Microprocessors are the most powerful consumers of energy in modern computers. The current consumption of a modern microprocessor can reach several tens of amperes. At the same time, the quality of the microprocessor supply voltage is the most important factor determining the stability of the entire system. How motherboard manufacturers solve the problem of providing the microprocessor with powerful and high-quality power is described in the article brought to your attention.
Preamble
The clock frequency of microprocessors is steadily growing and now reaches several GHz. Raise clock frequency microprocessor is accompanied by a significant increase in the power consumed by it, and, accordingly, leads to an increase in the temperature of the processor chip. In addition, the power consumption of microprocessors is also affected by an increase in the number of transistors on its chip (the more modern the processor, the more a high degree he has integration. Although CMOS transistors, which form the basis of microprocessors, consume scanty currents in the closed state, but when we are talking about several million transistors located on a processor chip, this is no longer necessary to neglect. The main energy consumption of CMOS transistors is carried out at the moment of its inclusion, and, naturally, the more often the transistors switch, the more energy they consume. As a result, millions of transistors switch from high frequency, are able to ensure the consumption of such a current by the microprocessor, the value of which already reaches 50 or more amperes. Thus, the processor crystal begins to heat up strongly, which leads to a significant deterioration in the switching processes of transistors and can disable them. At the same time, it is not possible to solve the problem solely by heat removal.
All this forces manufacturers to reduce the supply voltage of microprocessors, more precisely, the supply voltage of its core. Reducing the supply voltage can solve the problem of power dissipated on the microprocessor chip and lower its temperature. If the very first microprocessors of the 80x86 family had a supply voltage of +5V (and for the first time a voltage reduction to +3.3V was applied in the I80486), then the latest generation microprocessors can already work with a supply voltage of +0.5V (see the VR11 specification from Intel).
But the fact is that such low voltages are not produced by the system power supply. Recall that only voltages + 3.3V, + 5V and + 12V are formed at its output. Thus, the motherboard must have its own voltage regulator capable of lowering these "high-voltage" voltages to the level necessary to power the processor core, i.e. up to 0.5 - 1.6 V (fig.1).
Fig.1
Since this regulator provides the conversion of a constant voltage of + 12V to a constant voltage, but of a lower rating, the regulator was called DC-DC Converter (converter direct current into direct current). I would like to draw the attention of all specialists to the fact that the processor core voltage is now generated from the +12V voltage, and not from +5V or +3.3V, as it might seem more logical. The fact is that the +12V channel voltage is the highest, and therefore it is possible to create much more power in it at a lower current value. Thus, in modern computing systems+12V becomes the most important voltage, and it is in this channel that the largest currents flow. Incidentally, this is also reflected in the standards that describe the requirements for system blocks power supply, according to which, the load capacity of the +12V channel is maximum. In addition, the power supply output must have two +12V voltage channels (+12V1 and +12V2), and the current control in each of these channels must be carried out independently. One of these channels, namely +12V2, is intended just for powering the processor core, and it is subject to the most stringent stability requirements and the smallest tolerances for deviations from the nominal value.
It is also necessary to note the following point. Since the power consumed by the processors is quite large (it can reach almost 100 W), the voltage conversion must be carried out by the pulse method. Linear transformation is not capable of providing a sufficiently high efficiency at such a power, and will lead to significant losses, and, consequently, to heating of the converter elements. To date, only pulse conversion makes it possible to obtain an efficient and economical power supply with small dimensions and with an acceptable cost of execution. Thus, on the system board there is a DC-DC Converter, which is a step-down type step-down converter (Step Down or Trim).
DC-DC Step Down Converter
A basic circuit for a DC buck converter is shown in fig.2. I would like to note that regulators of this type in modern imported literature are called Buck Converter or Buck Regulator. Transistor Q1 in this circuit is a key that, by closing / opening, creates a pulsed voltage from a constant voltage.
Fig.2
In this case, the amplitude of the generated pulses is 12V. To improve conversion efficiency, Q1 must switch at a high frequency (the higher the frequency, the more efficient the conversion). In real motherboard regulator circuits, the switching frequency of the converter transistors can be in the range from 80 kHz to 2 MHz.
Further, the resulting impulse voltage is smoothed by the inductor L1 and the electrolytic capacitor C1. As a result, a constant voltage is created on C1, but of a smaller magnitude. In this case, the magnitude of the created DC voltage will be proportional to the width of the pulses received at the output of Q1. If transistor Q1 opens for a longer time, then the energy stored on L1 will also be greater, which, as a result, leads to an increase in the voltage on C1. Accordingly, and vice versa - with a shorter duration of the open state of the transistor Q1, the voltage across C1 decreases. This method of direct voltage regulation is called pulse-width modulation - PWM (PWM - Pulse Width Modulation).
A very important element of the circuit is the diode D1. This diode maintains the load current created by the inductor L1 during those periods of time when the transistor Q1 is closed. In other words, when Q1 is open, the inductor current and load current are provided by the power supply, while energy is stored in the inductor. After Q1 turns off, the load current is maintained by the energy stored in the inductor. This current flows through D1, i.e. the inductor energy is spent on maintaining the load current ( see fig.3).
Fig.3
However, in practical schemes ah buck regulators that generate high currents, there are some problems. The fact is that most diodes do not have sufficient speed, and also have a relatively large open resistance. p-n junction. All this is not of decisive importance at low load currents. But at high currents, all this leads to significant losses, strong heating of the diode D1, voltage surges and the occurrence of reverse currents through the diode when switching transistor Q1. That is why this scheme was finalized in order to increase performance and reduce losses, as a result of which, instead of diode D1, another transistor was used - Q2 (fig.4).
Fig.4
Transistor Q2, being a MOSFET, has a very low on-resistance and is very fast. Since Q2 performs the function of a diode, it works synchronously with Q1, but strictly in antiphase, i.e. at the moment of locking Q1, transistor Q2 opens, and, conversely, when Q1 is open, transistor Q2 is closed (see fig.5).
Fig.5
It is this solution that is the only possible one for organizing voltage converters on modern motherboards, where, as we have already said, very high currents are required to power the processor.
Having finished the review of the basic technologies for organizing switching voltage regulators, we turn to the consideration of practical schemes for their implementation.
