The Predictive Maintenance Tale - piezo element voltages

Hello Jan,

Not very suprsied by the result or the badness of some of the Arduino and Pi support stuff - it's attracted a lot of very low end suppliers.

There is a very old charge amp/filter fronmt end ciruit on my website and I've adapted it for your sounder sensor.

I've guessed your source capacity at 25nF.

This circuit gives real and fast protection and band pass filtering and the ability to drive an ADC properly.

I've modelled it with dual supplies but you can use and RRIO type CMOS amplifier with a single supply to match the ADC. (with suitable biasing of course)

Putting the protection diodes on the summing node of the charge amp means that there is no voltage (almost) across them in normal operation so they don't screw up the LF response (which the zener across the sensor can.)

C1 and V1 represent the sensor

R1 sets a HF rolloff (dependent on the source capacity as well (yes the source capacity))

C2 sets the charge gain (if C1 = C2 the output is the same as the open circuit outpuy of the sensor would have been))

Because R1 is between the sensor and the protection diodes they will clamp the maximum input to the amplifer effectively.

The blue trace is the voltage on the summing node.

MK

I installed a fresh LTSpice, updated my KiCad. I'll get going now.

In the meantime, I received a second piezo element. It was on backorder since begin July. This one is 34.5 nF.

Both the original one (10 nF) and this one are measured after calibrating the LCR meter.

Simulation results. I saw on your website that you used an LT071, so I downloaded the model from TI, and drew an opamp symbol for it.
changes:

• BAV99HM are no longer advised for new designs, suppliers suggest HY variant.
• 35 nF capacity for the piezo

It has been some time since I used LTSpice, the muscle memory was gone, but I got it working:

In volts:

KiCad schematic

edit: the charge capacitor is to be adapted to the capacity of the piezo element (see Michael's post above, and his paper). My element has 35 nF capacity. When I use a 33 nF value in the simulator, I get a response that's close to 0dB. With a 22 nF, it bops up to +4 dB

33 nF (match the input element)

22 nF:

michaelkellett, is it OK to have your name on schematic and pcb?

michaelkellett, is it OK to have your name on schematic and pcb?

Absolutely fine re name.

/members/scottiebabe makes a good suggestion re. single supply op amps.

The tricky part is getting the biasing right. I'll give it some thought and post something later today if I can.

MK

Here's a way of using a single supply. You need one big decoupling cap but its very hard to see any way round that. If you give the sensor a really big thump the voltage across C3 will chang so you need to pick the time constant of C3 and R3//R4 carefully.

A better performance can be obtained with an active voltage divider but thats more parts.

A really nice op amp for this would be the TI OPA375 but they are currently hard to get.

The MCP601 suggested by /members/scottiebabe would work but isn't rail to rail input so is kind of near its limit with a 3.3V supply (OK with 5V). It is noisey but this probably won't matter too much.

The NCV20166 from Onsemi looks good on paper and Farnell have stock (3528474).

I've never used them.

MK

As MK notes in their recommendations for selecting the integration capacitor, MLCC can also exhibit the piezo-electric effect and become sensors too. So if your sensor board is really going to rattle around, it's something to consider. Although when I tried it now on my o-scope with a 10uF 16V, I didn't see much.

Just food for thought.

BOM:

I selected components that are in stock, available on cut tape.
Actual quantity ordered: 10, 50, 100, 100, 100, 5.
Minimum is 10, 5, 5, 10, 10, 5. In that case the sum is approx. 8.5€ excl. VAT and shipping.

The goodies arrived (except the decoupling caps. I had them at home).