Current Matching Problems in Parallel Connection of Multiple Devices

0 Introduction

In large-scale conversion devices such as electrochemistry, nuclear fusion, and excitation, there is the problem of parallel connection of multiple power semiconductor devices (such as rectifiers, thyristors, and other new power semiconductor devices). From the perspective of line application, many successful experiences have been obtained [1]~[6]. Among them, references [1] and [6] also give requirements for devices from the perspective of current sharing coefficient. However, from the perspective of devices and their screening and matching, we believe that there is still a need for further discussion and research. Those engaged in device application focus on the intrinsic performance of the device; those engaged in device design focus on the requirements of the line for the device. The combination of the two aspects is the closest way to improve device performance. In recent years, we have made some attempts to solve the current sharing problem of devices in response to user requirements and gained some experience. These experiences are the result of the joint efforts of both parties. This article is an explanation of these little attempts and experiences.

1. The problem of current sharing in devices

When the output current capacity requirement is higher than the maximum available current of a single device, multiple devices must be connected in parallel. For some special applications, such as when power outages and equipment shutdowns due to quality problems are absolutely not allowed, multiple devices are often connected in parallel. In this way, even if 10% to 20% of the devices or branches have problems, normal operation can be ensured.

The on-state volt-ampere characteristics of bipolar devices such as rectifier diodes and thyristors show that the voltage drop curve decreases as the temperature increases, which is the so-called negative temperature coefficient of resistance. Devices with a negative temperature coefficient of resistance are not suitable for parallel connection, which makes parallel current sharing more difficult [7].

To connect multiple power semiconductor devices in parallel, the current sharing problem must be solved carefully. The device current sharing problem can be further divided into dynamic current sharing and steady-state current sharing.

The so-called dynamic current sharing refers to the current sharing from off to on, or from on to off. The former is the main one, and the latter can often be ignored. From off to on, the problem of triggering the opening at the same time is solved. Taking thyristors as an example, as long as they are from the same batch of devices, the error of the opening delay time is within 1 sec, and the entire opening delay time is only a few sec. Therefore, to ensure dynamic current sharing, we must pay attention to:

1) Select the threshold voltage VTO as low as possible [4];

2) Ensure the amplitude (for example, the given trigger current Igm is equal to 5 times the device trigger current Ig) and width (for example, 100 s), especially the steepness of the pulse leading edge (for example, 0.1 s)[5] of the gate trigger pulse, so that dynamic current sharing is guaranteed.

The so-called steady-state current sharing is the on-state current sharing, which is also the most important current sharing problem. From the application perspective, the main current sharing measures include resistor current sharing in small current applications and reactor current sharing in large current applications. In short, they are all passive and based on the premise of adding some additional electrical power.

It goes without saying that the reason for the uneven current is due to the different on-state parameters of the devices. Only by grasping the key on-state parameters of the devices can we grasp the main contradiction of parallel current sharing. This phenomenon is shown in Figure 1.

0 Introduction

In large-scale conversion devices such as electrochemistry, nuclear fusion, and excitation, there is the problem of parallel connection of multiple power semiconductor devices (such as rectifiers, thyristors, and other new power semiconductor devices). From the perspective of line application, many successful experiences have been obtained [1]~[6]. Among them, references [1] and [6] also give requirements for devices from the perspective of current sharing coefficient. However, from the perspective of devices and their screening and matching, we believe that there is still a need for further discussion and research. Those engaged in device application focus on the intrinsic performance of the device; those engaged in device design focus on the requirements of the line for the device. The combination of the two aspects is the closest way to improve device performance. In recent years, we have made some attempts to solve the current sharing problem of devices in response to user requirements and gained some experience. These experiences are the result of the joint efforts of both parties. This article is an explanation of these little attempts and experiences.

1. The problem of current sharing in devices

When the output current capacity requirement is higher than the maximum available current of a single device, multiple devices must be connected in parallel. For some special applications, such as when power outages and equipment shutdowns due to quality problems are absolutely not allowed, multiple devices are often connected in parallel. In this way, even if 10% to 20% of the devices or branches have problems, normal operation can be ensured.

