Testing

Sensitivity

Step 1: Perform a rough calibration to determine sensitivity.

The complete unit was placed into an insulated box, with a loosely fitting lid. It was left on the floor for an hour to thermally equilibrate. Continuous capacitance measurements were recorded. The box was then placed on a 1 m high table. It was left to equilibrate for a few minutes. Then it was placed back on the floor. This process was repeated a few times and the results were averaged.

A 1 m height change resulted in ~580,000 CDC counts, or roughly 3% of the full range of the CDC.

Full scale range then is: 2^24 / 38527 = +/- 217 Pascal.

The table has a wooden top and steel legs, and this could have contributed a parasitic capacitance change. So, try another method..

Step 2: Suspend the box on a rope and pulley.

A pulley at the ceiling was used to suspend the box on a rope. It was raised and lowered by 1 m keeping a roughly equal distance from floor and ceiling. The surrounding area is free of other objects that might contribute parasitic capacitance. Data was captured continuously.

The ambient barometric pressure changes are superimposed on the data, but the plot below shows a 1 m height agrees with Step 1 above. A similar value of ~524,000 CDC counts/m was measured.

Step 3: Calibrate against a water column.

A differential manometer was constructed and filled with water containing a drop of food coloring. Surfactant (dish soap) was added to reduce surface tension. One side of the manometer was left open to the atmosphere. The other side connected to a port on the sensor and a syringe. The other sensor port was open to the atmosphere.

This was surprisingly difficult to measure. Sensor full scale range was 3.6 mL of syringe travel. The water level changed ~0.05 inch, corresponding to ~15 Pascal, clearly at odds with Steps 1 and 2 above. The small tubing diameter may have contributed an excessive amount of surface tension. There was no apparent leak, and the CDC output remained constant after adjusting the syringe. With larger pressure changes the manometer responded as expected and remained steady. Needs follow-up..

Noise

Tubes were disconnected so both ports were open to the atmosphere. 1000 samples (110 seconds) were collected at the lowest sample rate of 9.1 Hz, and noise statistics calculated using the Analog Devices tool.

• RMS noise: 80 counts

• Peak-to-peak noise: 852 counts

• RMS resolution (ENOB): 17.68 bits

• Peak-to-peak resolution (noise-free): 14.27 bits

Considering the full scale of +/- 217 Pascal, this means:

• RMS resolution (ENOB): +/- 1.0 milliPascal (10 nanobar)

• Peak-to-peak resolution (noise-free): +/- 11.0 milliPascal (110 nanobar)

This noise floor then represents the effective measurement limit of the instrument: 2 milliPascal (0.02 microbar) RMS.

Expressed as membrane travel, this equates to 2.5 nm RMS. Hard to believe!

(When recording at a higher sample rate, the noise level will increase.)

Frequency Response

The slow leak on the reference volume acts as a high pass filter, allowing the sensor to equalize to slow barometric pressure changes. A syringe was connected to the reference volume, and moved by ~1.5mL to produce a step response. The data was logged and curve-fit to find the time constant. The high pass corner frequency is 0.25 mHz at 78 F.

The low pass corner frequency has not been determined yet.