Fundamentals of the organization of processor core voltage regulators
It’s worth mentioning right away that for quite a long time, manufacturers of the element base began producing specialized microcircuits designed to build switching voltage regulators for motherboards. personal computers. The use of such specialized microcircuits makes it possible to improve the characteristics of regulators, ensure their high compactness and reduce the cost of both the regulators themselves and the cost of their development. To date, there are three types of microcircuits used in motherboard voltage regulators designed to power the processor core:
- the main controller (Main Controller), which is also called as a PWM controller (PWM-Controller) or a voltage regulator (Voltage Regulator);
- MOS transistor control driver (Synchronous-Rectifier MOSFET Driver);
- a combined controller that combines the functions of both a PWM controller and a MOSFET driver.
Taking into account the variety of microcircuits used, in modern motherboards we can find two main options for building switching voltage regulators for powering the processor core.
I option. This option is typical for entry-level motherboards with low performance, i.e. it is most often used on motherboards that do not provide for the use of high-performance and powerful processors. In this version, the control of the power transistors of the converter is carried out by a microcircuit of the combined controller. This chip provides the following functions:
- reading the state of the processor supply voltage identification signals (VIDn);
- generation of PWM signals for synchronous control of power MOSFETs;
- control of the value of the generated supply voltage;
- implementation of current protection of power MOSFETs;
- generating a signal confirming the correct operation of the regulator and the presence of the correct voltage at its output to power the processor core (PGOOD signal).
An example of such a variant of the voltage regulator is shown in fig.6. In this case, as we can see, the power transistors are directly connected to the outputs of the combined controller chip. The HIP6004 chip was often used as such a controller.
Fig.6
II option. This option is typical for motherboards designed to work with high-performance processors. Since a high-performance processor implies the consumption of high currents, the voltage regulator is made multi-channel (Fig. 7).
Fig.7
The presence of several channels allows you to reduce the amount of current for each channel, i.e. reduce the currents switched by MOSFETs. This, in turn, increases the reliability of the entire circuit and allows the use of less powerful transistors, which has a positive effect on the cost of both the regulator itself and the motherboard as a whole.
This version of the regulator is characterized by the use of two types of microcircuits: the main PWM controller and MOS transistor drivers. Synchronous control of MOSFETs is carried out by drivers, each of which can control one or two pairs of transistors. The driver provides anti-phase switching of transistors in accordance with input signal(most often referred to as PWM), which determines the switching frequency and the open state of the transistors. The number of driver chips corresponds to the number of switching regulator channels.
All drivers are managed by the main controller (Main Controller), the main functions of which include:
-pulse shaping to control MOSFET drivers;
- changing the width of these control pulses in order to stabilize the output voltage of the regulator;
- control of the output voltage of the regulator;
- providing current protection of MOSFETs;
- reading the state of the processor supply voltage identification signals (VIDn).
In addition to these functions, other auxiliary functions may be performed, the presence of which will be determined by the type of main controller used.
The general scheme of such a voltage regulator is shown in fig.8. Most modern master controllers are 4-channel, i.e. have 4 PWM output signals to drive transistor drivers.
Fig.8
So, at the current time, voltage regulators for the processor core can be 2-channel, 3-channel and 4-channel.
An example of the implementation of a 2-channel regulator is presented on fig.9. This regulator is built using the HIP6301 type Main Controller chip, which, in principle, is four-channel, but two channels were left unused.
Fig.9
HIP6601B chips are used as key drivers in this scheme.
An example of the implementation of a 4-channel controller using the same Main Controller is presented in fig.10.
Fig.10
The HIP6301 controller decodes the processor core voltage based on a 5-bit identification code (VID0 - VID4) and generates output PWM pulses with a frequency of up to 1.5 MHz. In addition, it generates a PGOOD (good power) signal if the processor core voltage generated by the voltage regulator matches the value set using the VIDn signals.
Features of multi-channel regulators
When using multi-channel voltage regulators, there are several problems that motherboard designers have to solve. The fact is that each channel is a switching regulator, which, switching at a high frequency, creates current pulses at its output. These pulses, of course, must be smoothed out, and electrolytic capacitors and chokes are used for this. But the fact is that due to the large current load, the capacitance of the capacitors and the inductance of the inductors, nevertheless, is not enough to create a truly constant voltage, as a result of which ripples are observed on the processor power bus (fig.11). Moreover, neither an increase in the number of capacitors, nor an increase in the capacitance of capacitors and inductance of inductors, nor an increase in the conversion frequency (unless we talk about increasing the frequency by several times) saves from these ripples. Naturally, these ripples can lead to unstable operation of the processor.
Fig.11
The way out of the problem, just found in the use of a multi-channel architecture of the voltage regulator. But only using several parallel channels to solve the problem, anyway, will not succeed. It is necessary to make sure that the keys of different channels switch with a phase shift, i.e. they should open one by one. This will make sure that each channel will maintain the output current of the regulator for a strictly allotted period of time. In other words, smoothing capacitors will be constantly charged, but from different channels at different times. So, for example, when using a 4-channel regulator, the output capacitors are recharged four times in one clock cycle of the controller, i.e. pulsed currents of individual channels are out of phase with respect to each other by 90° (see fig.12). This corresponds to a 4-fold increase in the conversion frequency, and if the switching frequency of the transistors of each channel is 0.5 MHz, then the pulse frequency on the smoothing capacitor will already be 2 MHz.
Fig.12
Thus, the PWM pulses that are generated at the output of the main controller chip (PWM output signals) must follow with a certain phase shift, and this phase shift is determined by the internal architecture of the chip and is usually set already at the stage of chip design. But some controllers allow you to configure them according to different modes operation: 2-phase, 3-phase or 4-phase control (how this is done can be found in the descriptions for the controllers themselves).
With this lesson, I begin a series of articles on switching regulators, digital regulators, and output power control devices.
The goal that I set is the development of a controller for a refrigerator on a Peltier element.
We will do an analogue of my development, only implemented on the basis of the Arduino board.
- This development interested many, and letters rained down on me with requests to implement it on Arduino.
- The development is ideal for studying the hardware and software of digital controllers. In addition, it combines many of the tasks studied in previous lessons:
- measurement of analog signals;
- working with buttons;
- connection of indication systems;
- temperature measurement;
- work with EEPROM;
- connection with a computer;
- parallel processes;
- and much more.
