Measuring Output Dynamic Response on Power Supplies: Oscilloscope Grounding Issues

Measuring oscilloscope output ripple and dynamic response of low voltage (<1V) / high current (30-150A) power supplies has always been a challenge, with each new setup bringing its own errors. Using the oscilloscope "tip-and-barrel" approach or specialized impedance-matched voltage sense cables addresses errors introduced by probe lead grounding. However, even with the best probing methods, you can get distorted output measurements, especially when dynamic loads are applied or removed. I've noticed two sources of error:

• Ground loop caused by current flowing through the ground side of the voltage probe to the oscilloscope ground and the oscilloscope's AC plug ground connection.

  
  

• When measuring multiple signals simultaneously on the same oscilloscope, the oscilloscope may ground the test setup at multiple points, creating errors in all channels. This is especially true when both output voltage and output current are displayed simultaneously on the same oscilloscope.

Let’s look at the first error source in more depth. If the oscilloscope is grounded to the same building ground as the power supply or output load, then load changes can drive currents in the grounded casing of the oscilloscope probe. This current multiplied by the impedance of the casing will appear as a voltage on the oscilloscope itself, potentially swamping the actual ripple you are trying to measure. Other sources of this grounded casing current include the noisy lab power supply itself (even in static load situations) and external signal generators.

The current methods to eliminate or reduce ground shell current are:

• Wrap the AC wire around the toroidal ferrite several times and connect it to the signal generator.

• Dynamic loads were placed on the test board without using external electronic loads.

Other options include using a battery-powered oscilloscope or one with isolated inputs.

The second error source, oscilloscope ground conflict, is less well known than the first. It is also the cause of distortion in power supply dynamic load response measurements when using an oscilloscope to display both output voltage and dynamic load current changes or multiple output voltages simultaneously. When measuring multiple low voltages, the oscilloscope ground will be connected to multiple points in the test setup via multiple probes. The actual oscilloscope ground will be the “average” of the ground connections.

For example, if the voltage difference between the power ground at the voltage monitoring point and the current monitoring point is +40mV, and the ground connection quality of the two monitoring points is similar, the voltage monitoring point will have a +20mV error and the current monitoring point will have a -20mV error. Current monitors typically have signals of several hundred millivolts, while low-voltage applications such as computer core power supplies allow output voltage overshoot/undershoot of 50mV or less.

Figure 1 is an example of a power supply output test setup where the output response to a large step load is monitored, and the oscilloscope is also monitoring the dynamic load current across a low resistance. I use a signal generator to get a step load with the desired rise and fall times and sense voltage, connected to J502 with a 50Ω cable. The 50Ω R527 suppresses any reflections on the cable. I use the tip-and-barrel method with a 10x oscilloscope probe to detect the current across R500.

Figure 1: Test setup

Figure 2 shows the sensed voltage and sensed dynamic load current on the same oscilloscope when a 59A pulse load is applied and removed.

The application requirement is that VOUT (red curve on oscilloscope/J502 in Figure 1) remain within 855-945mV during load step and load dump conditions. The dynamic current is measured across a 10mΩ resistor to ground (R500 in Figure 1) and is the green curve in Figure 2.

Figure 2: 2A to 61A step load and load dump response with dynamic current sensing when connected to an oscilloscope

Observing the two red curve channels showing Vout in Figure 2, the voltage output drops to 861mV when the load step is applied, and then stabilizes at 889mV at the higher load. When the same increased load is removed (dumped), the voltage output peaks at 940mV before stabilizing at 900mV. Therefore, the voltage output remains within the 855-945mV limit and the test "passes". Observing the three green curve channels showing the voltage across the 10mΩ current sense resistor, the dynamic load increases from 0A to 593mV/10mΩ (or 59A), and then back to 0A.

Disconnect the current sensing probe from the oscilloscope and the output terminal will show different voltage waveforms. See Figure 3.

Figure 3: Same dynamic response of VOUT with current sensing disconnected from the oscilloscope

Looking at the two red curve channels showing the output voltage in Figure 3, the voltage output drops to 863mV when the load step is applied, and then stabilizes at 896mV at the higher load. When the same increased load is removed (dumped), the voltage output peaks at 949mV before stabilizing at 900mV. Therefore, the output voltage is above the 945mV limit and the test "fails".

Off-device testing experience

Testing with an external "load slammer" and trying to monitor the output voltage on the motherboard and the dynamic load on the slammer, the dynamic response was very poor. When I removed the current sense connection from the oscilloscope, I saw good dynamic response. In the above case, the current sense connection caused the false failure.

If you want to monitor current and voltage on the same oscilloscope, you will need either an oscilloscope with fully isolated inputs or a dynamic load with the sense resistor properly grounded at the voltage ripple monitor. For the first option, you will also need two sets of differential probes with good input isolation.