Crystal Oscillator Types: SPXO, TCXO, OCXO and VCXO Compared

A bill of materials lists a frequency, and a substitute part with the same frequency is ordered. Sometimes the board boots, sometimes it starts intermittently, sometimes the frequency error grows with temperature, and sometimes the current consumption and warm-up time are nothing like what the system was designed for. The cause is that the word crystal covers several different components with different interfaces and different electrical models.

A passive crystal needs an oscillator circuit inside the chip to make it work. A crystal oscillator already contains that circuit and simply presents a clock output. Even among oscillators, the way stability is achieved and the way the output can be controlled differ, and those differences decide which device the design actually needs.

Passive Crystal and Crystal Oscillator Are Not Interchangeable

A passive crystal is connected to the oscillator pins of a microcontroller or a system on chip and works with the amplifier inside the device, the external load capacitors and the crystal itself to form the resonant circuit. It has no output pin of its own, and its behaviour depends on the chip it is paired with.

An active oscillator contains the loop internally and presents supply, ground and clock output pins, and often an enable or control pin. The packages can be similar in size, but the interface, the electrical model and the layout requirements are not. Substituting one for the other is a redesign of the clock interface, not a component swap.

SPXO: A Fixed Frequency Output

The simplest active device is the standard oscillator, which provides a fixed frequency without temperature control or external tuning. It suits systems that need an independent clock source in a benign environment.

Selection still requires more than a frequency: frequency stability, supply voltage, output level, load capability, enable logic, start-up time and jitter or phase noise all have to match. Two devices with the same nominal frequency and a similar package can have entirely different output stages, and the interface compatibility is decided by those, not by the number on the lid.

TCXO and OCXO: Two Ways to Fight Temperature

A temperature compensated oscillator corrects the frequency variation of the crystal with a compensation network, balancing power, size and stability. It is the usual choice where the environment varies but the budget for power and volume is finite.

An oven controlled oscillator holds the crystal at a controlled temperature, which removes much of the external influence. The gain in stability comes with higher power consumption, a larger package and a warm-up period during which the output is not yet within specification. The actual warm-up time and stability figures have to come from the datasheet of the specific device, because the range between parts in this class is wide.

VCXO: Control, Not Accuracy

A voltage controlled oscillator shifts its output frequency in response to a control voltage. It is used for phase locking, synchronisation and systems that need to trim the clock, and its purpose is controllability rather than precision.

Design review for this device includes the control voltage range, the pull range, the tuning linearity and the noise present on the control input, because that node becomes a path into the clock. A control line that picks up digital noise will modulate the output, and the resulting jitter appears in the system even though the oscillator itself is behaving as specified.

Comparing the Four

It is a mistake to read the list as a hierarchy. Each type addresses a different system objective: a fixed reference for a processor, a stable reference for communications, a precision reference for timekeeping, and a tunable reference for locking. The priorities for stability, power, warm-up and controllability differ, and so does the cost of getting the choice wrong.

The comparison should be made against requirements, not against categories. Where the system needs a clock that starts in microseconds, an oven controlled device is wrong regardless of its stability. Where the system must hold a frequency over a wide temperature range, the uncompensated part is wrong regardless of its cost.

Seven Checks When Selecting a Clock Source

Confirm first whether the design needs a passive crystal or an active clock input, and check the clock pin mode of the device. State the nominal frequency and the allowed deviation, and include temperature, ageing and load variation in the budget rather than treating them as separate concerns. Then verify the supply voltage, output level, load capability and enable logic, and confirm that the start-up time or warm-up behaviour fits the system power-up sequence.

For interfaces and radio functions, check jitter, phase noise or short term stability against what the link requires, and review power consumption, package, mechanical environment and temperature range together. If a voltage controlled device is used, review the control input noise, the pull range and the loop strategy as a separate exercise with its own criteria.

Finally, verify the assumptions on the board. The load capacitors of a passive crystal must match the specification of the crystal and the parasitics of the layout, and the placement and grounding of the components determine whether the measured frequency matches the calculation. An oscillator that meets its datasheet specification can still produce a poor system clock if its supply is noisy or if the output trace shares a return path with a switching converter.

Where the Clock Meets the Layout

Whatever type is chosen, the clock is one of the more sensitive nets on the board and should be treated as such during the layout review. Keep the loop between the crystal, the load capacitors and the device pins as short as the placement allows, keep the components on the same side of the board, and keep a solid reference plane beneath the whole structure. A clock trace that shares its return path with a switching converter will carry the converter noise into every device that uses the clock.

The supply side matters just as much. Oscillators translate supply noise into output jitter, so the decoupling and the supply filtering belong to the clock design rather than to the general power budget, and a ferrite or an RC filter in front of a sensitive oscillator is a deliberate choice with a stated cutoff. Where several devices share one clock, the fan-out and the trace topology decide whether the edges arrive together, which is the same problem the address and data buses have.

Component sourcing is part of the same conversation, because the exact part number determines the interface. A substitute with a compatible frequency but a different output level, enable polarity or start-up behaviour is not a substitute at all, and the procurement process needs the tolerance and the interface constraints recorded alongside the frequency, plus a verification step on the first assembled board.

FAQ

Can any crystal be used with any microcontroller? No. The oscillator circuit inside the device has a drive level and a load capacitance requirement, and the crystal has to be chosen for it.

Is a TCXO always better than an SPXO? It is more stable and costs more, uses more power and usually has a longer lead time. Better depends on the requirement.

Does a VCXO improve accuracy? No. It makes the frequency adjustable. The accuracy still comes from the reference and the loop.

What should be checked first on a new clock design? The interface type and the start-up behaviour, because both are decided before the stability figures matter.

Summary

The right question is not how many megahertz are required, but what stability, start-up behaviour, power budget and control capability the system needs. Answer that first, and the four common oscillator options stop being unfamiliar abbreviations and become a comparison of engineering trade-offs, each with its own check list and its own layout requirements.