I will develop the development sequentially, step by step, explaining my actions. What will be the result - I do not know. I hope for a full-fledged working project of the refrigerator controller.
I don't have a finished project. I will write lessons according to the current state, so during the tests it may turn out that at some stage I made a mistake. I will correct. This is better than me debugging the development and issuing ready-made solutions.
Differences between development and prototype.
The only functional difference from the prototype development on the PIC controller is the absence of a fast voltage regulator that compensates for the ripple of the supply voltage.
Those. this version of the device must be powered by a stabilized power supply with a low level of ripple (no more than 5%). These requirements are met by all modern impulse blocks nutrition.
And the power supply option from an unstabilized power supply (transformer, rectifier, capacitive filter) is excluded. The speed of the Arduino system does not allow for a fast voltage regulator. I recommend reading about the power requirements of the Peltier element.
Development overall structure devices.
At this stage, you need to understand in general terms:
- what elements the system consists of;
- on which controller to execute it;
- are there enough conclusions and functionality controller.
I imagine the controller as a “black box” or “garbage pit” and connect everything I need to it. Then I look if, for example, the Arduino UNO R3 board is suitable for these purposes.
In my interpretation it looks like this.
I drew a rectangle - the controller and all the signals necessary to connect the elements of the system.
I decided that I need to connect to the board:
- LCD indicator (for displaying results and modes);
- 3 buttons (for control);
- error indication LED;
- fan control key (to turn on the hot side radiator fan);
- switching stabilizer key (for adjusting the power of the Peltier element);
- analog input for measuring the load current;
- analog input for measuring load voltage;
- temperature sensor in the chamber (accurate 1-wire sensor DS18B20);
- radiator temperature sensor (have not yet decided which sensor, rather DS18B20 too);
- computer communication signals.
There were 18 signals in total. The Arduino UNO R3 or Arduino NANO board has 20 pins. There are still 2 conclusions left in reserve. Maybe you want to connect another button, or an LED, or a humidity sensor, or a cold side fan ... We need 2 or 3 analog inputs, the board has 6. That is. everything suits us.
You can assign pin numbers immediately, you can during development. I appointed immediately. Connection occurs through connectors, you can always change. Keep in mind that pin assignments are not final.
impulse stabilizers.
For accurate temperature stabilization and the operation of the Peltier element in the optimal mode, it is necessary to adjust the power on it. Regulators are analog (linear) and pulse (key).
Analog regulators are a regulating element and a load connected in series to a power source. By changing the resistance of the regulating element, the voltage or current on the load is adjusted. As a regulating element, as a rule, a bipolar transistor is used.
The control element operates in linear mode. It is allocated "extra" power. At high currents, stabilizers of this type are very hot, have a low efficiency. A typical linear voltage regulator is the 7805 chip.
This option does not suit us. We will make a pulse (key) stabilizer.
Switching stabilizers are different. We need a step-down switching regulator. The load voltage in such devices is always lower than the supply voltage. The circuit of the step-down switching regulator looks like this.
And this is a diagram of the regulator.
Transistor VT operates in the key mode, i.e. it can only have two states: open or closed. The control device, in our case, the microcontroller, switches the transistor with a certain frequency and duty cycle.
- When the transistor is open, current flows through the circuit: power supply, transistor switch VT, inductor L, load.
- When the key is open, the energy stored in the inductor is supplied to the load. Current flows through the circuit: inductor, VD diode, load.
Thus, the constant voltage at the output of the regulator depends on the ratio of the open time (topen) and private key(tclose), i.e. on the duty cycle of the control pulses. By changing the duty cycle, the microcontroller can change the voltage at the load. Capacitor C smooths out the output voltage ripple.
The main advantage of this method of regulation is high efficiency. The transistor is always on or off. Therefore, little power is dissipated on it - always either the voltage across the transistor is close to zero, or the current is 0.
This classical scheme step-down regulator. In it, the key transistor is torn off from the common wire. The transistor is difficult to drive, requiring special bias circuits to the supply voltage rail.
So I changed the schema. In it, the load is disconnected from the common wire, but a key is attached to the common wire. This solution allows you to control the transistor switch from the microcontroller signal using a simple current driver-amplifier. 
- When the key is closed, the current enters the load through the circuit: power supply, inductor L, key VT (the current path is shown in red).
- When the key is open, the energy accumulated in the inductor is returned to the load through the regenerative diode VD (the current path is shown in blue).
Practical implementation of the key regulator.
We need to implement a switching regulator node with the following functions:
- the actual key controller (key, choke, regenerative diode, smoothing capacitor);
- load voltage measurement circuit;
- regulator current measurement circuit;
- hardware overcurrent protection.
I, with virtually no changes, took the regulator circuit from.
Scheme of a switching regulator for working with an Arduino board.
I used MOSFET transistors IRF7313 as a power switch. In an article on increasing the power of the Peltier element controller, I wrote in detail about these transistors, about a possible replacement, and about the requirements for key transistors for this circuit. Here is a link to the technical documentation.
On transistors VT1 and VT2, a key MOSFET transistor driver is assembled. This is just a current amplifier, in terms of voltage it even attenuates the signal to about 4.3 V. Therefore, the key transistor must be low-threshold. There are different ways to implement MOSFET drivers. Including using integrated drivers. This option is the easiest and cheapest.
To measure the voltage at the load, a divider R1, R2 is used. With such resistor values and a reference voltage source of 1.1 V, the measurement range is 0 ... 17.2 V. The circuit allows you to measure the voltage at the second load terminal relative to the common wire. We calculate the voltage at the load, knowing the voltage of the power source:
Uload = Usupply - Umeasured.
It is clear that the measurement accuracy will depend on the stability of maintaining the voltage of the power source. But we do not need high accuracy in measuring voltage, current, load power. We need to accurately measure and maintain only the temperature. We will measure it with high accuracy. And if the system shows that the Peltier element has a power of 10 W, but in fact it will be 10.5 W, this will not affect the operation of the device in any way. This applies to all other energy parameters.
The current is measured using a resistor-current sensor R8. Components R6 and C2 form a simple low pass filter.
