Discussion on the causes and protection measures of IGBT tube explosion in inverter welding machine

Due to the limited understanding of the working principles of switching devices and main circuit structures and the lack of detection methods, the reliability of the switching devices of high-power inverter welding machines is the top priority in the design of the entire machine. It is also the main reason why the return rate of domestic IGBT welding machines remains high and cannot be promoted on a large scale.

Key: Detection and quantitative analysis of IGBT current and voltage waveforms (the specific circuit takes a half-bridge inverter manual 400A welding machine as an example)

The opinion of the hero shery888:

1. Inherent problems of overcurrent protection of voltage-type PWM controller

At present, the common IGBT inverter arc welding machine PWM controller in China usually adopts voltage-type integrated chips such as TL494 and SG3525, and the current feedback signal is generally taken from the rectifier output end. When the output current signal is compared with the given current detected by the shunt, it is amplified by the proportional integral amplifier to control the output pulse width. After the IGBT is turned on, even if an overcurrent is generated, the PWM control circuit cannot turn off the overcurrent pulse being turned on in time. Due to the delay link in the system, the overcurrent protection time will be extended.

2. Current type overcurrent protection

The feedback current signal of the current PWM control circuit is obtained from the primary end of the high-frequency transformer through the current transformer. Since the current signal is taken from the primary end of the transformer, the response speed is fast, and the protection signal is synchronized with the current flowing through the IGBT. Once an overcurrent occurs, the PWM immediately turns off the output pulse, and the IGBT is protected in time. The control method of the current-type PWM controller, which detects the instantaneous current value pulse by pulse, responds quickly to changes in input voltage and load, and has good system stability. 3. Voltage-type PWM does account for the majority of applications.

Overcurrent protection sampling can also be taken from the primary of the transformer, and the overcurrent signal can be obtained through the mutual inductance coil or the Hall sensor, such as controlling the 8th pin of 3525. Shenzhen Ruiling's welding machine does a good job in this regard, and can well protect the switch tube from overcurrent.

4. How to determine whether the main circuit of an inverter power supply is reliable through detection means?

I think we can analyze the current and voltage waveforms of the switching devices and the main transformer under no-load and load conditions, and then adjust the switching device parameters, overcurrent and overvoltage buffer component parameters, and high-frequency transformer parameters in a targeted manner. The difficulty lies in how to choose the match.

Lyticast's opinion:

I had a current inverter welding machine nine years ago, but I haven't seen many companies have it until now. Domestic researchers only know how to copy and paste, how many of them are doing independent research and development? The products from nine years ago are still among the best in performance when compared now. In fact, they all use very common components. The key is the circuit design and excellent manufacturing process to ensure the quality. This welding machine has been used in a security door factory for nine years, two shifts a day for 16 hours. The small machine with a nominal 130A is better than the current nominal 200A, and there is very little spatter. The welding rod can burn to 4mm, the no-load voltage is only 48V, the temporary load rate is 100%, and the weight is only 10-5KG. When I designed it, I was very conservative. The heat sink alone weighed 4.5KG, and the input filter inductor weighed 1.6KG, and there was no interference with the power grid.

The PWM IC used at that time was the rare UC3846J in China, with a ceramic package and an operating frequency of 100KHz. The circuit board was quite difficult to make. The current feedback used a transformer to sample the peak current and a Hall to sample the average current, with dual-loop feedback. There are many advantages of current-type control. The peak current is not only used for protection, but more importantly, it participates in the control of the large-loop feedback. In simple terms, the output of the error amplifier is used to control the peak current, so that it can respond within half a cycle (5 microseconds). The response speed of the amplifier is not so important, although the error amplifier of UC3846 is very fast. Sometimes, in order to obtain a slower response speed, the response speed of the amplifier is deliberately slowed down. For example, when performing argon arc welding, a too fast response speed will make the arc characteristics harder. However, the quality of an inverter welding machine is not only about what kind of IC is used for control, but also the parameters of the drive circuit. This parameter should be adjusted according to the characteristics of the main switch element and the output rectifier diode, and the configuration of the buffer circuit is also very important. Every link of a successful welding machine must be perfect, and it does not necessarily cost a lot of money. The key is still a coordination problem. Domestic engineers have too narrow a knowledge base and lack technical exchanges, which will continue to widen the gap with imported products. I am willing to share everything I know with everyone to promote the development of my country's power electronics technology.

Here is a typical design method, for example a 400A manual welding machine:

The manual welding machine is the most difficult to make among all inverter welding machines, and its load dynamic range is the largest.

Basic design ideas: Engineering estimation of circuit limit values

1. Determine the capacity of the welding machine. Calculate the loaded voltage according to the formula = 20 + 0.04 * 400 = 36V. Take into account the rectifier tube voltage drop and cable voltage drop to take 40V, and the no-load voltage to take 60V. In this way, the main transformer turns ratio is 9 (calculated based on 380V three-phase input).

2. Estimate the primary peak current to determine the capacity of the main switch element. Take the maximum current/turn ratio*120%=53A. According to the parameter manual, 75A, 1200VIGBT (calculated based on the full bridge of the main circuit) should be selected. The operating frequency of IGBTs from different manufacturers can be selected between 22-28KHz.

