RF Cavity Breakdown Localization: Sensor and Signal Studies on Al - PowerPoint PPT Presentation

RF Cavity Breakdown Localization: Sensor and Signal Studies on Al Disk Peter Lane Pavel Snopok 1

Overall Setup ● Aluminum disk from mock cavity as medium ● Ultrasonic “ping” generator ● Sensors and speaker around disk circumference 2

Ultrasonic Ping Generator ● 555 for pulse width and pulse frequency ● 555 for signal frequency ● LM386 audio amplifier driving the transducer ● Scavenged piezoelectric transducer ● Separate 9V wall adapter powering the circuit – signal pollution and battery voltage issues 3

Sensor Calibration ● Check that sensor polarities are the same ● Check that sensor response is similar ● Input Signal – One pulse every 3.6 s – Pulse width of 0.48 ms – Signal frequency of 79 kHz ● Positioned speaker and sensor next to each other on side of Al disk 4

Sensor Calibration Setup 5

Sensor Polarity 6

Sensor Response 7

Sensor Calibration Conclusions ● Polarities are all the same. – Any polarity flips would have to come from reflections at metal/air boundary. ● Amplitudes can vary among sensors by ~2V. – Makes relative peak finding more difficult. – Hypothesis: Any signal which is observed having a uniform amplitude on all channels is EMI. 8

3-Channel Signal Analysis ● Use a consistent acoustic signal to better understand what is happening inside the cavity walls. ● Input Signal – One pulse every 270 ms – Pulse width of 32 µ s – Signal frequency of 120 kHz ● Senors placed at 90 degree intervals around the circumference 9

3-Channel Signal Setup 10

3-Channel Signal Overview 11

2D Modeling of the Disk ● Model a pressure wave with the same symmetric arrangement of sensors as before 12

Asymmetric Source Signal Simulation 13

3-Channel Signal Conclusions ● Signal envelope mean lifetime about 9 ms. ● Sensors close to the source are clean. ● Sensor further from the source is dirty due to reflection interference. ● Envelope oscillations due to reflections ● Time domain wave front detection is rather difficult even with nice input signals 14

Alternative Ideas for Wave Front Detection ● Integrate sensor data backwards in time to reconstruct the source position – Too computationally intensive for an online tool ● Use signal database or neural network – Requires the luxury of training on known outcomes ● Sensor symmetry correlations – Try to match signals from symmetrically placed sensors 15

Alternative Ideas for Wave Front Detection ● Rolling FFT idea from Keith Pedersen – Density plots show visible bands at points in time related to the beginning of the pulse – Need to see if this has sufficient resolution and can be turned into an detection algorithm – Will experiment with this if there are no other suggestions ● Need DAQ card to experiment in realtime 16