Piezo amplifiers

geo_leeman

Hey all,

I'm working up a sensor that measures the resonant frequency of a bit of wire using a piezo element (see attached photos). The wire is "plucked" by the electromagnet below it and freely resonates. The idea is that as mass accumulates on the wire (freezing water) the resonant frequency will be reduced. This type of sensing methodology has been used in the past, but I'm afraid I'm overcomplicating the signal conditioning here. All I'm interested in is the frequency response to the plucking, so greatly amplifying the signal to the point of a square wave, then counting crossings seems more sensible that full FFT, etc.

That being said, I've considered and played with instrumentation amps, JFETs, and JFET based op-amps like the TL071 and a circuit like this. Any thoughts on the best/cheapest approach here (disposable sensors)? I know I'm overthinking/complicating the situation as it's not really precision instrumentation and I'm fighting the urge to properly DC offset, etc. For what it's worth - I hooked it up to an INA128 setup on a dual supply and got nice results with gains of 1-5k. This circuit will be signal supply in it's final state though.

Any thoughts appreciated!

Phil Pilgrim (PhiPi)

'Sounds like a fun project! Which micro are you using to count the vibrations?

-Phil

geo_leeman

Hi Phil - I'm not sure yet. Since they are disposable maybe an M0 in the end. During development I'll likely use my prop dev board since it's easy to get setup and quicker to program that trudging through Atmel Studio for SAM21D or similar.

geo_leeman

Still haven't got the signal conditioning nailed down, but someone over at the EEVBlog had the interesting idea of using a least-squares fit for the frequency determination vs. FFT or zero-crossings. That could be a very interesting idea, anyone played with it before?

Bean

If the wire was magnetic (or a small magnet attached to the wire above the electromagnet) then you could use the electromagnet to both "pluck" and sense.

Bean

erco

Bean wrote: »

If the wire was magnetic (or a small magnet attached to the wire above the electromagnet) then you could use the electromagnet to both "pluck" and sense.

Bean

Ah, the genius of Bean-ius! I like that!

geo_leeman

It is a steel piece of "music" wire - that is a good idea! The piezo gets a tiny signal since the wire is clamped to it, so simplifying it even more would be great. I've seen some people continuously drive the with with a PLL, but I need to stay as low power as possible, so the pluck and sense method fits better. I'll give it a shot and see what the pickup looks like.

erco

@Geo: Is this for some high-flying apparatus?

"Pluck and sense" reminds me of the use-every-letter-typing-practice-sentence:

The job requires extra pluck and zeal from every young wage earner.

geo_leeman

@erco - yep! Planning on adding it to a weather balloon.

Phil Pilgrim (PhiPi)

If the wire ices up, is there any chance that the solenoid might ice up, too, causing the plunger to get stuck? I suppose you could leave it "on" long enough to heat up, though.

-Phil

Tracy Allen

In this case it seems like a simple AC amplifier would suffice. No need for DC accuracy (Fight that urge!). How about a super-simple expedient, CMOS inverters connected in linear mode. AC coupled with a feedback resistor of say 20MΩ. A Hex inverter like the CD4069 or the 74C04 give you up to 6 stages at potentially +40dB per stage with a 5V supply.

What is the frequency range observed for the ice sensor?

Vibrating wire instrumentation is commonly used in high accuracy geotechnical instrumentation. GEOKON is one major supplier. It does use the electromagnet both to pluck the wire and to return the signal, and the electronics can be sophisticated.

Phil Pilgrim (PhiPi)

Tracy Allen wrote:

In this case it seems like a simple AC amplifier would suffice. ...

Yup. That's the reason I asked what micro he was using. If it were a Propeller, he could do that with a couple resistors and a cap -- no external amplification needed.

-Phil

geo_leeman

That is a nice solution! I'm expecting the wire to resonate between 40-60 Hz depending on exactly where I end up cutting it length wise. Here's a screenshot of some results from a similar sensor in the literature. They used a servo with a magnet on the end of the arm to pluck the wire.... which seems awful.

Phil Pilgrim (PhiPi) wrote: »

If the wire ices up, is there any chance that the solenoid might ice up, too, causing the plunger to get stuck? I suppose you could leave it "on" long enough to heat up, though.