The simplest hardware protection is assembled on the R7 and VT3 elements. If the current in the circuit exceeds 12 A, then the voltage across resistor R8 will reach the transistor opening threshold of 0.6 V. The transistor will open and close the RES (reset) pin of the microcontroller to ground. Everything should turn off. Unfortunately, the threshold for such protection is determined by the base-emitter voltage of the bipolar transistor (0.6 V). Because of this, the protection only works at significant currents. You can use an analog comparator, but this will complicate the circuit.
The current will be measured more accurately with an increase in the resistance of the current sensor R8. But this will lead to the release of significant power on it. Even with a resistance of 0.05 ohms and a current of 5 A, 5 * 5 * 0.05 = 1.25 watts is dissipated on the resistor R8. Note that resistor R8 has a power of 2 watts.
Now, what current are we measuring. We measure the current consumption of the switching regulator from the power supply. The circuit for measuring this parameter is much simpler than the circuit for measuring the load current. Our load is “untied” from the common wire. For the system to work, it is necessary to measure the electric power on the Peltier element. We calculate the power consumed by the regulator by multiplying the power supply voltage by the current drawn. Let's assume that our regulator has an efficiency of 100% and decide that this is the power on the Peltier element. In fact, the efficiency of the regulator will be 90-95%, but this error will not affect the operation of the system in any way.
Components L2, L3, C5 are a simple RFI filter. It may not be necessary.
Calculation of the throttle of the key stabilizer.
The throttle has two parameters that are important to us:
- inductance;
- saturation current.
The required inductance of the inductor is determined by the PWM frequency and the allowable inductor current ripple. There is a lot of information on this topic. I will give the most simplified calculation.
We applied voltage to the inductor and the current through it began to increase the current. Increase, but did not appear, because some current was already flowing through the inductor at the moment I was turned on).
The transistor is open. The voltage is connected to the throttle:
Uchoke = Usupply - Uload.
The current through the inductor began to increase according to the law:
Ichoke = Uchoke * topen / L
- topen - public key pulse duration;
- L - inductance.
Those. the value of the ripple current of the inductor or how much the current has increased during the time of the open key is determined by the expression:
Ioff - Ion = Uchoke * topen / L
The load voltage may change. And it determines the voltage at the throttle. There are formulas that take this into account. But in our case, I would take the following values:
- supply voltage 12 V;
- minimum voltage on the Peltier element 5 V;
- Means maximum voltage on the throttle 12 - 5 \u003d 7 V.
The duration of the pulse of the public key topen is determined by the frequency of the PWM period. The higher it is, the less inductance the inductor needs. The maximum PWM frequency of the Arduino board is 62.5 kHz. I will tell you how to get such a frequency in the next lesson. We will use it.
Let's take the worst case - PWM switches exactly in the middle of the period.
- Period duration 1/62500 Hz = 0.000016 sec = 16 µs;
- Public key duration = 8 µs.
Current ripple in such circuits is usually set to 20% of the average current. Not to be confused with output voltage ripple. They are smoothed out by capacitors at the output of the circuit.
If we allow a current of 5 A, then we take a current ripple of 10% or 0.5 A.
L = Uchoke * topen / Ipulsation = 7 * 8 / 0.5 = 112 μH.
Inductor saturation current.
Everything in the world has a limit. And the throttle too. At some current, it ceases to be an inductance. This is the saturation current of the inductor.
In our case, the maximum inductor current is defined as the average current plus ripple, i.e. 5.5 A. But it is better to choose the saturation current with a margin. If we want hardware protection to work in this version of the circuit, then it must be at least 12 A.
The saturation current is determined by the air gap in the inductor's magnetic core. In articles about Peltier element controllers, I talked about the design of the throttle. If I start to expand this topic in detail, then we will leave Arduino, programming, and I don’t know when we will return.
My throttle looks like this.
Naturally, the inductor winding wire must be of sufficient cross section. The calculation is simple - the determination of heat losses due to the active resistance of the winding.
Active winding resistance:
Ra = ρ * l / S,
- Ra is the active resistance of the winding;
- Ρ – resistivity of the material, for copper 0.0175 Ohm mm2/m;
- l is the length of the winding;
- S is the cross section of the winding wire.
Thermal losses on the active resistance of the inductor:
The key regulator draws a decent current from the power supply and this current should not be allowed to pass through the Arduino board. The diagram shows that the wires from the power supply are connected directly to the blocking capacitors C6 and C7.
The main pulse currents of the circuit pass through the circuit C6, load, L1, D2, R8. This chain must be closed by links with a minimum length.
The common wire and power bus of the Arduino board are connected to the blocking capacitor C6.
The signal wires between the Arduino board and the key regulator module must be of the minimum length. Capacitors C1 and C2 are best placed on the connectors to the board.
I have assembled the circuit board. Soldered only the necessary components. My assembled circuit looks like this.
I set the PWM to 50% and checked the operation of the circuit.
- When powered from a computer, the board formed a given PWM.
- With autonomous power from an external power supply, everything worked great. Pulses with good fronts were formed on the throttle, there was a constant voltage at the output.
- When I turned on the power from both the computer and the external power supply at the same time, the Arduino board burned out.
My stupid mistake. Let me tell you so no one will repeat it. In general, when connecting an external power supply, you must be careful, ring all connections.
The following happened to me. There was no VD2 diode in the circuit. I added it after this trouble. I figured that the board can be powered from an external source through the Vin pin. He himself wrote in lesson 2 that the board can be powered from an external source through the connector (RWRIN signal). But I thought it was the same signal, only on different connectors.
0 Category: . You can bookmark.The device has a menu. Entering the menu, moving in it and exiting is carried out by simultaneously pressing the "H" and "B" buttons. In the process of this, the corresponding mnemonic appears on the indicator, "H-U", "B-U" (lower and upper voltage limits), "H-I", "B-I" (lower and upper current limits), "P-0" , "P-1" - manual or automatic mode, switching on the relay after the return of voltage or current within the specified limits. "-З-" indicates that the set parameters are written to non-volatile memory and the menu mode is exited. In the menu mode, the "H" and "B" buttons allow you to change the parameters in one direction or another, and holding the button for about 3 seconds accelerates the parameter change. The change occurs in a circle, 99.8-99.9-0.0-0.01, etc. When the set limits are exceeded, the relay turns off, the indicator starts flashing, signaling an accident. That. the device allows both charging and discharging the battery up to a certain voltage. Moreover, auto mode allows you to keep the battery constantly charged, and manual, to control the capacity of the battery, in A / hours.
