Availability of pfc. PFC: Operating a power supply with active PFC in conjunction with cheap UPSs that produce a step signal when operating on battery power can lead to computer malfunctions

Choosing a power supply for your computer is not as easy as it might seem. The stability and service life of computer components will depend on the choice of power supply, so it is worth taking this issue more seriously. In this article I will try to list the main points that will help you decide on choosing a reliable power supply.

Power.
At the output, the power supply provides the following voltages: +3.3 v, +5 v, +12 v and some auxiliary -12 v and + 5 VSB. The main load falls on the +12 V line.
Power (W - Watt) is calculated by the formula P = U x I, where U is the voltage (V - Volts) and I is the current (A - Ampere). Hence the conclusion, the greater the current along each line, the greater the power. But not everything is so simple, for example, with a heavy load on the combined line +3.3 v and +5 v, the power on the +12 v line may decrease. Let's look at an example based on the labeling of the Cooler Master RS-500-PSAP-J3 power supply - this is the first photo I found on the Internet.

It is indicated that the maximum total power on the +3.3V and +5V lines = 130W, and it is also indicated that the maximum power on the +12V line = 360W. Please note that two virtual lines +12V1 and +12V2 of 20 Amps each are indicated - this does not mean that the total current is 40A, since with a current of 40A and a voltage of 12V, the power would be 480W (12x40=480). In fact, the maximum possible current on each line is indicated. The real maximum current can be easily calculated using the formula I=P/U, I = 360 / 12 = 30 Ampere.
Also pay attention to the line below:
The +3.3V&+5V&+12Vtotaloutputshallnotexceed 427.9W– it turns out that the total power on all lines should not exceed 427.9W. As a result, we get not 490W (130 + 360), but only 427.9. Again, it is important to understand that if the load on the +3.3V and 5V lines is, say, 100W, then subtracting 100W from the maximum power, i.e. 427.9 – 100 = 327.9. As a result, we get 327.9W remaining on the +12V line. Of course, in modern computers the load on the +3.3V and +5V lines is unlikely to be more than 50-60W, so we can safely assume that the power on the +12V line will be 360W and the current 30A.

Power supply power calculation.
To calculate the power of the power supply, you can use this calculator http://www.extreme.outervision.com/psucalculatorlite.jsp, the service is in English, but I think you can figure it out.
From my own experience I can say that for anyone office computer A 300W power supply is quite sufficient. For a gaming game, a power supply of 400 - 500W is enough, for the most powerful gaming ones with a very powerful video card or two in the mode SLI or Crossfire– a 600 - 700W unit is required.
The processor usually consumes from 35 to 135W, video card from 30 to 340W, motherboard 30-40W, 1 memory stick 3-5W, HDD 10-20W. Also keep in mind that the main load falls on the 12V line. Yes, and don’t forget to add a margin of 20-30% for the future.

Efficiency
The efficiency of the power supply will not be unimportant. Efficiency (coefficient useful action) is the ratio of output power to consumed power. If the power supply could convert electrical energy without losses, then its efficiency was 100%, but this is not yet possible.
Let me give you an example: in order for a power supply with an efficiency of 80% to provide an output power of 400W, it must consume no more than 500W from the network. The same power supply, but with 70% efficiency, will consume about 571W. Again, if the power supply is not heavily loaded, for example at 200W, then it will also consume less from the network, 250W at 80% efficiency and approximately 286 at 70% efficiency.
There is an organization that tests power supplies to meet a certain level of certification. Certification 80 Plus was carried out only for a 115V electrical network common, for example in the USA. Starting from level 80 Plus Bronze, power supplies are tested for use in a 230V power supply. For example, to pass level certification 80 PlusBronze The efficiency of the power supply should be 81% at 20% load, 85% at 50% load and 81% at 100% load.

The presence of one of the logos on the power supply indicates that the power supply meets a certain level of certification.
Pros of a power supply with high efficiency:
Firstly, less energy is released in the form of heat, so the cooling system of the power supply needs to remove less heat, and therefore there is less noise from the fan. Secondly, small savings on electricity. Thirdly, the quality of the BP data is high.

Active or passive PFC?

PFC (Power Factor Correction) – Power factor correction. The power factor is the ratio of active power to total power (active + reactive).

Since the real load usually also has inductive and capacitive components, reactive power is added to the active power. The load does not consume reactive power - received during one half-cycle of the mains voltage, it is completely returned to the network during the next half-cycle, wasting the supply wires. It turns out that reactive power is of no use, and it is dealt with whenever possible using various corrective devices.

PFC - can be passive or active.

Advantages of active PFC:

Active PFC provides a power factor close to ideal (active 0.95-0.98 versus 0.75 passive).
Active PFC stabilizes the input voltage of the main stabilizer, the power supply becomes less sensitive to low mains voltage.
Active PFC improves the response of the power supply during short-term dips in mains voltage.

Disadvantages of active PFC:

Reduces the reliability of the power supply, as the design of the power supply itself becomes more complicated. Additional cooling required. Overall, the benefits of active PFC outweigh its disadvantages.

In principle, you can ignore the PFC type. In any case, when you buy a power supply with a lower power, it will most likely have a passive PFC; when you buy a more powerful unit from 500 W, you will most likely get a unit with an active PFC.

Power supply cooling system.
The presence of a fan in the power supply is considered normal; its diameter is most often 120, 135 or 140 mm.

Cables and connectors.
Pay attention to the number of connectors and the length of the cables coming from the power supply; depending on the height of the case, you need to select a power supply with cables of the appropriate length. For a small body, a length of 40-45 cm is sufficient.

A modern power supply has the following connectors:

	24-pin power connector motherboard. Usually separate 20 and 4 contacts, sometimes solid.
	CPU socket. Usually 4-pin, for more powerful processors 8-pin is used.
	Connector for additional food video cards. 6 and 8 pin. 8-pin sometimes combined 6+2 contacts.
	SATA connector for connecting hard disks and optical drives.
	4-pin connector (Molex) for connection old IDEs hard drives and optical drives are also used to connect fans.
	4-pin connector for connecting FDD drives.

Modular cables and connectors.
Many higher-power power supplies now use modular cable connections with connectors. This is convenient because there is no need to keep unused cables inside the case, and there is less confusion with wires; we simply add them as needed. The absence of unnecessary cables also improves air circulation in the case. Typically, these power supplies only have non-removable connectors for powering the motherboard and processor.

Manufacturers.
Power supply manufacturers are divided into three groups:

1. They produce their own products - these are brands such as FSP, Enermax, HEC, Seasonic, Delta, Hipro.
2. They manufacture their products, partially shifting production to other companies, for example Corsair, Antec, Silverstone, PC Power & Cooling, Zalman.
3. They resell under their own brand (some influence the quality and choice of components, some do not), for example Chiftec, Cooler Master, Gigabyte, OCZ, Thermaltake.

You can safely purchase products from these brands. On the Internet you can find reviews and tests of many power supplies and navigate through them.
I hope this article will help you answer the question “ how to choose a power supply for a computer?».

A little about power

Don't worry, you don't need any university knowledge of physics to understand how it works. We will simply explain the difference between a good power supply and a bad one. If you know the basic principles of operation, you are unlikely to make a bad purchase. So, let's move on.

Reactive current and reactive power

One of the important issues regarding power consumption when using switching power supplies is the "reactive" current caused by inductance. Please note that standby power consumption has nothing to do with idle mode. In addition, the load in this case does not overlap in any way with the power consumption at full load, but uses the same components. Reactive power must be significantly reduced (in the best case, it should not exist at all) so that it does not lead to loss of energy in the resistance, which will be released in the form of heat. Such wasteful energy consumption should be reduced to almost zero internal circuits switching power supplies.

Effective power and apparent power

Effective power is the opposite of reactive power in that it reflects actual power consumption. Apparent power is the sum of active and reactive power.

Power factor

This indicator is calculated as the ratio between effective power and apparent power and is between 0 (worst result) and 1 (ideal result). So, when buying a power supply, you need to make sure that it has a high power factor: this is one of the key quality indicators for power supplies.

Active PFC

Active Power Factor Correction (PFC) means active power factor correction. The power factor is important characteristic for the power supply, since it reflects the ratio between active and apparent power.

Advantages:

• An active power of about 99% can be considered ideal;
• High efficiency (less at low loads);
• Very stable power supply;
• Less energy consumption;
• Less heat generation;
• Less weight.