The on-state volt-ampere characteristics of bipolar devices such as rectifier diodes and thyristors show that the voltage drop curve decreases as the temperature increases, which is the so-called negative temperature coefficient of resistance. Devices with a negative temperature coefficient of resistance are not suitable for parallel connection, which makes parallel current sharing more difficult [7].

To connect multiple power semiconductor devices in parallel, the current sharing problem must be solved carefully. The device current sharing problem can be further divided into dynamic current sharing and steady-state current sharing.

The so-called dynamic current sharing refers to the current sharing from off to on, or from on to off. The former is the main one, and the latter can often be ignored. From off to on, the problem of triggering the opening at the same time is solved. Taking thyristors as an example, as long as they are from the same batch of devices, the error of the opening delay time is within 1 sec, and the entire opening delay time is only a few sec. Therefore, to ensure dynamic current sharing, we must pay attention to:

1) Select the threshold voltage VTO as low as possible [4];

2) Ensure the amplitude (for example, the given trigger current Igm is equal to 5 times the device trigger current Ig) and width (for example, 100 s), especially the steepness of the pulse leading edge (for example, 0.1 s)[5] of the gate trigger pulse, so that dynamic current sharing is guaranteed.

The so-called steady-state current sharing is the on-state current sharing, which is also the most important current sharing problem. From the application perspective, the main current sharing measures include resistor current sharing in small current applications and reactor current sharing in large current applications. In short, they are all passive and based on the premise of adding some additional electrical power.

It goes without saying that the reason for the uneven current is due to the different on-state parameters of the devices. Only by grasping the key on-state parameters of the devices can we grasp the main contradiction of parallel current sharing. This phenomenon is shown in Figure 1.

2 On-state theory and basic characteristic parameters

The on-state volt-ampere characteristic curves of most power semiconductor devices can be characterized by the developed Herlet relationship [8], that is, the instantaneous on-state voltage VTM represents the sum of the on-state junction voltage drop, the on-state body voltage drop, and the contact voltage drop. For a well-made device, the contact voltage drop can generally be ignored (the contact voltage drop conforms to Ohm's law, and it is easy to screen it out even if the manufacturing level is poor). The formula for the on-state junction voltage drop Vj and the on-state body voltage drop Vm is:

Formula (3) is a complex function, and the significant influence of on-state current on on-state voltage is implicit in its various parameters.

Although the function form of formula (3) is very complicated, after fully considering the inter-carrier scattering effect, Auger recombination effect, and end region recombination effect, VTM can be calculated according to a certain procedure, and the theoretical on-state volt-ampere characteristic curve is consistent with the actual one. If the parallel devices all have similar on-state volt-ampere characteristic curves, then the current sharing problem will be solved very well.

The function represented by the complex formula (3) can also be approximated using the simplest function form, such as 0, 0.5, 1 power exponents and a simple logarithm, and written as follows:

The four constants A, B, C, and D in the formula can be completely substituted with the data of the four test points, and the VTM can be obtained by solving the determinant.

Obviously, the precise data and measured results of the on-state parameters can be obtained by using formula (3) or (4), but it is not convenient for parallel current matching.

To this end, a straight line approximation is made near the rated operating point of the device to find a standard solution. Using Figure 2, this standard method of dealing with the problem is briefly introduced.

In Figure 2, V1 is the peak voltage under 0.5ITM, V2 is the peak voltage under 1.5ITM, VTM is the peak voltage under ITM, ITM is the on-state peak current, ITM=3 (or 3)ITAV, and ITAV is the positive half-wave average current.

From Figure 2, it is easy to get the following relationship.

On-state peak voltage

Here, the on-state threshold voltage VTO is the on-state voltage value determined by the intersection of the on-state approximate straight line and the voltage axis; the on-state slope resistance rT is the resistance value calculated from the slope of the on-state approximate straight line. The on-state threshold voltage VTO and the on-state slope resistance rT are the characteristic parameters for measuring the quality of the on-state characteristics and reflect the essence of the on-state characteristics.