A few notes. Don't forget to power 74HC595, 16n-+5V, 8n-ground. On buttons, it is better to use a pair of 3K3 and 10K resistors. The polarity of the indicator does not matter, it is selected by a resistor on the 11th leg of the controller (as in the diagram).
Application example for charging/discharging AB:
Hex file for PIC16F676 microcontroller, with control functions.
You do not have access to download files from our server- firmware file for voltammeter with parameters Umax=99.9V; Imax=9.99A; Pmax=99.9/999W; Cmax=9.99 A/h.
You do not have access to download files from our server- voltammeter hex_file with truncated functions, only Umax=99.9V and Imax=9.99A
Creation motherboards with an increased number of processor power phases, it is gradually becoming a kind of competition between motherboard manufacturers. For example, quite recently, Gigabyte produced boards with 12-phase processor power supplies, but in the boards it currently produces, the number of phases has grown to 24. But is it really necessary to use such a large number of power phases and why some manufacturers constantly increase them, trying to it is reasonable to prove that the more the better, while others are content with a small number of power phases? Maybe a large number of processor power phases is nothing more than a marketing gimmick designed to attract consumers' attention to their products? In this article, we will try to give a motivated answer to this question, and also consider in detail the principles of operation of multi-phase switching power supplies for processors and other elements of motherboards (chipsets, memory, etc.).
A bit of history
As you know, all components of motherboards (processor, chipset, memory modules, etc.) are powered by a power supply that is connected to a special connector on the motherboard. Recall that on any modern motherboard there is a 24-pin ATX power connector, as well as an additional 4-pin (ATX12V) or 8-pin (EPS12V) power connector.
All power supplies generate a constant voltage of ±12, ±5 and +3.3 V, however, it is clear that different motherboard microcircuits require a constant voltage of other denominations (moreover, different microcircuits require different supply voltages), and therefore the problem arises of converting and stabilizing the constant voltage received from the power supply into the DC voltage required to power a specific motherboard chip (DC-DC conversion). To do this, motherboards use appropriate voltage converters (converters), which lower the nominal voltage of the power supply to the required value.
There are two types of DC-DC converters: linear (analog) and pulse. Linear voltage converters on motherboards are no longer found today. In these converters, the voltage is reduced by dropping part of the voltage on the resistive elements and dissipating part of the power consumed in the form of heat. Such converters were supplied with powerful radiators and were very hot. However, with the growth of power (and, accordingly, currents) consumed by motherboard components, linear voltage converters were forced to be abandoned, since there was a problem of their cooling. All modern motherboards use switching DC-DC converters, which heat up much less than linear ones.
A step-down DC/DC converter for powering a processor is often referred to as a VRM (Voltage Regulation Module) or VRD (Voltage Regulator Down). The difference between VRM and VRD is that the VRD module is located directly on the motherboard, while the VRM is an external module installed in a special slot on the motherboard. Currently, external VRM modules are practically not found, and all manufacturers use VRD modules. However, the name VRM itself has taken root so much that it has become common and now it is even used to refer to VRD modules.
Switching voltage regulators used for the chipset, memory and other microcircuits of motherboards do not have their own specific name, however, they do not differ in principle from VRD. The difference is only in the number of power phases and output voltage.
As you know, any voltage converter is characterized by input and output supply voltage. As for the output supply voltage, it is determined by the specific microcircuit for which the voltage regulator is used. But the input voltage can be either 5 or 12 V.
Previously (during Intel processors Pentium III) used 5 V input voltage for switching voltage regulators, but subsequently motherboard manufacturers began to use 12 V input voltage more often, and now all boards use 12 V supply voltage as the input voltage of switching voltage regulators.
The principle of operation of a single-phase switching supply voltage regulator
Before proceeding to the consideration of multi-phase switching supply voltage regulators, let's consider the principle of operation of the simplest single-phase switching voltage regulator.
Switching Voltage Regulator Components
The switching power supply voltage buck converter basically contains a PWM controller (PWM controller) - an electronic key that is controlled by a PWM controller and periodically connects and disconnects the load to the input voltage line, as well as an inductive-capacitive LC filter to smooth out output voltage ripples . PWM is an abbreviation for Pulse Wide Modulation (pulse width modulation, PWM). The principle of operation of a pulsed step-down voltage converter is as follows. The PWM controller creates a sequence of control voltage pulses. A PWM signal is a sequence of rectangular voltage pulses characterized by amplitude, frequency, and duty cycle (Fig. 1).
Rice. 1. PWM signal and its main characteristics
The duty cycle of a PWM signal is the ratio of the time interval during which the signal has high level, to the period of the PWM signal: = / T.
The signal generated by the PWM controller is used to control the electronic key, which periodically, at the frequency of the PWM signal, connects and disconnects the load to the 12 V power line. The amplitude of the PWM signal must be such that it can be used to control the electronic key.
Accordingly, the output electronic key there is a sequence of rectangular pulses with an amplitude of 12 V and a repetition rate equal to the frequency of the PWM pulses. It is known from the course of mathematics that any periodic signal can be represented as a harmonic series (Fourier series). In particular, a periodic sequence of rectangular pulses of the same duration, when presented as a series, will have a constant component inversely proportional to the duty cycle of the pulses, that is, directly proportional to their duration. By passing the received pulses through a low-pass filter (LPF) with a cutoff frequency much lower than the pulse repetition rate, this constant component can be easily isolated, obtaining a stable constant voltage. Therefore, pulse voltage converters also contain a low-frequency filter that smoothes (rectifies) a sequence of rectangular voltage pulses. Structural block diagram such a pulsed step-down voltage converter is shown in fig. 2.
Rice. 2. Structural block diagram of such a pulsed step-down
voltage converter
Well, now let's look at the elements of a pulsed buck supply voltage converter in more detail.
