I ask the engineer at the EPA, “So, you are OK if I report 101% efficiency for my gearbox!?”

Efficiency is just Power Out / Power In. How hard can that be to measure? If your product is in the high-90% efficiency range – extremely difficult.

True story – the EPA was working on the 2024 regulations for commercial vehicle fuel economy certification, and my team was asked to review the draft for what become 40 CFR § 1037.565, the method to certify transmission efficiency. The original draft referenced a diesel engine test procedure that called for a torque meter with a 1% accuracy – which was acceptable to measure the power output of the engine when testing standard pollutants, but for a transmission that was 99% efficient in overdrive, if the input torque meter read 1% low, and the output torque meter read 1% high, the indicated efficiency would be 101%. If a gearbox, power converter, or motor reads over 100% efficient, at least you know your measurement was rubbish!

Differential power measurement (DPM) - measuring the input and output power directly is theoretically straight forward, but you are really trying to measure the power LOSS, or the system inefficiency – you are trying to measure the 1% of power that is missing, not the full power entering the system. At 90% efficiency, the error on the missing power is multiplied by 10x. At 99% efficiency, the error on the missing power is multiplied by 100x when referenced to the input or output; thus, a sensor with a 0.1% accuracy will have a 10% error in power loss, assuming it is properly scaled to the test. In the real world, the uncertainty is typically worse.

Node variables like Voltage, speed, or pressure are typically not too difficult to measure with sufficient accuracy, but through-variables like current, torque, or flow are often very difficult and expensive to achieve sufficient accuracy for efficiency measurement. Accuracy as a percent of point vs percent of full-scale range (%FSR) is a topic for another blog, but using multiple sensors to measure at low, medium, and high power is typically necessary.

If you want to show that your product is 0.5% more efficient than your competitors, prepare for a long and expensive ride if you want to use DPM. There are a couple other options to consider.

Calorimetry, or directly measuring the waste heat from a system is a direct measurement of power loss and can be a powerful tool to accurately measure system losses. If you wrap your product in a thermal blanket, measure the cooling flow rate (air, oil, or water flow rate) and the temperature rise of the coolant, you can easily calculate power loss knowing the heat capacity of the coolant. You can achieve decent results with inexpensive measurement equipment that scales over a wide power range. A 1% error in flow rate measurement is a 1% error in power loss, so testing over a wide power range with a single setup is reasonable. The limitations of calorimetry are that you need to operate a steady state for a LONG time to get a measurement, waiting for the system to achieve steady-state. Eliminating or qualifying the thermal leakage can be difficult. On any liquid-cooled product, I would always include calorimetry, at least as an independent check. If DPM and calorimetry agree, you should have confidence in your results.

Idle power measurement is another powerful method. If you disconnect the shaft from a motor, or unplug the load from a converter, the output power is exactly ZERO. The input power is exactly equal to the power loss for that special case. If you need to generate an efficiency map over a wide operating range, using one setup to measure idle power over the full input operating range, and a second setup to measure power loss at full load. Assuming the losses are linearly independent, interpolating between the models will give great results. Bonus points for proper application of a Kalman filter to combine DPM, calorimetry and idle losses.

I hope this wasn’t Too Much Information - Sincerely, Tom