The supply and utilization of system power is a critical factor in the design of every type of electronic product. Personal computers, telecommunications and broadcast equipment, and military equipment are all pressuring the humble power supply to provide ever more current in smaller and smaller packages. Due to their compactness and lower cost, high-frequency switching supplies (switchers) have become the preferred solution for most applications. Although more complex than traditional linear supplies, switchers are without peer in delivering current all out of proportion to their size.
The behavior of any power supply must be well understood before it is designed into a product or released to the market. After all, it will be the foundation of an end product's efficiency, safety, and reliability. The prudent designer will characterize a power supply's behavior thoroughly during the design stage. Components in the supply, particularly active components, may encounter momentary conditions that far exceed their average operating levels. The engineer must be aware of these peaks and account for them when choosing components like power MOSFETs (metal oxide semiconductor field effect transistors) and IGBTs (insulated gate bipolar transistors).
A broad repertoire of accepted power supply measurement procedures exists, but several basic methodologies are essential in power supply design evaluation. Historically these approaches have relied on static current and voltage measurements taken with digital multimeters. However, meaningful data about dynamic performance is needed to detect hidden problems in switching components. Consequently, the oscilloscope has become a cornerstone of switching power supply evaluation. With switching frequencies and edge speeds on the increase, a full-featured wideband instrument is needed to capture subtle signal details.
This article will discuss some of the tools and techniques used to observe a power MOSFET's in-circuit behavior in a power supply, emphasizing the tools and preparations needed to ensure accurate measurements. The article will show how these readings can detect instantaneous power peaks that elude static measurement techniques.
The Unit Under Test
Figure 1 shows a simplified circuit for the input side of a switcher. The MOSFET power transistor configuration is typical to switching power supplies and most power conversion products-PWM motor drives, electronic ballasts, and many others. The MOSFET is floating-it has no reference to either the input AC ground or the output ground terminal. Simplistic ground-referenced measurements with a conventional oscilloscope setup are not possible here because connecting the scope probe's ground lead to any of the transistor's terminals would short-circuit that point.
Figure 1. The circuit under test, showing the converter MOSFET and the test points for the drain-to-source voltage measurement , Vds.
Defining the Measurement and the Tools
The power measurement across the MOSFET is made up of two constituent waveforms: voltage and current. The product of these two variables at any instant in time is the instantaneous power. The graph of these products over time is the power waveform. The shape, amplitude, phase, and timing of the power waveform combine to tell a story about the real-world stresses on the MOSFET.
A lab-quality DSO, the Tektronix TDS 510A, has been chosen as the measurement platform for two reasons: 1.) Its ability to display not only voltage and current waveforms, but also to compute and display power waveforms with direct readout in watts; 2.) Its fully integrated interface to precision differential voltage and current probes. Moreover, the scope's 500 MS/s sample rate can capture very fast switching transients faithfully.
True Differential Voltage Measurements Make a Difference
The solution of choice for measuring the MOSFET voltage waveforms is a differential measurement. The voltage excursion is measured between two points (for example, the voltage between source and drain, Vds in Figure 1), neither of which need be at ground potential. Depending on the range of the power supply, these voltage waveforms may be riding on top of a voltage ranging from tens of volts to hundreds of volts. There are several ways to accomplish this measurement, in ascending order of preference:
- Elevate the scope's chassis ground. This extremely unsafe method endangers the operator, instrument, and unit under test. Moreover, it yields very imprecise measurements. This approach doesn't merit further discussion.
- Use two conventional scope probes (with their ground leads connected only to each other) and the built-in channel summing capability of an oscilloscope. This is known as a quasi-differential measurement. Unfortunately, the passive scope probes in combination with the scope's amplifiers lack the CMRR (common mode rejection ratio) to block the common mode voltage adequately. This setup cannot capture the measurement with good accuracy.
- Use a commercially-available probe isolator to isolate the scope's chassis ground. Thus the probe's "ground" lead is no longer at ground potential and can be connected directly to a test point. Probe isolators are an effective solution but are very costly, on the order of 2- to 5 times the cost of good differential probes.
- Use a battery-operated scope with individually isolated inputs, for example the Tektronix THS 720 TekScopeTM. When used with carefully chosen probes, this method delivers good results, especially in field service applications.
- Use a true differential probe on a wideband oscilloscope. This is the most appropriate method for critical measurements like those used to predict power supply component reliability and performance.
A true differential voltage probe (the Tektronix P5205) was chosen for this measurement because of its high CMRR (common mode rejection ratio), low circuit loading (only 7 pF input capacitance), and 100 MHz bandwidth.
Picking Up the Current Waveform
Of course, making the voltage measurement is only half the job. Acquiring current waveforms is a discipline all its own, with specialized tools and techniques. The common digital multimeter, though suitable for static current readings, lacks the ability to display the waveform properties in an AC environment. Here again the oscilloscope is the best tool for examining amplitude, timing, and phase characteristics.
The Tektronix TCP202 current probe was chosen for this application. This is a "non-invasive" probe; that is, it doesn't require breaking into the circuit to connect the probe. Its clip-on pickup acquires the signal by induction. Like the P5205 probe, the TCP202 relies on the scope's TekProbeTM interface to provide automatic ranging, scaling, and readout of the measurement in engineering units.