Flaws:

• Costs more;
• High probability of failure.

Passive PFC

With passive power factor correction, reactive currents can be reduced by using large inductors. This method is simpler and cheaper, but it is not the most effective.

Advantages:

• Costs less;
• No electromagnetic interference.

Flaws:

• Better cooling required;
• Not suitable for high loads;
• High energy consumption (energy loss);
• Heavier;
• Low active power (approximately 70% to 80%).

How to determine the efficiency of a power supply?

Basic principles, rules and regulations

One of the key performance indicators of a power supply is whether it meets Energy Star 5.0 and 80 PLUS standards. The latter will be a priority for computing and is a standard recognized throughout the world. In addition, if we are talking about European countries, then it is also necessary to check compliance with CE and ErP standards.

80 PLUS power supplies are more efficient.

The principles and specifications naturally influence the efficiency and quality of the food. A power supply marked with 80 PLUS certification will meet certain requirements, which are established through a set of tests. We would like to mention that the 80 PLUS stress testing conditions do not directly correspond to the ATX specification, however, they are performed under lower voltage US power grid conditions. In the conditions of Russia and Europe, with 230 V networks, the efficiency of 80 PLUS power supplies will be slightly higher than in the USA.

The 80 PLUS concept has been expanded to include several performance levels, Platinum, Gold, Silver and Bronze, and each of these standards has its own set of specifications. Thus, an 80 PLUS Platinum or 80 PLUS Gold power supply will be more efficient than a regular power supply. At the same time, these power supplies are more expensive.

Using the table below, you can see how a device's specification level affects its performance under a given load and evaluate each specific specification level.

	Efficiency at 20% load	Efficiency at 50% load	Efficiency at 100% load
80Plus	80,00%	80,00%	80,00%
80 Plus Bronze	82,00%	85,00%	82,00%
80 Plus Silver	85,00%	88,00%	85,00%
80 Plus Gold	87,00%	90,00%	87,00%
80 Plus Platinum	90,00%	92,00%	89,00%

Power consumption when the computer is turned off

When turning off the computer? The power supply usually continues to work. This is necessary to support some features like Wake-on-LAN. The power supply will waste some power even when the computer is turned off. Modern power supplies, especially those sold in Europe, according to manufacturers, consume no more than 1 W in this mode. If saving is really important to you, then this decision will be right.

CONTENT

Already long time Our laboratory tests ATX standard power supplies. All this time, the testing methodology has been continuously developing and improving, pursuing two goals at once - not only to be able to objectively compare different power supplies, but also to do it quite clearly.

Unfortunately, one of the main tests of our method - measuring voltage stability - could not boast of clarity, because it used its own load patterns for almost each unit, which made it impossible to discuss and compare the results of different power supplies without constant reference to the features applied to them patterns. In other words, the results of each of the blocks carried with them a bunch of conventions and reservations - of course, comparison was possible in the end, otherwise there would be no point in testing at all, however direct Comparison of figures or graphs, alas, was made very difficult by these reservations.

With this article, I present to you a new method for testing power supplies, which has replaced the old method of measuring voltage stability and gives an extremely visual and at the same time very accurate and objective result, equally well suited for comparing different power supplies, both in specific numbers and simply " by eye", based on the appearance of the resulting graphs. The basis is the methodology for constructing the so-called cross-load characteristics of power supplies, developed and applied by our colleagues from the ITC Online publication, but it has been significantly improved in order to further increase both information content and clarity.

Also in the article I will describe in more or less detail various aspects of the operation of computer power supplies, so that readers who do not understand the circuitry of switching power supplies will understand what these or those power supply parameters measured during testing mean and where they come from. Those of you who are quite familiar with the design and operation of switching power supplies can immediately scroll through the first two sections of the article to a description of the test equipment and testing methodology that we actually use.

Linear and switching power supplies

As is known, an electronic power source is a device that in one way or another solves the problem of changing, controlling or stabilizing the electrical power supplied to the load.

The simplest and still widely used control method is the absorption of excess power in the control device, that is, its banal dissipation in the form of heat. Power supplies operating on this principle are called linear.


Above is a diagram of such a source - a linear voltage stabilizer. The 220V household voltage is reduced by transformer T1 to the required level, after which it is rectified by diode bridge D1. Obviously, the rectified voltage must be higher than the output voltage of the stabilizer under any conditions - in other words, excess power is required; This follows from the very principle of operation of a linear stabilizer. In this case, this power is released in the form of heat on the transistor Q1, which is controlled by some circuit U1 so that output voltage Uout was at the required level.

This scheme has two significant drawbacks. Firstly, the low frequency of alternating current in the supply network (50 or 60 Hz, depending on the country) determines the large overall dimensions and weight of the step-down transformer - a transformer with a power of 200-300 W will weigh several kilograms (not to mention the fact that in linear stabilizers it is necessary to use transformers with a power twice greater than the maximum load power, because the efficiency of a linear stabilizer is about 50%, and the transformer must be designed for full power, including that which goes into heat on the stabilizer itself). Secondly, the voltage at the output of the transformer must in all cases exceed the sum of the output voltage of the stabilizer and the minimum voltage drop across the control transistor; This means that, in general, the transistor will have to dissipate quite a noticeable excess power, which will negatively affect the efficiency of the entire device.

To overcome these shortcomings, so-called switching voltage stabilizers have been developed, in which power control occurs without power dissipation in the control device itself. In its simplest form, such a device can be represented as an ordinary switch (the role of which can also be played by a transistor), connected in series with the load. In such a scheme average the current flowing through the load depends not only on the load resistance and supply voltage, but also on the switching frequency of the switch - the higher it is, the higher the current. Thus, by changing the switching frequency, we can regulate the average current through the load, and ideally, no power will be dissipated at all on the switch itself - since it is in only two states: either completely open or completely closed. In the first case, the voltage drop across it is zero, in the second case, the current flowing through it is equal to zero, and then the power released on it, equal to the product of the current and the voltage, is also always zero. In reality, of course, everything is a little different - if transistors are used as a switch, firstly, even in the open state, a small voltage drops on them, and secondly, the switching process does not occur instantly. However, these losses are a consequence of side effects, and they are much less than the excess power allocated to the linear stabilizer control device.

If we compare the numbers, then Typical efficiency a linear stabilizer is 25...50%, while the efficiency of a pulsed stabilizer can exceed 90%.

In addition, if we put a switch in the pulse stabilizer before the step-down transformer (obviously, in general, it makes no difference whether to regulate the input or output voltage of the transformer - they are inextricably linked with each other), then we get the opportunity to determine the operating frequency of the transformer regardless on the frequency of the supply network. And since the dimensions of the transformer decrease with an increase in its operating frequency, this makes it possible to use step-down transformers in pulse stabilizers that are literally toy-sized compared to their linear counterparts, which gives a colossal gain in the size of the finished device. For example, a transformer with a frequency of 50 Hz and a power of 100 W weighs just over two kilograms, while a transformer with the same power, but with a frequency of 35 kHz weighs only about 35 grams. This, of course, radically affects the dimensions and weight of the entire power source - if we calculate the ratio of the output power of the source to its volume, then for a switching power supply operating at a frequency of several tens of kilohertz, it will be approximately 4-5 W/cu. inch, while for a linear stabilizer this figure is only 0.3...1 W/cu. inch. Moreover, with increasing frequency, the power density of a switching power supply can reach up to 75 W/m3. inch, which is completely unattainable for linear sources even with water cooling (figures are given from the book by Irving M. Gottlieb “Power Supplies. Inverters, Converters, Linear and Switching Stabilizers”).

In addition, with this design, the pulse stabilizer depends much less on the value of the input voltage - after all, the step-down transformer is primarily sensitive to this, and when we turn on the switch before it, we can control the voltage and frequency of its operation as we need it. Accordingly, switching stabilizers can withstand supply voltage drops of up to 20% of the nominal value without any problems, while linear ones can only achieve operation at reduced network voltage by further reducing the already low efficiency.

In addition to the transformer, the use high frequency allows you to greatly (tens of times) reduce the capacitance and, accordingly, the dimensions of the smoothing capacitors (C1 and C2 in the above diagram). True, this is a double-edged sword - firstly, not all electrolytic capacitors are able to operate normally at such a frequency, secondly, in spite of everything, in a switching power supply it is technically very difficult to obtain an output ripple range below 20 mV, while in linear ones, if necessary, without special costs, the ripple level can be reduced to 5 mV, and even lower.

