It is in every computer, laptop and set-top box. It does not affect your frame rate or Bitcoin mining. It doesn't have billions of transistors, and it doesn't use the latest semiconductor processes. Sound boring? Not at all! Without this thing, our computers would do absolutely nothing.
PSUs, aka power supply units (English PSU, Power Supply Units), do not blow up the headlines like the latest processors, but they are interesting technologies that deserve our attention. So put on your white coats, masks, gloves and let's start opening up our humble little guy - the power supply, let's take it apart and look at what each of its organs does.
And yes, quite recently we figured out how to choose the right power supply. Recommended reading.
What is it and what is it eaten with?
Many computer components have names that require a bit of technical knowledge to understand what they are and why (for example, a solid-state drive), but in the case of a power supply, everything is pretty obvious. This is the unit that provides power.
But we cannot put an end to this, proudly declaring “the article is ready.” Our series of articles is devoted to the internal structure, and on the operating table we have a test subject - the Cooler Master G650M. This is a fairly typical representative, with characteristics similar to dozens of other models, but it has one feature that is not found in all power supplies.
Official photo of the Cooler Master power supply.
This is a standard size power supply that fits the ATX 12V v2.31 form factor, making it suitable for many computer cases.
There are other form factors - for example, for small cases, or completely unique ones upon special order. Not every block corresponds to the exact dimensions established by standard form factors - they may be the same width and height, but differ in length.
This power supply from Cisco is specially designed for server racks
PSU markings usually indicate their main parameter - the maximum power provided. In the case of our Cooler Master, this is 650 W. Later we will talk about what this actually means, but for now we will just note that there are also less powerful power supplies, since not all computers require exactly that much, and some need even an order of magnitude less. But still, most desktop computers are provided with power in the range from 400 to 600 W.
Power supplies like ours are assembled in rectangular, often unpainted, metal cases, which is why they are quite heavy. Laptops almost always have an external power supply, in a plastic case, but its internals are very similar to what we will see in the power supply we are considering.
Photo source nix.ru
Most typical power supplies are equipped with a power switch and a cooler for active thermal regulation, although not all power supplies need it. And not all of them have a ventilation grill - in server versions, in particular, this is rare.
Well, as you can see in the photo above, we are already armed with a screwdriver and are ready to start opening our copy.
A little theory
But before we start digging into the internals, let's ask ourselves, is a power supply really that necessary? Why can't you plug your computer directly into a power outlet? The answer is that computer components are designed for a completely different voltage than the mains voltage.
The graph below shows what the grid electricity should be (US = blue and green curves; UK = red curve). The X axis represents time in milliseconds, and the Y axis represents voltage in volts. The easiest way to understand what voltage is is by looking at the energy difference between two points.
If a voltage is applied to a conductor (such as a metal wire), the difference in energy will cause electrons in the conductor material to flow from a higher energy level to a lower one. Electrons are the constituent atoms that make up a conductor, and metals have many electrons that can move freely. This flow of electrons is called current and is measured in amperes.
A good analogy is with a garden hose: voltage is akin to the pressure you use, and water flow is akin to current. Any restrictions or obstacles in the hose are essentially like electrical resistance.
We see that the electricity in the network varies over time, which is why it is called alternating current voltage (AC, alternating current). In the US, line voltage varies 60 times per second, peaking at 340 V or 170 V, depending on location and connection method. In the UK the peak voltages are lower and the frequency of these fluctuations is also slightly different. Most countries have similar mains voltage standards, with only a few countries having lower or higher peak voltages.
The need for a power supply is that computers do not work with alternating current: they need a constant voltage that never changes, and also much lower. On the same graph it will look something like this:
But a modern computer requires not one constant voltage, but four: +12 volts, -12 volts, +5 volts and +3.3 volts. And since these values do not change, such a current is called constant (DC, direct current). Converting current from AC to DC (so-called rectification) is one of the main functions of the power supply. It's time to open it up and see how he does it!
Converting current from AC to DC is one of the main functions of PSU. It's time to see how he does it!
Here we must warn you that the power supply contains elements that accumulate electricity, including deadly electricity. Therefore, disassembling the PSU is potentially dangerous.
Official photo of the Cooler Master power supply.
The principle of operation of this power supply is similar to many others, and although the markings on various parts inside will be different, this does not make any fundamental differences.
The power cord connector is in the top left corner of the photo, and the current essentially flows clockwise until it reaches the output of the power supply (bundle of colored wires, bottom left corner).
Photo source
techspot.com
If we turn the board over, we can see that compared to the motherboard, the conductors and connections on it are wider and more massive - this is because they are designed to handle higher currents. Also, a wide stripe in the middle catches your eye, like a river flowing across the plain.
This again suggests that all power supplies have two clearly separated nodes: primary and secondary. The first is to adjust the input voltage so that it can be effectively stepped down; the second is all the settings for the already rectified and reduced voltage.
Circuit design of SMPS power supplies
So we got to practice. Depending on the required output power, different types of power supplies are used. Let's consider the types of transformer circuits
Flyback converter
The diagram below shows the basic current and voltage waveforms for a flyback transformer.
Basic flyback circuit with transformer
In the first phase of the cycle, the switch connects the inductor L directly to the input voltage. Due to the constant input voltage Ue, a linearly increasing current flows through the inductor.
In this phase, diode D is blocked. When the S button opens, the polarity on the inductor is reversed so that the diode conducts and the energy stored in the inductor is transferred to the load capacitor CLi R1. The throttle acts as a source of energy. Thus, by adjusting the charging time at a given frequency, you can change the energy stored in the inductor.
To obtain galvanic isolation between the input and output of the circuit, the inductor is replaced by a transformer. This element acts as an intermediate energy storage device so that the load circuit can use the energy stored in the transformer and there is no direct load on the power supply.
