Saber simulation-assisted design of topological inductors in switching power supply circuits

1. The position of the (input and output) filter network in the circuit.

The topological inductor (transformer) is required by the topology, and the filter inductor is required by the ripple. The filter inductor is used (LC filter network is added) only when the topological inductor is not enough to meet the ripple requirements.

This means:

1. If the topological inductor meets the ripple requirements, the filter circuit can be omitted.

2. When the topological inductor cannot meet the ripple requirements, the filter circuit is considered separately.

3. The main task of the topological inductor is to deal with the energy transfer required by the topology, not to deal with ripple.

4. The only task of the filter circuit is filtering, nothing else.

2. The relationship between the filter network and the topology.

All voltage-type topologies can always be expressed like this:

Among them, the input capacitor Cin and the output capacitor Cout are allowed by the topology, or even required by the topology.

At the same time, Cin and Cout can also be understood as the topology's own, built-in filter circuit.

Here, the filter network in the dotted line is now a capacitor, that is, a two-terminal filter network, but it can also be a three-terminal or even a four-terminal network.

Note: There is no inductor in the figure, and the inductor (or transformer) of the topology is not drawn in the topology module. 3. Output filter network

For most voltage-type topologies, there is always a capacitor Cout at the output end, and this capacitor means filtering.

In general, we can always meet any required ripple requirements by adjusting the size of Cout.

However, in some cases, we cannot obtain the required output ripple by adjusting the size of Cout, such as:

1. When meeting the required ripple, the required Cout is too large, and the cost and volume do not allow it.

2. When operating close to short circuit (such as electric welding machine or spot welding machine), the current index of ordinary capacitors cannot meet the requirements.

3. Some applications do not allow too large Cout to exist, such as inverter systems, and too large Cout will lead to control difficulties.

4. For reliability reasons, electrolytic capacitors are not used at the output end.

5. Due to the existence of capacitor ESR, high-precision power supplies cannot always meet the required output ripple indicators.

What should we do?

In fact, it is very simple:

1. Find an acceptable capacitor

2. Divide this capacitor into two

3. Put an appropriate inductor in the middle

4. Adjust this inductor until the output ripple requirements are met.

A few points of explanation:

1. Generally, power supplies output active power, that is, resistive loads. In this case, we directly take Co1 = Co2 for the best filtering effect.

2. Even if the load has some inductive components, because Co2 is generally large, its capacitive reactance is sufficient to cope with the impact of large inductive loads, and generally there is no need to consider increasing Co2.

3. When there is a capacitive load (such as an electrolytic power supply and a charging power supply), you can consider reducing Co2 (that is, highlighting Co1), and it doesn't matter if it is greatly reduced.

4. Co2 can (should) be cancelled for welding power supplies. 5. This method

should be used with caution for resonant loads (such as ultrasonic power supplies and induction heating power supplies).

6. Filtering is filtering, don't mix it with the inductance in the topology, only in this way can the best effect be achieved.

7. Except for special circumstances, it is not recommended to use two-pole or multi-stage LC filtering. When the total capacitance and the total copper and magnetic flux are equivalent, the single-stage filtering ripple effect is the best, and no standing wave interference will be generated. 4. Design Example (Typical) Output Filter

Taking the 50KHz, 100W (120W) flyback power supply in the previous post as an example, the current ripple index is 30mV.

Now the requirement is to achieve a 2mV ripple accuracy.

Method 1: Increase the output filter capacitor:

Increase the current filter capacitor C2 of 2200uF by 15 times, that is, 33mF, and the output ripple will be reduced by 15 times (without considering ESR), which is equal to 2mV.

If you think that 33mF25V large-scale electrolytic capacitors are difficult to find or not cost-effective, then:

Method 2: Add a level of LC filtering:

When Co1 = Co2 = 470uF, with a 5A 1.3uH inductor, the output ripple (same frequency as PWM) can be reduced to below 1.6mV. Or:

When Co1 = Co2 = 330uF, with a 5A 2.2uH inductor, the output ripple can be reduced to about 2.0 mV. It can be seen that even if a little bit of LC filtering is added, the improvement in output ripple, cost and volume is very significant.

Let's look at the working condition of this filter inductor:

The DC component of the current is 5.0A, and the AC component is about 0.1A, which is only about 2%.

In other words, this inductor is basically a DC biased inductor with very little AC component. This means that you don't need to use advanced materials, and don't need to consider the skin effect. You can just use a single strand of ordinary iron powder core magnetic ring winding.

The following are the design parameters of this inductor:

Summary:

It is very simple to add an LC filter network at the output end. As long as the filter capacitor is divided into two and an inductor is (arbitrarily) inserted, the filtering effect can be significantly improved (not inserting the inductor is equivalent to the original circuit), and the effect is always better than the single capacitor filtering effect. Therefore:

1. Engineers should always think: "Should my filter capacitor be divided into two, with a small I-shaped insert in the middle?" And there is no need to calculate, it is definitely better than a single capacitor.

2. This method improves the filtering effect under the same conditions, or reduces the cost, reduces the volume, and even reduces the PCB area under the same filtering effect.

3. Since it can be achieved without increasing the cost (or even reducing the cost), it is a laborious and thankless task to reduce the ripple in the topology (on the inductor or control mode), and the problem of "such and such topology, such and such mode has large ripple" should no longer be a problem.