For circuit designers, choosing a timing device is no longer a matter of “just drop in a standard crystal.” Frequency stability, phase noise, jitter, and environmental ruggedness all have to be balanced against one another — and that means deciding not only which technology to use, but which grade. This article lays out the technical fundamentals of timing devices and the structural shift happening in the market, from a designer’s point of view.
The Physics of the Quartz Crystal — Why It’s So Stable
The stability of a quartz crystal comes from its extremely high Q factor (quality factor). An AT-cut crystal reaches a Q of 100,000 to several million — a region that LC oscillators and silicon oscillators cannot touch no matter how hard they try. The higher the Q, the sharper the resonance, and the less the oscillation frequency is pulled around by external disturbances. This is the source of low phase noise and low jitter.
The cut angle is another key parameter that determines the frequency-temperature characteristic. The widely used AT-cut has a third-order temperature characteristic, tracing a cubic curve whose frequency-temperature coefficient is nearly zero around room temperature. The “±10 ppm” and “±0.5 ppm” figures you read on a datasheet are the combined result of this crystal physics and the compensation circuitry around it.
The SC-Cut — The Trump Card for High-Stability Applications
If the AT-cut is the workhorse of the general-purpose range, the SC-cut (Stress Compensated cut) is what underpins the world of ultimate stability. If you work with OCXOs or high-precision references, understanding its characteristics is well worth your time.
Whereas the AT-cut is a single-rotation cut, the SC-cut is a doubly rotated cut, rotated about two axes. The choice of those two angles is designed to cancel the mechanical stress that develops inside the quartz blank as temperature changes. That is the origin of the name “stress compensated,” and the practical benefits it brings are substantial.
The figure above schematically compares the frequency-temperature characteristics of the AT-cut and the SC-cut (C-mode). The AT-cut traces a cubic (S-shaped) curve with an inflection point near 25 °C, flattening out around room temperature. The SC-cut, by contrast, has its inflection (turnover) point up at roughly 90 °C. An OCXO sets its oven temperature to this upper turnover point. At the turnover point the frequency-temperature coefficient (the slope of the curve) is zero, so even if the oven temperature wobbles slightly the frequency barely moves. This physics is exactly why the SC-cut is meant to be operated hot.
First, it is robust against dynamic thermal response. When the temperature of an AT-cut changes abruptly, the thermal gradient across the blank generates stress, causing a large transient frequency excursion (thermal transient). The SC-cut’s thermal-transient sensitivity is more than an order of magnitude smaller, so the frequency settles quickly even right after power-on or during a rapid ambient change. In oven-stabilized applications like an OCXO, this benefits both the warm-up time and the post-settling stability.
Second, it has low drive-level dependence (DLD) and small aging. Because stress-induced frequency variation is suppressed, it tends to be favorable for long-term stability (aging).
Third, making use of the C-mode. The SC-cut supports multiple vibration modes (B-mode and C-mode). The lower-frequency C-mode you primarily use benefits from stress compensation, but the more temperature-sensitive B-mode coexists with it, so the oscillator circuit needs a mode-selection design that reliably picks the C-mode. This is a design consideration the AT-cut does not have.
The trade-off is that the SC-cut demands tighter machining-angle precision and loses out to the AT-cut on yield and cost. That is precisely why it is reserved for applications that need ppb-class (parts-per-billion) stability, or where severe thermal transients are a problem — base-station reference clocks, instrumentation, satellites. There is basically no reason to choose an SC-cut for general-purpose equipment. Just remember: “its value shows in the high-stability range where transient response and warm-up matter.”
Choosing the Grade — SPXO / TCXO / OCXO
Oscillators split clearly by stability grade, and the grade is tied to which cut is used.
- SPXO (standard): No compensation. Around ±10–100 ppm. For cost-first consumer devices. AT-cut.
- TCXO (temperature compensated): Temperature sensed by a thermistor, etc., and compensated. ±0.5–2 ppm. The mainstay for smartphones, GNSS, and automotive comms. AT-cut.
- OCXO (oven controlled): The crystal is held at a constant temperature by a heater. ±1–100 ppb. Larger in power and size, but indispensable for reference applications that demand high stability. This is where the SC-cut shines.