A condition for energy conservation would be to have an air gap in the transformer core or an insulating pad between both halves of the core (which has the same effect as a mid-core air gap), but using a mid-core air gap provides better feedback between the windings.
Forward type converters
The figure shows the basic circuit of a forward-type converter. When switch S is closed, a linearly increasing current flows through the coil directly to the capacitor Ca and to the load R1. At this stage, energy is simultaneously transferred to the inductor and the load. Diode D is blocked.
Basic Forward Power Supply Circuit
When the key opens, the magnetic field of the throttle is interrupted. The polarity of the inductor changes, opening the diode. The energy from the inductor goes through the diode to the capacitor and to the load. Since energy transfer to the output circuit also occurs when the switch is closed, the type of this transformer is called forward-running. As with flyback transformers, the energy stored in the inductance in this type of power supply can be varied by varying switching times.
Forward power supply with transformer
This diagram shows a direct type power supply with a transformer to separate and convert the mains voltage. When using a core without an air gap, constant magnetic contact is maintained between the primary and secondary windings. But collecting and smoothing the output current must be implemented in a separate inductor Ls, for each output voltage separately. The energy stored by the transformer during the conduction phase is transferred to L1, Dl, Ce in the blocking phase. The diode opens when the polarity of the energy storage inductor changes.
Push-pull converters
In fact, push-pull transformers consist of two single transformers connected to each other.
Basic circuit of a push-pull power supply
Switches S1 and S2 alternately connect the primary winding to the source Ue. Compared to a forward and reverse transformer, this configuration allows for a full hysteresis loop. Thanks to the bipolar system, you can get twice the power with the same core size.
Push-pull converter
Even with large load changes, the push-pull transformer generates a symmetrical output voltage, allowing direct use of AC voltage without prior rectification, such as in halogen lighting.
Filtration
The first thing the power supply does with mains electricity is not rectification or reduction, but equalization of the input voltage. Since our homes, offices and businesses have many electrical devices and appliances that constantly turn on and off, as well as emitting electromagnetic interference, alternating current in the network is often “lumpy” and with random surges and drops (the frequency is also not constant). Not only does this make it difficult for the power supply to carry out conversions, but it can damage some of the components inside it.
Our power supply has two stages of so-called input filters (transient filter), the first of which is built directly at the input using three capacitors. It performs a role similar to that of a speed bump on the road - only instead of speed, this filter dampens sudden surges in input voltage.
Photo source techspot.com
The second filter stage is more complex, but essentially does the same thing.
The yellow bricks are again capacitors, but the green rings wrapped in copper wire are inductive coils (although when used in this way they are usually called chokes). The coils accumulate electrical energy in a magnetic field, but the energy is not lost and is smoothly returned due to self-induction. Thus, a suddenly appearing high pulse (jump) is absorbed by the magnetic field of the inductor in order to give an even voltage at the output without any jumps.
The two small blue disks are another representative of the capacitor variety, and just below them (green, with long legs covered with black insulators) is a metal oxide varistor (MOV). They are also used to protect against input voltage surges. You can read more about the different types of input filters here.
Photo source techspot.com
From this power supply unit you can often determine how much the manufacturer has saved, or what budget class the device belongs to. Cheaper ones will have simplified input filtering, and the cheapest ones will not have it at all (avoid these!).
Now that the tension has been leveled and combed, he is allowed to move on - in fact, to transformation.
Conversion
As we said, the power supply needs to change the AC voltage, which in American outlets is usually around 120 volts (technically, this is 120 volts rms, but we won’t break our tongue like that), giving the output a constant voltage of 12, 5 and 3. 3 volts.
The first step is to convert AC to DC, and our unit uses a bridge rectifier for this. In the photo below it is a flat black element glued to the radiator.
Photo source techspot.com
This is another place where the power supply manufacturer can cut costs, since cheaper rectifiers perform worse (eg run hotter). Now, if the peak input voltage is 170 V (which is the case for a 120 V network), then after passing through the rectifier bridge, it will become 170 V, but already DC.
In this form, it enters the next stage, and in our block it is an active power factor correction module (APFC or Active PFC, Active Power Factor Correction converter). This unit also stabilizes the voltage, smoothing out “dips” due to storage capacitors; In addition, it protects against output power surges.
Passive correctors (PPFC or Passive PFC) do essentially the same job. They are less efficient, but good for low-power power supplies.
Photo source techspot.com
The APFC in the photo above is represented by a pair of large cylinders on the left - these are capacitors that accumulate equalized current before sending it further along the chain of processes in our power supply.
Behind the APFC there is a PWM, pulse width modulator (PWM, Pulse Width Modulator). Its purpose is to convert direct current back to alternating current using several fast-switching field-effect transistors. This needs to be done because in the next step a step-down transformer awaits us. These devices, based on electromagnetic induction, consist of two windings with different numbers of turns on the metal core needed to step down the voltage, and operate transformers with alternating current only.
The frequency of the AC current (the rate at which it changes; in Hertz, Hz) greatly affects the efficiency of the transformer - the higher the better, so the frequency of the original 50/60 Hz supply increases by about a thousand times. And the more efficient the transformer, the smaller its size. This type of device that uses these ultra-fast DC frequencies is called a Switched Mode Power Supply (SMPS).
In the photo below you can see 3 transformers - the largest one has 12 volts at its only output, and the smaller one has 5 volts (we'll talk about it a little later). In other power supplies you can find one large transformer for all voltages at once, that is, with several outputs. And the smallest transformer is designed to protect PWM transistors and suppress its interference.
| Photo source techspot.com
You can implement in different ways obtaining the necessary voltages, PWM protection, and so on. It all depends on the budget segment and the power of the device. However, everyone equally needs to remove the voltage from the transformers and straighten them again.