It is obvious that a converter operating at a frequency of several tens of kilohertz is a source of interference not only into its own load, but also into the supply network, as well as simply into the radio air. Therefore, when designing switching power supplies, it is necessary to pay attention to both the filter at its input (contrary to popular belief, it does not so much protect the power supply from external interference, but rather protects other devices from interference created by this power supply), and the electromagnetic shielding of the power supply itself , which in the case of heavy-duty units means using a steel casing. Linear power supplies, as I noted above, although they are more sensitive to external interference, do not create any interference themselves, and therefore do not require any special measures to protect surrounding equipment.

In addition, switching power supplies require significantly more complex (and, therefore, expensive) electronics than their linear counterparts. The price advantage of pulse units is obvious for fairly powerful products, where the price is primarily determined by the cost power transformer and the necessary heat removal, and therefore linear sources with their large dimensions and low efficiency are a clear loser; however, as the components of switching power supplies become cheaper, they are increasingly crowding out low-power linear sources - for example, switching power supplies with a power of several watts are no longer uncommon (for example, mobile phone chargers), although a few years ago such powers had advantages line sources were obvious.

If we talk about tasks in which the determining parameter is dimensions, then switching power supplies are out of competition - with all the design tricks, it is simply impossible to obtain the same power density from a linear source as from a pulsed one.

Computer power supplies

Currently, all power supplies used in computers are switching. This is due to the fact that to ensure reasonable dimensions and heat dissipation, power density and efficiency are required, which are fundamentally unattainable for linear power supplies of such power - for example, the power density of a conventional ATX power supply is 2...5 W/cu. inch (depending on its output power), and efficiency is at least 68% when working with maximum load.

The figure above is a somewhat simplified block diagram of a typical computer power supply. Below, using the Macropower MP-300AR unit as an example, the typical arrangement of components in a real power supply is shown (in most units of other models there will be no significant differences):

The 220V supply voltage passes through a two- or three-section filter, which protects other devices connected to the network from interference generated by the power supply. After the filter, the voltage is supplied to rectifier D1, and from it to an optional (but increasingly common in new units) power factor correction circuit (PFC - Power Factor Correction). More details about what PFC is and why it is needed will be discussed below. Now I would like to dwell in more detail on the filter, because there are a couple of questions associated with it that are often asked by users.

Power supply without PFC

In the oscillogram above, the green “beam” is the mains voltage, and the yellow one is the current consumed by the mains power supply. With this picture, the power factor turns out to be approximately 0.7 - that is, almost a third of the power only heats the wires to no avail, without producing any useful work. And if for private users this figure is not of great importance, because residential electricity meters take into account only active power, then for large offices and in general any premises where many computers are working simultaneously, a low power factor represents a noticeable problem, because all electrical wiring and related equipment must be calculated based on the total power - in other words, with a power factor of 0.7, it should be a third more powerful than it could be if the power supply did not consume reactive power. The low power factor also affects the choice of uninterruptible power supplies - for them, the limitation is again the total, not the active power.

Accordingly, in Lately Power factor correction (PFC) devices are becoming increasingly popular. The simplest and therefore most common is the so-called passive PFC, which is a conventional inductor of relatively high inductance, connected to the network in series with the power supply.

Passive PFC Power Supply

Power supply with active PFC

As you can see, the shape of the current consumed by a power supply with an active PFC differs very little from that of a conventional resistive load - the resulting power factor of such a unit can reach 0.95...0.98 when operating at full load. True, as the load decreases, the power factor decreases, at a minimum dropping to approximately 0.7...0.75 - that is, to the level of units with passive PFC. However, it should be noted that the peak values ​​of current consumption for blocks with active PFC are still, even at low power, noticeably less than for all other blocks.

The graph below shows the result of an experimental measurement of the dependence of the power factor on the load on the power supply for three units - without PFC at all, with passive PFC and, finally, with active PFC.

Not only does active PFC provide near-ideal power factor, but unlike passive PFC, it improves the performance of the power supply. Firstly, it additionally stabilizes the input voltage of the main stabilizer of the unit - not only does the unit become noticeably less sensitive to low mains voltage, but also when using an active PFC, units with a universal power supply of 110...230V are quite easily developed, which do not require manual switching mains voltage. Secondly, the use of active PFC improves the response of the power supply during short-term (fractions of a second) dips in the mains voltage - at such moments the unit operates using the energy of the high-voltage rectifier capacitors C1 and C2, and this energy is proportional to the square of the voltage across them; as I noted above, when using an active PFC, this voltage reaches 400V versus the usual 310V - therefore, the efficiency of using capacitors more than doubles (due to the fact that the energy stored in capacitors is far from completely exhausted, the efficiency grows even faster than the square voltage on capacitors).

In fact, active PFC has only two drawbacks - firstly, like any design complication in general, it reduces the reliability of the power supply, and secondly, it also has an efficiency other than 100%, and therefore requires cooling (however, on the other hand , active PFC slightly reduces losses in the input filter and in the inverter itself, so that there is no overall drop in the efficiency of the unit). However, the benefits of using an active PFC outweigh these disadvantages in the vast majority of cases.

So, if you need a unit with power factor correction, then you should first of all pay attention to models with active PFC - only they provide a really good power factor, while also significantly improving other characteristics of the power supply. From the point of view of home users, units with active PFC will be useful for owners of low-power UPSs: let’s say you already have a UPS with a capacity of 500 VA, of which 50 VA is consumed by the LCD monitor, and 450 VA remains for the system unit, and you are going to upgrade the latter is up to the modern level - and a fairly serious modern configuration may well consume up to 300 W from the power supply at maximum load. In this case, on a power supply with a power factor of 0.7 and an efficiency of 80% (this is a fairly typical figure for. good block) we get the total power consumed from the network 300/(0.75*0.8) = 500 VA, and on the same block with a power factor of 0.95 - respectively, 300/(0.95*0.8) = 395 VA. As you can see, in the case of a power supply without PFC, replacing the UPS with a more powerful one is inevitable, otherwise in the event of a power outage at the wrong moment, the current one simply cannot cope with the load, and in the case of a unit with an active PFC, there is even still a small reserve of 55 VA In a good way, of course, in this calculation one must also take into account the fact that at the output of inexpensive UPS the voltage is not sinusoidal, but trapezoidal - however, only the absolute numbers obtained will change, while the advantage of the power supply with active PFC will remain.

And in conclusion of this section, I would like to dispel one myth associated with PFC: many users confuse power factor and efficiency, while they are completely different quantities. Efficiency, by definition, is equal to the ratio of the output power of the power supply to the active power it consumes from the network, while the power factor is the ratio of the active power consumed from the network to the total power consumed from the network. Installing a PFC circuit in the power supply affects the active power it consumes only indirectly - due to the fact that the PFC itself consumes some power, plus the input voltage of the main stabilizer changes; The main task of PFC is to reduce the reactive power consumed by the unit, which is not taken into account in the calculation of efficiency. Therefore, there is no direct connection between efficiency and power factor.

Power supply testing stand

The main stand for testing power supplies in our laboratory is a semi-automatic installation that allows you to set the required load on the +5V, +12V, +3.3V and +5V buses of the standby mode of the unit under test, while simultaneously measuring the corresponding output voltages.

The hardware of the installation is based on a 4-channel Maxim MX7226 DAC, the outputs of which are connected to current sources. The latter are made using LM324D operational amplifiers and powerful field effect transistors IRFP064N installed on radiators with forced air cooling.

Each of the transistors has a maximum power dissipation of 200 W, and since three such transistors are used in each of the most powerful load channels (+5V and +12V), the installation allows you to test any currently existing ATX power supplies, up to the most powerful – even taking into account the decrease in the permissible power dissipation of transistors as their temperature increases, the permissible load power for each channel is at least 400 W.

To measure the set load currents and output voltages of the unit under test, the installation uses two 4-channel Maxim MX7824 ADCs - one ADC is responsible for currents, the other for voltages.

All control of the installation, from turning on the power supply under test and ending with carrying out all possible tests, as well as recording and processing their results, is carried out from the computer via LPT port. A program was written specifically for these purposes, which allows you to manually set the load current independently on each of the buses, as well as perform some standard tests of power supplies (for example, constructing a cross-load characteristic, which will be discussed below) in full automatic mode.