Preparing for the Measurement
A little time spent setting up the scope/probe system in advance can help ensure stable, repeatable power measurements. Both the current and the voltage probes are affected:
- There is a simple "nulling" procedure that should precede any instantaneous power measurement. Both the P5205 and the TCP202, and other probes of their type, have built-in DC offset trimmers. With the unit under test turned off and the scope and probes fully warmed up, set the oscilloscope to measure the mean of both the voltage and current waveforms. Use the sensitivity settings that will be used in the actual measurement. With no signal present, adjust the trimmers to null the mean level for each waveform to OV, or as close as possible. This step ensures that the "quiescent" voltages and currents in the measurement system are not added to the levels at the test points.
- It is essential to use current and voltage probes with well-matched delay characteristics. Otherwise the power measurement-actually the product of instantaneous voltage and current readings-might be corrupted by delays that, for example, shift the current waveform relative to the voltage waveform. Such a shift would in turn displace the peaks in the power waveform, possibly leading to an incorrect assessment of the transistor's behavior.
The P5205 differential voltage probe and the TCP202 current probe are inherently matched to within ±2ns, close enough for most applications. In addition, some scopes itself provide an adjustment for further delay equalization (deskew) between the probes, if needed.
- In spite of the differential voltage probe's high (80 dB) CMRR, it's wise to verify the probe's performance in the actual measurement environment. To do so, simply connect both leads to the same test point, for example the drain of the MOSFET. Both probe tips see the same signal-a "common mode" signal. Ideally, the differential probe should reject the whole signal and display a flat trace on the scope screen. In reality a small amount of the signal is passed through, and the resulting trace reveals common mode error. While this simple test isn't definitive, it will expose gross CMRR problems that might affect the measurement outcome.
If problems do arise (for example, when using a lesser-quality differential probe) the common mode error can be subtracted mathematically by the scope. While triggering on the current waveform, capture the common mode error waveform as previously explained and save it in the scope's reference memory. Then subtract this fixed quantity from each measurement using the oscilloscope's built-in math function.
The Moment of Truth
After all the preparations, the measurements themselves are relatively simple. The object is to examine the nature of the switching transitions in the 40 kHz converter circuit depicted in Figure 1. For the differential voltage reading, the probe tips are connected to MOSFET source and drain terminals. The resulting Vds waveform is shown in Figure 2 (upper trace).
Figure 2. The TDS 510A display, showing voltage, current, an power waveforms, in addition to numerical readouts.
For the current measurement, the clamp-on probe must acquire a signal from a conductor passing through its inductive pickup core. If it isn't physically possible to clamp around the conductor of interest (in this case the lead coming from the MOSFET's drain), then it will be necessary to add a loop of wire in series with the signal as a test point. In fact, this technique can be used to increase the sensitivity of the current probe if necessary. Instead of just one loop of wire, use several turns-the sensitivity of the probe will be multiplied by the number of turns. Figure 3 illustrates the technique.
Figure 3. A non-invasive current probe (the Tektronix TCP 202) attached to a multi-turn wire loop in series with the signal path. This technique increases the sensitivity of the current measurement.
The current waveform from the MOSFET measurement is shown in Figure 2 (middle trace). This particular reading did not require the increased sensitivity technique mentioned above, and therefore gives a correctly scaled current waveform and readout.
At this point we can begin to see the direction this measurement is heading. The traces are almost complements of one another: on the voltage trace, voltage is at its maximum when no current flows, and at its minimum when current is at its peak. However, a brief transient in the current waveform disturbs an otherwise smooth switching transition. This transient occurs during the time when there is still approximately 60 V Vds voltage across the MOSFET.
The bottom trace in Figure 2 is the power measurement, automatically computed by the TDS 510A oscilloscope. It reveals just what the voltage and current traces promise: a single strong peak that coincides with the current transient. This is the reading that summarizes the circuit behavior. Assume the MOSFET was chosen for an average power capacity of, say, 20 watts. Conventional Digital Multimeter current and voltage readings would indicate that the transistor was operating well within safe limits. But can this MOSFET withstand a 30W peak in every switching cycle? Just as importantly, is this peak raising the average power dissipation of the circuit to unacceptable levels? Why is the trailing current transition so much cleaner than the leading edge? These questions can point to solutions for the problem, which may range from changing the switching characteristics to simply using a larger MOSFET. Conversely, the true power, as calculated by using the scopes' Mean function to determine the mean value of the instantaneous power readings, may indicate that the transistor operating within safe limits. Either way, the designer can make informed decisions about the components in the circuit.
Looking at the power measurement, it's easy to see why all the preparation before the measurement was important. For example, small DC offsets in the probing tools, when compounded by scale factors and multiplication, can lead to large numerical errors. Likewise, a delay difference between the current and voltage probes would change the relative positions of the two respective waveforms. As a result, the peak in the power waveform would be dislocated, or might disappear altogether! This could lead to a design that wouldn't be tested until it was in the marketplace-a situation that nobody likes to risk.
An accurate power measurement system using a high performance oscilloscope, a true differential probe, and a precision current probe is the best toolset for characterizing the active components in a switching power supply. The scope-based measurement methodology helps designers evaluate the variables that produce cost-effective, yet reliable and market-worthy power supply designs.