Almost all books on power electronics applications have a sentence about parallel current sharing: "Try to choose devices with consistent characteristic parameters." There are so many device parameters, and there are dozens of on-state parameters alone. How to choose them? Some choose to use the consistency of the on-state average voltage drop VT as the principle of current sharing matching, while others choose to use the consistency of the rated on-state peak voltage VTM as the principle of current sharing matching. Practice has shown that their limitations are very large. We believe that according to the actual working current of the user, selecting the consistent on-state threshold voltage VTO value and the on-state slope resistance rT value is the correct matching principle for current sharing of parallel devices.

Use formula (3) or (4) to calculate, or directly use a peak voltage tester to measure the V1 and V2 values, and substitute them into formulas (5) to (7) to immediately obtain the VTO value, rT value, and VTM value. The group with close VTO and rT is the group with successful current sharing matching.

3 Test results and calculation matching

At the request of the user, the grouping results of parallel current matching when ITM=300 A, i.e., ITAV=100 A, are also required. This is the real conventional situation in actual use. For this reason, with ITM=300 A as the center, 150 A and 450 A are used as the test or calculation currents of the new V1 and V2 values, respectively. The result of 450 A in Table 1 is used, and the result of 150 A is added, as shown in Table 2. The parameters in Table 2 are marked with a superscript O to distinguish them from the parameter symbols in Table 1. The data in Table 2 are clear at a glance and no further explanation is required.

4 Conclusion

1) From Table 1 to Table 2, from I TM=900 A to I TM=300 A, that is, from the high current section of the on-state volt-ampere characteristic curve to the low current section, the threshold voltage reflecting the junction voltage is reduced, which is exactly what is needed for parallel current sharing. The slope resistance reflecting the internal voltage drop is slightly increased, but this does not affect parallel current sharing. Due to the need for parallel current sharing, it is completely necessary to provide two sets of on-state threshold voltage and on-state slope resistance values ​​in the product sample, and some foreign samples do this.

2) The data in the two tables are obtained from the recent parallel current sharing test. They seem a bit idealized and not typical, but they are very clear as a method for screening matching devices for parallel current sharing test. It is generally believed that the devices for parallel current sharing must be selected from the same batch, so the test screening and matching work is much easier, that is, the matching screening with consistent opening time is eliminated. The selection of the test (or calculated) current in the parallel current sharing of devices must be determined by the actual current of the user, which is a prerequisite that must be paid special attention to.

3) The above parallel current sharing method is obtained after many failures. The methods we have adopted are:

(1) Focus on the turn-on parameters, such as ensuring the trigger parameters are consistent. However, when the amplitude and width of the gate trigger pulse are ensured in the application, the impact on parallel current sharing is small;

(2) Focusing on the average on-state current ITAV and the average on-state voltage drop VT, matching and screening are performed according to the current sharing coefficient method provided in the literature [1]. Although it meets the user's application requirements, it is still not ideal. This is because the average on-state voltage drop VT reflects the average effect after stabilization. In this way, there are more variable factors and it is more difficult to control. It is not as good as using the instantaneous value.

(3) The rated on-state peak voltage VTM value is used for matching screening. Only one point value, which is often not the actual operating point value, is used as the screening parameter, so it has greater limitations.

4) When replacing the current-sharing device of a whole machine, in principle, the same batch of devices should be used. If the device is replaced with a product from a different manufacturer, the threshold voltage of the on-state and the slope resistance value of the on-state must be completely similar under the same conditions, and the turn-on time must also be tested and matched, otherwise the parallel current-sharing will fail. There are many lessons in this regard.

5) When a large number of devices (e.g. 8 or more) are assembled in parallel on a line, the problem of edge current concentration often occurs due to the effect of the magnetic field. Since it is closely related to the site, the user will seriously solve this problem.

6) Practice has repeatedly proved that according to user requirements, it is a convenient, simple and practical way to use V1 and V2 value testing (or calculation) to determine the on-state threshold voltage and on-state slope resistance value, and then match the current-sharing devices. According to this current-sharing matching method, under certain conditions, the reactor can be omitted and the direct parallel connection of thyristors can be realized.