In addition to the main installation, two auxiliary devices are also used to test the blocks. First, it is a square pulse generator with a frequency discretely variable from 60 Hz to 40 kHz:

The generator is connected to the power supply under test in the form of a load - using a switch you can choose whether it will be connected to the +12V bus or to +5V, in both cases the peak current of the load it creates is about 1.3 A. This allows you to estimate how much a well-tested power supply responds to relatively powerful impulses rectangular loads, following with frequencies from tens of hertz to tens of kilohertz.

Secondly, to take oscillograms of the current consumed by the power supply and, at the same time, the supply voltage, a conventional shunt with powerful wirewound resistors total resistance of about 0.61 Ohm:

When testing the power supply, the probes of a digital two-channel oscilloscope are connected to this board - one of its channels records an oscillogram of the mains voltage, and the other - an oscillogram of the current consumed by the power supply. Next, the resulting oscillograms are processed by a specially written software for this purpose. small program, which immediately calculates all the parameters we are interested in - the active, reactive and apparent power consumed by it and, accordingly, the power factor and efficiency of the power supply.

To take oscillograms, a digital two-channel “virtual” oscilloscope is used (virtuality in this case means that this oscilloscope is a board installed in a computer and, unlike conventional oscilloscopes, cannot work without a computer, because it does not have its own hardware for controlling and displaying information ) M221 manufactured by the Slovak company ETC. The oscilloscope has an analog bandwidth of 100 MHz, a maximum arbitrary signal digitization speed of 20 million samples per second, and a sensitivity of 50 mV/div to 10 V/div. In addition to measuring the efficiency and power factor of the tested power supplies, the oscilloscope is used to evaluate the range, shape and frequency composition of the power supply output voltage ripple.

To quickly assess currents and voltages during the testing process, as well as for periodic testing of other measuring equipment, our laboratory uses the Uni-Trend UT70D multimeter, which allows you to measure currents and voltages with very good accuracy, including non-sinusoidal ones, which is very important when testing power supplies without power factor correction - many meters that are not marked "TrueRMS" are not able to adequately measure alternating currents and voltages whose shape is different from a sine wave.

To measure the temperature inside the power supply, we use a Fluke 54 Series II digital thermometer with thermocouples 80PK-1 and 80PK-3A (the names of all models are given in the Fluke catalog). Unfortunately, the non-contact infrared digital thermometer we had showed unsatisfactory measurement accuracy on shiny metal surfaces (for example, on aluminum heatsinks of power supplies), which forced us to switch to using a thermocouple thermometer.

To measure the speed of power supply fans, an optical tachometer Velleman DTO2234 is used. It allows you to measure the speed of a fan in a closed power supply without the slightest problem, that is, without disturbing its natural thermal regime - you just need to stick a thin strip of reflective material on one of the fan blades.

And finally, to provide all power supplies with the same mains voltage, regardless of its daily fluctuations, as well as to provide the ability to test units at high or low supply voltages, they are connected to the network through a Wusley TDGC2-2000 laboratory autotransformer with a permissible load power of up to 2 kW and voltage regulation limits from 0 to 250V.

Methodology for testing power supplies

The first and most important test for any power supply is the construction of the so-called cross-load characteristic. As I already said in the theoretical part of the article, each output voltage of the power supply depends on the load not only on the corresponding bus, but also on the loads on all other buses.

The ATX standard provides maximum permissible deviations output voltages from the nominal - this is 5% for all positive output voltages (+12V, +5V and +3.3V) and 10% for negative output voltages (-5V and -12V, of which, however, only the last one remains in modern units ). The cross-load characteristic (CLC) of a block is that area of ​​load combinations in which none of the output voltages goes beyond the permissible limits.

The power supply system is constructed in the form of an area on a plane, where the load on the +12V bus is plotted along the horizontal coordinate axis, and the total load on the +5V and +3.3V bus is plotted along the vertical coordinate axis. When building a power supply unit, an installation for testing power supplies in a fully automatic mode changes the load on these buses in steps of 5 W and, if all output voltages of the unit are at this step fit within the given framework, places a point on the plane, the color of which - from green to red - corresponds to the deviation of each of the voltages at a given point from the nominal value. Since the setup we use controls three main output voltages, for each power supply we get, respectively, three graphs (for each voltage), in which the same area will be shaded in different colors. The shape of the area on all three is the same, since it is determined not for each of the voltages separately, but for all of them together, and the deviation is beyond the permissible limits any of voltages means that the corresponding point will not be on the graphs for everyone stress; The shading of the area is different because it is constructed individually for each voltage. Below is an example of a power supply system for a Macropower MP-360AR Ver. 2, colored in accordance with voltage deviations +12V (in the articles I will provide animated pictures in which all three voltages will be shown in turn, the current voltage is indicated in the upper right corner of the graph, above the color scale):

On this graph, each point strictly corresponds to one measurement step, and for convenience during the measurement process, points where the voltages are outside the permissible limits are indicated in gray and smaller size– this is necessary for the convenience of the experimenter monitoring the progress of measurements in real time. After the measurements are completed, the obtained data is processed using bilinear interpolation - so instead of individual points, a shaded area with clear edges is more convenient for perception:

So what do we see in this graph? The tested power supply copes remarkably well with the load on the +12V bus - it is capable of delivering the required voltages at maximum load on this bus and only 5W on the +5V bus (5W is a typical initial value in our measurements; for powerful units that operate unstably at such light loads, it increases to 15 W or 25 W).

A smooth vertical border in the lower right part of the graph means that here the unit has reached the power limit of the +12V bus (for this unit it is 300W), and the installation did not increase the load current further in order to avoid failure of the power supply. Above, the vertical boundary turns into an inclined one (upper right corner of the graph) - this is the area where the installation reached the maximum power of the power supply (in this case it is 340W), and therefore, as the load increased by +5V, it was forced to reduce the load by +12V, to again prevent the power supply from failing or its protection from tripping.

We continue to go around the contour counterclockwise. At the top of the graph, the inclined line turns into a flat horizontal line - this is the area where the installation reached the maximum permissible load of +5V, and then did not increase the power on this bus any more, although the power supply produced voltages within normal limits.

And finally, in the upper left part of the graph we see an uneven sloping line, which is clearly not explained by the power limit - after all, the +12V load in this area is too small. But this line is perfectly explained by the red color of the graph - with a large load of +5V and a small load of +12V, the voltage on the +12V bus reached 5% deviation, thereby marking the boundary of the KNH.

Thus, from this graph we can say that this power supply maintains the level of output voltages well and allows you to get the declared power from it without any problems, but it will be preferable for the most modern systems powering both the processor and the video card from +12V, due to load imbalance in the direction of this bus it perceives better than a bias towards the +5V bus.

For comparison, let's look at the power supply of a significantly cheaper power supply - L&C LC-B300ATX with a stated power of 300W. The graph in this case is again built only for +12V voltage:

The differences from the MP-360AR are immediately obvious. Firstly, the bottom line of the contour is no longer horizontal - on the right side it begins to go up, and the red color shows that this was caused not only by the voltage going beyond +5V (which happens quite often with a heavy load of +12V), but also a voltage drop of +12V. Secondly, there is no upper horizontal “shelf” on the circuit; the top point of the graph corresponds to a +5V load of about 150W - which means that the maximum 180W promised by the manufacturer for this bus cannot be achieved in practice, in principle, under any load combinations. Thirdly, despite the higher declared power on the +5V and +3.3V rails compared to the MP-360AR (180W versus 130W), it is clearly visible that the inclined line in the upper left part of the graph for the MP-360AR began at the load power at +5V more than 80 W, while the LC-B300 only has about 50 W. This means that, despite the formally declared higher power on the +5V bus for the LC-B300 compared to the MP-360AR, in practice in many cases you can get more real power This bus will work just from a unit produced by Macropower.

I think attentive readers have already noticed that if both graphs are plotted on the same scale, the PCB of the Macropower block will turn out to be strongly elongated along the +12V axis compared to the PCB of the L&C block. This is explained by the fact that these two blocks belong to different versions of the ATX/ATX12V standard Power Supply, in which different load distribution between the power supply buses was considered preferable. For comparison, the figure below shows the CNCs that, according to Intel (as the compiler of the entire family of ATX standards), power supplies should have had in different years:

As you can see, initially the ATX standard assumed consumption mainly from the +5V and +3.3V buses - and indeed, almost the entire computer’s hardware was powered from these voltages; at +12V, a noticeable load was created only by the mechanics of hard drives and optical drives.

