, from 0 to 25 A;

monitor the voltage of the tested power supply on the specified buses in real time;
automatically measure and plot cross-load characteristics (CLC) for a specified power supply;
automatically measure and plot graphs of the efficiency and power factor of the unit depending on the load;
in semi-automatic mode, build graphs of the dependence of unit fan speeds on load;
calibrate the installation in semi-automatic mode in order to obtain the most accurate results.

Of particular value, of course, is the automatic construction of KNH graphs: they require measuring the output voltages of the unit for all combinations of loads permissible for it, which means a very large number of measurements - to carry out such a test manually would require a fair amount of perseverance and an excess of free time. The program, based on the passport characteristics of the block entered into it, builds a map of the permissible loads for it and then goes through it at a given interval, at each step measuring the voltages generated by the block and plotting them on a graph; the entire process takes from 15 to 30 minutes, depending on the power of the unit and the measurement step - and, most importantly, does not require human intervention.

Efficiency and power factor measurements

To measure the efficiency of the unit and its power factor, additional equipment is used: the unit under test is connected to a 220 V network through a shunt, and a Velleman PCSU1000 oscilloscope is connected to the shunt. Accordingly, on its screen we see an oscillogram of the current consumed by the unit, which means we can calculate the power it consumes from the network, and knowing the load power we have installed on the unit, its efficiency. The measurements are carried out in a fully automatic mode: the PSUCheck program described above can receive all the necessary data directly from the oscilloscope software, which is connected to a computer via a USB interface.

To ensure maximum accuracy of the result, the output power of the unit is measured taking into account fluctuations in its voltages: say, if under a load of 10 A the output voltage of the +12 V bus drops to 11.7 V, then the corresponding term when calculating the efficiency will be equal to 10 A * 11.7 V = 117 W.

Oscilloscope Velleman PCSU1000

The same oscilloscope is also used to measure the ripple range of the power supply's output voltages. Measurements are made on the +5 V, +12 V and +3.3 V buses at the maximum permissible load on the unit, the oscilloscope is connected using a differential circuit with two shunt capacitors (this is the connection recommended in ATX Power Supply Design Guide):

Peak-to-peak measurement

The oscilloscope used is a two-channel one; accordingly, the ripple amplitude can be measured on only one bus at a time. To get a complete picture, we repeat the measurements three times, and the three resulting oscillograms - one for each of the three monitored buses - are combined into one picture:

The oscilloscope settings are indicated in the lower left corner of the picture: in this case, the vertical scale is 50 mV/div, and the horizontal scale is 10 μs/div. As a rule, the vertical scale is unchanged in all our measurements, but the horizontal scale can change - some blocks have low-frequency ripples at the output, for which we present another oscillogram, with a horizontal scale of 2 ms/div.

The speed of the unit's fans - depending on the load on it - is measured in a semi-automatic mode: the Velleman DTO2234 optical tachometer we use does not have an interface with a computer, so its readings have to be entered manually. During this process, the load power on the unit changes in steps from 50 W to the maximum permissible; at each step, the unit is kept for at least 20 minutes, after which the rotation speed of its fan is measured.

At the same time, we measure the increase in temperature of the air passing through the block. Measurements are carried out using a Fluke 54 II two-channel thermocouple thermometer, one of the sensors of which determines the air temperature in the room, and the other - the temperature of the air leaving the power supply. For greater repeatability of results, we attach the second sensor to a special stand with a fixed height and distance to the unit - thus, in all tests, the sensor is in the same position relative to the power supply, which ensures equal conditions for all testing participants.

The final graph simultaneously displays the fan speeds and the difference in air temperatures - this allows, in some cases, to better assess the nuances of the operation of the unit’s cooling system.

If necessary, a Uni-Trend UT70D digital multimeter is used to control the accuracy of measurements and calibrate the installation. The installation is calibrated by an arbitrary number of measurement points located in arbitrary sections of the available range - in other words, for voltage calibration, an adjustable power supply is connected to it, the output voltage of which changes in small steps from 1...2 V to the maximum measured by the installation on a given channel. At each step, the exact voltage value shown by the multimeter is entered into the installation control program, based on which the program calculates the correction table. This calibration method allows for good measurement accuracy over the entire available range of values.

List of changes in testing methodology

10/30/2007 – first version of the article