Design

Specifications

  • Full scale: +/- 200 Pascal (+/- 2.0 millibar, +/- 0.8 inch water column)

  • Sensitivity: 40,000 digital counts / Pascal (4,000 counts / microbar)

  • Noise: 2 milliPascal (0.02 microbar) RMS 0.25 mHz - 2 Hz.

  • Frequency response: 0.25 mHz to ... (40 Hz? TBD)

  • Reference volume: 1 L thermos bottle, vented by 25 gauge (0.010 inch ID) needle, 0.65 inch long.

  • Sensor type: differential-capacitive. +/- 4 pF full scale, <50 pF common mode.

  • Membrane: 1 mil thick aluminized mylar, 4.5 x 4.5 inches

  • Front-end electronics: 24-bit AD7746 Capacitance to Digital converter

  • Interface: USB

Note: 1 Atmosphere = 14.7 psi = 406.493 inches H2O = 101,253 Pascal = 1,012,530 microbar.

Sensor

The sensor is a differential-capacitive air pressure sensor. It is fabricated from PCBs, aluminumized mylar film, and closed cell double-sided adhesive sheet. Dual air ports are provided to connect airline tubing. One side of the sensor connects to a reference volume, and the other side is open to the atmosphere.

The reference volume includes a calibrated slow leak, provided by a length of fine-gauge hypodermic needle. This functions as a pneumatic high-pass filter, allowing the reference volume to equalize to slow barometric pressure changes. The length and/or gauge of tubing controls the time constant, and hence the corner frequency of the filter.

Relative air pressure is determined by measuring the differential capacitance between the two capacitors formed by the mylar membrane and fixed plates on each PCB.

Electronics

Ignoring resistance of the aluminum coating, a simplified representative circuit for the sensor is shown below. With the membrane at the zero position, with no deflection, nominal values are:

  • C1 = C2 = 8.2 pF

  • C3 = C4 = 8.2 pF

  • C5 = 8.0 nF

C1-C4 represent the air gap between the mylar and plates. C5 represents the mylar itself, aluminumized on both sides. C5 is constant and much larger than the others. With deflection, C1 and C2 increase while C3 and C4 decrease, more or less symmetrically, so:

  • Cdiff = (CIN+ - CIN-) ~= (C1-C4) / 2

Microbarometer differential capacitive pressure sensor simulation.

Full scale range of +/- 4.096 pF equates to +/- 10.3 mils (260 microns) of membrane travel.

In the best case scenario, ignoring noise contributions from other sources and looking only at the AD7746 peak-to-peak noise floor of (an astoundingly low) 27 aF, this implies a membrane deflection measurement limit of ~2 nm. In reality, it probably won't be anywhere near that. But it does suggest a very high sensitivity for the sensor. See the testing section for more detail.

An Analog Devices AD7746 Capacitance to Digital Converter directly connects to the CIN+ and CIN- plates and measures their capacitance. It uses a delta-sigma charge-balancing technique to measure the charge transferred to an unknown capacitor compared to a known reference capacitor. This converter can provide as much as 18 bits of noise-free resolution. The AD7746 is configured to measure the differential capacitance between the two plates (Cdiff), ignoring their non-changing common-mode capacitance (C1, C2, Cparasitic) Therefore the dynamic range is maximized.

Eventually the CDC and a microcontroller will reside on a custom board, possibly attached directly to the sensor. In the short term, the Analog Devices EVAL_AD7746EB evaluation board will be used instead. It contains the AD7746 and Cypress FX2 that communicates over USB.

Software

Initially, the board will be controlled by a custom Linux app that issues I2C commands over USB using a Cypress-defined protocol. The board runs the Analog-provided firmware. The app configures the AD7746 and collects capacitance and temperature data at the sample rates of 9.1 Hz - 90 Hz.

The host will log the data, process it, and generate plots periodically.