These 8-point onboard power design specifications can not be ignored!

In the hot-swap system, the capacitor on the board starts to charge when the board is inserted. Because the voltage across the capacitor cannot be abruptly changed, it will cause the voltage of the entire system to drop instantaneously. At the same time, because the power supply impedance is very low, the charging current will be very large, and the fast charging will impact the capacitors in the system, which may easily lead to the failure of the tantalum capacitors.

Hot-pluggable systems must use a power slow-start design

In the hot-swap system, the capacitor on the board starts to charge when the board is inserted. Because the voltage across the capacitor cannot be abruptly changed, it will cause the voltage of the entire system to drop instantaneously. At the same time, because the power supply impedance is very low, the charging current will be very large, and the fast charging will impact the capacitors in the system, which may easily lead to the failure of the tantalum capacitors.

If a fuse is used for overcurrent protection in the system, the transient current may cause the fuse to blow, and choosing a fuse with a large current may not blow when the system current is abnormal, thus failing to provide protection. Therefore, in a hot-swap system, the power supply must adopt a slow-start design to limit the start-up current and avoid excessive transient currents that affect system operation and device reliability.

LDO

1. In the design of step-down power supply with large voltage difference or large current, it is recommended to use switching power supply and avoid using LDO

Using a linear power supply (including LDO) can obtain lower noise, and because of its simple use and low cost, it is widely used on single boards. The FPGA core power supply, the power supply for the RF clock section of some boards, etc. are all scaled from higher voltage power supplies using linear power supplies. The basic principle of linear power supply is shown in the figure.

After sampling, the output voltage is subtracted from the reference power supply (provided by the transistor bandgap reference source or Zener diode), and the difference is amplified and then controlled to control the voltage drop on the push tube V dropout =V output -V input , so that when V input When V output changes due to changes or load current changes, the stability of V output is ensured by the change of V dropout .

These 8-point onboard power design specifications can not be ignored!

It can be seen from the figure that all the load current flows through the adjustment tube, and the difference between the input voltage and the output voltage is all added to the adjustment tube. The power dissipated on the pass tube is V dropout *I. When the voltage difference is large, or when the load current is large, the regulator will suffer a large power dissipation.

LDOs must calculate heat dissipation and meet derating specifications

In addition, the power provided by the input power supply is V input *I, that is, when a linear power supply is used, the calculation of the power supply power cannot be calculated by the product of the load voltage and current, and the product of the linear power supply input voltage and load current must be used to calculate the power supply when using a linear power supply The calculation of power cannot be calculated by the product of the load voltage and current, but must be calculated by the product of the input voltage of the linear power supply and the load current. Calculations and thermal simulations must be performed to ensure the proper functioning of the system.

For example, an LDO in a TO-263 package reduces the voltage from 3.3V to 1.2V, the load current is 1.5A, and the power dissipated on the load is 1.8W. At this time, the LDO bears a 2.1V voltage drop, the power dissipated is 3.15W, and the power provided by the 3.3V power supply is 4.95W!

The thermal resistance of the package is about 40°C/W, and if no heat dissipation measures are taken, the temperature rise can reach about 120°C. Appropriate heat dissipation measures must be determined through thermal simulation for the LDO, and the 3.3V power supply must be able to provide 1.5A of current (or more than 5W of power) in the budget to ensure the normal operation of the system.

The use of switching power supply can achieve high efficiency, and it is recommended to use switching power supply for conversion in the case of large current and large voltage difference. If the circuit has high requirements on ripple, the method of using the switching power supply and the linear power supply in series can be used to suppress the noise of the switching power supply by using the linear power supply.

2. When selecting the filter capacitor at the LDO output, pay attention to the minimum capacitor required by the manual and the ESR/ESL requirements of the capacitor to ensure the stability of the circuit.It is recommended to use multiple equivalent capacitors in parallel to increase reliability and improve performance

The LDO output capacitor provides transient current for load changes, and because the output capacitor is in the voltage feedback regulation loop, in some LDOs, the capacitance is required to ensure the regulation loop is stable. The capacitance of this capacitor does not meet the requirements, and the LDO may oscillate, resulting in a large ripple in the output voltage.

Connecting multiple capacitors in parallel, as well as connecting large-capacity electrolytic capacitors in parallel with small-capacity ceramic capacitors, is beneficial to reduce ESR and ESL and improve the high-frequency performance of the circuit. However, for some linear regulated power supplies, the ESR of the output capacitor is too low, and It may induce loop stability margin drop or even loop instability.