Electronic key and control driver
A pair of n-channel MOSFETs (MOSFETs) is always used as an electronic key for switching power supply voltage converters of motherboard components, connected in such a way that the drain of one transistor is connected to the 12 V supply line, the source of this transistor is connected to the output point and drain of another transistor, and the source of the second transistor is grounded. The transistors of this electronic switch (sometimes called a power switch) work in such a way that one of the transistors is always in the open state, and the other is in the closed state.
To control the switching of MOSFETs, control signals are applied to the gates of these transistors. The control signal of the PWM controller is used to switch the MOSFETs, however, this signal is not fed directly to the gates of the transistors, but through a special chip called a MOSFET driver or power phase driver. This driver controls the switching of the MOSFETs at a frequency set by the PWM controller, applying the required switching voltages to the gates of the transistors.
When the transistor connected to the 12 V supply line is turned on, the second transistor, connected through its drain to the source of the first transistor, is turned off. In this case, the 12 V supply line is connected to the load through a smoothing filter. When the transistor connected to the 12V supply line is closed, the second transistor is turned on and the 12V supply line is disconnected from the load, but the load is connected to ground through a smoothing filter at this moment.
Low pass LC filter
The smoothing, or low-pass, filter is an LC filter, that is, an inductance connected in series with the load and a capacitance connected in parallel with the load (Fig. 3).
Rice. 3. Scheme of a single-phase pulse voltage converter
As is known from the physics course, if a harmonic signal of a certain frequency is applied to the input of such an LC filter U in (f), then the voltage at the filter output U out (f) depends on the reactances of the inductance (Z L = j2fc) and capacitor Z c = 1/(j2fc). The transfer coefficient of such a filter K(f) =(U out (f))/(U in (f)) can be calculated by considering a voltage divider formed by frequency dependent resistances. For an unloaded filter, we get:
K(f) = Z c /(Z c + Z L)= 1/(1 – (2 f) 2LC)
Or, if we introduce the notation f0 = 2/, then we get:
K(f) = 1/(1 – (f/f0) 2)
It can be seen from this formula that the transfer coefficient of an unloaded ideal LC filter increases indefinitely with approaching the frequency f0, and then, at f>f0, decreases proportionally 1/f2. At low frequencies (f
As already noted, smoothing voltage pulses using an LC filter is necessary so that the filter cutoff frequency f 0 = 2/ was significantly lower than the repetition rate of voltage pulses. This condition allows you to choose the necessary capacitance and inductance of the filter. However, let's digress from the formulas and try to explain the principle of the filter in a simpler language.
At the moment when the power switch is open (transistor T 1 is open, transistor T 2 is closed), energy from the input source is transferred to the load through the inductance L in which energy is stored. The current flowing through the circuit does not change instantly, but gradually, since the EMF that occurs in the inductance prevents the current from changing. At the same time, the capacitor installed in parallel with the load is also charged.
After the power switch closes (transistor T 1 is closed, transistor T 2 is open), the current from the input voltage line does not flow into the inductance, but according to the laws of physics, the emerging induction EMF maintains the current direction. That is, during this period, the current is supplied to the load from the inductive element. In order for the circuit to close and the current to flow to the smoothing capacitor and to the load, the transistor T 2 opens, providing a closed circuit and current flow along the path inductance - capacitance and load - transistor T 2 - inductance.
As already noted, using such a smoothing filter, you can get a voltage at the load that is proportional to the duty cycle of the PWM control pulses. However, it is clear that with this method of smoothing, the output voltage will have ripples of the supply voltage relative to some average value (output voltage) - fig. 4. The magnitude of the voltage ripple at the output depends on the switching frequency of the transistors, the value of the capacitance and inductance.
Rice. 4. Voltage ripple after smoothing with an LC filter
Output voltage stabilization and PWM controller functions
As already noted, the output voltage depends (for a given load, frequency, inductance and capacitance) on the duty cycle of the PWM pulses. Since the current through the load changes dynamically, the problem arises of stabilizing the output voltage. This is done in the following way. PWM controller that generates transistor switching signals is connected to the load in a loop feedback and continuously monitors the output voltage at the load. Inside the PWM controller, a reference supply voltage is generated, which should be on the load. The PWM controller constantly compares the output voltage with the reference voltage, and if a mismatch occurs U, then this error signal is used to change (correct) the duty cycle of the PWM pulses, that is, the change in the duty cycle of the pulses ~ U. Thus, the stabilization of the output voltage is realized.
Naturally, the question arises: how does the PWM controller know about the required supply voltage? For example, if we talk about processors, then, as you know, the supply voltage different models processor may be different. In addition, even for the same processor, the supply voltage can dynamically change depending on its current load.
The PWM controller learns about the required nominal supply voltage by the VID (Voltage Identifier) signal. For modern processors Intel Core i7 processors that support the VR 11.1 power specification, the VID signal is 8-bit, and for legacy processors that are compatible with the VR 10.0 specification, the VID signal was 6-bit. The 8-bit VID signal (a combination of 0 and 1) allows you to set 256 different levels of processor voltage.
Limitations of a single-phase switching supply voltage regulator
The single-phase circuit of the switching supply voltage regulator considered by us is simple in execution, but it has a number of limitations and disadvantages.
If we talk about the limitation of a single-phase switching supply voltage regulator, then it lies in the fact that MOSFETs, inductances (chokes), and capacitances have a limit on the maximum current that can be passed through them. For example, for most MOSFET transistors that are used in motherboard voltage regulators, the current limit is 30 A. At the same time, the processors themselves, with a supply voltage of about 1 V and a power consumption of more than 100 W, consume more than 100 A. It is clear that if at such a current strength a single-phase supply voltage regulator is used, then its elements will simply “burn out”.
If we talk about the disadvantage of a single-phase switching supply voltage regulator, then it lies in the fact that the output supply voltage has ripples, which is highly undesirable.
In order to overcome the current limitations of switching voltage regulators, as well as to minimize output voltage ripple, polyphase switching voltage regulators are used.
Multi-phase switching voltage regulators
In polyphase switching voltage regulators, each phase is formed by a MOSFET switching driver, a pair of MOSFETs themselves, and an LC smoothing filter. In this case, one multichannel PWM controller is used, to which several power phases are connected in parallel (Fig. 5).