-Phil

I don't think it'll be a problem as long as I leave an ~1/8" gap. We're not talking about massive accumulation (we think).

Tracy Allen wrote: »

In this case it seems like a simple AC amplifier would suffice. No need for DC accuracy (Fight that urge!). How about a super-simple expedient, CMOS inverters connected in linear mode. AC coupled with a feedback resistor of say 20MΩ. A Hex inverter like the CD4069 or the 74C04 give you up to 6 stages at potentially +40dB per stage with a 5V supply.

What is the frequency range observed for the ice sensor?

Vibrating wire instrumentation is commonly used in high accuracy geotechnical instrumentation. GEOKON is one major supplier. It does use the electromagnet both to pluck the wire and to return the signal, and the electronics can be sophisticated.

I'm liking the idea of a single pickup/excitation device. I dug through the bin tonight and all I've got is an M74HC4049 which doesn't look like what I want. I guess its time for a digikey order. I like that sort of solution though as I'm not so dependent on lots of resistor/cap values on the BOM and not dealing with temp coeffs of all that.

To be clear, your thinking a cap and resistor in series to the input of the first inverter, chain a few inverters, then have 20M from the output of the last inverter in the chain to the input of the first? Like figure 6: https://www.fairchildsemi.com/application-notes/AN/AN-88.pdf

geo_leeman

Doh! - Here's the results of the other paper.

Tracy Allen

The goal in figure 6 of AN-88 is linear AC gain of exactly x10, which requires that the open loop gain be so very high that it drops out of the equations, like an op-amp. But your goal is is simply high gain, squaring the signal out to the rails.

You could try that circuit, but instead of 1M at the input, just make it zero ohms--Feed the signal directly to the CMOS input though the capacitor. At 5V Vdd, the overall open loop gain could be over 100dB. A hitch is that the input impedance is reduced by the open loop gain, so the performance with a high impedance piezo pickup might be disappointing.

Better to make the gain in stages, with each stage built around a single inverter, and the stages capacitor coupled. The first stage should be lower gain so as to achieve higher input impedance. Subsequent stage or stages could drive toward the rails. Say x10 in the first stage, then full 40dB, x100 in the second stage, that gets you to the x1000 factor that you mentioned in the first post.

You have a 74HC4049. That should work. My first choice would probably be an CD4007, which is a kit of complementary mos transistors that you can hook up as inverters or in other ways that are good for special linear circuits. There is a nice writeup in this Nuts and Volts article,
fet_principles_and_circuits.
There is even a nice front end stage there, a high impedance source follower.

Of course high gain circuits can oscillate :~( so layout is important. Or close the loop, the wire as the resonant element in an oscillator.

geo_leeman

Thanks Tracy - that's a useful article! I'll play with this a bit and keep folks posted on how it goes.

Tracy Allen

Thinking about it more, the circuit model of a piezo sensor is a tiny current source in parallel with a relatively large capacitor. People often make a high impedance amplifier to detect the small voltage across the capacitor. However, it may be better to think of the Norton equivalent, go directly for the dQ/dt by driving the current into a virtual short circuit, so voltage across the capacitor does not have to change. The piezo might be connected directly to the input of the single or triple inverter circuit as in the attached diagram. No input resistor.

I know that this approach works with pvdf piezo sensors into the Propeller sigma-delta inverter; piezo directly to the summing junction with feedback thru a high-ohm resistor.

geo_leeman

I'll give that a shot. I was playing with this on the breadboard last night and getting some screwy results... looks like a worn out set of terminals as there was some mechanical effects of wiggling components. Whoops! Stay tuned.

geo_leeman

Hey Tracy et al. That circuit does the trick of squaring things up nicely! It's not exactly stable on a breadboard with flying scope ground lead, etc, but that's expected. I'll have it on a PCB next. While I like the idea of using the coil as a pickup as well, this is likely to work better as I can leave the coil further from the wire as that signal is really tiny as well. I suppose this circuit would be suitable for that as well though. Planning on laying it out so that 1-all stages can be used depending on which parts are populated.