Flexible Analog Signal Conditioning

Status
Not open for further replies.

brtaylor

Well-known member
Overview:
I'm working on a Teensy 3.6 project where I would like to create an analog signal conditioning circuit that can be flexibly configured by the Teensy to work with a variety of input signals. I'm anticipating this circuit to need to work with:
1. Potentiometers
2. Strain bridges
3. Analog instruments (i.e. pressure transducers, accelerometers, gyros, etc)

It would be nice to also accommodate RTD temperature sensors; however, this can be on it's own circuit if needed.

Analog data will be sampled at a fixed 400 Hz rate, so I would use a 200 Hz AAF stage as part of my analog filtering. Other aspects of this project will make the per-unit costs high, so I'm not budget limited for this design.

Initial Thoughts:
I already know that I'll likely need several reference voltage sources available for the customer to use. The customer has typically used 10V reference sources for their POTs and strain bridges. Most analog instruments I see are looking for a 3.3V source. So, my current thought process is to make several sources available: something like 3.3V, 5V, 10V, higher? I know that I'll need an AA filter stage. My biggest question is how to attenuate the incoming voltage signals? I could use a voltage divider set to the largest reference voltage, follow it with a trim stage (using the Teensy's DAC and a summing amplifier), and then a PGA to bring the potentially tiny signal up to meet the full scale range of the Teensy ADC. Are there better approaches like programmable voltage dividers, below unity op amps, or digital POT's that I should be considering instead of a voltage divider?
 
If you care about precision, an old-fashioned resistive divider can be made quite accurate if you use precision resistors, they don't notice ESD and are probably OK with brief overvoltage. I don't know the specifics of what parts you may have in mind, but if you go with programmable or digital devices I would be careful to check the specs on tempco, noise, and any nonlinear behavior.
 
I think this is a fairly open question, so consider a mux. One input to the mux is scaled for 10v signals, another for 5 volt signals, another for 3 volt signals, and a 4th with a gain stage ahead of the mux. The output of the mux feeds a simple low pass filter. To enable the low pass filter to be simple, the sample rate is increased to 3200 hz and a simple digital filter is used to clean up what the analog filter missed. The sample rate is then decimated to the desired 400 hz.
 
I think this is a fairly open question, so consider a mux. One input to the mux is scaled for 10v signals, another for 5 volt signals, another for 3 volt signals, and a 4th with a gain stage ahead of the mux. The output of the mux feeds a simple low pass filter. To enable the low pass filter to be simple, the sample rate is increased to 3200 hz and a simple digital filter is used to clean up what the analog filter missed. The sample rate is then decimated to the desired 400 hz.

Thanks! I kind of like this approach - seems like it could be lower noise than my original idea.
 
A question regarding voltage references. I'm planning on using the TI REF5025 (http://www.ti.com/lit/ds/symlink/ref5020.pdf) voltage reference to provide a 2.5V reference. Similarly, I'm planning on also using the REF5050 and REF5010 to provide 5V and 10V reference sources. These will be options to provide voltages for potentiometers, strain bridges, etc.

I decided to use a 2.5V reference instead of the 3.3V mentioned in post 1, because this gives me a nice 1, 2, and 4 gain on my PGA. I will be using a MK66FX1M0 (Teensy 3.6 MCU), but integrating it on a custom PCB, so I can alter the layout. I'm planning on connecting VDDA to VDD through a ferrite bead and VSSA to VSS through a ferrite bead, same as the Teensy 3.6. VREFH and VREFL will be connected to the TI REF5025 outputs. Should I use VDDA and VSSA to supply power and ground to the REF5025 or should I use VDD and VSS?
 
I think you have an unrealistic expectation for low pass filter performance.

If you set the corner frequency to 200Hz and use a 400Hz sample rate then the attenuation will never be better than 3dB at the Nyquist frequency. If aliasing really is a concern (you know that there is signal or noise above Nyquist) then the corner should be at 1/10th of the sample rate. This gives your filter a couple of octaves to do its thing.

Then you have to ask yourself just what is the minimum attenuation required above Nyquist? This will determine the filter order. Then you have to decide if the passband gain response is more important (Butterworth) or phase (Bessel).
 
Status
Not open for further replies.
Back
Top