Rice. 5. Structural scheme multiphase switching supply voltage regulator
Application N-phase regulator supply voltage allows you to distribute the current over all phases, and therefore, the current flowing through each phase will be in N times less than the load current (in particular, the processor). For example, if you use a 4-phase processor supply voltage regulator with a current limit of 30 A in each phase, then the maximum current through the processor will be 120 A, which is quite enough for most modern processors. However, if processors with a TDP of 130 W are used or the possibility of overclocking the processor is expected, then it is advisable to use not a 4-phase, but a 6-phase switching processor supply voltage regulator or use chokes, capacitors and MOSFETs designed for a higher current in each supply phase .
To reduce the output voltage ripple in multi-phase voltage regulators, all phases operate in synchronism with the time s m shift relative to each other. If T is the switching period of the MOSFETs (PWM signal period) and is used N phases, then the time shift for each phase will be T/N(Fig. 6). The PWM controller is responsible for synchronizing the PWM signals for each phase with a time shift.
Rice. 6. Timing shifts of PWM signals in a polyphase voltage regulator
As a result of the fact that all phases work with time s m shift relative to each other, the ripple of the output voltage and current for each phase will also be shifted along the time axis relative to each other. The total current passing through the load will be the sum of the currents in each phase, and the resulting current ripple will be less than the current ripple in each phase (Fig. 7).
Rice. 7. Current per phase
and resulting load current
in a three-phase voltage regulator
So, the main advantage of multi-phase switching supply voltage regulators is that they allow, firstly, to overcome the current limit, and secondly, to reduce the output voltage ripple with the same capacitance and inductance of the smoothing filter.
Discrete Multi-Phase Voltage Regulators and DrMOS Technology
As we already noted, each power phase is formed by a control driver, two MOSFETs, a choke and a capacitor. At the same time, one PWM controller simultaneously controls several power phases. Structurally, on motherboards, all phase components can be discrete, that is, there is a separate driver chip, two separate MOSFET transistors, a separate inductor and capacitance. This discrete approach is used by most motherboard manufacturers (ASUS, Gigabyte, ECS, AsRock, etc.). However, there is a slightly different approach, when instead of using a separate driver chip and two MOSFET transistors, one chip is used that combines both power transistors and a driver. This technology was developed by Intel and called DrMOS, which literally means Driver + MOSFETs. Naturally, separate chokes and capacitors are also used in this case, and a multi-channel PWM controller is used to control all phases.
Currently, DrMOS technology is only used on MSI motherboards. It is rather difficult to talk about the advantages of DrMOS technology in comparison with the traditional discrete way of organizing power phases. Here, rather, everything depends on the specific DrMOS chip and its characteristics. For example, if we talk about new MSI boards for processors of the Intel Core i7 family, then they use the Renesas R2J20602 DrMOS chip (Fig. 8). For example, on MSI board Eclipse Plus uses a 6-phase processor voltage regulator (Fig. 9) based on an Intersil ISL6336A 6-channel PWM controller (Fig. 10) and Renesas R2J20602 DrMOS chips.
Rice. 8. DrMOS Chip Renesas R2J20602
Rice. 9. Six-phase processor voltage regulator
based on 6-channel PWM controller Intersil ISL6336A
and DrMOS ICs Renesas R2J20602 on MSI Eclipse Plus board
Rice. 10. Six-channel PWM controller
Intersil ISL6336A
The Renesas R2J20602 DrMOS IC supports MOSFET switching frequencies up to 2 MHz and is very efficient. With an input voltage of 12 V, an output of 1.3 V and a switching frequency of 1 MHz, its efficiency is 89%. The current limit is 40 A. It is clear that with a six-phase processor power supply, at least a twofold current reserve is provided for the DrMOS microcircuit. With a real current value of 25 A, the power consumption (released as heat) of the DrMOS chip itself is only 4.4 watts. It also becomes obvious that when using Renesas R2J20602 DrMOS chips, there is no need to use more than six phases in the processor voltage regulators.
Intel in its Intel DX58S0 motherboard based on Intel chipset X58 for Intel Core i7 processors also uses a 6-phase, but discrete processor voltage regulator. A 6-channel PWM controller ADP4000 from On Semiconductor is used to control the power phases, and ADP3121 microcircuits are used as MOSFET drivers (Fig. 11). The ADP4000 PWM controller supports the PMBus (Power Manager Bus) interface and is programmable for operation in 1, 2, 3, 4, 5 and 6 phases with the ability to switch the number of phases in real time. In addition, using the PMBus interface, you can read the current values of the processor current, its voltage and power consumption. One can only regret that Intel did not implement these features of the ADP4000 chip in the processor status monitoring utility.
Rice. 11. Six-phase processor voltage regulator
based on ADP4000 PWM controller and ADP3121 MOSFET drivers
on an Intel DX58S0 board (two power phases shown)
Note also that each power phase uses On Semiconductor NTMFS4834N MOSFET power transistors with a current limit of 130 A. It is easy to guess that with such current limits, the power transistors themselves are not the bottleneck of the power phase. In this case, the current limit on the supply phase imposes a choke. In the voltage regulator circuit under consideration, PULSE PA2080.161NL chokes with a current limit of 40 A are used, but it is clear that even with such a current limit, six phases of the processor power supply are enough and there is a large margin for extreme overclocking of the processor.
Dynamic phase switching technology
Almost all motherboard manufacturers are now using the technology dynamic switching the number of processor power phases (we are talking about boards for Intel processors). Actually, this technology is by no means new and was developed by Intel a long time ago. However, as it often happens, having appeared, this technology turned out to be unclaimed by the market and for a long time lay in the "repositories". And only when the idea of reducing the power consumption of computers took possession of the minds of developers, they remembered the dynamic switching of the processor power phases. Motherboard manufacturers are trying to pass off this technology as their own and come up with various names for it. For example, Gigabyte calls it Advanced Energy Saver (AES), ASRock calls it Intelligent Energy Saver (IES), ASUS calls it EPU, and MSI calls it Active Phase Switching (APS). However, despite the variety of names, all these technologies are implemented in exactly the same way and, of course, are not proprietary. Moreover, the ability to switch the power phases of the processor is built into the Intel VR 11.1 specification, and all PWM controllers that are compatible with the VR 11.1 specification support it. Actually, motherboard manufacturers have little choice here. These are either PWM controllers from Intersil (for example, the 6-channel PWM controller Intersil ISL6336A), or PWM controllers from On Semiconductor (for example, the 6-channel PWM controller ADP4000). Controllers from other companies are used less frequently. Both Intersil and On Semiconductor VR 11.1 compliant controllers support dynamic power phase switching. The only question is how the motherboard manufacturer uses the capabilities of the PWM controller.
