Suggestions for accelerometer signal conditioning

I want to use an ADXl326 accelerometer to measure vibration in one plane only. The accelerometer is fed with 3.3 V and returns 1.65 V at 0 g +/- 57 mV/g

What I want to do is remove the 1.65 V bias to get an AC signal, then amplify that signal so it is close to -1.65 / +1.65 V so that I can add the bias back and have a DC signal again, which is close to the full ADC range of 0-3.3 V.

I considered using an AD8221 instrumentation amplifier with no gain resistance on pins 2 & 3 i.e. unity gain, followed by a PGA112 programmable gain amplifier with SPI control and 1.65 V on the reference pin.

However, I was wondering if anyone had any suggestions which would not need a -5 & +5 V supply?

• Very well asked. I would have edited your question for proper punctuation and voltage V units, but am on mobile now. What is the maximum G you are expecting? Input peak AC voltage... Mar 6, 2017 at 14:45
• There's no need to remove the bias and then put it back in. Just amplify the signal with the bias built into the amplifier. There are many examples of such "signal conditioning" (Google that phrase) circuits on the Internet. Mar 6, 2017 at 14:58
• Add an analog high pass filter. Jun 4, 2019 at 10:31

simulate this circuit – Schematic created using CircuitLab

The zero G null adjust pot only has to correct source null error away from 1.65/3.3=50% so ratio of 1k/20K is 5% can be reduced to 2% for example.

Higher Gain is possible but choice of low input bias current, rail to rail type, unity gain stable Op Amp is necessary.

This is just one of many examples, not final version... after you specify full scale g and frequency limit.

• The accelerometer measures -16/+16 but I wouldnt expect anywhere near that. The application is bearing condition monitoring so a healthy bearing with a very low vibration, say 0.1g's will need more amplification to use max adc range than a bad one with 5 or 6 g's hence the pga. Frequency limit would be 1600hz. Mar 6, 2017 at 18:03
• You have not accepted an answer yet. You don't need a bipolar supply or signal. Only need to convert 57mV/G for +/-0.1G to say 10V/G = +/1V but still centred at Vdd/2 to avoid offset error with a matched resistor pair. (0.2%) or use an offset fine error null. Then your gain becomes 10V/57mv = 17.5 x for R Ratio or boost to full scale with 16.5V/g . OK? Oct 7, 2019 at 1:11