- VCXO / VCTCXO: Frequency is voltage-tunable. Used to lock the reference inside a PLL loop.
In RF systems with demanding phase-noise requirements, you have to budget not only for the oscillator’s own noise floor but also for the degradation after multiplication (20 × log N). Whether to base the system on a high-frequency SC-cut OCXO or to multiply up a lower reference with a PLL comes down to whether you prioritize close-in phase noise or the far-out noise floor.
Phase-Noise Design — Pinning It Down with Worked Examples
Phase noise tends to stay abstract, but plugging in numbers makes the decisions concrete.
Example 1: Degradation due to multiplication
Suppose you generate a 10 GHz carrier by PLL-multiplying a 10 MHz SC-cut OCXO reference. The multiplication ratio is N = 10 GHz / 10 MHz = 1000. In ideal multiplication, phase noise degrades by:
Degradation = 20 × log10(N) = 20 × log10(1000) = 60 dBIf the reference phase noise is −150 dBc/Hz at a 1 kHz offset, the theoretical floor at the 10 GHz carrier is −150 + 60 = −90 dBc/Hz. The added noise of the multiplier and PLL stacks on top of that, so in practice it is a few dB worse. “Even with a great reference, you always degrade by the square of the multiplication ratio (N², i.e., 20 log N in power terms)” — miss this one point and the RF-stage phase-noise budget falls apart.
Example 2: Converting phase noise to RMS jitter
Integrating the phase-noise spectrum L(f) gives the integrated (RMS) jitter. Over an integration range f1 to f2:
Phase jitter (rad, RMS) = √( 2 × ∫ L(f) df )
Time jitter (s, RMS) = phase jitter / (2π × f0)Suppose that for a 156.25 MHz clock, the integrated phase noise over the 12 kHz–20 MHz band is −100 dBc (i.e., an integrated value of 10^(−100/10) = 1 × 10⁻¹⁰).
Phase jitter = √(2 × 1e-10) ≈ 1.414e-5 rad
Time jitter = 1.414e-5 / (2π × 156.25e6) ≈ 1.44e-14 s ≈ 14.4 fs (RMS)In the world of 10 GbE and high-speed SerDes, where “a few hundred fs or less” is the requirement, this reference quality on the order of ten-odd femtoseconds is what counts. Put another way, most of the jitter budget is decided the moment you select the reference source.
Another Option — MEMS
For reference, silicon MEMS oscillators have recently entered the picture as an option. The approach fabricates a resonator with a semiconductor process, converts its output to an arbitrary frequency with a PLL, and builds up stability with digital temperature compensation. Its strengths are shock and vibration resistance — thanks to the lightweight resonator — and the flexibility of an arbitrary frequency. However, because it goes through a PLL, its close-in phase noise can still be at a disadvantage versus a same-price-class crystal, and the crystal retains the edge where low close-in phase noise matters in an RF reference. Harsh vibration environment, or low phase noise — the two are increasingly splitting by application.
What the Market Is Telling Us
The market numbers also back up the importance of the high-stability, high-reliability segment. The crystal oscillator market was about US$3.1 billion in 2025 and continues to grow at a CAGR of roughly 4% toward 2030 — growth that persists despite being a mature market. The driver is automotive, showing the fastest growth on the back of rising ECU counts from EVs, ADAS, V2X, and LiDAR. The OCXO sub-segment is also expanding, supported by satellite and communications-edge demand for sub-ppm performance. Surface-mount packages already account for roughly 70% of the market, and the pressure for miniaturization is a constant.
Takeaways for Designers
Clock-source selection has entered a phase where you break the requirements down along the axes of phase noise, jitter, stability, ruggedness, size, and cost, and evaluate the optimum point case by case. In the high-stability range especially, the thermal-transient immunity and warm-up behavior that the SC-cut delivers directly govern the reference quality of the system. Whether you can read the “fine print of the datasheet” — degradation after multiplication, warm-up, holdover — determines the quality of your reference design going forward. It may be a part just a few millimeters on a side, but it remains the linchpin that holds the entire system’s timing budget.