Active Crystal Oscillator: Preparation, Placement and Process Control

An active crystal oscillator packages the quartz blank and the sustaining circuit in one component, so the designer gets a logic level output instead of a raw resonant element. That saves the load capacitors, the feedback resistor and the start up tuning that a passive crystal requires, and it makes the clock available as soon as power is applied. The convenience has a price, and it is paid in a small set of parameters that decide whether the clock survives an industrial temperature range.

This article works through those parameters using an audio board clock as the example, and it covers the layout and the reflow limits that the datasheet lists separately. The point is to select the part by the numbers that will actually matter in the field rather than by frequency and package alone.

What An Active Oscillator Brings

A passive crystal relies on the oscillator circuit inside the chip it drives, and that circuit has to be tuned with the right load capacitance and biased correctly. When it starts slowly or runs at the wrong frequency, the cause is often the external network rather than the crystal. A packaged active crystal oscillator removes that variable, because the manufacturer characterises the whole circuit and specifies the output directly.

The trade is a slightly higher cost, a larger footprint than a bare crystal, and a supply current that a passive part does not draw. Where a board has several clock domains that must be independent, the packaged part is usually the simpler choice, and where the board is extremely cost sensitive and the chip has a well characterised oscillator, a passive crystal remains adequate.

Frequency Tolerance Across Temperature

Frequency tolerance is quoted at a reference temperature, and a separate figure covers the deviation across the operating range. The two must be added, along with the ageing figure, to get the total error that the system sees over life. A part specified at a few parts per million at room temperature can drift far more at the cold end of an industrial range if the temperature stability is not stated properly.

For a serial audio interface the clock error translates directly into a sampling rate error, because the converter derives its timing from the same reference. A deviation of tens of parts per million is enough to break synchronisation between two devices that each run from their own oscillator, which is why the temperature stability figure deserves more attention than the room temperature accuracy when the product is specified over a wide range.

Supply Range And Supply Decoupling

A wide supply range lets one part number serve boards at 1.8 volts and at 3.3 volts, which simplifies inventory but does not remove the need to check the output level at each rail. An oscillator that swings to a level the receiver does not recognise is a clock source that appears to work on a bench supply and fails at the corner of tolerance. Confirm the output swing against the input threshold of the device being clocked.

Supply decoupling is not optional on these parts. A ceramic capacitor of about 0.1 microfarad in a small body should sit within a millimetre or two of the supply pin, with the ground connection made by a via directly beside the pin rather than by a trace to a distant plane. A clock device draws its current in short pulses at the output edges, and the impedance of that loop determines how much of the noise reaches the oscillator.

The Enable Pin And Output Control

Most packaged oscillators provide an enable pin that turns the output on and off. The pin usually has an internal pull up, so an unused input can be left open and the output remains active, but the datasheet should always be read because some devices invert the sense of the pin or require an external resistor. Driving the pin low places the output in a high impedance state, which is useful for multiplexing two clock sources or for power management.

Two cautions apply. First, the enable pin is not a power down pin; the oscillator may continue to run internally, so the supply current may not fall as much as expected. Second, switching the output on produces a transient at the load, and a receiver that is not expecting it may respond with a glitch. Where two sources share a net, the disable and enable sequence should be separated by a period during which neither drives.

Layout And Pad Geometry

The output trace should be short and routed over an unbroken reference plane, and it should not cross a plane split, because the return current has to follow the trace and a split forces it to detour. On a four pad package, the supply and ground are usually on adjacent pads, so the decoupling loop is naturally small if the capacitor is placed on the same side as the part.

The pad geometry should follow the recommended land pattern rather than being scaled from a similar body size, because the pad positions on these packages are defined by the internal lead frame. A footprint that is slightly too large allows the part to shift during reflow and produces a joint with an uneven fillet, and the resulting mechanical stress can crack the ceramic base of the package. Standard practice for pad design and the placement rules described for component placement apply to a part of this size.

Reflow And Handling Limits

Packaged oscillators are usually rated for a lead free reflow profile with a peak of about 260 degrees Celsius and a limited time above liquidus, and the datasheet states how many reflow cycles the part tolerates. Exceeding either the peak or the number of cycles shifts the frequency permanently, and the shift is small enough that a functional test may still pass. The oven profile should be verified on a test board rather than assumed from the paste datasheet.

Handling matters as well, since the package contains a suspended quartz blank. A part that is dropped or that is stressed by a deformed reel can shift frequency without any visible damage. Where a batch of boards shows a frequency shift that cannot be explained by the process, checking whether the components were damaged before reflow is worth the time, because the damage is not detectable by inspection after the joint is made. The soldering conditions are summarised in the comparison of leaded and lead free processes, and the defects that appear during reflow are covered under component shift during soldering.

FAQ

Is an active oscillator always more accurate than a passive crystal? No. Accuracy depends on the cut of the blank and the compensation that is designed in. A packaged part is easier to specify, but a well matched passive crystal with a good load network can be as accurate.

Should the enable pin be grounded when it is not used? Not unless the datasheet says so. Many devices have an internal pull up that keeps the output active, and grounding the pin on those parts disables the output.

Why does the frequency shift after assembly? The two common causes are thermal damage during reflow and mechanical stress applied to the package during handling. Both produce a permanent shift that a functional test may not detect.