Every year, billions of crystals are manufactured and are used in nearly every electronic device. Your digital and analog subsystems in phone networks, digital watches, and even your microwave need stable oscillators for precise timing and moving data around digital circuits. In addition to XTAL oscillator circuit layout, you need to select the right crystal that will provide stable timing.
The XTALs you’ll find on the market can be manufactured for frequencies ranging from ~10 kHz to ~100 MHz. These components have varying levels of intrinsic jitter and temperature sensitivity, and choosing the right oscillator for your system is critical to ensuring accurate timing for digital and analog circuits. As frequencies increase into the GHz regime, you’ll need to alter your strategy and choose a highly stable reference oscillator for your system. Here’s what you need to know about XTAL oscillator circuit design and component selection, and when you need to start thinking about more stable high frequency oscillators.
Your goal in laying out an XTAL oscillator circuit is to ensure your clock signal is properly isolated from other components, and that there is minimal drift and jitter/phase noise on the output. Some of these points are related to the chemical composition of your oscillator (ceramic vs. quartz), and some are related to your layout (insufficient PDN decoupling can be one major source of jitter). I’ve compiled some of the basic tips for laying out any XTAL oscillator circuit in the list below.
There are some other important recommendations you should follow, which relate to the type of clocking used and EMI suppression.
If you’re using an XTAL oscillator circuit as a system clock for a series of point-to-point functional blocks, you’ll have difficulties in routing your clock lines as you need to ensure clock signals arrive at the each component at the same time as the signals. The goal is to ensure data signals latch in the receiver at the correct time. Doing this properly requires accounting for propagation delay in your components, which becomes extremely difficult as the number of components in a point-to-point chain increases.
Using a system clock in this point-to-point topology requires ultra-precise timing, which can be rather difficult with conventional EDA tools. This problem with timing arises due to the propagation delay of each component in the chain.
This difficulty is the reason many components use embedded clocking, where the clock signal is embedded in the first few bits of a data stream. The other scheme is source-synchronous clocking, where the clock signal is routed in parallel with the data stream. Computer motherboards and their peripherals will use either of these clocking schemes as it becomes more difficult to properly route a single system clock as a board size increases. Signals routed between components, e.g., in a PCIe subsystem, will need to be routed with the same length and timing matching techniques that are used with parallel data buses. Fortunately, PCB EDA programs provide design rule features that make this type of routing rather easy.
An XTAL oscillator circuit is known to be a problematic source of EMI in many systems. This can manifest itself as radiated near-field or far-field EMI during EMC testing, as noise in a downstream functional block, or as jitter. Depending on the grounding strategy with a high frequency XTAL oscillator circuit, you may unintentionally create a broadband antenna beneath your clock. Be careful to not leave large sections of grounded copper pour beneath the clock; currents will capacitively couple between this copper pour and the nearest reference plane, which can emit like a strong patch antenna at particular resonances (see the top image for a good example of proper routing).
One option for reducing EMI from clocks is to use components that run with spread spectrum clocking. This technique spreads power out over a broad frequency range, but it reduces the peak energy of high-speed signals, which then reduces the strength of induced broadband noise. Spread spectrum clocking involves applying frequency modulation to the clock output, which then triggers a downstream circuit when the frequency and phase are at specific values (see below). This often eliminates the need for EMI filters, ferrite beads, coils, and chokes.
Power spectra and EMI reduction with spread spectrum clocking.
If your board will need to operate with highly precise timing at a range of temperatures, then you’ll need consider a temperature-compensated XTAL oscillator circuit (TCXO). These components are common in RF systems and high speed digital systems. These components use a varactor diode within a temperature compensation feedback loop. As the temperature changes, so does the resonance frequency of the crystal. The varactor diode allows the resonance frequency to be compensated back to the desired value. These components are also available as voltage-controlled devices with broad output range. An example of such a circuit is shown below.
Example temperature compensation circuit for an XTAL oscillator.
Using a voltage-controlled TCXO gives you an additional lever to pull when you need to compensate for temperature changes. The output can then be fed into a VCO for multiplication up to higher frequencies. These components are available up to ~100 MHz and allow control/compensation down to ppm levels.
Once you get to the GHz range and beyond, specialized oscillator ICs will outperform an XTAL oscillator circuit. There are a number of reasons for this, which have to do with integration of components onto a semiconductor die and the reduction of parasitics in the circuit layout. The typical voltage-controlled oscillator (VCO) circuit you could easily layout on a PCB is available as a low-cost IC, providing an easy way to adjust the clock output frequency. Components with output reaching ~22 GHz are currently available, and it is a simple matter to increase the oscillator frequency to an arbitrary higher frequency with a fractional PLL. In fact, a PLL contains a VCO, which is used for frequency multiplication.
Here, the major consideration in selecting an oscillator centers around the level of integration required, the particular application, and whether an SoC/ASIC is available for your application. If there is an SoC available for your particular application, and this SoC can run with a slower system XTAL oscillator circuit, then just use the SoC as you’ll likely save on cost. If you’re designing something truly unique, such as in an industrial application or with a variety of required functions, then an SoC is likely unavailable unless you design it yourself. Otherwise, you’ll need to use multiple components to build out a larger system, but you’ll have more control through the use of discrete components and integrated circuits.
If you’re unsure whether you need an XTAL oscillator circuit or RF oscillator IC for your next product, an experienced design firm can help you weigh your options and make the best decision for your new design. The PCB design team at NWES has experience creating custom PCBs for computer peripherals, industrial applications, IoT products, and wearables. Our passion is designing modern PCBs and creating cutting-edge technology. We've also partnered with advanced PCB manufacturers, and we'll make sure your next PCB layout can be fabricated and assembled at scale. Contact NWES for a consultation.