Author: Texas Instruments Richard Nowakowski and Brian King
The benefit of increasing the switching frequency is obvious, but there are some shortcomings, designers should understand the pros and cons, in order to choose the most suitable switching frequency to be applied. This practical article will explain these considerations one by one.
DC power converters with high switching frequencies are becoming popular because they can save board space with smaller output capacitance and inductance. On the other hand, the demand for point-of-load power is becoming more stringent as the processor core voltage drops below 1V, which makes the power supply subject to reduced load cycles, making it difficult to achieve the required lower voltage at higher frequencies.
Many suppliers of power components are aggressively promoting faster DC power converters and claiming that their products can save space. A DC power converter switching at 1 or 2MHz sounds great, but designers should be aware of other factors that can impact the power supply system, in addition to their size and efficiency. This article will provide several design examples to illustrate the various advantages and disadvantages of increasing the switching frequency.
Select an app
In order to illustrate the pros and cons of high switching frequency, this paper designed and implemented three different power supply, their input voltage is 5V, the output voltage is 1.8V, and the output current is 3 A, these are DSP, ASIC or FPGA and other high-performance processor common power requirements. With the limitation of filter design and performance, these designs allow up to 20mV ripple voltage, approximately equal to 1% of the output voltage, and the peak-to-peak inductance current is set to 1 a.
This article will compare the design of different frequencies such as 350, 700 and 1600kHz to illustrate their pros and cons. These examples are the Texas Instruments (TI) TPS54317 as a voltage regulator, it is a built-in MOSFET 1.6MHz, low voltage, 3A synchronous DC buck converter, with programmable frequency and external compensation circuit, dedicated to high-density processor power load point applications. selecting inductors and capacitors
Inductors and capacitors are selected according to the following simple formula:
V = Lxdi/dt
After finishing, you can get:
where Δil = 1 a peak-to-peak; D = 1.8v/5v = 0.36.
I = Cxdv/dt
After finishing, you can get:
Where: ΔV = Mv,i = 1 A peak-to-peak.
Equation 2 assumes that the series impedance of the capacitor is negligible, such as ceramic capacitors, so the three designs in this article choose to use ceramic capacitors with very small impedance and volume. In the reorganization of Equation 2, the multiplier 2 represents a decrease in the capacitance value caused by DC bias, because most ceramic capacitor data sheets do not consider this effect.
This paper uses the circuit in Figure 1 to evaluate the effectiveness of three different designs.
Some of the components in Figure 1 do not indicate values because they differ in the values of the three designs. The output filters consist of L1 and C2, and their values in three designs are listed in Table 1, which are calculated from the previous formula.
Table 1: Capacitance and inductance values selected at 350kHz, 700kHz and 1600kHz frequencies
Note that the higher the frequency, the smaller the number of laps required for the inductor, so the lower the DC impedance. the compensating parts for these error amplifiers are designed for the three switching frequencies in this article, but there is no discussion of how to calculate and select these component values. minimum on-time
The minimum on-time controlled by a digital DC power converter is determined by the minimum pulse width that can be generated by the pulse-width modulation (PWM) circuit. In a buck converter, the ratio of FET conduction time over the entire switching cycle is called the load cycle (duty cycle), which is equal to the ratio of the output voltage to the input voltage.
In the figure 1 circuit, for example, the load cycle of the TPS54317 can be found as 0.36 (1.8v/5.0v) from the data table, and the minimum on-time is 150ns (maximum). As long as the minimum pulse width that the component can control, the designer can easily calculate the minimum load period that the circuit can achieve by using equation 3, and then use Equation 4 to calculate the lowest output voltage that the converter can provide (refer to table 2). It is important to note that the minimum output voltage of the converter is also limited by the reference voltage, for example, the TPS54317 reference voltage is 0.9V.
Minimum load cycle = min on time x switching frequency (3)
Minimum output voltage = input voltage x minimum load cycle (no less than TPS54317 reference voltage) (4)
Table 2: Minimum output voltage at minimum on-time 150ns
In this example, the minimum output voltage of the 1.6MHz switching frequency is limited to 1.2V (the original text here is mistakenly 1.8V). However, if the frequency rises to 3MHz, the minimum output voltage limit will increase to 2.3V. If a DC power converter is to provide a lower output voltage, it must omit partial pulses, reduce the input voltage, or reduce the switching frequency. Before choosing the switching frequency of a DC power converter, it is best to check the data table to ensure that the minimum on-time control of the component meets the design requirements.
