Cheap, innocent looking piezo elements do okay as greetings cards sounders. They seem to cause grief when used as contact mics. They seem to promise a lot but easily sound rough as commonly applied. Two solutions are proposed - the cheap and cheerful FET buffer which probably meets all the requirements of 80% of users who are using the devices on outside structures, and the high performance version for those who want higher performance and lower noise. First of all though, why do these things sound rotten the way they are most commonly used.
The sound to voltage conversion isn't noted for its high quality - most piezo contact mics are tuned speaker elements used in reverse. The brass disc on which the element is glued is designed to resonate at the design frequency of 2-4kHz so that a large audio ouptut is achieved with a small power input. This will tend to lead to a peakiness at mid-frequencies, which is why I glue these to a magnet to try and spread this peak and make the coupling to the source better.
However, the main reason these have gotten a bad rap is that many people couple them into a standard audio load, which loses low frequencies.
A piezo element can be thought of as a sound-dependent voltage source in series with a large capacitance of about 15nF. Here the trouble starts. You need to put this into a load which is higher than the impedance of the series capacitor at the lowest frequency of interest. If that is 20Hz, since the impedance of the capacitor is 1/2*pi*freq*C then you want that to be > 320k. I've terminated mine in 3.9M because that's what I had to hand.
So what to people do? They go and stick this into the line input of their recorder, typical impedance 50k, or the mic input of their minidisc, typical impedance about 7k, and they start to grouch that this damn thing sounds tinny. Which is does - from the frequency response you can see that the mic input starts losing low frequencies below 1kHz (red line) and the line starts to go below 200Hz (blue line).
We don't really like that very much. This tends to be compounded with the problem that these things aren't really designed as mics and resonate at 2-4kHz anyway unless you stick them down so you get a harsh ugly result. What you need is a high impedance input, which gets you the blue line in the plot above
You could try and equalise in post production. The trouble with that approach is you need the specs of the mic and amplifier input to know where to start your 6dB/octave LF boost, so you will probably have to tune that by ear if you don't know these. Plus you will boost all the noise and hum at low frequencies. With the mic input you are looking at puuting in a 25dB boost by the time you get to 50Hz which is just asking for hum problems, and even the 8dB boost on the line input version is more than doubling any hum you may have.
These things are very much like the crystal mics people used to use with valve open reel recorders. Valves do high impedance really well, and inputs were often in the area of 1 megohm input impedance. The move to transistor circuitry meant consumer audio inputs tended to end up around 50k which is reasonably easy to do. Crystal mics also used the piezoelectric principle, and were valued for their low cost and high output voltage which overcame the preamplifier noise better. Transistorising audio gear killed off the crystal mic as they sounded tinny and harsh with the lower input impedances for the same reason.