The power measurement of electronic devices usually refers to the measurement of switching power supplies (of course, there are linear power supplies). There are a lot of materials about switching power supply. The content discussed in this article is PWM switching power supply, and it is only a summary of test experience, to briefly describe some factors that are likely to cause system failure. Therefore, before reading this article, it is assumed that you have a certain understanding of switching power supplies.
1 Brief description of switching power supply
Switching Mode Power Supply (Switching Mode Power Supply, often abbreviated as SMPS), is a high-frequency power conversion device. Its function is to convert the voltage into the voltage or current required by the client through different forms of architecture.
The topology of the switching power supply refers to the structure of the switching power supply circuit. Generally, it is divided into isolated and non-isolated converters according to whether the output ground wire is electrically isolated from the input ground wire. Non-isolation means that the input terminal is connected to the output terminal, and there is no isolation measure. Most of the common DC/DC converters are of this type. The so-called isolation means that the input terminal and the output terminal are not directly connected in the circuit, and the isolation transformer is used to transfer energy through electromagnetic conversion, and the input terminal and the output terminal are completely electrically isolated.
For switching converters, there are only three basic topological forms, namely:
● Buck
● Boost (boost)
● Buck-Boost (buck-boost)
The three basic topological forms are determined by the way the inductor is connected. If the inductor is placed at the output, it is a Buck topology; if the inductor is placed at the input, it is a Boost topology. When the inductor is connected to ground, it is the Buck-Boost topology.
2 Test of key parameters that easily cause system failure
The following test items refer to the results of the test under static load, only the noise test needs to use dynamic load.
2.1 Jitter of phase point
For a typical PWM switching power supply, if the phase point jitter is too large, the system will usually be unstable (related to the phase margin mentioned later). For a 200~500K PWM switching power supply, the typical jitter value should be below 1ns.
2.2 Collapse of Phase point
Sometimes engineers measure the following waveform, which is a typical phenomenon of inductor saturation. For engineers who are not experienced enough, they are often ignored. Inductance saturation will cause the inductance value to drop sharply, similar to a short circuit, which will cause a sharp increase in current, and the MOS tube will often burn out due to the sharp increase in temperature. At this time, it is necessary to replace the inductor with a larger saturation current.
2.3 Shoot through test
The purpose of the test is to see whether the lower tube is turned on at the same time when the MOS tube is turned on, which causes the power supply to be directly turned on to the ground and cause a short circuit. As shown in Figure 3, the blue curve (Vgs_Lmos) means that the lower tube is brought up while the upper tube is on. If the peak of the blue curve exceeds the Vth requirement of the MOS tube, the duration (Duration ) It also exceeds the datasheet requirements, so there is a risk of simultaneous conduction. Of course, this is a common situation that everyone sees.
Many people will ignore the following situation, even some experienced power supply test engineers. Figure 4 below is the waveform when the lower tube is opened and the upper tube is closed (Figure 4-1 is a schematic diagram, and Figure 4-2 shows the actual test diagram). Although it is not brought up at the same time, please note that the upper and lower tubes are crossed, and the level of the cross point is much higher than the specified Vth value of the MOS tube. This is a serious shoot through phenomenon. The direct consequence is that the MOS tube burns out!
2.4 Phase margin and bandwidth
Phase margin and bandwidth are items that many companies have not tested (especially smaller companies are limited by instruments), but this is a very important test item. Whether the power system is stable, whether it can work effectively for a long time (3 years or more), the phase margin and bandwidth can play a decisive role to a large extent. Many companies completely rely on the recommended values in the reference design schemes given by the power chip manufacturers, but there are often not small differences with your design, so there will be great potential risks.
If the system is an unstable system, which is reflected in some power test items, the following main problems will be seen.
● The noise test of the power supply passes, but the power supply is still unstable. Behaves as a functional test failure. When debugging, engineers often say that the noise of my power supply is already very small and I have added a lot of capacitors. Why is it still not running? In fact, his closed-loop system is inherently unstable.
● The phase point jitter is too large. This is a typical unstable phenomenon.
● The transient response is too large. The stupid way is to add a lot of capacitors to meet the requirements of transient response. For low-cost products, this costs money.
If you have not tested the Bode plot of the loop gain of the system with the correct method, how do you start debugging these items and let him pass the test? Only keep experimenting back and forth. Then run functional tests back and forth. Oh, my god, a huge amount of work. Moreover, for some low-cost products, low-cost solutions such as aluminum electrolytic capacitors and MLCC capacitors are often used (the inductance and resistance are basically unchanged). The capacitance of these capacitors will decrease over time. Such as MLCC, if the system runs at normal temperature for two to three years, the capacitance value will change to half of the original value. The change of this half-capacitance will have a great impact on the stability of the system, which is an important reason why the quality of many low-priced products is unreliable. Does that mean that the higher the price, the more capacitors you use, the better? Of course not. This is why the phase margin is tested. You need to adjust a set of reasonable values that can cover the requirements of full capacitance and half capacitance at the same time. This can also achieve low price and high quality.
According to the Nyquist theorem for system stability requirements, the specification requires that the phase margin of a closed-loop system is less than 60 degrees, and 45 to 60 degrees can be considered as a limit requirement. For bandwidth, the requirement of 200~500K switching power supply is 10%~30% switching frequency. From the perspective of the stability of the switching power supply, the lower the bandwidth, the easier it is to stabilize the power supply. From the perspective of the dynamic index of the switching power supply, the higher the bandwidth, the better the dynamic performance of the power supply.
Figure 5 below is a typical Bode plot:
Another very important point is that in addition to the PWM switching power supply, there are many linear power supplies (LDO) whose compensation network is outside the chip, and similar loop gain Bode plot tests should be done to ensure their stability. LDO testing is easily overlooked by most manufacturers. For example, the circuit shown in Figure 6 below, many people will directly measure the noise and finish.
We may see that the phase margin cannot meet the requirements. As shown in Figure 7 below, it is only about 30 degrees. At this time, only by debugging different parameters can you get better results. So as to meet the requirements of system stability.
2.5 Power supply ripple (ripple) and noise (noise)
Power supply ripple and noise seem to be simple items in power supply testing. But it may also have a greater impact on your test results and functions.
The first is ripple. When we test, we just see if it meets the specification requirements, such as 30mV and so on. Sometimes, the ripple is related to the PLL of the system. If your PLL is not jitter, you can consider further reducing the ripple.
Noise, some people will ask, why my system noise and his system noise are basically in the same range, but my system will fail? First of all, we have to eliminate the reasons for system stability mentioned above. Then, dear, have you ever done FFT with an oscilloscope to see the difference in the frequency domain of the same noise?