3. The calculation process of the main transformer is omitted and everyone knows it.

4. Determination of the main control circuit. As mentioned earlier, to ensure the safety of the main switch components and the output dynamic characteristics, current type control should be used. Please search online for information on UC3846 or UC3825. Feedback is still the same old way, current transformer + Hall.

5. Determination of drive parameters. Everyone may use a driver IC. In fact, it is not necessary when the output voltage is not very high. It is sufficient to use a pulse transformer unipolar drive, which is both cheap and reliable. The negative voltage of the driver IC is mainly used in occasions such as frequency converters to prevent the diode from recovering. This problem does not exist in welding machines. Using negative voltage can easily cause the IGBT to self-lock and fail.

6. The value of gate resistance. Pass a wire through the main transformer and connect it to an oscilloscope for observation. At this time, use a voltage regulator to reduce the input voltage and short-circuit the output. See if the voltage spike is on the leading edge or the trailing edge. A high peak on the trailing edge indicates that the recovery speed of the rectifier tube is slow, and the IGBT conduction speed needs to be reduced. A high peak on the leading edge indicates that the IGBT is turned off too quickly, and the turn-off speed should also be reduced.

7. Snubber circuit: The primary side uses an RC circuit to directly connect the two ends of the main transformer, and the access point should be as close to the IGBT as possible. The secondary side also uses an RC circuit to connect the two ends of the diode.

Generally speaking, the full-bridge circuit is the best choice for high power, as the switching stress of the main switch element is minimal. The unidirectional magnetic bias no longer exists under the control of the current-type IC, and no DC blocking capacitor is required when connecting the main transformer.

Note: Current mode control cannot be used for half-bridge circuits! Determination of inductance: Under normal circumstances, it is calculated as 3000/f(KHz)=microhenry. For example, 100KHz, 30 microhenry, 25KHz, 120 microhenry. When making inductors, pay attention to the inductor current capacity and whether the magnetic flux will be saturated. Once the magnetic flux is saturated, the IGBT will not be burned, but the arc characteristics will be significantly deteriorated. In severe cases, the arc will be frequently broken. The 120-170 microhenry, 400A inductor uses a 60*60*200 rectangular iron core and is wound vertically with 4*10 silk-wrapped flat copper wire. When it is fully wound, the inductance is about 170 microhenry. The arc of the manual welding machine made by this method is stable, easy to start, and the current does not overshoot. The stability of the welding process is guaranteed to the greatest extent. We can develop other types of welding machines on this basis, such as CO2 welding machines. As long as the speed control of the wire feeder is changed to arc length feedback, a variable speed wire feeding CO2 welding machine can be obtained. It will have all the advantages of the descending characteristics, the most obvious of which is the extremely small spatter, which is obtained because there is no overshoot current during short circuit. By changing the parameters of the UC3846 amplifier, the current can even be a very small value during short circuit transition, and the arc will start immediately after the short circuit is restored, and enter the next process.

The printed wiring of current-type PWM is very particular. Pay attention to the direction of the ground wire and the choice of the grounding point. Generally speaking, avoid parallel routing of the power ground and the signal ground. For UC3846, the grounding point should be connected to the ground end of the high-frequency decoupling capacitor. Use the star grounding method to make the ground wire spread radially. In addition, because the power ground and signal ground of UC3846 are shared, the high-frequency decoupling capacitor should be installed as close to the ground end of the IC as possible. This high-frequency decoupling capacitor is usually best to use a 1 microfarad polypropylene laminated capacitor. In high-frequency applications, UC3846 can directly drive the pulse transformer. The circuit is relatively simple. If it is to be used with power expansion, it is best to reversely connect a Schottky diode to the ground at its output end to prevent the ground potential from becoming negative.

Note: UC3846 is a highly sensitive IC with multiple ultra-high-speed amplifiers inside. It should be installed away from interference sources. If necessary, using a silicon steel shield is also a good choice.

The turn-off speed of the drive circuit of the insulated gate type switching element must be very fast. The switching speed of the switching element is adjusted by adjusting the gate resistance. For its typical drive circuit, please refer to the monograph of MOTOROLA - TMOS power field effect tube. It uses diode unidirectional rectification and PNP transistor discharge and turn-off, and the speed is very fast. The typical value can reach 100ns. It is like it is easy for you to let BMW run 160KM/h, but it is difficult to let Xiali do it. Only on this basis can we talk about adjusting the drive speed. In terms of cost, the cost of the entire full-bridge pulse transformer drive circuit is not even half the price of driver ICs such as M57962, and the circuit is simple, so why not? This drive circuit is common to both MOS and IGBT.

The newly released UC3825 is easier to wire. Its power ground and signal ground are separated. When using UC3846, the ground copper foil should be wide rather than narrow, and the shorter the power ground to the ground point, the better. In addition, the copper foil of the circuit board needs to be thickened to reduce the ground resistance and ground inductance as much as possible. If conditions permit, a three-layer board can be used, with the middle layer as the ground layer, which can greatly improve performance.

