June 18, 2021-
3-axis accelerometer.
I rigged up the Silicon designs 3-axis accelerometer to the Graphtec GL 240 data logger. Each of the three channels was fed into two A/D inputs, differentially. One path was DC coupled and the other was fed through a 1 uF capacitor. The input impedance of the datalogger is around 1 MOhm. That gives a 1 second time constant. Output impedance of the sensor is 1 Ohm so it can drive a long cable.
The accelerometer provides 2000 mV/g, or 2 mV per milli-g. Noise is reputed to be 10 ug per root Hz, so in 100 Hz bandwidth we expect 100 ug rms which corresponds to 0.2 mV rms at the output.
Ran the sensor off a 9V Li ion battery, but it presumably has a built-in DC to DC converter and I didn't filter the power supply at all. Similarly, didn't put a low pass filter on the accelerometer output either.
With 3 differential inputs running, the datalogger can scan at 20 Hz. Any higher frequency vibrations contribute to the rms, just aliased.
Very crude noise determination (with no low pass filter on accelerometer) gives 0.3 mV rms, which rather close to sensor spec, and corresponds to about 0.2 millig's.
With sensor off, rms is around 0.01 mV so we're totally sensor dominated. It might benefit from a low pass filter that rolls off at sampling freq/2, or else a faster A/D.
Temperature: C_T^2.
Plan is to use a pair of 150KOhm fast-response thermistors, separated by 1m. Variance should grow as separation^(2/3). The trick is to not be led astray by air currents. This means the thermistors must have minimal self-heating. For the ones we have, the 'dissipation constant' is 0.050mW/C. So dumping 50 uW into the thermistor produces a 1C change in temperature. Need to calibrate Wheatstone bridge output with one in a known breeze and one shielded.