# Gas station without pumps

## 2012 July 1

### Instrumentation amp, try 2

I did some more playing around with the 2-op-amp instrumentation amplifier and finally got it to sort-of work.

The gain should be $(R_1+R_2)/R_1 + 2 R_2/ R_{gain}$, that is 30.8kΩ/6.8kΩ + 48kΩ/330Ω or 150. I put in C1 to block the DC component of input, then used Rup and Rdown as a voltage divider to get a small difference signal (about 1/101th of the Vplus signal).  I needed to add C2 to eliminate some high-frequency noise that was only apparent at Vout. I still have a huge 60Hz noise signal at the output of the amplifier, (about 70mV peak-to-peak, so about 0.5mV on the input). I can change that by doing things like touching my laptop case, which reduces the 60Hz noise by about 10%.  Bring my hand close to the circuit input increases the 60Hz noise.  I’m not going to be able to do an EKG unless I can make an amplifier with better rejection of common-mode noise.  (I’ve bought a couple of the TI INA126PA-ND amplifiers—I think that may be the only way I’ll get enough common-mode rejection.)

I tested to see whether I was getting amplification of the common-mode signal or the difference signal by looking at what happens if I reverse the connections of Vin_plus and Vin_minus to the signal source.  This negates the different signal, but leaves the common-mode signal alone.

With the wires hooked up as shown, Vout is about equal to Vin_plus, but swapping the wires makes Vout about –2.5 Vin_minus (in both cases, the input is taken as the AC voltage from the top of C2 to ground).  The reversal of polarity is a good sign that the differential signal is being amplified, but there is also an amplification of the common-mode signal.  I tried combining the two measurements to get the common-mode and differential gains—it looks like a common-mode gain of about -1.25 and a differential gain of about 125.  Since my voltage estimates were not very accurate (maybe ±30%), this is compatible with the instrumentation amplifier having the right differential gain. If I could measure the AC voltages accurately, instead of eyeballing them on the scope,I might be able to get better estimates.  Since I’m just whistling into the mic to get a signal, it is difficult to get more accurate measurements.

To determine whether the gain was positive of negative, I used the dual-trace scope and triggered on one channel.  If the two signals were in phase (positive gain) they superimposed.  If the two signals were in opposite phases (negative gain), one went up when the other went down.  When I tried doing an X-Y plot, by putting the input signal on the x axis of the scope, I could see that I was not getting exactly 0° and 180°, because the pattern on the screen was an ellipse rather than a straight line.  The 180° was closer to being right than the 0°.  I suspect that the poor common-mode rejection is due to my badly mis-matched resistors—I’ve not tried computing what the effect of a 1% or 5% difference in the supposedly matched resistors would be.

I can also see the common-mode amplification by connecting both Vin_plus and Vin_minus to the same signal.  When I do that, I see a common-mode gain of about 0.2, with a phase shift of about -90°.

I think I need to come up with a better signal source for testing common-mode gain and differential gain.  I did finally order myself the FG-500K function generator kit, to replace the function generator that died.  I decided that the digital readout of the VC2002 function generator was not worth the extra \$80.  I may come to regret that decision.