However, over time, the situation began to change - processors became more and more powerful, and powering them from +5V created a number of problems for motherboard developers. Firstly, at that time it was already clear that the increase in power consumption of processors would continue further, which would lead to a large current consumption of +5V, and therefore there would be a problem with supplying such currents to the motherboard - a standard connector may simply not cope. Secondly, the motherboard's power connector will either have to be squeezed in next to the processor's VRM, or a bus designed for high currents, which again is difficult...

In this regard, Intel proposed the ATX12V standard, according to which the processor must be powered from the +12V bus - obviously, with the same power consumption this means 2.4 times less current. However, since the main ATX connector has only one +12V wire, it was necessary to introduce an additional 4-pin ATX12V connector... however, with this Intel killed two birds with one stone - not only did it solve in advance the problem of burnt connector contacts due to too high load currents, but also simplified PCB design for motherboard manufacturers, because placing a small 4-pin connector directly next to the VRM is much easier than placing a larger 20-pin connector.

Unfortunately, AMD did not support Intel's initiative, and therefore many owners of motherboards for Socket A, of which, even among those currently on sale, 20-25% still do not have an ATX12V connector, fully experienced the problems reported by Intel spoke four years ago - with the advent of powerful processors for this platform, the first reports appeared about burnt-out contacts of the power supply, and about a strong imbalance in its output voltages (as you can see from the above-mentioned power supply voltages, even cheap units cope better with +12V loads )...

In fact, the only technical disadvantage from the introduction of ATX12V is a slight decrease in the efficiency of the VRM, because the efficiency of any pulse converter decreases with an increase in the difference between the input and output voltages. However, this was more than offset by an increase in the efficiency of the power supply itself - as for motherboard developers, for power supply developers, the decision to focus on the main consumption on the +12V bus greatly simplified the design of the units.

As you can see from the graphs, ATX12V versions up to and including 1.2 differed from regular ATX only in the increased permissible consumption on the +12V bus. More serious changes occurred in version 1.3 - for the first time in the entire history of the development of computer power supplies, it introduced the required permissible load on the +5V bus decreased, while the load on the +12V bus increased even more - in fact, the adaptation of power supplies to the most modern systems, in which fewer and fewer consumers remain on the +5V bus (processors have long been powered by +12V, and now video cards have followed suit). Unlike previous models, ATX12V 1.3 power supply is no longer required to support stable voltages with a large load at +5V and at a small load at +12V.

And finally latest version today is ATX12V 2.0. As you can easily see, the power of the power supply on the +5V bus has decreased even more - now it is only 130W; but the permissible load power at +12V has increased significantly. In addition, the ATX12V 2.0 units acquired a 24-pin motherboard power connector instead of the old 20-pin - if four years ago the old connector was no longer enough to power the processor, which is why ATX12V was invented, now the permissible current of the connector is no longer enough to power the processor PCI Express cards. Also, two +12V sources appeared in the ATX12V blocks, but in fact inside the block they are one source, only the protection trip current limits are separate - according to the safety requirements of the IEC-60950 standard, currents of more than 20A are not allowed on the +12V bus, which is why it is necessary split this tire into two parts. However, in cases where compliance with this standard is not required, manufacturers may simply not install the corresponding circuit - then an ATX12V 2.0 power supply with currents on the +12V buses, say, 10A and 15A, can be safely considered as a power supply with one +12V bus with current 25A.

So, if we return to the units discussed above, we can say that MP-360AR Ver. 2 complies with the ATX12V 2.0 standard, and LC-B300 complies with the ATX12V 1.2 standard, hence the difference in their PCBs. However, the reason, of course, is not only in formal compliance with different versions of the standard - remember how I complained that in practice it is impossible to obtain the declared +5V power from the LC-B300... and now let's superimpose the recommended Intel KNH on its graph for 300-watt ATX12V 1.2 blocks:

As you can see, the unit simply does not fit into the standard requirements for 300-watt models regarding the permissible load of +5V, so it can be considered as 300-watt only with the caveat that these watts are not very fair. For comparison, you can look at the graph of the same MP-360AR, but with the recommended PCB for 350-watt ATX12V 2.0 blocks:

As you can see, the match is almost perfect. I think there is no need to comment on the comparative quality of these two blocks.

Generally speaking, it is quite difficult to meet Intel’s very stringent requirements for power supply - there are not many units that can boast of this, however, such a gross violation of the recommendations, as in the case of the LC-B300, is rare.

Regarding the colors of KNH, we can say that the ideal, of course, is a uniform green color... however, the ideal, as we know, is usually unattainable. The situation is quite normal when each voltage, except for the fairly stable +3.3V, goes through the entire range from green or yellow-green at one end of the graph to red at the other; it also happens that there is no green color on the KNKh at all - this means that the voltage was initially overestimated. The worst situation is when any voltage passes through the entire range of colors twice - from red at one edge through green in the middle to red at the other edge of the circuit. This situation, for example, is visible in the LC-B300 discussed above and means that at one edge of the PCB the voltage has dropped significantly (obviously, with a small load of +5V and a large load of +12V, the latter can only drop), and on the other edge - vice versa , has grown a lot; in other words, its stability leaves much to be desired...

And, to complete the description of KNKh, I will give an example of an ideal power supply. Above, I already mentioned in passing the Antec and OCZ power supplies with separate auxiliary stabilizers on each of the main buses; below I bring to your attention the experimentally measured power supply voltage of the OCZ Technology PowerStream OCZ-470ADJ unit (this is already a full-fledged picture with all three voltages, the frame change period is 5 sec.):

As you can see, not only is all The KNH circuit is determined only by the permissible maximum load of the power supply, so not a single voltage even came close to a 5 percent deviation. Unfortunately, such power supplies are still relatively expensive...

Of course, testing power supplies does not end with the construction of power supply units. Firstly, all units are tested for stability under constant load from zero to maximum in 75 W increments. This way it is determined whether the block is even able to withstand the full load.
Secondly, as the load increases, the temperature of the diode assemblies of the unit and the fan rotation speed are measured, which in almost all modern power supplies depends on temperature in one way or another.

However, the results of temperature measurements should be treated with some skepticism - most power supplies have different designs of radiators and the location of diode assemblies on them, and therefore temperature measurements have a rather large error. However, in critical cases, when the power supply is on the verge of death from overheating (and this sometimes happens in the cheapest models), the thermometer readings can be interesting - for example, in my practice there were units in which the radiators became hotter under full load hundreds of degrees.

More interesting are the fan speed measurements - despite the fact that all manufacturers claim them temperature regulation, practical implementation may vary greatly. As a rule, for blocks of the lower price range the initial fan speed is already about 2000...2200 rpm. and as it warms up it changes only by 10...15%, while for high-quality blocks the initial speed can be only 1000...1400 rpm, doubling when warmed up at full power. Obviously, if in the first case the power supply will always be noisy, then in the second case users are not too powerful systems, lightly loading the power supply can count on silence.

Also, when the power supply is operating at full power, the ripple amplitude of its output voltages is measured. Let me remind you that, according to the standard, the ripple range in the range up to 10 MHz should not exceed 50 mV for the +5V bus and 120 mV for the +12V bus. In practice, noticeable ripples of two frequencies may be present at the output of the unit - about 60 kHz and 100 Hz. The first is the result of the operation of the unit's PWM stabilizer (usually its frequency is about 60 kHz) and is present to one degree or another on all power supplies. Below is an oscillogram of fairly typical ripples at the PWM operating frequency, green– bus +5V, yellow – +12V:

As you can see, this is exactly the case when the ripple on the +5V bus has gone beyond the permissible limits of 50 mV. The oscillogram shows exactly the classic shape of such ripples - triangular, although in more expensive power supplies the switching moments are usually smoothed out by chokes installed at the output.

The second frequency is double the frequency of the supply network (50 Hz), which reaches the output usually due to insufficient capacitance of the high-voltage rectifier capacitors, errors in the circuitry, or poor design of the power transformer or printed circuit board of the unit. As a rule, these fluctuations (in articles they are given with a time base of 4 ms/div) are observed in many units in the lower price range and are quite rare in middle-class models. The range of these ripples grows in proportion to the load on the power supply and, at maximum, can also sometimes go beyond the permissible limits.