Naturally, the question arises: why is the technology of dynamic switching of power phases called energy-saving and what is the efficiency of its application?
Consider, for example, a motherboard with a 6-phase processor voltage regulator. If the processor is not heavily loaded, which means that the current consumed by it is small, it is quite possible to get by with two power phases, and the need for six phases arises when the processor is heavily loaded, when the current consumed by it reaches its maximum value. Indeed, it is possible to make the number of power phases involved correspond to the current consumed by the processor, that is, so that the power phases are dynamically switched depending on the processor load. But isn't it easier to use all six power phases at any processor current? To answer this question, you need to take into account that any voltage regulator itself consumes part of the electricity it converts, which is released in the form of heat. Therefore, one of the characteristics of a voltage converter is its efficiency, or energy efficiency, that is, the ratio of the power transferred to the load (to the processor) to the power consumed by the regulator, which is the sum of the power consumed by the load and the power consumed by the regulator itself. The energy efficiency of the voltage regulator depends on the current value of the processor current (its load) and the number of power phases involved (Fig. 12).
Rice. 12. Dependence of energy efficiency (efficiency) of the voltage regulator
on the processor current with a different number of power phases
The dependence of the energy efficiency of the voltage regulator on the processor current with a constant number of power phases is as follows. Initially, with an increase in the load current (processor), the efficiency of the voltage regulator increases linearly. Further, the maximum efficiency value is reached, and with a further increase in the load current, the efficiency gradually decreases. The main thing is that the value of the load current, at which the maximum efficiency value is reached, depends on the number of supply phases, and therefore, if the technology of dynamic switching of the supply phases is used, then the efficiency of the supply voltage regulator can always be maintained at the highest possible level.
Comparing the dependences of the energy efficiency of the voltage regulator on the processor current for a different number of power phases, we can conclude: at a low processor current (with a slight processor load), it is more efficient to use a smaller number of power phases. In this case, less energy will be consumed by the voltage regulator itself and released as heat. At high processor currents, the use of a small number of power phases leads to a decrease in the energy efficiency of the voltage regulator. Therefore, in this case, it is optimal to use a larger number of power phases.
From a theoretical point of view, the use of the technology of dynamic switching of the processor power phases should, firstly, reduce the overall power consumption of the system, and secondly, heat dissipation on the supply voltage regulator itself. Moreover, according to motherboard manufacturers, this technology can reduce system power consumption by as much as 30%. Of course, 30% is a number taken from the ceiling. In reality, the technology of dynamic switching of power phases can reduce the total power consumption of the system by no more than 3-5%. The fact is that this technology allows you to save electricity consumed only by the voltage regulator itself. However, the main consumers of electricity in a computer are the processor, video card, chipset and memory, and against the background of the total power consumption of these components, the power consumption of the voltage regulator itself is quite small. Therefore, no matter how you optimize the power consumption of the voltage regulator, it is simply impossible to achieve significant savings.
Marketing "chips" of manufacturers
Motherboard manufacturers go to great lengths to attract the attention of buyers to their products and motivatedly prove that they are better than those of competitors! One of these marketing "chips" is the increase in the power phases of the processor voltage regulator. If earlier six-phase voltage regulators were used on top motherboards, now they use 10, 12, 16, 18 and even 24 phases. Do you really need so many power phases, or is this just a marketing gimmick?
Of course, polyphase voltage regulators have their undeniable advantages, but there is a reasonable limit to everything. For example, as we have already noted, a large number of power phases allows the use of low current components (MOSFETs, chokes and capacitances) in each power phase, which, of course, are cheaper than high current limiting components. However, now all motherboard manufacturers use solid polymer capacitors and ferrite core chokes, which have a current limit of at least 40 A. MOSFETs also have a current limit of at least 40 A (and recently there has been a trend towards MOSFETs). with a current limit of 75 A). It is clear that with such current limitations, it is sufficient to use six power phases on each phase of the wave. Such a voltage regulator is theoretically capable of providing a processor current of more than 200 A, and therefore a power consumption of more than 200 watts. It is clear that even in the extreme overclocking mode, it is almost impossible to achieve such current and power consumption values. So why do manufacturers make voltage regulators with 12 phases or more, if a six-phase voltage regulator can also provide power to the processor in any mode of its operation?
If we compare 6- and 12-phase voltage regulators, then theoretically, when using dynamic power phase switching technology, the energy efficiency of a 12-phase voltage regulator will be higher. However, the difference in energy efficiency will be observed only at high processor currents, which are unattainable in practice. But even if it is possible to achieve such a high current value at which the energy efficiency of 6- and 12-phase voltage regulators will differ, then this difference will be so small that it can be ignored. Therefore, for all modern processors with a power consumption of 130 W, even in the mode of their extreme overclocking, a 6-phase voltage regulator is enough for the wave. The use of a 12-phase voltage regulator does not provide any advantages even with dynamic phase switching technology. Why manufacturers started making 24-phase voltage regulators is anyone's guess. There is no common sense in this, apparently, they expect to impress technically illiterate users, for whom "the more the better."
By the way, it would be useful to note that today there are no 12- and even more so 24-channel PWM controllers that control the power phases. Maximum amount channels in PWM controllers is six. Therefore, when voltage regulators with more than six phases are used, manufacturers are forced to install several PWM controllers that work in synchrony. Recall that the PWM control signal in each channel has a certain delay relative to the PWM signal in the other channel, but these signal timing offsets are implemented within the same controller. It turns out that when using, for example, two 6-channel PWM controllers to organize a 12-phase voltage regulator, the supply phases controlled by one controller are combined in pairs with the supply phases controlled by another controller. That is, the first power phase of the first controller will operate synchronously (without time shift) with the first power phase of the second controller. The phases will be dynamically switched, most likely, also in pairs. In general, this is not an "honest" 12-phase voltage regulator, but rather a hybrid version of a 6-phase regulator with two channels in each phase.
Loading...