In the photo below we see an aluminum heatsink for the low-voltage diodes that perform this rectification. And also, specifically in this PSU, we see a small additional board in the center of the photo - this is a node of voltage regulation modules (VRM, Voltage Regulation Modules), providing outputs of 5 and 3.3 volts.
Photo source techspot.com
And here we should talk about what pulsation is.
In an ideal world, with ideal power supplies, alternating current will be converted into absolutely smooth, without the slightest fluctuation, direct current. In reality, such 100% accuracy is not achieved, and the DC voltage has, albeit minor, fluctuations.
This effect is called ripple voltage, and in our power supplies we like to keep it as low as possible. Cooler Master does not provide information about the ripple voltage value in the specification for our experimental PSU, so we resorted to third-party testing results. One such analysis was performed by JonnyGuru.com and they found that the maximum ripple voltage of the +12V output was 0.042V (42 millivolts).
The graph below shows the deviation of the actual voltage received (blue curve; its shape, of course, is not such an ideal sine wave - after all, the ripple itself is not constant) from the required flat voltage of +12 V DC (red straight).
This deviation is largely the responsibility of the capacitors throughout the PSU. Low-quality, cheap capacitors lead to an increase in this unnecessary ripple. If it is too large, then some electronic components of the computer that are most sensitive to power quality may begin to work unstable. Fortunately, in our example, 40-plus millivolts is normal. Not great, but not bad either.
But the matter does not end with obtaining acceptable output voltages. It is necessary to ensure control of the outputs so that the power at each of them is always full and stable, regardless of the power of the loads at other outputs.
Photo source techspot.com
The chip you see in this photo is called a supervisor and it makes sure that the pins do not have too high or low voltage and current. It works simply - it simply turns off the power supply when such problems occur.
More expensive PSUs can be equipped with a digital signal processor (DSP), which not only monitors voltages, but can also adjust them if necessary, and also send detailed data about the state of the power supply to the computer using it. For the average user, this function is quite controversial, but for servers and workstations it is highly desirable.
The principle of operation of a computer power supply
The article was written based on the book by A.V. Golovkov and V.B Lyubitsky “POWER SUPPLY FOR SYSTEM MODULES OF THE IBM PC-XT/AT TYPE” Material taken from the interlavka website. The alternating mains voltage is supplied through the mains switch PWR SW through the mains fuse F101 4A, noise suppression filters formed by elements C101, R101, L101, C104, C103, C102 and chokes I 02, L103 to: • three-pin output connector to which the power cable can be connected display; • two-pin connector JP1, the mating part of which is located on the board. From connector JP1, alternating mains voltage is supplied to: • bridge rectification circuit BR1 through thermistor THR1; • primary winding of the starting transformer T1.
At the output of rectifier BR1, smoothing filter capacitances C1, C2 are included. The THR thermistor limits the initial surge of charging current for these capacitors. The 115V/230V SW switch provides the ability to power a switching power supply from both a 220-240V network and a 110/127V network.
High-ohm resistors R1, R2, shunt capacitors C1, C2 are baluns (equalize the voltages on C1 and C2), and also ensure the discharge of these capacitors after turning off the switching power supply from the network. The result of the operation of the input circuits is the appearance on the rectified mains voltage bus of a direct voltage Uep equal to +310V, with some ripples. This switching power supply uses a starting circuit with forced (external) excitation, which is implemented on a special starting transformer T1, on the secondary winding of which, after the power supply is turned on, an alternating voltage with the frequency of the supply network appears. This voltage is rectified by diodes D25, D26, which form a full-wave rectification circuit with a midpoint with the secondary winding T1. SZO is a smoothing filter capacitance on which a constant voltage is generated, used to power the control microcircuit U4.
The TL494 IC is traditionally used as a control chip in this switching power supply.
The supply voltage from the SZO capacitor is supplied to pin 12 of U4. As a result, the output voltage of the internal reference source Uref = -5B appears at pin 14 of U4, the internal sawtooth voltage generator of the microcircuit starts, and control voltages appear at pins 8 and 11, which are sequences of rectangular pulses with negative leading edges, shifted relative to each other by half the period. Elements C29, R50 connected to pins 5 and 6 of the U4 microcircuit determine the frequency of the sawtooth voltage generated by the internal generator of the microcircuit.
The matching stage in this switching power supply is made according to a transistorless circuit with separate control. The supply voltage from the capacitor SZO is supplied to the middle points of the primary windings of control transformers T2, TZ. The output transistors of IC U4 perform the functions of matching stage transistors and are connected according to the circuit with the OE. The emitters of both transistors (pins 9 and 10 of the microcircuit) are connected to the “case”. The collector loads of these transistors are the primary half-windings of the control transformers T2, T3, connected to pins 8, 11 of the U4 microcircuit (open collectors of the output transistors). The other halves of the primary windings T2, T3 with diodes D22, D23 connected to them form demagnetization circuits for the cores of these transformers.
Transformers T2, TZ control powerful transistors of the half-bridge inverter.
Switching the output transistors of the microcircuit causes the appearance of pulsed control EMF on the secondary windings of control transformers T2, T3. Under the influence of these EMFs, power transistors Q1, Q2 alternately open with adjustable pauses (“dead zones”). Therefore, alternating current flows through the primary winding of the T5 power pulse transformer in the form of sawtooth current pulses. This is explained by the fact that the primary winding T5 is included in the diagonal of the electrical bridge, one arm of which is formed by transistors Q1, Q2, and the other by capacitors C1, C2. Therefore, when any of the transistors Q1, Q2 is opened, the primary winding T5 is connected to one of the capacitors C1 or C2, which causes current to flow through it as long as the transistor is open. Damper diodes D1, D2 ensure the return of energy stored in the leakage inductance of the primary winding T5 during the closed state of transistors Q1, Q2 back to the source (recuperation).