Also, at a load of 150 W, the rectangular pulse generator already mentioned above in the previous section of the article is connected to the power supply, after which the amplitude of the pulses is measured using an oscilloscope. friend power supply wire, that is, not on the one to which the generator is connected. In this way, the overall response of the unit to such a pulsed load is checked, and, in particular, how well it will suppress interference from each of the devices connected to it. However, due to the presence of sharp voltage surges when the generator switches, the measurement accuracy is not too high, but sometimes interesting conclusions can be drawn from these measurements.

And finally, measurements of the efficiency and power factor of the units. Perhaps this is the least important and interesting section - as experience has shown, these parameters are quite close for various blocks, and since for the vast majority of users they do not matter, since their small fluctuations do not have any effect on the operation of the computer (and large fluctuations among different models blocks of the same type are not observed), then measurements are carried out only in fairly rare cases. Thus, the power factor is measured for units for which its correction is declared, and the efficiency is either together with the power factor (in fact, the efficiency value is obtained automatically, no additional measurements are required for this), or if for one reason or another there are suspicions that this block it goes beyond the permissible limits, which happens extremely rarely.

In the end, I would also like to say that I do not measure and will not measure, despite the potential possibility. I have a very negative attitude towards tests that measure the absolute maximum power output by a power supply - when during the test the load on the unit increases until the protection is triggered or the unit simply burns out. Such tests give too much scatter in the results, not only depending on the specific instance of the block, but also depending on exactly how the experimenter loads it - that is, how the load is distributed across the block buses. In addition, for the normal functioning of a computer, what is needed is not a certain rated ability of the power supply to hold such and such power, but the ability to produce voltages and ripples within the tolerance established by the standard, which, unfortunately, is not usually paid attention to in such tests. Therefore, the numbers obtained in such tests, although very beautiful, but, alas, do not have much to do with reality.

So, the methodology we have currently developed for testing power supplies allows us not only to study the behavior of the power supply in great detail, but also to clearly compare different power supplies - and this has become especially clear thanks to the construction of cross-load characteristics, which can be used very objectively and without additional reservations say what a particular block is.
Hello again!..
Unfortunately, my article was delayed because... an urgent work project arose, and interesting difficulties when implementing a power factor corrector ( further KKM). And they were caused by the following - in our production, to control the cash register, we use a “custom-made” microcircuit, which for our purposes is produced by Austria, which was friendly especially in 1941 and, accordingly, cannot be found on sale. Therefore, the task arose to convert this module to an accessible elementary base and my choice fell on a PWM controller chip - L6561.
Why her? Banal availability, or rather I found it in "Chip & Dip", I read the datasheet and liked it. I ordered 50 pieces at once, because... cheaper and in my amateur projects I already have several tasks for it.

Now about the main thing: in this article I will tell you how I remembered designing single-ended converters almost from scratch ( it would seem that they have something to do with it), why I killed a dozen keys and how you can avoid it. This part will tell you the theory and what happens if you neglect it. The practical implementation will come out in the next part, as I promised, along with charger, because they are essentially one module and they need to be tested together.
Looking ahead, I’ll say that for the next part I’ve already prepared a couple of dozen photos and videos where my memory won’t last long "retrained" first into the welding machine, and then into the power supply for "goat". Those who work in production will understand what kind of animal this is and how much it consumes to keep us warm)))

And now to our sheep...

Why do we even need this cash register?

Main trouble “classical” rectifier with storage capacitors (this is the thing that turns 220V AC into +308V direct current), which operates from a sinusoidal current, is that this same capacitor charges (takes energy from the network) only at moments when the voltage applied to it is greater than on itself.

Do not read in human language, for the faint of heart and with scientific degrees

As we know, electric current completely refuses to flow if there is no potential difference. The direction of current flow will also depend on the sign of this difference! If you freaked out and decided to try to charge your mobile phone with a voltage of 2V, where the Li-ion battery is designed for 3.7V, then nothing will work out for you. Because The current will be given by the source that has the higher potential, and the energy will be received by the one with the lower potential.
Everything is just like in life! You weigh 60 kg, and the guy on the street who came up to ask you to call 120 kg - it’s clear that he will hand out the pussy, and you will receive it. So here too - a battery with its 60 kg 2V will not be able to supply current to a battery with 120 kg 3.7V. It’s the same with a capacitor, if it has +310V and you apply +200V to it, then it will refuse to receive current and will not charge.

It is also worth noting that based on the “rule” described above, the time allotted to the capacitor for charging will be very short. In our case, the current changes according to a sinusoidal law, which means the required voltage will be only at the peaks of the sinusoid! But the capacitor needs to work, so it gets nervous and tries to charge. He knows the laws of physics, unlike some, and “understands” that time is short and therefore begins to consume a huge current at these very moments, when the voltage is at its peak. After all, it should be enough to operate the device until the next peak occurs.

A little about these “peaks”:

Figure 1 - Peaks in which the capacitor is charged

As we can see, the portion of the period in which the EMF takes on a sufficient value for charging (figuratively 280-310V) is about 10% of the total period in the AC network. It turns out that instead of constantly gradually taking energy from the network, we pull it out only in small episodes, thereby “overloading” the network. With a power of 1 kW and an inductive load, the current at the time of such “peaks” can easily reach values ​​of 60-80A.

Therefore, our task comes down to ensuring uniform energy extraction from the network so as not to overload the network! It is the cash register that will allow us to implement this task in practice.

Who is this KKM of yours?

Power corrector- This is a regular step-up voltage converter, most often it is single-ended. Because we use PWM modulation, then at the moment public key The voltage across the capacitor is constant. If we stabilize the output voltage, then the current taken from the network is proportional to the input voltage, that is, it changes smoothly according to a sinusoidal law without the previously described consumption peaks and jumps.

Circuitry of our PFC

Here I decided not to change my principles and also relied on the datasheet of the controller I had chosen - L6561. Company engineers STMicroelectronics have already done everything for me, and more specifically, they have already developed the ideal circuit design for their product.
Yes, I can recalculate everything myself from scratch and spend a day or two on this matter, that is, all my already rare weekends, but the question is why? Prove to myself that I can, fortunately this stage has long been passed)) Here I remember a bearded joke about the area of ​​red balls, they say a mathematician applies a formula, and an engineer takes out a table with the area of ​​red balls.... So it is in this case.

I advise you to immediately pay attention to the fact that the circuit in the datasheet is designed for 120 W, which means we should adapt to our 3 kW and extreme work stress.

Now some documentation for what was described above:
Datasheet for L6561

If we look at page 6, we will see several diagrams, we are interested in the diagram with the signature Wide-range Mains, which means in Basurmanian “for operation in a wide range of supply voltage” . It is this “mode” that I had in mind when speaking about extreme voltages. The device is considered universal, that is, it can operate from any standard network (for example, in the states 110V) with a voltage range of 85 - 265V.

This solution allows us to provide our UPS with the function of a voltage stabilizer! For many, this range will seem excessive and then they can make this module taking into account the supply voltage of 220V + - 15%. This is considered the norm and 90% of devices in price category up to 40 thousand rubles are completely devoid of cash register, and 10% use it only with the calculation of deviations of no more than 15%. This undoubtedly allows us to somewhat reduce the cost and dimensions, but if you haven’t forgotten yet, we are making a device that must compete with ARS!

Therefore, for myself I decided to choose the most correct option and make an indestructible tank that can be pulled even in a country house where there is 100V in the network, a welding machine or a pump in a well:


Figure 2 - Standard circuit design offered by ST

Adaptation of standard circuitry to our tasks

a) When I look at this diagram from DS, the first thing that comes to mind is you need to add a common mode filter! And this is correct, because. on high power they will start to drive electronics crazy. For currents of 15 A or more, it will have a more complicated appearance than many are accustomed to seeing in the same computer power supplies, where only 500-600 W. Therefore, this revision will be a separate item.

B) We see capacitor C1, you can take a tricky formula and calculate the required capacitance, and I advise those who want to delve into this to do this, at one time remembering 2nd year electrical engineering from any polytechnic. But I won’t do this, because... Based on my own observations from old calculations, I remember that up to 10 kW this capacity grows almost linearly relative to the increase in power. That is, taking into account 1 µF per 100 W, we find that for 3000 W we need 30 µF. This container is easily filled from 7 film capacitors of 4.7 µF and 400V each. Even a little with reserve, after all The capacitance of a capacitor is highly dependent on the applied voltage.

