Evaluating a DC to DC converter in the lab
Switch mode power supplies are used in most electronic systems. Especially in applications where linear voltage regulators were used, switching power supplies offer the advantage of higher power efficiency. Evaluating a switch mode power supply circuit in the lab requires some special knowledge. Though measuring output ripple voltage, conversion efficiency, inductor current and loop stability seem straight forward, not applying the correct procedures will yield useless measurement results. In this article some of the most important procedures are introduced.
A good setup
The most important part of the evaluation is a decent test setup. Figure 1 shows a poor setup and Figure 2 shows an optimized setup. The difference is the length and the geometry of the cables. In Figure 3 we see the device under test in the middle and on the left some power source, usually a lab power supply and on the right we see the load. This load can be a passive resistive load, a reactive load such as a capacitor or Inductor or it can be an electronic load.
Since our power supply evaluation is targeted at testing the device under test and not the setup, we need to make sure that parasitic influences such as the cable inductances from the power source to the device under test as well as the cable inductances from the device under test to the load are as small as possible. Long cables with a large area in the current flow loop, indicated with red arrows in Figure 3, will increase the impedance of the cable. Load transients or line transients can easily be influenced by a setup with long cables. As a conclusion we use short cables and we reduce the area the current flow is surrounding by twisting the cable together as shown in Figure 2.
In the photographs from the bench, we use regular cable clips to attach to the power supply board. For an even better setup we would solder the ends of the thick cables directly on the board. Very often doing so is not practical and so we can also use clips as long as the expected currents are rather small.
Measuring the output voltage
Since the most basic purpose of a DC to DC converter is to generate a certain voltage, measuring the output voltage is quite important. One might think that attaching an oscilloscope voltage probe to the output voltage is all you need to do. If you do this as shown in Figure 4 however, you will see an output voltage waveform as shown in Figure 5. Besides the average output voltage of 3.3 V, you also see periodic voltage spikes with a peak to peak voltage of roughly 1.1 V (0.5 V per division). This can be quite astonishing if you consider that the whole purpose of this power converter is to generate a regulated output voltage. Who would like to accept a 3.3 V voltage rail with 1.1 V of noise on top of it?
Actually the 3.3 V are much cleaner than what has been measured. The trouble was actually the measurement setup and not the device under test. The trouble is the long ground connection of the voltage probe. This little cable has ground potential, but it actually is a high impedance node. All sorts of EMI (electro magnetic interference) is being picked up by this ‘antenna’. A better measurement setup is shown in Figure 6. Here we removed the gnd cable from the voltage probe and replaced it with a very short gnd connection right from the shielding of the probe tip to a ground connection close to Vout, the point where we connect the scope probe.
Figure 7 shows the measurement result with this improved setup. Now we see a clean 3.3 V with much smaller periodic spikes. The resolution is now 20 mV per division resulting in only 75 mV of measured noise spikes.
We can conclude that the spikes were not real on the output voltage but just a measurement pickup. Another thing to add is the naming convention of output ripple voltage. With the term ‘output ripple’ power supply engineers refer to the periodic ripple which repeats itself with the switching frequency and typically has the shape of either a triangle type or a sine type waveform. Since figure 7 has a higher voltage resolution, we can see the ripple waveform. The exact shape is defined by the output capacitor with its ESL (equivalent series inductance), ESR (equivalent series resistance) and capacitance value. In our example we can see that the ripple looks like a sine wave, we can tell that the output capacitance seems to have very little ESR and ESL.
Measuring the inductor current
It is also quite important to measure the inductor current. All switch mode regulators, except for charge pumps, use some sort of inductive element to store and release energy. Every inductive component such as a coil or a transformer can only store a limited amount of energy. If this limit is exceeded, the inductor saturates and the inductance goes down. This saturation can happen very quickly with inductors that have one air gap in the core, or it can happen quite gradually with inductors with a composite core. In such a core the air gap is distributed within the core material and the resulting saturation happens slowly. This means that the inductance value of the inductor versus the current running through the inductor is only slowly decreasing.
