How to improve the reliability of inverter power supply?

First, let's talk about the electrolytic capacitor in the input circuit. As we know, the DC input current of the inverter is usually very large. The maximum input current of a 12V 1000W inverter can reach more than 120A. At this time, the selection of the electrolytic capacitor at the input end is very critical. If the selection is inappropriate, the failure of the electrolytic capacitor will become commonplace.

The second thing to talk about is the adaptability to different load characteristics. This includes two issues: 1. The power margin of the inverter itself, the maximum allowed load startup output current and overcurrent protection measures; 2. The adaptability to loads with different characteristics such as inductive, capacitive, negative resistance, etc. Generally, if these issues are not handled well technically, various problems are likely to occur when the product is used.    

The next issue is the heat dissipation. In addition to the main power switch components, high-frequency rectifier diodes, main power transformers and other components, the heat dissipation of electrolytic capacitors cannot be taken lightly.....

When it comes to the reliability of the inverter, there is an important issue that must be mentioned, which is the parallel connection of MOS tubes. Of course, this also includes the parallel drive problem and the PCB wiring problem. The simple four words "equal current and equal voltage" not only include balanced drive, PCB wiring balance (the DC and AC resistance of the wiring are equal), but also include the heat dissipation and temperature uniformity of the tube body, and the dynamic and static matching of the MOS tube Ron (tube selection) and other issues.

Apart from grid connection, another issue that has a significant impact on operational reliability is the inverter's "self" protection, including current limiting protection mode (mentioned earlier), thermal shutdown protection, user operation abnormality protection, load abnormality protection, startup protection, etc.

This statement is not wrong. In fact, if the rear stage is turned off in time, the overcurrent of the front stage can be automatically relieved. Of course, in practical applications, both the front stage high-frequency high-power DC/DC and the rear stage 50Hz/60Hz inverter should have a good current limiting control loop.

It goes without saying that the parameter setting and selection of the original device will affect the reliability of the product. However, for MOS tubes and ultra-fast rectifier diodes, the impact of different packaging forms on reliability is sometimes very obvious! We have to pay serious attention to it.

 Before talking about the driving problem, here is a measured waveform of the G pole of the MOS tube on one side of the push-pull inverter circuit (1:1 blue) and the secondary voltage wave of the step-up transformer (15:1 yellow). This is the measured waveform when the circuit is at full load of 1000W DC+24V input. It can be seen that when the other MOS tube is turned on, the interference spike waveform is connected in series to the G pole of the cut-off MOS tube.

Since the power supply of the MOS tube driving part of most inverters is the same as that of the main oscillator IC, which is powered by a single power supply (SG3525 is not uncommon for direct drive tube MOS output), the driving waveform is mostly 0V~+15V square wave. At this time, if the driving waveform is disturbed (see the spike part in the above figure), if it is close to the Vth value of the MOS tube, it is needless to say that it will have a negative impact on the system, at least it will affect the efficiency and temperature rise. If general means cannot effectively reduce or avoid this interference, it is necessary to use negative pressure shutdown. This issue should be given enough attention in professional mass production plans.

This figure shows the G-pole waveform of the push-pull A-phase and B-phase MOS tubes when the inverter is fully loaded (1:10). Due to the use of a +15V turn-on and -5V turn-off driving method and the selection of low Qgs power MOS tubes, the "peak" interference of the driving waveform is greatly reduced. It can also be seen that due to the use of negative voltage shutdown, the "burrs" from the other phase crosstalk at full load are effectively controlled within the 0V line (red circle), ensuring that the MOS tube can be absolutely reliably cut off during the cut-off period.

Before talking about loop feedback and overcurrent protection, let's continue with the heat dissipation topic on the 4th floor and talk about the structural design and heat dissipation of the main power tube. Let's take an example: a small copycat company copied a mature inverter circuit. This circuit was well received by others, but the proportion of main power MOS tubes exploding in their own products was relatively high...

Later I was told to send a sample to take a look... After I took the sample and opened it, I found that 8 TO-220 packaged main power MOS tubes were densely packed on one side, and the thickness of the aluminum shell wall was only 3mm~4mm... Although there was a thermal probe, I was still speechless.

 In an environment of 25 degrees Celsius, the output is fully loaded at 1000W. After 10 minutes, the temperature at B (the center of the 8 MOS tubes) in the picture is 6~8 degrees higher than that at A! The temperature at C (green circle) is the lowest, 14~15 degrees lower than that at B! (C is the air inlet, D is the fan, and the prototype is designed for air inlet, which is said to extend the life of the oil-containing bearing). The same model of power MOS tubes working in parallel are actually

The temperature difference in actual operation is so large that it is extremely unfavorable for "current sharing"! So it is no wonder that the reliability is not high.

Combined with the heat dissipation design, for the parallel connection of MOS tubes, from parameter selection and matching (such as Ron, Qgs, etc., the error should be less than 5%) to the similar PCB routing parameters of each MOS tube (AC and DC impedance of PCB wiring), strict same driving waveform, synchronous and consistent temperature rise changes during operation (more on this later), etc., of course, the reasonable selection of current limiting protection points, and every detail of the assembly welding process cannot be taken lightly! Only in this way can high reliability be guaranteed when working in parallel.

In the attached figure, A is the primary transformer sampling, and B is the constantan wire sampling. In the inverter, A is mostly used for peak current limiting, and B is mostly the average current limiting mode.

When using constantan wire sampling, in order to reduce losses, the output voltage is generally very low and needs to be amplified before being used as a feedback signal. It is mostly used for average current limiting control. Although the response speed is slow, it has the advantages of high and stable current limiting accuracy. When the battery voltage drops from 14.5V to 10.5V, combined with the compensation of the current limiting value, a more ideal constant output power can be obtained, which will not affect the output power of the inverter due to the drop in battery voltage.