Chain C4, R7, which shunts the primary winding T5, helps suppress high-frequency parasitic oscillatory processes that arise in the circuit formed by the inductance of the primary winding T5 and its inter-turn capacitance when transistors Q1, Q2 are closed, when the current through the primary winding abruptly stops.
Capacitor SZ, connected in series with the primary winding T5, eliminates the direct current component through the primary winding T5, thereby eliminating unwanted magnetization of its core.
Resistors R3, R4 and R5, R6 form basic dividers for powerful transistors Q1, Q2, respectively, and provide optimal switching mode from the point of view of dynamic power losses on these transistors.
The flow of alternating current through the primary winding T5 causes the presence of alternating rectangular pulse EMF on the secondary windings of this transformer. The T5 power transformer has three secondary windings, each of which has a terminal from the middle point. Winding IV provides an output voltage of +5V. The diode assembly SD2 (half bridge) forms a full-wave rectification circuit with a midpoint with winding IV (the midpoint of winding IV is grounded). Elements L2, СУ, С11, С12 form a smoothing filter in the +5V channel. To suppress parasitic high-frequency oscillatory processes that occur when switching the diodes of the SD2 assembly, these diodes are shunted by calming RC circuits C8, R10nC9, R11.
The diodes of the SD2 assembly are diodes with a Schottky barrier, which achieves the required speed and increases the efficiency of the rectifier.
Winding III together with winding IV provides an output voltage of +12V together with the diode assembly (half bridge) SD1. This assembly forms, with winding III, a full-wave rectification circuit with a midpoint. However, the middle point of winding III is not grounded, but is connected to the +5V output voltage bus. This will make it possible to use Schottky diodes in the +12V generation channel, because the reverse voltage applied to the rectifier diodes with this connection is reduced to the permissible level for Schottky diodes.
Elements L1, C6, C7 form a smoothing filter in the +12V channel.
Resistors R9, R12 are designed to accelerate the discharge of the output capacitors of the +5V and +12V buses after turning off the UPS from the network. The RC circuit C5, R8 is designed to suppress oscillatory processes occurring in the parasitic circuit formed by the inductance of winding III and its interturn capacitance. The five-tap AND winding provides negative output voltages of -5V and -12V. Two discrete diodes D3, D4 form a half-bridge of full-wave rectification in the -12V generation channel, and diodes D5, D6 - in the -5V channel. Elements L3, C14 and L2, C12 form anti-aliasing filters for these channels. Winding II, as well as winding III, is shunted by a damping RC circuit R13, C13.
The middle point of winding II is grounded.
Stabilization of output voltages is carried out in different ways in different channels. Negative output voltages -5V and -12V are stabilized using linear integrated three-terminal stabilizers U4 (type 7905) and U2 (type 7912). To do this, the output voltages of the rectifiers from capacitors C14, C15 are supplied to the inputs of these stabilizers. The output capacitors C16, C17 produce stabilized output voltages of -12V and -5V. Diodes D7, D9 ensure the discharge of output capacitors C16, C17 through resistors R14, R15 after turning off the switching power supply from the network. Otherwise, these capacitors would be discharged through the stabilizer circuit, which is undesirable. Through resistors R14, R15, capacitors C14, C15 are also discharged.
Diodes D5, D10 perform a protective function in the event of breakdown of the rectifier diodes.
If at least one of these diodes (D3, D4, D5 or D6) turns out to be “broken”, then in the absence of diodes D5, D10 a positive pulse voltage would be applied to the input of the integrated stabilizer U1 (or U2), and through electrolytic capacitors C14 or C15 alternating current would flow, which would lead to their failure. The presence of diodes D5, D10 in this case eliminates the possibility of such a situation occurring, because the current closes through them. For example, if diode D3 is “broken”, the positive part of the period when D3 should be closed, the current will be closed in the circuit: to D3 - L3 -D7- D5 - “case”. Stabilization of the +5V output voltage is carried out using the PWM method. To do this, a measuring resistive divider R51, R52 is connected to the +5V output voltage bus. A signal proportional to the output voltage level in the +5V channel is removed from resistor R51 and fed to the inverting input of the error amplifier DA3 (pin 1 of the control chip). The direct input of this amplifier (pin 2) is supplied with a reference voltage level taken from resistor R48 included in the divider VR1, R49, R48, which is connected to the output of the internal reference source of the microcircuit U4 Uref=+5B. When the voltage level on the +5V bus changes under the influence of various destabilizing factors, the magnitude of the mismatch (error) between the reference and controlled voltage levels at the inputs of the error amplifier DA3 changes. As a result, the width (duration) of control pulses at pins 8 and 11 of the U4 microcircuit changes in such a way as to return the deviated output voltage +5V to the nominal value (as the voltage on the +5V bus decreases, the width of the control pulses increases, and when this voltage increases, it decreases) . Stable (without parasitic generation) operation of the entire control loop is ensured by a chain of frequency-dependent negative feedback surrounding the error amplifier DA3. This chain is connected between pins 3 and 2 of the control chip U4 (R47, C27).
The +12V output voltage in this UPS is not stabilized.
Adjustment of the output voltage level in this UPS is carried out only for the +5V and +12V channels. This adjustment is carried out by changing the level of the reference voltage at the direct input of the error amplifier DA3 using trimming resistor VR1. When changing the position of the VR1 slider during the UPS setup process, the voltage level on the +5V bus will change within certain limits, and therefore on the +12V bus, because voltage from the +5V bus is supplied to the middle point of winding III.