Filter capacitor

1. The power supply filter can use RC, LC, π type filter. For power supply filtering, it is recommended to prefer magnetic beads, and then inductors.At the same time resistance, inductance and magnetic beads must consider the voltage drop caused by their resistance

Passive filter circuits can be used in occasions with high power requirements and where noise needs to be isolated in local areas. When a passive filter circuit is used, it is recommended to use magnetic beads for filtering.

The main difference between magnetic beads and inductors is that the Q value of the inductance is higher, while the magnetic beads are resistive at high frequencies and are not prone to resonance and other phenomena.

Inductors have high machining accuracy, while magnetic beads have relatively low machining accuracy and lower cost. When choosing a filter device, magnetic beads are preferred. Resistors and capacitors were chosen to form a resonant-free first-order RC low-pass filter, but this circuit can only be used with very small currents. The load current will create a voltage drop across the resistor, causing the load voltage to sag. No matter what kind of filter is used, it is necessary to consider the voltage drop of the load current on the Inductor, magnetic bead or resistor, and confirm that the filtered voltage can meet the requirements of the subsequent circuit.

For example, a first-order RC filter is used in the design of a single-board phase-locked loop, and the filter resistance is 12 ohms. The working current of the VCXO in the phase-locked loop is about 30mA, and a voltage drop of 300mV is generated on the filter resistor. The actual working voltage of the VCXO with a rated voltage of 3.3V is only less than 3V, which is prone to vibration stoppage and other phenomena. On a certain optical port daughter card, the SD (optical detection) signal of a certain type of optical module rose slowly when the optical fiber was inserted, which could not correctly reflect the actual situation.

After inspection, it is found that the DC resistance of the filter inductor is about 3 ohms, the operating current of the optical module is about 100mA, and the voltage drop on the inductor causes the operating voltage of the optical module to be only about 2.9V. There will be a slow rise in SD on this type of optical module. Fault. In addition, for the filter circuit, it should be ensured that the capacitor network behind the inductor, magnetic beads or resistor can ensure low impedance at all frequencies of interest. If necessary, capacitors of various capacities should be used in parallel, and the target impedance should be achieved by local copper plating. (Refer to the clock driver chip filter circuit design section).

On a single board, magnetic beads and 0.1u capacitors are used to provide filtering for the clock driver chip. After testing, the ripple on the pins of the clock driver chip is as high as 1V or more. Using multiple capacitors in parallel can effectively provide decoupling for the clock chip.

2. Large-capacity capacitors should be used in parallel with small-capacity ceramic chip capacitors

Large-capacity capacitors are generally electrolytic capacitors, which are larger in size and have longer pins, and are often wound structures (tantalum capacitors are sintered carbon powder and manganese dioxide). The large equivalent series inductance of these capacitors results in poor high frequency characteristics of these capacitors, with resonant frequencies ranging from a few hundred KHz to several MHz (see Sanyo’s OSCON Device Handbook and AVX’s Tantalum Capacitor Device Handbook).

Small-capacity ceramic chip capacitors have low ESL and good frequency characteristics, and their resonance points can generally reach tens to hundreds of MHz (see the reference “High-speed Digital Design” and AVX and other companies’ ceramic capacitor device manuals), It can be used to provide a low-impedance return path for high-frequency signals to filter out high-frequency interference components on the signal. Therefore, when applying large-capacity capacitors (electrolytic capacitors), small-capacity ceramic capacitors should be used in parallel with the capacitors.

3. Input capacitance

Calculate the ripple current of the input capacitor. This derivation process uses the integral formula. Through analysis and derivation, we can have a more thorough understanding of the working principle of the circuit.

If considering the output ripple current. Then the waveform of the ripple current on the capacitor is:

These 8-point onboard power design specifications can not be ignored!

Since the upper tube is turned on, the size of the input current can be approximately regarded as the size of the output current. Therefore, it is only necessary to superimpose the waveform of the output current on the waveform of the input capacitor, and the waveform in the above figure can be obtained.

Then according to the definition of effective current, we can calculate the current squared in time

These 8-point onboard power design specifications can not be ignored!

In order to simplify the calculation, we split the energy into the ripple part and the DC part. The original DC part, we directly use multiplication to calculate. The DC part, we can get it according to the approximate calculation method. The power consumption of the AC part can be calculated according to the formula:

These 8-point onboard power design specifications can not be ignored!

