Accurately Test Magnetics Carrying DC Bias Current
In power supplies, dc-dc converters and many other power electronics circuits, chokes and transformers form an essential part of the filter network that provides low-ripple dc power. In many designs, such as the forward converter, the main output smoothing choke is also an essential part of the complete switching converter design topology. For this reason, it is vital to the overall performance, reliability and efficiency of the converter that the smoothing choke behaves predictably at all possible operating points of the switching converter.
During the manufacture of this critical component, it is possible that manufacturing errors or a combination of design tolerances may produce a choke that will not perform satisfactorily in-circuit. It is common practice then to test the choke before it is fitted to the power circuit.
During design, the power-circuit design engineer will confirm design calculations and margins by measuring inductance (Q), ac resistance and turns ratio over a range of frequencies, and will confirm overload capacity (or margin) by raising dc bias current to the point at which the transformer or choke saturates as detected by a fall in inductance. In addition, during manufacture, spot measurements are made to confirm that the component has been correctly assembled using the right core material and air gap, as well as winding the right number of turns of the specified wire.
Whether during component development or power-supply development, there are certain pitfalls engineers must avoid when testing power inductors. These include the use of common test methods that may produce wildly inaccurate results or that require cumbersome test setups. A new versatile method described here helps designers avoid such pitfalls and achieve highly accurate inductance measurements.
Tests for a DC Choke
A common example of an inductor that carries dc current is the output storage choke of a forward converter (Fig. 1). In this example, the current in the output choke (IL) consists of two parts: the dc load current and a saw-tooth current (iL) that is determined by the voltage across L, Vl and the value of inductance:
The dc current is often the most significant part of IL.
In terms of the B-H loop of the choke's core, this means that the magnetic flux density (B) is also shifted from its normal loop around zero (Fig. 2). The magnetic design of the choke must ensure that there is a sufficient flux density margin to avoid saturation with the dc bias current being applied.
To avoid core saturation caused by high levels of dc bias, it is common to design this type of choke by introducing an air gap into the core, either physically with a spacer for ferrite cores or by using powdered core type. The additional complexity of the design further increases the need for comprehensive testing at the operating point. At this point, the choke must provide the desired inductance to the operating frequency, and as low as possible an impedance to the dc current.
The inductance measured at the operating point is often different than the inductance measured without bias, because the slope of the B-H curve changes over the range of H. It is important to test for the correct inductance in the presence of the maximum-rated dc current (Table 1).
In the design environment, the design engineer is making sure that the design calculations were correct and that the part will work as intended. The ultimate test is to study the part's behavior in circuit — checking waveforms, power loss and overall circuit performance — but making measurements first will usually save time.
In addition to the production tests described in Table 1, the engineer may perform the tests outlined in Table 2.
Applying a DC Bias Current
To measure inductance, an LCR or impedance meter is connected across the choke (Fig. 3). The meter applies an ac voltage at the desired frequency and determines the imaginary part of the impedance across its terminals. The inductance (L) equals XL/(2πf). If the dc bias current is applied using a conventional bench power supply, or even a battery, then it's impedance must be significantly higher than that of the choke being measured. Unfortunately, this is never the case for a simple supply (Fig. 4).
The output capacitance of the power supply will dominate the impedance measurement and give a wildly inaccurate result. For example, if the inductance of the choke is 100 µH and its operating frequency is 100 kHz, then it has an impedance of:
XL = 2πfL = 2 × π × 100 × 10-6 × 100 × 103 = 62.83 Ω.
If the output capacitance of the power supply is 10,000 µF, then its impedance is:
The impedance measured by the LCR meter is that of the power-supply capacitor and choke impedance in parallel, and the impedance measurement is swamped by the capacitor. Consequently, the result is 100% inaccurate.
A common solution to this problem, as used by most commercial apparatus, is to feed the dc power-supply output to the choke via a coupling inductor (Fig. 5). A typical solution uses a coupling inductor up to 10 times the value of the choke under test to reduce errors due to the source appearing as a load to the meter. However, to measure different values of chokes, several different values of coupling inductors must be fitted inside the dc bias supply and relay switched into the power circuit. Still, the resulting measurement will be fairly accurate with a measurement error of ±10%.
This solution may provide a reasonable measurement at optimum conditions, but it is large, heavy and often un-reliable because of the relay switches. Practical considerations of size, and particularly self-resonant frequency, limit the performance of this method in terms of both accuracy and the highest frequency that can be measured.
A New DC Bias Supply Topology
A better method of supplying the constant current from a power supply to the choke under test is to use a transistor (Fig. 6). The transistor is not employed as a constant-voltage device as you would expect in a power supply, but is positioned after the output smoothing capacitor (C) and operates as follows.
An ideal bipolar transistor exhibits the following characteristic for a given value of base current (Fig. 7). As long as the transistor has sufficient current gain (B = IC/IB) and is above the saturation for collector-emitter voltage (VCE), then the collector current (IC) is constant for any value of VCE. For an ideal transistor, the slope of the characteristic is zero:
The impedance of the bias supply seen by the LCR meter and the choke is given by:
In other words, this circuit, using an ideal transistor, provides infinite impedance to the LCR meter and has no effect on the recorded measurement. In practice, the slope is not zero and there is some variation of IC with VCE:
IC = B × IB + VCE s,
where s is the slope in amps per volt. For a typical power transistor, the slope could be 1 mA/V. This represents an impedance of 1/s = 1 kΩ for the basic circuit. By careful use of ac current feedback and compensation, this value can be further increased from 1 kΩ to 100 kΩ or more. Typically, this impedance will introduce a measurement error that is well below the error of the LCR meter itself. In practice, the resulting measurement error with this technique is just 0.1% or less.
In Voltech's DC1000 dc bias supply, the use of this topology provides users with small, robust and easy-to-use supply that avoids all of the common testing pitfalls described earlier. Inside the DC1000, microprocessor control provides users with either a simple manual rotary dial to set the current or comprehensive remote control via RS232. Sweep software is provided for measuring inductance at increasing levels of bias current to confirm the saturation point of the part under test.
Each DC1000 can supply up to 25 A, and 10 units may be paralleled to supply up to 250 A if required. A key feature of the DC1000's topology is that errors are well defined and independent of the type of LCR meter being used. This means that the power supply can be used with confidence with any manufacturer's LCR meter.
Wrong wire gauge used.
Wrong core material or incorrectly gapped.
Wrong core material or incorrectly gapped.