In Pressure sensor assembly yesterday, I talked about putting together the pressure sensor circuit board. What I didn’t mention is that I also wired up a differential amplifier with a DC gain of about 1072, using the instrumentation amp prototyping board that I designed earlier. I didn’t mention it, because when I powered it up, I got a signal stuck at one rail, and I was afraid I’d have to rethink the entire lab, because we wanted to measure DC pressures, and if there was a DC offset large enough to saturate the differential amplifier, we would not be able to.
This evening I decided to try debugging the circuit. I began by looking at the output voltages of the sensor, which should be almost identical at about half the power supply voltage. They weren’t. So I was afraid that I’d damaged the sensor somehow.
My next step was to disconnect the sensor from the power supply and amplifier and try measuring the resistances of the various connections on the breakout board. I was looking to see whether there was a short (which would be hard to fix now that I’ve glued the board to the PVC plug) or an open. The circuit for the pressure sensor is a simple one.
The bridge resistors are silicon piezoresistors with a positive temperature coefficient, and the zero temperature coefficient series resistors provide a voltage increase with temperature that compensates for the decreased sensitivity of the piezoresistors with increased temperature. According to Freescale Application Note AN1318, “in a typical MPX2100 sensor, the bridge resistors are nominally 425 ohms; RC1 and RC2 are nominally 680 ohms.” So I was expecting to see about 425Ω between the S+ and S- connectors, about 1100Ω from either output to either power rail, and about 1790Ω between the power rails.
I measured the 6 pairs of contacts and got
Seeing this made my error obvious: I’d miswired the PC board, so that the connectors weren’t connected to what they should have been! I went back to the data sheet for MPX2300DT1 (Rev 8, 04/2010) and realized that no where on the sheet did they give the mapping of the pins. I had guessed the mapping of the pins from their other packages, but they are not the same!
On the breakout board, the power pins are labeled to S- and 0, while the sensor outputs are labeled S+ and +5.
So I rewired the amplifier to the pressure sensor, leaving 0 connected to Gnd, but connecting “S-” to the 5v supply. I then connected “+5” to the positive amplifier input and “S+” to the negative amplifier input. Now the sensor correctly gave a reading in the middle of the voltage range for both outputs, and the instrumentation amplifier provides and output that is midrange also. Blowing into the plug to increase the pressure pins the output at the highest voltage and inhaling with my mouth pressed against the plug pins the output at the low end.
I think that the gain on the amplifier is too high, since I’m pinning the output too easily. The sensor is supposed to provide 5µV/V/mmHg, so with a 5V power supply, it should provide a 25µv/mmHg differential signal, and a gain of 1072 should provide 26.8mV/mmHg. It should take a pressure of 93mmHg (1.8psi, 12.4kPa, 50″H2O) to pin the output of the amplifier. I tried sucking water through a long tube, and found that I could support about 50″ of water (not carefully measured), so this calibration seems about right.
The amplifier also provides a lot of low-frequency noise, about ±1V and with pulse lengths around 3–5msec. I suppose that this could be capacitive or inductive pickup from the leads, since I was using untwisted speaker cable for the connection. In fact, disconnecting the pressure sensor from the amplifier inputs, but leaving the speaker cable in place provides a large enough signal to swing the amplifier from rail to rail, even with the ends of the leads shorted together. Replacing the speaker cable with a twisted pair does not seem to reduce the noise, and the noise is not synced to the AC power supply (using “line” triggering on the oscilloscope, so this does not seem to be simply AC noise coupling.
I’m going to have to think some more about whether this experiment is going to be worthwhile, since it looks like I’ll have to redesign the PC breakout board with a proper pinout. I may not be able to sense small changes in pressure without better noise control, and I doubt that I’ll get big enough pressure changes from the shaker table. (The bursting membrane approach is still possible, though.)