Some essential equipment is needed to effectively measure torque ripple. A torque sensor with really high accuracy and bandwidth is required, and noise immune torque communication is equally as important. Analog signals are extremely prone to noise in a Pulse-Width Modulation (PWM) environment and so, if you have a ten volt signal, you are very likely to pick up inverter switching noise which can distort signals, especially if you are looking at torque ripple due to inverter switching.
HBM’s torque cells use a frequency output and frequency modulation. They output a square wave at a given frequency, and modulate plus and minus frequency based on torque. This makes them highly noise immune on the sensor side. On the data acquisition side, you‘ll need a system that can record and translate that noise immune frequency output at a sufficient bandwidth. If you are not acquiring the signal and conditioning it correctly, you are going to lose all the bandwidth information that is available to you. Furthermore, you want an acquisition system that is going to correlate that high bandwidth torque to electrical signals such as voltage or current, power signals, or other mechanical signals such as position, vibration, and displacement.
To be highly effective, engineers need two things: a high-end data acquisition system with high-end torque sensors; and the knowledge of how to combine them for detailed analysis.
Accuracy, bandwidth, and time alignment are key aspects of a larger issue: system efficiency. Take an internal combustion engine that is 30% to 40% efficient. A 3% error gives you 39% efficiency instead of 36%. This is great. When looking at electric machines that have much higher efficiencies, range is significantly more important because of the battery life. You have a motor efficiency of 85% to 98%. A 3% error here gives 101% instead of 98%, but this is impossible.
Torque ripple fits into this context because high accuracy torque accounts for really small disturbances in the average. To truly understand this, look at an equation for a somewhat high-speed machine: 80 kW @ 20k RPM → 2093 Rad/sec x 38.22 Nm → .25 Nm offset is 500 W → .625%
This example comes out with 0.5% inefficiency. The torque is an important measurement that must be obtained precisely and accurately, especially at higher speeds. The bandwidth for our signal also needs to be time-aligned in phase with voltages and currents because, for efficiency, an average of torque and speed over time is taken and divided by an average of voltage and current over time.
Furthermore, time alignment is needed so that one of these high peaks is not added into an efficiency measurement, because having half a ripple instead of a full ripple could make a difference. Filtering this out would cause your test to be slower, and you would lose some of that bandwidth and phase information. There is real value in accounting for accuracy and bandwidth in your efficiency measurements.