The combined protection of this UPS includes:
• limiting circuit for controlling the width of control pulses; • complete circuit protection against short circuit in loads; • incomplete output overvoltage control circuit (only on the +5V bus).
Let's look at each of these schemes.
The limiting control circuit uses current transformer T4 as a sensor, the primary winding of which is connected in series with the primary winding of the power pulse transformer T5. Resistor R42 is the load of the secondary winding T4, and diodes D20, D21 form a full-wave rectification circuit for alternating pulse voltage removed from load R42.
Resistors R59, R51 form a divider. Part of the voltage is smoothed out by capacitor C25. The voltage level on this capacitor proportionally depends on the width of the control pulses at the bases of power transistors Q1, Q2. This level is fed through resistor R44 to the inverting input of the error amplifier DA4 (pin 15 of the U4 chip). The direct input of this amplifier (pin 16) is grounded. Diodes D20, D21 are connected so that capacitor C25, when current flows through these diodes, is charged to a negative (relative to the common wire) voltage.
In normal operation, when the width of the control pulses does not exceed acceptable limits, the potential of pin 15 is positive, due to the connection of this pin through resistor R45 to the Uref bus. If the width of the control pulses increases excessively for any reason, the negative voltage on capacitor C25 increases and the potential of pin 15 becomes negative. This leads to the appearance of the output voltage of the error amplifier DA4, which was previously equal to 0V. A further increase in the width of the control pulses leads to the fact that the switching control of the PWM comparator DA2 is transferred to the amplifier DA4, and the subsequent increase in the width of the control pulses no longer occurs (limitation mode), because the width of these pulses no longer depends on the level of the feedback signal at the direct input of the error amplifier DA3.
The short circuit protection circuit in loads can be conditionally divided into protection of channels for generating positive voltages and protection of channels for generating negative voltages, which are implemented in approximately the same circuitry. The sensor of the short-circuit protection circuit in the loads of channels generating positive voltages (+5V and +12V) is a diode-resistive divider D11, R17, connected between the output buses of these channels. The voltage level at the anode of diode D11 is a controlled signal. In normal operation, when the voltages on the output buses of the +5V and +12V channels have nominal values, the anode potential of diode D11 is about +5.8V, because through the divider-sensor current flows from the +12V bus to the +5V bus along the circuit: +12V bus - R17-D11 - +56 bus.
The controlled signal from the anode D11 is fed to the resistive divider R18, R19. Part of this voltage is removed from resistor R19 and supplied to the direct input of comparator 1 of the U3 microcircuit of the LM339N type. The inverting input of this comparator is supplied with a reference voltage level from resistor R27 of the divider R26, R27 connected to the output of the reference source Uref=+5B of the control chip U4. The reference level is selected such that, during normal operation, the potential of the direct input of comparator 1 would exceed the potential of the inverse input. Then the output transistor of comparator 1 is closed, and the UPS circuit operates normally in PWM mode.
In the case of a short circuit in the load of the +12V channel, for example, the anode potential of diode D11 becomes equal to 0V, so the potential of the inverting input of comparator 1 will become higher than the potential of the direct input, and the output transistor of the comparator will open. This will cause the closing of transistor Q4, which is normally open by the base current flowing through the circuit: Upom bus - R39 - R36 - b-e Q4 - “case”.
Turning on the output transistor of comparator 1 connects resistor R39 to the "case" and therefore transistor Q4 is passively turned off by zero bias. Closing transistor Q4 entails charging capacitor C22, which serves as a delay element for the protection. The delay is necessary for the reasons that during the process of the UPS entering mode, the output voltages on the +5V and +12V buses do not appear immediately, but as the high-capacity output capacitors are charged. The reference voltage from the source Uref, on the contrary, appears almost immediately after the UPS is connected to the network. Therefore, in the starting mode, comparator 1 switches, its output transistor opens, and if the delay capacitor C22 were missing, this would lead to the protection triggering immediately when the UPS is turned on to the network. However, C22 is included in the circuit, and the protection operates only after the voltage on it reaches the level determined by the values of resistors R37, R58 of the divider connected to the Upom bus and which is the base for transistor Q5. When this happens, transistor Q5 opens, and resistor R30 is connected through the low internal resistance of this transistor to the “case”. Therefore, a path appears for the base current of transistor Q6 to flow through the circuit: Uref - e-6 Q6 - R30 - k-e Q5 - “case”.
Transistor Q6 is opened by this current until saturation, as a result of which the voltage Uref = 5B, which powers transistor Q6 along the emitter, is applied through its low internal resistance to pin 4 of the control chip U4. This, as was shown earlier, leads to the stop of the digital path of the microcircuit, the disappearance of output control pulses and the cessation of switching of power transistors Q1, Q2, i.e. to protective shutdown. A short circuit in the +5V channel load will result in the anode potential of diode D11 being only about +0.8V. Therefore, the output transistor of the comparator (1) will be open, and a protective shutdown will occur. In a similar way, short-circuit protection is built in the loads of channels generating negative voltages (-5V and -12V) on comparator 2 of the U3 chip. Elements D12, R20 form a diode-resistive divider-sensor, connected between the output buses of the negative voltage generation channels. The controlled signal is the cathode potential of diode D12. During a short circuit in a -5V or -12V channel load, the potential of cathode D12 increases (from -5.8 to 0V for a short circuit in a -12V channel load and to -0.8V for a short circuit in a -5V channel load). In any of these cases, the normally closed output transistor of comparator 2 opens, which causes the protection to operate according to the above mechanism. In this case, the reference level from resistor R27 is supplied to the direct input of comparator 2, and the potential of the inverting input is determined by the values of resistors R22, R21. These resistors form a bipolarly powered divider (resistor R22 is connected to the bus Uref = +5V, and resistor R21 is connected to the cathode of diode D12, the potential of which in normal operation of the UPS, as already noted, is -5.8V). Therefore, the potential of the inverting input of comparator 2 in normal operation is maintained lower than the potential of the direct input, and the output transistor of the comparator will be closed.