A switching regulator usually has the feature of a current limit. This current limit is supposed to protect the circuitry during over load or short circuit conditions. This current limit senses the inductor current during the on-time or the off-time. Ideally this current limit function reacts instantaneously. In reality there is a certain propagation delay of a few nano seconds however. During this time, excessive currents can flow and damage the circuitry. As a protection for these initial few nano seconds, the circuit designer selects an inductor which has enough inductance left even at currents as high as the current limit of the switching regulator.
To evaluate the inductor in the circuit one needs to measure the inductor current. The best way of doing this is to unsolder one side of the inductor, lift it up on one side and attach a short cable in between the uplifted inductor terminal and the printed circuit board. This cable loop can then be used to attach a current probe. Such a measurement setup has the advantage of being able to sense AC as well as DC currents. Also the current measurement will not pick up much switching noise.
When lifting one side of the inductor up, make sure to select the side away from the switch node. In a buck regulator this is the output voltage side of the inductor, in a boost regulator this will be the input voltage side of the inductor. For the current measurement itself it does not matter which side of the inductor is lifted up. However, attaching the measurement cable loop on the noisy side, will couple switching noise capacitively onto other parts of the circuit and the circuit function itself might be negatively influenced.
Figure 8 shows a board with the ADP2164 and the cable loop for sensing the inductor current. The ADP2164 is a fully integrated 4 A step down switching regulator. Figure 9 shows in the solid line how the inductor current looks in normal operation below the saturation level of the inductor. The dotted line shows the inductor current ripple with a slightly saturating inductor. Towards the maximum currents, the slop rises since the inductance is reduced and the characteristic peaks can be seen. Inductors with a very sharp saturation behavior might not allow a measurement such as the dotted line in Figure 9. Those inductors either show a waveform as the solid line or run excessive currents breaking the switching MOSFET of other parts of the circuit.
Evaluating startup and line transient response
The startup of a circuit can be quite challenging. A power supply sees in the first instance a short circuit on the output. The reason for this is that the output capacitor is discharged and the power supply has to charge this capacitor in addition to providing current to the load at the output. With certain loads and especially with large output capacitors, it can happen that the power supply goes into over current protection during startup and never comes up. One way of preventing this is to use a soft-start function. With such a function, the output voltage is gradually increased and by that the peak currents during startup are reduced to values below the current limit. Also most systems prefer a slowly ramping output voltage to prevent any unintended voltage spikes.
Besides startup, line transient response should also be tested. What happens to the output voltage of a power supply as the input voltage changes quickly between different values? For such a test a programmable lab power supply with adjustable slopes is very useful. If such a lab supply is not available. Two different lab power supplies can be used with different adjusted voltages. Then they can be attached to the input of the power supply under test via diodes. These diodes in the Vin path protects the lower voltage power supply from seeing the higher voltage from the higher voltage power supply. Figure 10 shows a setup with two lab power supplies.
Is the circuit stable?
A stable circuit is necessary for fixed frequency operation, lowest output ripple voltage and best line and load transient behavior. Stability can easily be checked by looking at the falling edge jitter. If the falling edge jitters more than about 5% of a switching period it is usually considered to be instable. Stability should be evaluated at different line and load conditions. Figure 11 shows excessive falling edge jitter on a switch node voltage measurement.
A common test to find out how stable a circuit is and how fast the control loop reacts is a load transient test. Figure 12 shows three different plots of how the output voltage recovers back to nominal value after the load performed a step from low load to high load. The x axis is time, the y axis is output voltage.
Figure 12: Different load transient responses
The bottom plot shows a typical underdamped system. Here the circuit tends to be close to instability with phase margins, an indication for stability, far below 40 degrees. In the middle plot, we see an overdamped system. These systems tend to be very stable but unfortunately relatively slow in recovering the output voltage after a load transient. The plot on the top is the ideal situation. It has enough phase margin of about 50 degrees and a fast recovery of the output voltage after a load transient. These systems are called critically damped, since there is no voltage overshoot after recovery of the output voltage but also the output voltage comes up as quickly as possible.
