MORE ON QUARTZ
Last time we were
discussing quartz crystals, Nazis and the jungles of
If you've ever ordered a crystal from a manufacturer, you've discovered that there's a world of difference between the free and easy theory of what crystals do--plunk one into a circuit and it controls the frequency--and actual practice, which revolves around series or parallel models, different cuts, drive level, load capacitance, tolerance and operating temperature. For a two-terminal device, the quartz crystal sure can get complicated in a hurry.
A crystal's oscillating frequency drifts ever so slightly with temperature, so in order to tighten tolerances, they're often made and calibrated at an elevated temperature. That way, they can be operated in a temperature-controlled oven, and the variance due to changes in ambient temperature is removed. Tolerance is just a quality-control or calibration issue… obviously, the more accurate you want your crystal to be, the more you're going to have to pay. Different cuts have different characteristics, but 99% of the crystals we see in communications are AT-cut, so at least that's one specification that's easy to deal with. Drive level doesn't matter very much, but if you overdrive your crystal, you may damage it.
Series resonant crystal oscillators are simpler than their parallel brothers, but there's a price for simplicity--you can't trim the frequency to get exactly what you want. And most series resonant oscillator circuits will "take off" and still oscillate without the crystal… at a frequency more or less of their own choosing. The parallel resonant circuit is a bit more complicated, but it is better behaved. It can be made to shut down if the crystal isn't plugged in. And it can have a small reactance added to "pad" the frequency up and down a bit: maybe 100 Hz per MHz of oscillating frequency. In either case, the crystal is much the same, but it is specified differently. Remember from last month that a crystal is much like a series R-L-C circuit, where L and C are motional reactances, and they resonate at a frequency. This frequency is the series resonant crystal frequency. A parallel resonant crystal will be cut to a slightly lower resonant frequency (offset below the desired frequency a bit), but will be specified while operating into a particular capacitance, which will always act to increase the frequency. This load capacitance is the effective extra capacity that the crystal sees externally between its two terminals. Sometimes you have to calculate a bit, using the formula for series capacitors, to figure out this value. But it's essential if you're trying to order a parallel resonant crystal. If you try and use a series resonant crystal in a parallel circuit, it will always be too high in frequency. Padding just speeds the oscillator up more.
Radio amateurs figured out a long time ago that quartz crystals of similar frequencies can be hooked up in networks that provide very narrow bandwidth. The crystal filter was born. In recent years, ceramic filters, made by a process similar to ceramic capacitors, have made IF filters for AM and FM radios very inexpensive and compact. The performance doesn't match the crystal filter, but neither does the price, either…
One high-priced high-performance filter that has come along is the SAW, or surface acoustic wave filter. A piezoelectric transducer excites the surface of a plate of glass that has been etched with aluminum traces in a such a way that some frequencies are reinforced, others are cancelled out. Since acoustic waves travel much more slowly than electromagnetic waves, a small device can be many acoustic wavelengths long. At the other end of the plate, another transducer picks up what's left of the wave and converts it back into electricity. This is followed by a big preamplifier, because the transducer losses will probably be more than 50 dB. Because thermal expansion would cause the filter to drift, an oven is likely to be used. Altogether, a very elegant, very smooth, high performance filter is possible, with a price to match!