Crystal Oscillator Supplier Guide: Specs, Lead Times, Pitfalls

Timing components look simple on a BOM. One line item, a few cents to a few dollars, done. Then production stops because a 25 MHz oscillator in a 3225 package went to 16-week lead time, and the broker stock you found in a panic turns out to be remarked parts from an unknown reel. Any buyer who has been through that once starts treating their crystal oscillator supplier as seriously as they treat their MCU source. This guide is about what to actually check before you commit, with the mistakes we see most often in the field.

Why Timing Parts Deserve More Attention Than They Get

Oscillators and crystals are the classic "cheap part that stops the line." A $0.40 crystal failure takes down a $40 processor just as dead as a power IC failure does. And unlike a lot of passives, timing components have real parametric spread: frequency tolerance, aging, load capacitance matching, drive level. Get the match wrong and you get intermittent startup failures that only show up at temperature extremes, usually after you have shipped.

There is also a quiet shift happening on the silicon side. A recent Semiconductor Engineering piece on purpose-built ring oscillators describes how chipmakers now embed oscillator structures on-die to monitor process behavior. That is a design-tool story, but the sourcing angle is this: on-chip oscillators are getting better, and some engineers assume they can drop the external crystal entirely. For anything involving USB, Ethernet, CAN, or wireless, you still can't. Internal RC oscillators drift 1% or more over temperature. A UART link starts getting bit errors somewhere around 2% total error budget split between both ends. The discrete timing part isn't going away, which means neither is the sourcing problem.

Match the Part Type to the Application Before You Call a Supplier

Half the bad purchases we see start with an under-specified RFQ that just says "25 MHz oscillator, 3.3V." That describes thousands of parts with wildly different behavior. Here is the practical breakdown:

Two traps in this table deserve comment. First, a passive crystal is not a drop-in for an oscillator even at the same frequency and footprint class. The crystal needs the MCU's internal Pierce oscillator and correct load capacitors; the XO is a powered device that outputs a clock. Buyers mix these up on shortage substitutions more than you'd think. Second, MEMS parts are legitimately good now, especially for shock and vibration, but the phase noise and jitter numbers differ from quartz at the same price point. Check your PHY's jitter requirement before swapping.

The 32.768 kHz Problem Nobody Plans For

While everyone worries about the main clock, the humble 32.768 kHz watch crystal causes its own share of pain. These parts keep RTCs and sleep-mode timers alive, and they matter more every year as designs go battery-powered. EE Times recently covered an LED-based wireless power approach for indoor IoT devices, the kind of perpetually-powered sensor node that runs for years on a coin cell or harvested energy. Every one of those nodes needs a low-ESR 32.768 kHz crystal that will actually start with the tiny drive current a power-sipping RTC provides.

The pitfall: maximum ESR on cheap 32.768 kHz crystals can run 70 to 90 kΩ, while some RTC oscillator circuits only guarantee startup at 50 kΩ or below. The board works fine on the bench at 25°C and fails to keep time in the field at -20°C. When you brief your crystal oscillator supplier on these parts, state the ESR limit and the drive level explicitly, not just the frequency and footprint.

What to Ask a Crystal Oscillator Supplier Before the First PO

A real qualification conversation goes past price and stock count. Work through this list:

• Is the stock factory-sealed with date codes visible? Ask for photos of the actual reels. A supplier who hesitates on this is telling you something.
• Which franchises or authorized lines do they hold? For timing, that means lines like Epson, NDK, TXC, Abracon, SiTime, Rakon. Open-market sourcing is fine for shortage buys, but the supplier should say plainly which stock is franchised and which is brokered.
• What are real lead times right now, by package? Standard MHz oscillators in 3225 have run anywhere from 6 to 20 weeks in the past few years depending on frequency. Common frequencies (8, 12, 16, 24, 25, 26, 40 MHz) sit in stock. Oddball values like 14.7456 MHz or 27.12 MHz often don't.
• Can they support custom-frequency programming? MEMS suppliers program to order in days. Some quartz suppliers keep blank-programmable oscillator programs too. This changes your shortage options completely.
• How do they handle PCN and EOL notices? Timing makers quietly discontinue older through-hole and large-package lines. You want notice 12 months ahead, not after the last-time-buy window closes.
• MOQ and reel quantities. 3225 oscillators typically come 3,000 per reel; some 2016 parts are 2,000 or 3,000. If your build is 500 boards a month, cut-tape availability matters.

