I am designing a ADC module for torque-based motor control application.
- The input source is Wheatstone bridge load cell with excitation voltage of 3 V.
- The loop time of controller would be around 300 Hz.
- Since the torque source also have bandwidth lower than 300 Hz, the frequency of interest would be about <300 Hz.
- 24-bits Delta-sigma ADC MCP3561 would be operated with output data rate of 300 Hz.
- The excitation voltage and ADC reference voltage is supplied by REF3330 voltage reference.
- I have noticed that, since the impedance of load cell is quite low(<300 Ohm), current drained from voltage reference is quite significant(>10 mA, REF3330 output current: 5 mA, REF3030 output current: 25 mA).
- Now I noticed that voltage-follower (or buffer, ADC driver) is needed to add large impedance and reduce current.
- Considering the noise characteristic, rail-to-rail feature, offset voltage, production issue and etc., the op amp OPA376(7.8 nV/sqrt Hz) has been chosen for ADC driving and possibly active filtering.
And here are my questions:
In this kind of application, where should the passive low pass filter be placed? In some design it is placed before the buffer:
BRIDGE SENSOR – PASSIVE LPF – BUFFER – ADC
In another, after the buffer:
BRIDGE SENSOR – BUFFER – PASSIVE LPF – ADC
I believe in the former case the op amp will not suffer for driving quite large capacitive load (~ a few uF in this case) but have some disadvantages on voltage drop before the signal enters the buffer.
On the contrary, in the latter case, a quite clean signal comes in to the module without the possibility of any other load distorting the signal. However, a large capacitor after the buffer might be a problem for the op amp and settling time must be dealt with carefully.
If it is okay to drive a ADC directly with voltage follower, I guess
BRIDGE SENSOR – BUFFER – PASSIVE LPF – BUFFER – ADC
could also be a solution.
Or is there any other superior active filter based solution with similar complexity?
Best Answer
I would recommend:
How I decided the order is based on input and output impedance.
Generally you want to feed circuits with a low output impedance, and you want to load circuits with a high input impedance. Below is an example of why.
In circuit 1, the LPF is loading V1's output impedance R1. Because of that, the 3dB frequency has shifted by 10%. In circuit 2, a buffer has a large (ideally infinite) input impedance, so the source resistance R3 has little/no effect. The buffer has a small (ideally zero) output impedance, so the filter does not load it, and the theoretical 3dB frequency is as expected.
Looking at your suggestions:
^ Passive LPF loads the Bridge sensor. Would also need a differential filter.
^ Probably ok with a delta-sigma ADC (relatively high input impedance). If you were using a SAR ADC and the Passive LPF have a high-ish resistor I'd stick a buffer in-between LPF and ADC.
^ Most conservative option. Again, second buffer may or may not be needed depending on ADC and LPF.
A more complicated, but more compact version could be:
Look up Differential Active LPF for implementations.