So the effective current on the total capacitor is:

These 8-point onboard power design specifications can not be ignored!

If the capacitor of 220uF is selected, the effective current that each can withstand is 3.8A. . If we calculate that the effective current value of the input capacitor is 7A, we need to choose 2 220uF capacitors. The effective current value that polymer electrolytic capacitors can withstand is limited. The capacity of the capacitor needs to be fully considered in the design.

boost circuit

1. When using a boost power supply (BOOST), a fuse must be added to prevent the entire board from being powered down when the power supply is connected to when the load is short-circuited.The size of the fuse is determined by the maximum output current of the module or the maximum current of the load

The basic topology of a boost power supply (Boost) is shown in the following figure:

These 8-point onboard power design specifications can not be ignored!

When Q1 is turned on, the resistance at both ends is very small, the power supply voltage is applied to both ends of L, and the electrical energy is converted into a magnetic field and stored in L. At this time, D1 is turned off to prevent the voltage on C0 from flowing to Q1. When Q1 is turned off, the current in L cannot be abruptly changed, and the power supply and L together charge C0 through D1 and supply power to the load, resulting in an output voltage higher than the input voltage.

As can be seen from the topology in the figure, we cannot cut off the path between the input and output or control the output current by controlling the on-off of Q1. When the output power supply is short-circuited, the input power supply (usually the main power supply of the single board) is directly short-circuited to ground through L and D1.

The result will be a burnout of L or D1 and a failure mode of open circuit. Before L or D1 is burnt out, the power supply of the board is in a short-circuit state. If the currents of L and D1 are greatly derated, the power supply of the board may be protected and cannot be powered on. In order to avoid the above problems, it is recommended to add a fuse to the boost power supply to prevent short circuit of the load. The size of the fuse depends on the maximum output current of the module or the maximum current of the load.

Anti-reverse

1. The power supply should have anti-reverse connection treatment. If the input current exceeds 3A, the reverse connection of the input power supply is only allowed to damage the fuse; if it is less than or equal to 3A, the reverse connection of the input power supply is not allowed to damage any device.

The power supply must be protected against reverse connection. If the input current exceeds 3A, the reverse connection of the input power supply is only allowed to damage the fuse; if it is less than or equal to 3A, the reverse connection of the input power supply is not allowed to damage any device. When the loop current is large, the reverse connection of the DC power supply can be handled as follows. The schematic is shown below:

These 8-point onboard power design specifications can not be ignored!

When the DC power supply is connected normally, due to the reverse bias of the input diode of the optocoupler D1, the output C-E cannot be turned on. At this time, the parallel NMOS transistor will stabilize the voltage of G-S to 12V, so that DS is turned on. In this way, the power loop will be formed smoothly. Capacitor C1 plays a role in slow start, which can play the purpose of anti-surge. Resistor R6 and diode VD3 form the discharge loop of capacitor C1.

When the power supply is reversed, due to the positive bias of the optocoupler input diode, the output CE is turned on, so that the parallel NMOS transistor is turned off. In this way, the circuit is cut off, which plays the role of anti-reverse connection protection. Since the R DS of the parallel NMOS transistor is relatively small and the loss is small, it is more suitable for the occasion of low voltage and high current. When the loop current is small, a diode can be directly connected in series with the input loop. When reversed, power is blocked due to the unidirectional conductivity of the diode.

inductance

1. Disable the magnetic saturation circuit; it is forbidden to select the power module using the magnetic saturation circuit to disable the magnetic saturation circuit, because:

a. The magnetic saturation circuit is sensitive to temperature because of the magnetic ring used, and it is easy to be unstable when working at high temperature.

b. The dynamic load capacity is poor, the work is the worst when the magnetic saturation circuit load is the smallest, and the output is easy to be unstable.

power-on sequence

1. For devices with multiple power supplies, the power-down sequence requirements must be met.

For devices with multiple power sources such as core voltage, IO voltage, etc., they must meet the requirements of their power-on and power-off sequences. If these conditions are not met, it is very likely that the device cannot work normally, or even trigger a latch and cause the device to burn. For example, TMS320C6414T DSP, described in Errata after May 2005, when DVDD is powered on earlier than CVDD, the problem of PCI/HPI data error may occur.for

QDR, DDR memory, its power-on sequence is also required, otherwise it may lead to latch-up, resulting in the consequences of device burnout. When there are multiple power supplies, a dedicated power-up sequence control device can be used to ensure the power-up sequence if necessary. In the design, it should be ensured that the power supply is in a shutdown state when the device is not loaded with a sintering file. It is also possible to ensure that the power-up and power-down sequence requirements are not violated during power-up and power-down by connecting Schottky diodes between different power supplies.