One more thing. Ask how they test incoming stock on brokered buys. A competent distributor runs X-ray and decap sampling on high-risk lots, or at minimum solderability and marking inspection. If the answer is "we trust our sources," that's your cue to walk.

Counterfeit and Remark Risk in Timing Parts

Oscillators don't get counterfeited as aggressively as MCUs or power MOSFETs, but remarked parts show up whenever lead times stretch. The common pattern: a cheap ±50 ppm commercial-grade oscillator (0 to 70°C) gets remarked as the ±25 ppm industrial version (-40 to +85°C) of the same series. Visually identical. Electrically functional at room temperature. It passes your incoming test, ships in your product, and drifts out of spec in a cold warehouse application.

Defenses are unglamorous but they work. Buy industrial-grade parts from franchised stock when the application is industrial. Check laser marking depth and font against a known-good sample. Pull date codes and verify them against the maker's code format, which you can usually get from the manufacturer rep. And if a broker's price is 40% under the franchised quote during a shortage, the discount is the risk premium, not a bargain.

At XingHuan International, timing components go through the same incoming inspection as the rest of our stock: marking inspection, packaging and label verification against manufacturer formats, and traceability records on franchised lines. It's not glamorous work, but it's what keeps a remarked ±50 ppm part out of a ±25 ppm bin.

Lead Time Reality and How to Plan Around It

Quartz crystal blanks are a long-lead raw material, and oscillator makers build to forecast rather than to a deep catalog. The practical consequences for buyers:

Standard frequencies in 3225/2520: usually available from stock at distributors, 4 to 8 weeks from factory if not. Dual-source these; pin-compatible alternates exist across Epson, TXC, Abracon, and SiTime for most common values.

Non-standard frequencies: 10 to 16 weeks is normal for quartz. MEMS programmable parts cut this to 1 to 3 weeks, which is why many shortage-era designs qualified a MEMS alternate on the BOM as insurance. If your PCB layout keeps the oscillator footprint generic (4-pad 3225 with NC pads), a MEMS part often drops in with a BOM change only.

TCXOs and OCXOs: treat these as allocation-prone. RF-heavy demand cycles (handsets, GNSS modules) tighten TCXO supply fast. If your product ships in volume with a TCXO, negotiate a bonded-stock or scheduled-order arrangement with your supplier rather than buying spot.

Through-hole HC-49 crystals: still made, still cheap, but the line closures keep coming. If a legacy design depends on one, do a lifetime buy calculation now rather than after the EOL notice.

A Quick Word on Load Capacitance (Because Suppliers Get Blamed for This)

Roughly a third of "bad crystal" complaints we hear about trace back to load capacitance mismatch, not defective parts. A crystal speced at 12.5 pF CL running on a board with 18 pF of effective load will sit tens of ppm off frequency, and the MCU's oscillator may be marginal on startup. Before you RMA a batch, calculate the actual load from your cap values and stray capacitance: CL ≈ (C1 × C2)/(C1 + C2) + Cstray, with Cstray usually 3 to 5 pF. A good supplier will walk through this math with you. A great one asks about it before you order.

FAQ

Q: What's a realistic lead time for a crystal oscillator in 2026, and when should I worry?

A: Common frequencies in 3225 or 2520 packages typically ship from distributor stock or within 4-8 weeks. Non-standard frequencies run 10-16 weeks from quartz makers. Start worrying when your frequency is non-standard and your volume is under 10k/year; that's the profile that gets deprioritized when capacity tightens, so qualify a programmable MEMS alternate early.

Q: How do I tell a remarked oscillator from a genuine industrial-grade part?

A: Compare laser marking font and depth against a known-good sample from franchised stock, and verify the date code format with the manufacturer or their rep. Electrically, test frequency tolerance at temperature extremes (-40°C and +85°C), since remarked parts usually fail at the edges rather than at room temperature. When in doubt, buy through a distributor that provides traceability documentation, like XingHuan International does on its timing lines.

Q: Should I switch from quartz to MEMS oscillators to avoid supply problems?

A: MEMS parts offer 1-3 week lead times on programmed frequencies and better shock/vibration tolerance, which makes them strong shortage insurance. The trade-offs are higher phase noise on some families and single-vendor lock-in (SiTime dominates), so check your interface's jitter budget and keep a quartz alternate on the AVL if your volumes justify it.

Related Resources

• Crystals & Oscillators
• Electronic Components Distributor