Protection against output overvoltage on the +5V bus is implemented on elements ZD1, D19, R38, C23. Zener diode ZD1 (with a breakdown voltage of 5.1V) is connected to the +5V output voltage bus. Therefore, as long as the voltage on this bus does not exceed +5.1 V, the zener diode is closed, and transistor Q5 is also closed. If the voltage on the +5V bus increases above +5.1V, the zener diode “breaks through”, and an unlocking current flows into the base of transistor Q5, which leads to the opening of transistor Q6 and the appearance of voltage Uref = +5V at pin 4 of the control chip U4, i.e. . to protective shutdown. Resistor R38 is a ballast for the zener diode ZD1. Capacitor C23 prevents the protection from triggering during random short-term voltage surges on the +5V bus (for example, as a result of the voltage settling after a sudden decrease in the load current). Diode D19 is a decoupling diode.
The PG signal generation circuit in this switching power supply is dual-functional and is assembled on comparators (3) and (4) of the U3 microcircuit and transistor Q3.
The circuit is built on the principle of monitoring the presence of alternating low-frequency voltage on the secondary winding of the starting transformer T1, which acts on this winding only if there is a supply voltage on the primary winding T1, i.e. while the switching power supply is connected to the mains. Almost immediately after the UPS is turned on, the auxiliary voltage Upom appears on the capacitor SZO, which powers the control microcircuit U4 and the auxiliary microcircuit U3. In addition, the alternating voltage from the secondary winding of the starting transformer T1 through diode D13 and current-limiting resistor R23 charges capacitor C19. The voltage from C19 powers the resistive divider R24, R25. From resistor R25, part of this voltage is supplied to the direct input of comparator 3, which leads to the closing of its output transistor. The output voltage of the internal reference source of the microcircuit U4 Uref = +5B, which appears immediately after this, powers the divider R26, R27. Therefore, the reference level from resistor R27 is supplied to the inverting input of comparator 3. However, this level is chosen to be lower than the level at the direct input, and therefore the output transistor of comparator 3 remains in the off state. Therefore, the process of charging the holding capacity C20 begins along the chain: Upom - R39 - R30 - C20 - “body”. The voltage, which increases as capacitor C20 charges, is supplied to the inverse input 4 of the U3 microcircuit. The direct input of this comparator is supplied with voltage from resistor R32 of the divider R31, R32 connected to the Upom bus. As long as the voltage across the charging capacitor C20 does not exceed the voltage across resistor R32, the output transistor of comparator 4 is closed. Therefore, an opening current flows into the base of transistor Q3 through the circuit: Upom - R33 - R34 - 6th Q3 - “case”. Transistor Q3 is open to saturation, and the PG signal taken from its collector has a passive low level and prohibits the processor from starting. During this time, during which the voltage level on capacitor C20 reaches the level on resistor R32, the switching power supply manages to reliably enter the nominal operating mode, i.e. all its output voltages appear in full. As soon as the voltage on C20 exceeds the voltage removed from R32, comparator 4 will switch and its output transistor will open. This will cause transistor Q3 to close, and the PG signal taken from its collector load R35 becomes active (H-level) and allows the processor to start. When the switching power supply is turned off from the network, the alternating voltage disappears on the secondary winding of the starting transformer T1. Therefore, the voltage on capacitor C19 quickly decreases due to the small capacitance of the latter (1 µF). As soon as the voltage drop across resistor R25 becomes less than that across resistor R27, comparator 3 will switch and its output transistor will open. This will entail a protective shutdown of the output voltages of the control chip U4, because transistor Q4 will open. In addition, through the open output transistor of comparator 3, the process of accelerated discharge of capacitor C20 will begin along the circuit: (+)C20 - R61 - D14 - k-e of the output transistor of comparator 3 - “case”.
As soon as the voltage level at C20 becomes less than the voltage level at R32, comparator 4 will switch and its output transistor will close.
This will cause transistor Q3 to open and the PG signal to go to an inactive low level before the voltages on the UPS output buses begin to decrease unacceptably. This will initialize the computer's system reset signal and reset the entire digital part of the computer to its original state. Both comparators 3 and 4 of the PG signal generation circuit are covered by positive feedback using resistors R28 and R60, respectively, which speeds up their switching. A smooth transition to mode in this UPS is traditionally ensured using the forming chain C24, R41, connected to pin 4 of the control chip U4. The residual voltage at pin 4, which determines the maximum possible duration of the output pulses, is set by the divider R49, R41. The fan motor is powered by voltage from capacitor C14 in the -12V voltage generation channel through an additional decoupling L-shaped filter R16, C15.
Exits
All power supplies come with long bundles of wires sticking out from the back. The number of wires and available connectors for powering devices will vary from model to model, but some standard connections should be provided by all power supplies without exception.
Since voltage is the magnitude of the potential difference, each output has two wires: one for the specified voltage (for example, +12 V) and a wire against which the potential difference is measured. This wire is called the ground, "earth", "reference wire" or "common" wire, and these two wires form a loop: from the power supply to the consumer device, and then back to the power supply.
Because some of these closed loops carry small currents, they may share common ground wires.
Official photo of the Cooler Master power supply.
The main required connector is the 24-pin ATX12V v. 2.4, which provides main power via several pins of different voltages, and also has a number of special pins.
Of these special ones, we note only the “+5 standby” pin – standby power supply for the computer. This voltage is always supplied to the motherboard, even when the computer is turned off, provided that it remains plugged in and its power supply is working. The motherboard needs standby power in order to remain active.
