Piezoelectric energy harvesting

One common use of energy harvesting systems is in those places where the access to conventional sources of energy is difficult due to availability, space constraints, environmental hazards or sealed equipment. This article explores the possibilities of piezoelectrics to extract electrical energy and store it in capacitors to supply power of ultra-low power microprocessors.

The piezoelectric under test is a Lead Zirconate Titanate PZT-5J with reference S128-H5FR-1808YB manufactured by MIDE. The tests I am going to conduct are aimed at a specific application in which the piezo has to extract energy at 100 Hz with an acceleration as low as 2 ms^{-2} or 0.2 g.

From right to left, the setup consists of a signal generator, signal conditioner and amplifier (top), signal amplifier (bottom), oscilloscope and vibrator or shaker with the piezoelectric (behind the oscilloscope).The signal generator feeds a 100 Hz sinusoid with an amplitude of 50 mV to the amplifier which is connected to the shaker. The signal conditioner is connected to an accelerometer and converts its output from picocoulombs to mV so 10 mV corresponds to 1 g or 9.8 ms^{-2}. The output of the signal conditioner is connected to the oscilloscope to visualize the frequency and amplitude of the vibration.

The next picture shows a detailed view of the piezoelectric and a piece printed with a 3D printer to change the oscillation frequency by changing the free length of the cantilever. A screw with nut is attached to the tip of the piezoelectric to add a mass which helps in the selection of the resonant frequency. This figure shows the setup that sets the resonance to 100 Hz.

The first test was to set an acceleration of 2 ms^{-2}, blue plot or channel 2 in the figure below and measure the output of the piezoelectric, magneta plot or channel 3. Channel 2 is connected to the signal conditioner and amplifier to have the output of the accelerometer and it shows a peak value of 23 mV which corresponds to 2.3 ms^{-2}. Notice that both the signal generator and the amplifier are set to their minimum values so getting accelerations below this is difficult.

The peak value of the piezoelectric is 1.2 V. I need at least 5 V to drive the energy harvesting power supply LTC3588-1 so it is necessary to add a voltage multiplier at the output of the piezo electric show in the figure below.

The four capacitances can be adjusted to the type of voltage source, or in my case, the piezoelectric. It was found that using 10 microfarads the current drawn from the piezo was inside limits so its output voltage won’t drop much. It is true that reaching the steady state at the output would be faster using lower capacitance but this is not critical for this application. This circuit multiplies by four and rectifies the sinusoid given by the piezoelectric, therefore, the output would be close to 1.2 x 4 = 4.8 V minus the voltage drop in the diodes. I selected Schottky barrier diodes 1SS384 with a forward voltage as low as 0.18 V at 1 mA. Even with this multiplier circuit the resulting voltage, 4.8 V, is not enough to drive the LTC3588-1 so another stage would be necessary. Since I didn’t have that new stage ready, I rised the vibration amplitude to 4.1 ms^{-2} to have a peak close to 1.6 V at the output of the piezo. Then, the output of the multiplier in this case is 5.79 V which can drive the LTC3588-1 power supply and I can have a steady value of 3.3 V at its output as shown in the yellow plot of the next figure:

I used the LTC3588-1 breakout by Sparkfun which has a capacitance at the output of 10 microfarads and added another capacitor of 1000 microfarads. The charging time with an acceleration of 4.1 ms^{-2} is close to 25 minutes.

The next step is to test how long I can use an ultra low power microprocessor with a charged 1000 microfarads capacitor. The program cycle of the microprocessor would be: Enter shutdown mode -> Wake up with the LTC2588-1 PGOOD signal as interruption -> Read data from a sensor -> Store data in an external EEPROM -> Enter shutdowm mode. All the cycle would take a few milliseconds so I am confident that the charge in the capacitor would be more than enough.