If the converter stops the brake pole drive pulse is not fast enough to achieve the required load cycle, the converter will omit the partial pulse (pulse skipping) to provide the required low output voltage. At this point, although the power supply will still strive to maintain the output voltage stability, but ripple voltage will still increase due to the pulse interval. Due to the omission of the pulse, the output ripple will appear some sub-harmonic components, which may lead to noise problems. The current-limiting circuit may also not operate properly because the component may not respond to a large current surge. Sometimes even the controller does not work properly, causing the control loop to become unstable. The fastest controllable guidance pass time is an important parameter of a DC power converter, and the designer should check the specifications listed in the component data sheet to ensure that the switching frequency and minimum on-time are met. efficiency and power consumption
The efficiency of DC power converters is one of the most important considerations for power supply design. Low efficiency equals high power consumption, the need to install heat sinks on the board or expand the copper foil area to eliminate heat. In addition, high power consumption will also cause a great burden on the upstream. Power sources are available in the following ways:
Impact Factor • Power source
function of gate charge, driving voltage and frequency · FET Drive Power consumption
Input voltage, output current, Fet function of rise/fall time and frequency of FET switching power consumption
i2x Conduction Impedance · FET Impedance
I2X DC impedance + AC core power consumption • Inductive power consumption
IRMS2 capacitance power consumption x equivalent tandem group resistance
Querying data tables to find out IQ component Power Consumption (IQ) during component operation
In these three examples, the main power sources include FET drive power dissipation, FET switching power consumption, and inductive power dissipation. The FET impedance is no different from the component power consumption, as these three designs use the same component. Capacitive power dissipation can also be ignored, as they all use ceramic capacitors with a small equivalent series impedance. To demonstrate the effect of the high frequency switch, Figure 2 shows the efficiency values of these design measurements.
Figure 2: Efficiency with 5 V inputs and 1.8V outputs at different frequencies
Figure 2 clearly shows that when the open-off frequency increases, the efficiency decreases . To improve efficiency at a variety of frequencies, designers should choose a DC power converter with low on-resistance, low gate voltage, and a small quiescent current at full load, or inductors and capacitors with a smaller equivalent impedance. Component Dimensions
Table 3 is the inductance and capacitance values and the pad area they require on the board.
Table 3: Component dimensions and total area requirements
The recommended pads area for both capacitance and inductance is slightly larger than the individual components, but this is also included in the design considerations for three circuits. Then just add the area of the individual parts (including the ICS, filters, and other small resistors and capacitance pads) and multiply by 2 to accommodate the component spacing to get the desired total area. As shown in table 2, when the frequency increases from 350kHz to 1600kHz, the filter size is reduced by half and the board area is reduced by 30%, so that the area can be saved approximately 100 square millimeters.
However, this approach has its limitations because inductance and capacitance cannot be reduced to zero, and space efficiency is also subject to the law of diminishing returns. In other words, due to the size limitations of the mass production of inductors and capacitors, the practice of reducing the total area by increasing the frequency is unlikely to continue indefinitely. instantaneous response
Instantaneous response is a good indicator of power supply performance. In this paper, three power supply Bode plots are plotted to compare their performance at a higher frequency. From Figure 3, it can be seen that the phase edge limits of these power supplies are between 45-55 degrees, indicating that they both provide a good damping instantaneous response.
Figure 3: Bode plot with 350kHz, 700kHz, and 1600kHz frequencies
cross-over frequency is approximately the 1/8 of the switching frequency, so when using a high switching frequency DC power converter, you should confirm that the Power module error amplifier is wide enough to support the high cross-over frequency, such as the TPS54317 error amplifier gain frequency bandwidth typical Is 5MHz. Table 4 is the actual instantaneous response time and the associated voltage peak overshoot value.
Table 4: Instantaneous response
The higher the switching frequency is seen in table 4, the lower the overshoot value, because the bandwidth of these designs can become larger. A smaller transient voltage overshoot is advantageous for new high-performance processors, as they typically require a voltage regulator accuracy of up to 3%, including instantaneous voltage peaks.
Designed to require a larger output current, TI also offers multi-phase parallel, dual-channel, 1MHz and DC power conversion controller TPS40140 with external MOSFETs. Designers can bring the advantages of high switching frequency to the application design as long as multiple power stage circuits are connected in parallel, allowing them to operate in different phases.
For example, designers can connect the output of 4 sets of 500kHz switching frequencies to get the effective frequency of 2MHz. The benefit of this approach is to reduce ripple, reduce input current capacity, accelerate instantaneous response, and distribute power dissipation across the entire board to provide better thermal management. The designer can connect up to 8 TPS40140 components through a digital bus and operate synchronously with different phases, making the effective frequency up to 16MHz. Conclusion
Switching power converters with high switching frequency have advantages and disadvantages, the benefits mentioned herein include smaller size, faster transient response, and lower voltage overshoot and undershoot values, the main disadvantage being reduced efficiency and increased heat.
Increasing the switching frequency also poses some potential problems, such as omitting pulses (pulse skipping) and noise, so when selecting a DC power converter for high-frequency applications, check the manufacturer's datasheet to confirm certain important specifications such as minimum on-time, error amplifier gain bandwidth, FET Impedance and switching power consumption. Components that perform well on these specifications may be costly, but they can bring more benefits and are easier to use when encountering design challenges.