Most PSUs also have an additional 8-pin motherboard connector with two +12V lines, and at least one 6 or 8-pin power connector for PCI Express.
The PCI Express slot can handle a maximum of 75W of graphics cards, so this slot provides additional power for modern GPUs.
Our particular power supply in question actually uses two PCI Express power connectors on the same lane for cost reasons. Therefore, if you have a really powerful video card, try to allocate an independent power line for it, do not share it with other devices.
The difference between 6 and 8-pin PCI Express connectors is two additional ground wires. This allows you to increase the current, satisfying the needs of the most power-hungry video cards.
Over the past few years, we have increasingly begun to notice power supplies proudly labeled “modular” (modular PSU). This simply means that they have detachable cables, allowing you to use only the number of cables and connectors you need without connecting anything unnecessary, thereby freeing up space inside the unit.
Photo source nix.ru
Our Cooler Master, like most, uses a fairly simple system for connecting modular cables.
Each connector has one +12V, +5V and +3.3V wire, as well as two ground wires, and depending on what device the cable is connected to, the connector at the other end will use either the corresponding or simplified wiring.
The Serial ATA (SATA) connector pictured above is used to connect power to hard drives, solid-state drives, and peripherals such as DVD drives.
This familiar connector is called intricately: “AMP MATE-N-LOK power connector 1-480424-0”. But everyone simply calls it Molex, despite the fact that this is just the name of the company that developed this connector. It provides one +12V and +5V terminal, and two ground wires.
On output wires, manufacturers can also save money or increase the price by using brighter or softer wires. The cross-section of the wire also plays an important role, since thicker wires have less resistance than thin ones, so they heat up less when current passes through them.
What to look for when choosing
At the beginning of our article, we said that most power supplies have the value of their maximum power in their name. In simple terms, electrical power is voltage times current (for example, 12 volts x 20 amps = 240 watts). Although this statement is not entirely technically accurate, for our purposes it is satisfactory.
Like most models, our power supply has a nameplate that contains basic information about how much power each voltage line can provide.
Photo source nix.ru
Here we see that the total maximum power of all +12 V lines is 624 W. Adding all the other powers, we end up with 760 W, not 650. What's wrong with that? The simple fact is that the +5 V lines (except for the standby line) and +3.3 V are created through the VRM using one of the +12 V lines.
And of course, all output voltages come from one source: the wall outlet. Thus, a power of 650 W is the maximum that the power supply can provide overall across all lines. That is, if you have a load of 600 W on the +12 V lines, then you have only 50 W left on all other lines. Fortunately, most equipment takes most of its power from 12V lines anyway, so the problem of an incorrectly selected power supply is rare.
To the right of the table with power specifications, there is an “80 Plus Bronze” icon on the nameplate. This is the efficiency rating used by the industry to meet requirements for power supply manufacturers. Efficiency also reflects the amount of total load the power supply can handle.
20%, 50% and 100% – percentage of load in relation to maximum power for standard systems
If our Cooler Master is loaded at exactly half of its maximum power, that is, 325 W, then its expected efficiency will be in the range of 80-85% depending on the network voltage (115/230 V).
This means the actual load of the power supply on the network is from 382 to 406 W. A higher 80 PLUS rating doesn't mean the power supply will give you more power, it's just more energy efficient - it wastes less power in all stages of filtering, rectification and conversion.
Also note that maximum efficiency is achieved between 50 and 100% load. Some manufacturers provide graphs showing what efficiency can be expected from their device at various loads and line voltages.
Official image from Cooler Master.
Efficiency graph for the Cooler Master V1300 Platinum power supply. The vertical scale is efficiency (efficiency), the horizontal scale is % of load relative to maximum power.
Sometimes it is useful to pay attention to this information, especially if you are going to fork out for a kilowatt power supply. If your computer draws close to this power limit, the efficiency of the power supply will be slightly reduced.
You may come across some “single-channel” and “multi-channel” (or combined - equipped with a switch) power supplies. The term "channel" in this case is simply another word for the specific voltage output by the PSU. Our Cooler Master has one 12V channel and various power connectors providing +12V line from this channel. A multi-channel power supply has two or more systems providing 12 volt lines, however there is a big difference in how this is implemented.
Multi-channel power supplies are widely used for servers or data centers for fault tolerance purposes - if one of the channels fails, the system’s functionality will not be affected. For regular computers, multi-channel PSUs may also be offered, but most likely you will encounter pseudo-multi-channel, when the manufacturer simply splits a single channel into two or three supposedly independent channels. For example, our test subject produces up to 52 amps along the +12V line, which is equivalent to 624 W of electricity. A cheap “multi-channel” version of such a power supply will supposedly have two +12 V channels in the specification, but in reality these are only two half-channels, each of which will provide only 26 A (or 312 W).
A good desktop power supply that uses quality components doesn't require multi-channel +12V at all, so don't worry about it!
↑ Final version of the scheme
with changed denominations:
Fragment excluded. The full version is available to patrons and full members of the community.
Forward!
Now the voltage is raised normally, there is a margin upward - I begin to torment him. At first, as a load on the 30+30V bus, there was a garland of 220V*60W light bulbs. Three jokes. When the voltage across them is ~60V, the total load on the power supply is only 18W, so a “reservoir” was added to cool the PEVs used as a load. PEVs are connected with a 10+5+5+5+5+10 Ohm garland. Oscillograms for different loads: Blue – lower key gate. Yellow – current shape in the primary (conversion using current protection resistors) Load 18W. (three lamps 220V*60W)
I launch the whole garland. Load 96W. (40 ohms)
I tilt one section to 10 ohms. Load 127W. (30 ohms)
Then the experiment failed - the protection worked. From overvoltage. At the controller power supply, 23.3V is almost the threshold. The water in the bucket managed to heat up to 45 degrees. The diodes of the rectifier bridge on the 30+30V bus also became very hot. There are SF56 switched on in pairs. Apparently Schottkys are asking here. The oscillograms show that the power supply is trying to tighten the voltage that “falls” under load, reducing the frequency. At the same time, the secondary voltage on the controller power supply also increases. The voltage sags at 30+30, from minimum to maximum – 3V. With a light load – 63.2V, at 127W – 60.2V. The result is a drawdown of 1.5V per shoulder - quite good. I thought it would be worse.