The pulse transformer drive circuit has an advantage that other circuits do not have, that is, the bridge arm direct conduction phenomenon will never occur. The pulse transformer cannot output all four levels high, but can only output high levels alternately. As long as the dead time is sufficient, the common conduction phenomenon will never occur. When driven by a driver chip, once the PWM is wrong, it is very likely that the two channels will be all high, causing the bridge arm to be direct (a common occurrence), which will instantly cause the IGBT to self-lock. At this time, the on-chip protection circuit is powerless and can only watch the IGBT explode. Even if it does not self-lock, secondary breakdown will occur (IGBT also has secondary breakdown, but the tolerance is much higher than that of GTR, and it is essentially GTR. MOSFET also has secondary breakdown, but it is much wider than SOA, and generally only occurs during extremely high voltage transients, with a typical value of 30V/ns, which is generally not considered), and the result is the same. Below I will write down my experience with the electroplating power supply. The capacity of my electroplating power supply is not too large. It is an AC/DC square wave output with a voltage of 12V, a current of 200A, and a modulation frequency of 400-1500Hz. I have changed it to a 100A aluminum welding machine and the effect was good.

The control is done by UC3846, which directly drives the pulse transformer.

The main circuit is still the same old one. Due to the small capacity, four IRF360 (25A/400V) are used as the four arms of the full bridge. The feedback sampling is still the same. The main transformer uses the TDK/EI70 magnetic core, the rectifier tube uses IR's Schottky tube 400A/100V, full-wave rectification, and the operating frequency is 110KHz.

The difference is that the output filter inductance is very large, reaching 120 microhenries. The secondary inverter uses full-bridge output, and the switch tube is the IXFN75N10 field effect tube of IXYS company, 6 in each arm. The full-load voltage drop is only 1.3V, so the heat sink is not too large.

Why use a low voltage FET full bridge as the secondary inverter? Instead of the usual half bridge IGBT inverter?

This is because:

1. Calculated from the static power consumption, although the low-voltage field effect tube full bridge is two tubes connected in series, the on-resistance of the low-voltage field effect tube is extremely low. In comparison, the voltage drop is still lower than that of IGBT. In addition, the low-voltage field effect tube full bridge only needs full-wave rectification, which reduces the voltage drop of a diode. Not only that, a much higher driving voltage than the normal voltage is used, reaching 18V. Under normal circumstances, the field effect tube only needs about 7V. However, it should be noted that there is a difference between low-voltage field effect tubes and high-voltage field effect tubes, and their on-resistance component ratios are different. The low-voltage is mainly the channel resistance and the high-voltage is mainly the body resistance. The channel resistance will continue to decrease with the increase of VGS.

2. The reliability of field effect tube is not comparable to that of IGBT.

3. Since there is no filter capacitor in the primary inverter output, if the secondary inverter uses an IGBT half-bridge, the inductor current will have nowhere to be released, and high voltage will be generated during commutation. This high voltage will break through the rectifier tube. In order to prevent such things from happening, a large-capacity RC absorption circuit must be used, and the inductance must be limited, which will significantly reduce the efficiency of the whole machine. This will not happen with a low-voltage field-effect tube full bridge. It can use the method of instantaneous conduction of four tubes at the same time during commutation to provide a current channel (extend the drive pulse width so that the two pulses overlap for a short time), so no high voltage will be generated. At the same time, the continuity of the current is maintained. This is not particularly good for electroplating, but it is different if it is used for welding. At the moment of commutation, high voltage will be generated between the workpiece and the welding gun, and this high voltage will play the role of automatic arc maintenance.

4. The field effect tube is simple to drive. Four Toshiba TLP250 are used here. The disassembled machine only costs 1.9/piece and can be powered by a single power supply.

5. Low-voltage FET full bridge can use unlimited filter inductance and unlimited welding cable. As mentioned in the previous post, too small inductance, including inductance saturation, will cause arc breaking, but because large inductance and commutation high voltage arc maintenance can be used, this circuit can still perform perfect welding when the welding current is only 6A.

The low-voltage field effect tube with a withstand voltage of 100V can be used in any specification of square wave aluminum welding machine. Theoretically, it can withstand an input of 200V. Even the no-load voltage of a 630A aluminum welding machine is only about 70V, and there is still a large margin. The full-bridge secondary inverter only needs a single power supply instead of the positive and negative power supplies that must be used like the IGBT half-bridge.

You can combine a 400A manual welder and a low-voltage field effect tube full bridge to easily get a 400A square wave aluminum welder.

Note: The start of the square wave modulation circuit must be automatically triggered and maintained by the welding current, and it must stop immediately once the current disappears. In other words, the working process should be: output unidirectional voltage - arc start - modulation start - square wave welding - full stop.

This machine also has disadvantages, that is, too many field effect tubes increase the difficulty of assembly (the cost is not necessarily high). In addition, the output must use four terminals, it has no common terminal, and the four terminals can jump out of the secondary inverter and output directly, which is more efficient.