C) We will need a serious power transistor, because... The current consumed from the network will be calculated as follows:


Figure 3 - Calculation of rated current for PFC

We got it 41.83A. Now we honestly admit that we will not be able to maintain the temperature of the transistor crystal in the region of 20-25 o C. Or rather, we can handle it, but it will be expensive for such power. After 750 kW, the cost of cooling with freon or liquid oxygen is eroded, but so far this is far from the case))) Therefore, we need to find a transistor that can produce 45-50A at a temperature of 55-60 o C.

Considering that there is inductance in the circuit, I would prefer IGBT transistor, because they are the most durable. The maximum current must be selected for the search first about 100A, because this is the current at 25 o C; with increasing temperature, the maximum switching current of the transistor decreases.

A little about Cree FET

Literally on January 9, I received a parcel from the States from a friend with a bunch of different transistors for testing, this miracle is called - CREE FET. I won’t say that this is a new mega technology; in fact, transistors based on silicon carbide were made back in the 80s, they just brought it to mind only now. As a primary materials scientist and composer in general, I am scrupulous about this industry, so I was very interested this product, especially since 1200V was stated at tens and hundreds of amperes. I couldn’t buy them in Russia, so I turned to my former classmate and he kindly sent me a bunch of samples and a test board with forward.
I can say one thing - it was my most expensive fireworks!
8 keys were so fucked up that I was upset for a long time... In fact, 1200V is a theoretical figure for the technology, the declared 65A turned out to be only a pulsed current, although the documentation clearly stated that it was nominal. Apparently there was a “rated pulse current” or whatever the Chinese come up with. In general, it’s still bullshit, but there is one BUT!
When I finally did it CMF10120D 300 W corrector, it turned out that on the same radiator and circuit it had a temperature of 32 o C versus 43 for the IGBT, and this is very significant!
Conclusion on CREE: the technology is crude, but it is promising and definitely SHOULD BE.

As a result, after looking through the catalogs from the exhibitions I visited (a convenient thing, by the way, ala parametric search), I chose two keys, they were - IRG7PH50 And IRGPS60B120. Both are 1200V, both are 100+A, but upon opening the datasheet, the first key was eliminated immediately - it is capable of switching a current of 100A only at a frequency of 1 kHz, which is disastrous for our task. The second switch is 120A and the frequency is 40 kHz, which is quite suitable. Look at the datasheet at the link below and look for a graph showing the dependence of current on temperature:

Figure 4.1 - Graph showing the dependence of the maximum current on the switching frequency for the IRG7PH50, let’s leave it for the frequency converter

Figure 4.2 - Graph with operating current at a given temperature for IRGPS60B120

Here we see the treasured figures that show us that at 125 o C both the transistor and the diode can easily handle currents of just over 60A, while we can implement conversion at a frequency of 25 kHz without any problems or restrictions.

D) Diode D1, we need to choose a diode with an operating voltage of at least 600V and a rated current for our load, that is 45A. I decided to use the diodes that I had on hand (I recently purchased them to develop a welder for an “oblique bridge”): VS-60EPF12. As can be seen from the markings, it is 60A and 1200V. I put everything with a reserve, because... This prototype is being made for myself and my loved one, and it makes me feel better.
You can actually get a 50-60A and 600V diode, but there is no price between the 600V and 1200V versions.

D) Capacitor C5, everything is the same as in the case of C1 - just increase the value from the datasheet in proportion to the power. Just keep in mind that if you are planning a powerful inductive load or a dynamic load with rapid increases in power (ala a 2 kW concert amplifier), then it is better not to skimp on this point.
I'll put it in my option 10 electrolytes at 330 uF and 450V, if you plan to power a couple of computers, routers and other small things, then you can limit yourself to 4 electrolytes of 330 uF and 450V each.

E) R6 - it is also a current shunt, it will save us from crooked hands and accidental errors, it also protects the circuit from short circuit and excess load. The thing is definitely useful, but if we act like the engineers from ST, then at currents of 40A we will end up with an ordinary boiler. There are 2 options: a current transformer or a factory shunt with a drop of 75 mV + op-amp ala LM358.
The first option is simpler and provides galvanic isolation of this circuit node. I gave how to calculate a current transformer in a previous article, it is important to remember that the protection will work when the voltage on leg 4 rises to 2.5V (in reality up to 2.34V).
Knowing this voltage and current of the circuit, using the formulas from parts 5 you can easily calculate the current transformer.

G) And the last point is the power choke. More about him below.

Power choke and its calculation

If someone carefully read my articles and has an excellent memory, then he should remember article 2 and photograph no. 5, it shows 3 skein elements that we use. I'll show you again:

Figure 5 - Frames and core for power winding products

In this module we will again use our favorite toroidal rings made of pulverized iron, but only this time not just one, but 10 at once! What did you want? 3 kW is not Chinese crafts...

We have the initial data:
1) Current - 45A + 30-40% of amplitude in the inductor, total 58.5A
2) Output voltage 390-400V
3) input voltage 85-265V AC
4) Core - material -52, D46
5) Gap - distributed

Figure 6 - And again, dear Starichok51 saves us time and considers it as a program CaclPFC

I think the calculation showed everyone how serious this design will be)) 4 rings, a radiator, a diode bridge, and an IGBT - horror!
Winding rules can be read in the article “Part 2”. Secondary winding dangles on the rings in quantity - 1 turn.

Throttle summary:

1) as you can see, the number of rings is already 10 pieces! This is expensive, each ring costs about 140 rubles, but what will we get in return in the following paragraphs
2) the operating temperature is 60-70 o C - this is absolutely ideal, because many set the operating temperature at 125 o C. In our production we set it to 85 o C. Why is this done - for a restful sleep, I calmly leave home for a week and know that nothing will flare up, nothing will burn, and everything will be icy. I think the price for this at 1500 rubles is not so lethal, is it?
3) I set the current density to a meager 4 A/mm 2 , this will affect both heat and insulation and, accordingly, reliability.
4) As you can see from the calculation, the recommended capacity after the inductor is almost 3000 uF, so my choice with 10 electrolytes of 330 uF fits perfectly here. The capacitance of capacitor C1 turned out to be 15 uF, we have a double reserve - you can reduce it to 4 film capacitors, you can leave 7 pieces and it will be better.

Important! The number of rings in the main choke can be reduced to 4-5, simultaneously increasing the current density to 7-8 A/mm 2. This will save a lot of money, but the current amplitude will increase somewhat, and most importantly the temperature will increase to at least 135 o C. I think this good decision for a welding inverter with a duty cycle of 60%, but not for a UPS, which operates around the clock and probably in a rather limited space.

What can I say - we have a monster growing)))

Common Mode Filter

To understand the difference between the circuits for this filter for currents of 3A (the computer power supply mentioned above) and for currents of 20A, you can compare the circuit from Google on ATX with the following:


Figure 7 - Schematic diagram of a common mode filter

Several features:

1) C29 is a capacitor for filtering electromagnetic interference and is marked "X1". Its nominal value should be in the range of 0.001 - 0.5 mF.

2) The choke dangles on the core E42/21/20.

3) Two chokes on rings DR7 and DR9 are wound on any spray core with a diameter of more than 20 mm. I wound the same D46 from -52 material until filled in 2 layers. Noise in the network even when rated power practically none, but this is actually redundant even in my understanding.

4) Capacitors C28 and C31 are 0.047 µF and 1 kV each and they must be of class "Y2".

According to the calculation of the inductance of the chokes:

1) The inductance of the common mode inductor should be 3.2-3.5 mH

2) Inductance for differential chokes is calculated using the formula:


Figure 8 - Calculation of the inductance of differential chokes without magnetic coupling

Epilogue

Using the competent and professional developments of ST engineers, I was able to minimal costs make, if not perfect, then simply excellent active power factor correction with parameters better than any Schneider. The only thing you should definitely remember is how much do you need it? And based on this, adjust the parameters for yourself.

My goal in this article was precisely to show the calculation process with the possibility of adjusting the initial data, so that everyone, having decided on the parameters for their tasks, could calculate and manufacture the module themselves. I hope I was able to show this and in the next article I will demonstrate the joint operation of the cash register and the charger from part No. 5.

