HOW-TO GUIDE
How to Use an Oscilloscope to Check a Crystal Oscillator
By the OhmSavvy Research Team ·
Before you probe
- Crystal pins are close together and close to other pins. Probe with the board unpowered first to position the tip, and use a spring ground so the clip cannot short anything.
How a microcontroller crystal oscillator works
Most microcontroller and clock circuits use the Pierce arrangement. Inside the chip, an inverting amplifier sits between two pins, XTAL_IN and XTAL_OUT, with a high-value resistor across it that biases it into its linear region. Outside, the crystal connects across the two pins and a load capacitor runs from each pin to ground. The crystal and capacitors form a filter that only passes the crystal's resonant frequency back to the amplifier's input, so it oscillates there.
Load capacitance
Why the probe can stop or pull the crystal
A probe is not invisible. Its tip capacitance appears in parallel with the load capacitor on whichever pin you touch, and on a crystal circuit that is a large change. Extra load shifts the frequency slightly, usually downwards, and reduces the loop gain. If the oscillator had little margin, it simply stops, and the scope shows a flat line that was caused by the measurement.
1Set up a compensated 10x probe
Switch the probe to 10x and set the channel to match. At 10x the probe presents around 10 MΩ and the lower capacitance shown above; at 1x it loads the circuit far more heavily and usually has too little bandwidth for a megahertz clock anyway. Compensate the probe on the scope's calibration output first, as described in how to compensate an oscilloscope probe, and swap the long ground clip for a spring ground tip.
2Probe the output pin, or a buffered clock
The best place to look is a buffered clock output, if one exists. Many microcontrollers can route the system or oscillator clock to a general-purpose pin through a firmware setting, and some clock chips have a dedicated output. A buffered output is isolated from the crystal, so the probe cannot disturb it.
Without one, probe XTAL_OUT. Avoid XTAL_IN: it is the amplifier's high-impedance input and the most sensitive node in the circuit. Ground the spring tip to the nearest ground pad, ideally the ground side of one of the load capacitors.
3Check that the waveform looks right
Set DC coupling, a timebase that shows a few cycles (about 50 ns/div for 8 MHz), and trigger on the rising edge at the middle of the swing. On XTAL_OUT a healthy oscillator usually shows a roughly sinusoidal waveform centred near the amplifier's bias point, often around half the supply. Amplitude depends on the chip and its drive setting, from a fraction of a volt to nearly rail to rail, so compare with a known-good board if you can. A buffered output looks like a square wave.
4Measure the frequency with the scope's counter
Add a frequency measurement, or use the hardware frequency counter if your scope has one, which usually gives more digits. The reading should match the crystal marking closely, for example 8.000 MHz. This confirms the right crystal is fitted and running at its fundamental rather than at a wrong frequency. It does not verify the crystal's ppm tolerance, because a hobby scope's timebase is not that accurate and the probe pulls the frequency slightly. For the measurement method itself, see how to measure frequency with an oscilloscope.
Signs of no oscillation and what causes them
A dead oscillator shows a steady DC level on both pins and no swing, often sitting at the bias point or pulled to a rail.
| Check | Why it matters |
|---|---|
| Has firmware enabled the external oscillator? | Many microcontrollers boot on an internal RC clock and only start the crystal when configured to |
| Does it start with the probe removed? | Test via a clock output or the chip's behavior; the probe may be what stops it |
| Load capacitors fitted and the right value? | Missing or wrong capacitors change the load and can stop oscillation |
| Crystal correct for the circuit? | Wrong frequency, wrong load capacitance or an overtone part can fail to start |
| Solder joints and board clean? | Cracked joints, a damaged crystal or flux residue leaking between pins all stop it |
Start-up problems
A crystal oscillator does not start instantly: the amplitude grows from noise over many cycles. Megahertz crystals typically start within milliseconds, while 32.768 kHz watch crystals can take hundreds of milliseconds or more. To watch it, put the scope in Single mode triggered on the supply rising or the reset line releasing, using a second channel, and set a timebase that covers the expected start-up. If the amplitude builds slowly, or the oscillator only starts when warm or with a finger near it, the circuit has marginal gain. Our triggering guide explains Single mode.
Oscilloscope bandwidth needed for a crystal oscillator
A rule of thumb is at least three times the frequency to read a sine's amplitude within a few percent, and about five times for a square clock output so its edges look square. Below the frequency itself the scope barely shows the signal. Our bandwidth explainer covers the roll-off.
| Crystal | Sine at XTAL_OUT (≈ 3×) | Square clock out (≈ 5×) |
|---|---|---|
| 32.768 kHz | ≈ 100 kHz | ≈ 165 kHz |
| 8 MHz | ≈ 24 MHz | ≈ 40 MHz |
| 16 MHz | ≈ 48 MHz | ≈ 80 MHz |
| 25 MHz | ≈ 75 MHz | ≈ 125 MHz |
Which oscilloscope to use
By that rule, a 200kHz pocket scope only suits watch crystals, a 50MHz scope covers common microcontroller crystals up to about 16 MHz, and 100MHz leaves headroom for faster clocks. Compare all the options in our best beginner oscilloscopes roundup.
Scopes in our reviews
50MHz, 4 channels, 1 GSa/s. Our best overall pick for general electronics.
100MHz, 2 channels, 1 GSa/s, with a built-in waveform generator.
200kHz, 1 channel, 2.5 MSa/s. Pocket-sized; suits only slow signals such as watch crystals.
As an Amazon Associate I earn from qualifying purchases. Prices and availability are shown on Amazon.
Frequently asked questions
Should I probe XTAL_IN or XTAL_OUT?
Probe XTAL_OUT, or better still a buffered clock output if the chip has one. The input pin is a high-impedance node at the start of the amplifier, so the probe's capacitance and resistance disturb it most and are more likely to stop the oscillation. The output is driven by the amplifier and tolerates the extra load better, though it is still affected.
Why does my crystal stop when I touch it with the probe?
The probe adds capacitance, typically around 10 to 15 pF for a 10x probe and much more for 1x, in parallel with a load capacitor that may itself be only 10 to 20 pF. That can push the circuit past the point where the amplifier has enough gain to sustain oscillation. If it stops, the oscillator may have little margin even without the probe, which is worth investigating.
Can a multimeter test a crystal oscillator?
Only crudely. A resistance check can find a crystal shorted by damage or contamination, since a good one reads open, and a DC voltage reading on the pins shows whether the chip is biasing them. A multimeter cannot show whether the crystal is actually oscillating or at what frequency with any confidence, which is the job of an oscilloscope or a frequency counter.
Why does the scope show a slightly different frequency from the crystal marking?
Two reasons. The probe's capacitance pulls the crystal slightly, usually lower. And the scope's own timebase has a tolerance, often tens of parts per million on hobby scopes, which is similar to or larger than the crystal's own tolerance. A reading within a small fraction of a percent confirms the right crystal is running, not that it is within its ppm specification.
How do I check a 32.768 kHz watch crystal?
Carefully, because these low-power oscillators run on very little drive and are easily stopped by probe loading. Use a 10x probe on the output pin, a slow timebase and a sensitive V/div. If the chip can route the real-time clock to an output pin, measure there instead. These crystals also start slowly, often taking hundreds of milliseconds or longer.
