Results

Under Construction... This page will document ongoing data collection and analysis.


Real World Data

Pressure data was collected overnight and plotted on helicorder plots. Each horizontal line is one hour of data.

The same data was filtered into different frequency ranges, for visualization. Pressure data was captured at ~90 Hz. The first plot shows lightly filtered data, from 0.001 - 40.0 Hz:

Microbarometer data from 0.001 to 40 Hz

The first part of the plot was taken during a breezy day. The middle part of the plot is nighttime, which is much quieter.

The plot below is the same data filtered from 1.0 Hz - 40 Hz at 20X magnification.

This being the 4th of July, distant fireworks were going off between 02:00 and 06:00 UTC. Some large spikes during this time correspond to sporadic local neighborhood fireworks.

There are some interesting features at 14:18, 16:18, and 17:18 UTC which need closer examination. They might be related to a microwave oven that was active around those times.

Microbarometer data from 1.0 to 40 Hz.

The Power Spectral Density is plotted against the noise models for worldwide infrasonic stations. If my data corresponds to reality, it should have a shape that matches the expected shape for infrasound observations. It should have an amplitude that falls somewhere between the low-noise and high-noise models (probably not too close to the low-noise model since this is residential neighborhood). In the PSD plot below, the sensor output (blue) agrees with the noise model and has the expected shape.

The sensor and electronics self-noise (orange) is measured by opening both ports to the atmosphere (disconnecting the reference volume).. The 1/f corner frequency (with AD7746 CAPCHOP=0) is observed at ~0.12 Hz. Below this frequency, the noise floor increases due to flicker noise. Enabling the chopper amplifier (CAPCHOP=1) would reduce this noise, at the expense of cutting the sample rate in half.

The spurs at 1.87, 3.74, 29.1 and 30.9 Hz are the aliased 3rd, 6th, 2nd harmonics and fundamental of the 60 Hz AC grid. If the output data rate is reduced to 16.1 Hz, then these frequencies would fall into the digital filter notches and be reduced by at least 55 dB. The band-limited RMS noise shown in the plot below is inflated due to the spurs.

Microbarometer power spectral density.

For comparison, here is the plot of self-noise for a professional infrasound sensor, the Chaparral Physics Model 21. It does ~15 dB better across its passband of 0.1 - 200 Hz. I think they are plotting against a different global noise model. (Bowman 2007 maybe?)

Further tests will be performed:

  1. Set CAPCHOP=1 and repeat the 24 hour acquisition, for both real data and self-noise.

  2. Replace the sensor with a stable fixed capacitor pair and repeat the acquisition to determine the CDC noise floor.


More to follow...