In many cases these stability tests are good enough for evaluating the behavior of the control loop of the switching power supply. If more detailed evaluation is necessary, a Bode plot, gain and phase diagram, of the control loop can be taken using a network analyzer.
Measuring the efficiency
For efficiency measurements we hook up a voltage meter and current meter to the input and output of the device under test. Never trust the voltage or current reading on an electronic load and on a lab power supply for an efficiency measurement. On most power supplies and loads this reading is very inaccurate. Especially the voltage measurement of such devices tends to be poor. This has to do with the fact that many of such devices do not have separate Kelvin sense cables which would need to be attached right at the device under test. Typically accurate multimeters do a good job. Attaching the multimeter for voltage sensing right at the device under test inputs and outputs is key to prevent inaccurate measurements due to voltage drops on the power lines.
When the circuit has been turned on and a measurement is performed, one can see constant changing values on the multimeter’s voltage and current readings. When is the best time to take the actual reading? The change of the measurements comes from the circuit heating up during operation. As the board temperature increases, the efficiency of the circuit is reduced. For accurate efficiency data, one needs to run the board for a certain time, usually many minutes, until the board is at steady state temperature. This is usually where the board efficiency measurements are most useful.
One other fact to pay attention to especially at low power DC to DC converters is that the currents and voltages on the input and on the output of the device under test are not all DC. Especially the input current on a buck regulator is an AC current. The input capacitor of the circuit is supposed to filter the pulsed input currents but there is usually still quite a lot of AC content on the current. For really accurate efficiency measurements we need to find out how well the multimeter averages this AC current. High quality multimeters have a very good precision at DC and also give a good precision when certain waveforms are measured at AC. However, the multimeter specification only specifies accuracy at well defined waveform shapes such as saw tooth, sine and triangular. The input current of a buck regulator has a different waveform. For very accurate efficiency measurements, a multimeter should be tested with a signal generator supplying a waveform close to what the DC to DC converter will generate on the input, a known resistance and an oscilloscope. Figure 13 shows such a setup.
Figure 13:
Test setup to evaluate how well a multimeter averages changing currents
For most applications, the power supplies do not need to be tested at such a great level of detail.
Thermal behavior
Other key evaluations include a thermal test. This test needs to be performed at the steady state operating condition but also at overload and short circuit conditions. The most critical circumstance is during overload conditions. Then the components see the highest thermal stress. Often in short circuit conditions, the power supply protects itself and shuts down operation periodically. Good over current protection schemes such as with the ADP2164 make sure that the thermal stress during current limit operation is low. To find out which the most stressful over current condition is, slowly increase the load current until the point where the current limit in the device under test is triggered. Typically the switch node will start to skip pulses. Right below this threshold is the operation with the most thermal stress for the components.
During such evaluations, it is very nice to have a thermal camera to measure the temperatures of the different components on the circuit. For accurate thermal measurements with such a camera, the board would need to be painted in a matt finish black color. Otherwise, there will be some false measurements especially on shiny reflective surfaces such as solder joints and shiny copper.
Without a thermal camera a thermo sensor can also be used. This is quite cumbersome however to place the sensor onto each component on the board. For such placement thermo conducting paste needs to be used to make the temperature measurement accurate.
Often it is a good idea to use a thermal sensor on some point on the board under evaluation even if a thermal camera is available for the measurement. This way the quality of the thermal image can be double checked.
Are we done yet?
This article discussed many of the basic ways of evaluating a DC to DC converter. Besides these, there are other aspects that can be tested such as EMI, special safety tests and evaluations of moisture sensitivity, vibration and others. For many of these more specialized tests, dedicated equipment other than standard lab equipment is needed. This is why very often such evaluations are outsourced to dedicated test labs.
About the author
Frederik Dostal studied Microelectronics at the University of Erlangen, Germany and started to work in the Power Management Business in 2001. Dostal has been active in various applications positions including four years in Phoenix, Arizona working on switch mode power supplies. Dostal joined Analog Devices in 2009 and works as Power Business Technical Manager for Europe.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