Conversion technology

Introduction

In recent decades, the number of electronics used in homes, offices and factories has increased dramatically, and most devices use switching power supplies. Such sources generate harmonic and nonlinear current distortions, which negatively affect the electrical wiring and electrical appliances connected to it. This influence is expressed not only in various types interference, affecting the operation of sensitive devices, but also in overheating of the neutral line. When currents flow in loads with significant harmonic components that are out of phase with the voltage, the current in the neutral wire (which is practically zero with a symmetrical load) can increase to a critical value.

The International Electrotechnical Commission (IEC) and the European Organization for Electrotechnical Standardization (CENELEC) have adopted standards IEC555 and EN60555 that set limits on harmonic content in the input current of secondary power supplies, fluorescent lamp electronic loads, DC motor drivers and similar devices.

One of the effective ways to solve this problem is the use of PFC (Power Factor Correction) power factor correctors. In practice, this means that a special PFC circuit must be included in the input circuit of almost any electronic device with pulse converters, which provides reduction or complete suppression of current harmonics.

Power factor correction

A typical switching power supply consists of a mains rectifier, a smoothing capacitor and a voltage converter. Such a source consumes power only at those moments when the voltage supplied from the rectifier to the smoothing capacitor is higher than the voltage across it (the capacitor), which occurs for about a quarter of the period. The rest of the time, the source does not consume power from the network, since the load is powered by a capacitor. This leads to the fact that power is taken by the load only at the voltage peak, the consumed current has the form of a short pulse and contains a set of harmonic components (see Fig. 1).

A secondary power source with power factor correction consumes current with low harmonic distortion, takes power from the network more evenly, and has a crest factor (the ratio of the amplitude value of the current to its rms value) lower than that of an uncorrected source. Power factor correction reduces the RMS value of current consumption, which allows you to connect more different devices to one outlet of the electrical network without creating overcurrent in it (see Fig. 2).

Power factor

Power Factor PF is a parameter characterizing the distortion created by the load (in our case, the secondary power source) in the AC network. There are two types of distortions - harmonic and nonlinear. Harmonic distortion is caused by a reactive load and represents a phase shift between current and voltage. Nonlinear distortions are introduced into the network by “nonlinear” loads. These distortions are expressed in the deviation of the current or voltage waveform from a sinusoid. When harmonic distortion The power factor is the cosine of the phase difference between current and voltage or the ratio of active power to total power consumed from the network. For nonlinear distortion The power factor is equal to the share of the power of the first harmonic current component in the total power consumed by the device. It can be considered an indicator of how uniformly the device consumes power from the mains.

In general power factor is the product of the cosine of the phase difference angle between voltage and current and the cosine of the angle between the fundamental harmonic vector and the total current vector. The reasoning given below leads to this definition. The effective current flowing in the active load has the form:

I 2 eff =I 2 0 +I 2 1eff +SI 2 neff,

where I 2 neff is the constant component (in the case of sinusoidal voltage it is zero), I 2 1eff is the main harmonic, and under the sum sign are the lower harmonics. When working with a reactive load, a reactive component appears in this expression, and it takes the form:

I 2 eff =I 2 0 +(I 2 1eff(P) +I 2 1eff(Q))+SI 2 neff. Active power- this is the average value of the power allocated to the active load over the period.

It can be represented as the product of the effective voltage and the active component of the current P=U eff H I 1eff(P). Physically, this is the energy released in the form of heat per unit time per active resistance. Under reactive power understand the product of the effective voltage and the reactive component of the current: Q = U eff H I 1 eff (Q). The physical meaning is energy that is pumped twice per period from the generator to the load and twice from the load to the generator. Total power is the product of the effective voltage and the total effective current: S = U eff H I eff (total). On the complex plane, it can be represented as the sum of vectors P and Q, from which the dependence I 2 =I 1eff(total) cos j is visible, where j is the angle between vectors P and Q, which also characterizes the phase difference between the current and voltage in the circuit.

Based on the above, we derive the definition for power factor:

PF=P/S=(I 1eff cos j)/(Ieff(total)).

It is worth noting that the ratio (I 1eff)/(Ieff(total)) is the cosine of the angle between the vectors corresponding to the effective value of the total current and the effective value of its first harmonic. If we denote this angle as q, then the expression for the power factor takes the form: PF=cos j Х cos q. The task of power factor correction is to bring the phase difference angle j between voltage and current, as well as the harmonic distortion angle q of the consumed current, closer to zero (or, in other words, to bring the shape of the current curve as close as possible to a sinusoid and compensate for the phase shift as much as possible).

Power factor is expressed as a decimal fraction, the value of which ranges from 0 to 1. Its ideal value is one (for comparison, a typical switching power supply without correction has a power factor value of about 0.65), 0.95 is a good value; 0.9 - satisfactory; 0.8 - unsatisfactory. Applying power factor correction can increase a device's power factor from 0.65 to 0.95. Values ​​in the range of 0.97...0.99 are also quite realistic. Ideally, when the power factor equal to one, the device consumes a sinusoidal current from the network with zero phase shift relative to the voltage (which corresponds to a fully resistive load with a linear current-voltage characteristic).

Passive power factor correction

The passive correction method is most often used in inexpensive low-power devices (where there are no strict requirements for the intensity of lower current harmonics). Passive correction allows you to achieve a power factor of about 0.9. This is convenient in the case when the power source has already been designed, all that remains is to create a suitable filter and include it in the circuit at the input.

Passive power factor correction consists of filtering the current consumption using an LC bandpass filter. This method has several limitations. An LC filter can be effective as a power factor corrector only if the voltage, frequency and load vary within a narrow range of values. Since the filter must operate in the low frequency region (50/60 Hz), its components are large in size, weight and low quality factor(which is not always acceptable). Firstly, the number of components with a passive approach is much smaller and, therefore, the time between failures is longer, and secondly, with passive correction, less electromagnetic and contact interference is created than with active one.

Active power factor correction

An active power factor correction must satisfy three conditions:

1) The shape of the consumed current should be as close to sinusoidal as possible and “in phase” with the voltage. The instantaneous value of the current consumed from the source must be proportional to the instantaneous network voltage.

2) The power taken from the source must remain constant even if the network voltage changes. This means that when the network voltage decreases, the load current must be increased, and vice versa.

3) The voltage at the output of the PFC corrector should not depend on the load size. As the voltage across the load decreases, the current through it must increase, and vice versa.

There are several schemes that can be used to implement active power factor correction. The most popular currently is the “boost converter circuit”. This circuit satisfies all the requirements for modern power supplies. Firstly, it allows you to work in networks with different supply voltages (from 85 to 270 V) without restrictions or any additional adjustments. Secondly, she is less susceptible to deviations electrical parameters networks (voltage surges or short-term power outages). Another advantage of this scheme is that it is more simple implementation surge protection. A simplified diagram of a “boost converter” is shown in Fig. 3.

Principle of operation

The standard power factor corrector is an AD/DC converter with pulse width modulation (PWM). The modulator controls a powerful (usually MOSFET) switch, which converts direct or rectified mains voltage into a sequence of pulses, after rectification of which a constant voltage is obtained at the output.

Timing diagrams of the corrector's operation are shown in Fig. 4. When the MOSFET switch is turned on, the current in the inductor increases linearly - while the diode is locked, and capacitor C2 is discharged to the load. Then, when the transistor is turned off, the voltage across the inductor “opens” the diode and the energy stored in the inductor charges capacitor C2 (and simultaneously powers the load). In the above circuit (unlike a source without correction), capacitor C1 has a small capacitance and serves to filter high-frequency interference. The conversion frequency is 50...100 kHz. In the simplest case, the circuit operates with a constant duty cycle. There are ways to increase the effectiveness of correction dynamic change duty cycle (matching the cycle with the voltage envelope from the mains rectifier).

The "boost converter" circuit can operate in three modes: continuous , discrete and the so-called " critical conductivity mode" IN discrete mode, during each period the inductor current manages to “fall” to zero and after some time begins to increase again, and in continuous- the current, not having time to reach zero, begins to increase again. Mode critical conductivity used less frequently than the previous two. It is more difficult to implement. Its meaning is that the MOSFET opens at the moment when the inductor current reaches zero value. When operating in this mode, adjusting the output voltage is simplified.

The choice of mode depends on the required output power of the power supply. Devices with a power of more than 400 W use continuous mode, while low-power devices use discrete mode. Active power factor correction allows you to achieve values ​​of 0.97...0.99 with a THD (Total Harmonic Distortion) coefficient of 0.04...0.08.