In general, I decided to continue the experiment. Reduced the voltage on the monitored bus to +24V. I had 24.5V and made it 23V. At this voltage, the 24V relays latched confidently, but the voltage on the self-supply bus did not exceed the permissible limits. At the same time, I accidentally tested the short circuit protection (aka current overload). The fact is that the wires to the load lamp are simply soldered, as can be seen in the photo. And there were scissors lying nearby. I began to pull the probe of the device towards me - the light bulb moved with its base towards the scissors - a click and silence. The power supply clearly turned off. It has trigger protection, so until the controller’s power is removed, or rather drops below 11V, it will not start again. I waited for the condenser to discharge in the primary and restart. The restart was successful, I loaded it at 25 Ohms and briefly at 20 Ohms. Everything starts and works. I was waiting for the current to be triggered, the voltage on the CS leg is growing, but still does not reach the level of the beginning of the limitation of -0.6V. I'm more worried about the rectifier - it starts to get very hot. We urgently need to find Schottky, Volt, something like 100.
But they tell the truth. The resonator works well under load. If, without load, some rustling is heard from the trance, and the power switches heat up slightly, then under load, a complete idyll sets in - the radiators are at room temperature, the trance does not rustle. True, I haven’t impregnated it with anything yet - maybe it won’t rustle after impregnation.
Something needs to be done with the rectifier. There are two options for alteration - increase the number of diodes or still install Schottky. The second option won.
I replaced the radiators on the power section and the rectifier. At the output - Schottky 20A 200V - connected by bridge in pairs. There are two more of them on the back side of the right radiator in the photo. I impregnated the transformer with NC varnish. Well, you need to check what happened:
Now the situation has improved dramatically. This is what happens with a load of 125W:
And at a load of 145W.
The most interesting thing is that the surges on the current-measuring resistor disappeared. What this is connected with - I cannot explain. I'm chasing The load floats in a bucket of water. Electrolysis processes on exposed leads can be observed almost immediately.
The radiator of the power switches did not change the temperature. The straightener heats up slightly, but not as quickly as on ultrafast. After a few minutes, the water in the bucket begins to heat up noticeably. It’s no longer comfortable to dip your finger, and bubbles form on the surface of the resistors - it’s obviously already hot there.
The rectifier radiator heats up to 40-45 degrees, the radiator of the power switches is cold, as if they are not working... For a power supply without airflow at an active load of almost 150 W, this is a good result... I wonder how a computer power supply without airflow would feel under similar conditions? Well, for the sake of familiarization, I tried out all sorts of nonsense along the way. Voltage waveform before rectifier with minimum load
Also, with maximum
Ripple at the outputs + - 30V. After the bridge, there are two Nichicon PL(M) 470uF 63V capacitors in the arm with an unknown resource (they were in a high-quality power supply unit that had worked for several years in 24*7*365 mode), bridged with a 1uF 250V film. With minimum load:
And the same thing, with the maximum.
Pulsations with a conversion frequency, so what the device appears in the lower corner is from the flashlight. The location of the probes relative to the power supply and each other slightly influences the scope and shape, so the result is approximate. The “needles” seem to be from switching diodes, we need to think about snubbers...
Surely readers will have questions. Where is the kilowatt? Give me a welder! Why didn’t I try it at a load above 150W? But I'm still just learning! (c) Besides, I don’t need load power higher than 60W, and even then the UMZCH radiators will drain onto the floor, and the neighbors will throw tomatoes at me. So in reality it will work at 10-15W per channel, and only on holidays. The current protection resistors are already set to limit the current to 2.5A for power switches, and I don’t see the need to select a different value yet.
Just for fun, here are the oscillograms of the launch:
And stops:
Blue – lower key gate. Yellow – controller power supply.
Well, now, actually, what the board and power supply were made for.
The joint came out right away. The power supply refused to start on banks of 10000 µF + 2200 µF in each arm of each channel. A total of 24400 uF per shoulder. The current is simply triggered. I had to delay the soft start even more. Now the capacitor Css=47uF. But this is not noticeable to the eye.
There is ringing silence in the speakers. At idle, the power switches, transformer, and resonant circuit capacitor heat up more. Everything is about 40 degrees. Iced schottky. Well, it’s quite logical, the efficiency of the resonator is higher at rated power, as is directly stated in the apnot.
What did you like in general?
1. Interesting. Informative. 2. The operation of the controller protection is flawless. It is unlikely that you will be able to burn the power switches. Unless you specially put nails on the board. 3. Well-chewed documentation. 4. Good efficiency for resonant topology.
Of the minuses.
1. Without instruments, by eye, nothing will work. 2. Winding with stranded wire. 3. The trans must be sectioned. (although you can section it yourself, but I ordered a ready-made one)
But I think many people have an LC meter and an oscilloscope? Yes, the oscillator must be galvanically isolated from the power supply - otherwise the boom is guaranteed... For example, I used the TS-180, switched on with Ktr = 1. There, all the windings will follow and we will get 220-230V.
We are planning to try the FSFR2100 - it is already on the road. Trying a resonator for lamps just for fun is purely experimental.
Well, that's all for now.
