Analog or digital double integration

[Jaguarjoe]

I need to double integrate an accelerometer's output to get velocity and displacement. I can do it analog or digital and I wonder which way is best. The accelerometer will be a MEMS device. The displacement is about 3/8" and its frequency is about 1-5 Hz.

Thanks for all of your input!

Joe

 

[schmitt trigger]

It depends how one defines "best".

Lowest cost?
Highest performance?
Easiest implementation?
Simplest to design?
Smallest size?
Lower current consumption?
Other?

I would guess that digital circuit could be the "best" in many categories...but at least for me, an analog integrator would be far easier to design and troublesoot.....my analog skills are superior to my digital skills, but that is only me.

But I also suspect that a "hybrid" approach could also work the best. An analog front end doing some preliminary signal conditioning could substantially reduce the ADC performance requirements and lead to a simpler processing algorithm.

 

[jpanhalt]

I faced a somewhat similar dilemma with a thermal accelerometer. That device operated at 100 Hz, so sampling time could be a factor in your type of application with that device. Your frequency may he high enough that that is not an issue.

What sort of g will this device experience or, alternatively, how much time for the 3/8". Will acceleration be constant? What device are you considering?

John

 

[misterT]

What is the application? Why are you measuring acceleration/speed/displacement?

 

[cachehiker]

When I was last called on to measure dynamic deflection of a trampoline, we did the integration in software but used an analog high pass filter with its offset nulled prior to the A/D converter. It wasn't hard but we were already doing an FFT and finding a few local maxima in code as well.

The high pass filter wasn't 100% required but the amplified, piezo-based accelerometer would return to a slightly different "zero" between jumps, causing the integrator output to gradually rise or fall and make the running chart drift off the screen quicker without it. We didn't mind but the middle management suck-ups wanted real time feedback so they could make their uninformed yes-man decisions faster.

 

[crutschow]

For stability and accuracy as well as ease of response modification, a digital approach would be preferred if the maximum frequency is 5 Hz.

 

[Jaguarjoe]

What is the application? Why are you measuring acceleration/speed/displacement?

I suffer with "Essential Tremors", I want to instrument a glove with 2 or 3 axis MEMs accerometers to quantify how badly my fingers shake.

 

[Jaguarjoe]

It depends how one defines "best".

Lowest cost?
Highest performance?
Easiest implementation?
Simplest to design?
Smallest size?
Lower current consumption?
Other?

.

Yes! All of the above

 

[crutschow]

Yes! All of the above

Unfortunately, Mother Nature generally doesn't allow you to have all of those at the same time. You need to put them in some order of importance. That's where the engineering comes in.

 

[schmitt trigger]

And there may be other tradeoffs too, in the list.

For instance, let's assume that we decide for an all digital approach.
But then, the software engineer comes back to us and tell us that the microcontroller is already too busy......That he will require a more capable ($$$) controller.

Or that we decide for an all analog approach......but the required high stability film caps are backordered and have long lead times.


Decisions, decisions, decisions.

 

[cachehiker]

Just my 2¢.

Consider an amplified, lower noise 3-axis accelerometer. Noise was the big issue with our piezo based device. It was fortunate that we were out to measure relatively high amplitude signals.

smallest size - all digital except for the A/D and an amplifier
simplest to design - off the shelf microcontroller like an Arduino
lower current consumption - consider an TI MPS 430 with a serial LCD and no backlight

easiest implementation & lowest cost - analog and measure its output with a DMM
highest performance - probably an analog front end with digital post processing, Cypress PSoC and an LCD?
There is a Freesoc development board available with configurable Gain Controlled Amplifiers and Filter Blocks built in but as far as I'm concerned, they've proven themselves to have a substantial learning curve compared to the Arduinos. I've yet to play with the TI chips.

We used an analog front end and an Intel Dual Core with a 16-bit Data Acquisition Card. The total cost for the entire system was between $8000 and $10000 but it gets used for much, much more.