These 8-point onboard power design specifications can not be ignored!

Because the power module and the capacitor on the power supply will affect the power-on sequence of the power supply, the voltage requirements may be violated during the power-on process, as shown in the right figure above, so it must be tested and verified.

2. When multiple chips work together, the operation must be started after the initialization of the slowest power-on device is completed.

When multiple chips work together, the initialization must be completed in the slowest period before the operation can be started, otherwise unpredictable results may be caused.

For example, the LVT16244 driver has a power-on 3-state function. Even if the OE terminal is pulled down to the ground, it needs to wait until the power supply voltage rises to a certain threshold before leaving the high-impedance state, and the EPLD and other devices may have already started to work, which may cause the EPLD to read to the wrong state. See previous instructions. For some devices such as ROM, it can start to work after a period of time after power-on. If reading is started before this time, it may also cause data errors.

PCB Design

1. The Kelvin method should be used in the layout of the power module/chip sensing terminal

Many power modules and power chips are designed with independent Sense pins as the feedback input for the output voltage. This Sense signal should be led to the power module from the position where the power is taken, and should not be directly led to the power module at the output end of the power module, so that the path from the power module output to the actual power supply can be compensated by the feedback inside the power module. Attenuation caused. The white traces are shown in the figure below.

These 8-point onboard power design specifications can not be ignored!

For power monitoring circuits, etc., the same principle should also be followed, that is, the power supply should be led to the monitoring circuit from the point where monitoring is actually required, rather than from the closest point of the monitoring circuit to the monitoring circuit to ensure accuracy.

2, Buck power supply PCB design points

1. The input capacitor and the output capacitor should share the ground as much as possible;

2. The number of output current vias ensures that the current capacity is sufficient, and the current is the set overcurrent value;

3. If the output current is greater than 20A, it is best to distinguish between the control circuit AGND and the power ground GND, and the two are grounded at a single point. If no distinction is made, ensure that the AGND is well grounded;

4. The input capacitor is placed close to the D pole of the upper tube;

5. Because of its strong current, high voltage characteristics and large radiation, Phase pins need to do the following treatments

a: The S pole of the upper tube connected to the Phase, the D pole of the lower tube and one end of the inductor are flattened, and no holes are drilled, that is, try to ensure that the three and the power chip are on the same plane, and it is best to place them on the top noodle;

b: On the premise of ensuring sufficient flow capacity of the Phase plane, the area should be reduced as much as possible;

c: The key signal is far away from the Phase plane;

d: The Phase network with small current is directly drawn, and it is forbidden to pull the plane;

6. The GND of the input capacitor and the power input are due to high noise, and sensitive signals should be kept away from this plane. Following the 3W principle, it is forbidden to route high-speed signals in the middle of the vias punched in the above ground plane, and pay special attention to the high-speed signals on the backplane;

7. GATE, BOOT capacitor traces should be as thick as possible, generally 15mil~40mil;

8. The voltage sampling is easy to be interfered due to the small current. If it is near-end feedback, try to get as close to the power chip as possible. If it is far-end feedback, it needs to go through the differential line and stay away from the interference source;

9. The DCR current sampling network requires differential wiring. The entire sampling network should be as compact as possible, and should be placed close to the power chip, and the temperature compensation resistor should be placed close to the inductor;

10. The loop compensation circuit should be as small as possible, reduce the loop, and place it close to the power chip;

11. It is forbidden to drill holes under the inductor. On the one hand, some inductors are metal surface layers to prevent short circuits; on the other hand, because the radiation of the inductors is large, if the holes are drilled below, the noise will be coupled;

12. Via holes need to be drilled under the MOS tube for heat dissipation. The number of via holes is calculated according to the maximum output current, not the overcurrent value;

13. Drill holes at the bottom of the power chip to the back for heat dissipation. The larger the copper coating, the better the heat dissipation, and it is best to partially brighten the copper;

These 8-point onboard power design specifications can not be ignored!

The Links:   NL6448BC26-30D 2MBI300NK-060