Crystal-Less BLE Chips: When You Can Skip the External Crystal and When You Can't
Most BLE designs ship with two crystals nobody questions. They’re on the reference schematic, so they end up on your schematic, then on your board, eating 4 to 10 mm² of space you don’t have. On a disposable medical patch or a coin-sized asset tag, that’s real estate you’d kill for. But vendor datasheets now promise “crystal-less” operation, and you’re wondering: can I actually rip these out?
Sometimes yes. Sometimes it’ll cost you months of debugging intermittent disconnections at temperature corners you never tested. The difference comes down to understanding exactly what each crystal does, what replaces it when it’s gone, and whether your application can absorb the consequences.
Here’s a vendor-agnostic framework for making that call.
Two Crystals, Two Different Bets
A typical BLE SoC relies on two external crystals. The high-frequency crystal (usually around 32 MHz) drives RF synthesis and baseband clocking. The low-frequency crystal (32.768 kHz) runs the sleep timer and schedules connection events. Each one needs a pair of load capacitors and a chunk of board area.
┌─────────────────────────────────────────────┐
│ BLE SoC │
│ │
│ ┌──────────┐ ┌──────────┐ │
│ │ HF Osc │ │ LF Osc │ │
│ │ (32 MHz) │ │(32 kHz) │ │
│ └────┬─────┘ └────┬─────┘ │
│ │ │ │
│ ▼ ▼ │
│ RF Synthesis Sleep Timer / │
│ & Baseband Connection Events │
│ │
└──────┬──────────────────────────────────────┘
│ │
┌───┴───┐ ┌───┴───┐
│HF XTAL│ │LF XTAL│ ← One or both
│+2 caps │ │+2 caps│ may be removed
└───────┘ └───────┘When vendors say “crystal-less,” they almost always mean the LF crystal is optional. The SoC’s internal RC oscillator takes over sleep timing, periodically calibrating itself against the HF source. A handful of parts also support eliminating the HF crystal, replacing it with an on-chip oscillator that calibrates against incoming BLE packets. Eliminating both is rare and carries the tightest constraints.
The BOM and Footprint Case for Going Crystal-Less
Let’s put numbers on what you’d save.
Component | Footprint (mm²) | BOM Cost (10k) | Notes
-----------------+------------------+-----------------+-------------------
32 kHz crystal | ~1.6–2.4 | $0.05–$0.15 | + 2 load caps
32 MHz crystal | ~3.2–5.0 | $0.10–$0.30 | + 2 load caps
Load caps (each) | ~0.5 | $0.005 | 2 per crystal
-----------------+------------------+-----------------+-------------------
Total savings | ~4–10 mm² | $0.17–$0.60 | If both removedAt 100k units, dropping both crystals saves $17,000 to $60,000 in component cost alone. Factor in pick-and-place charges (typically $0.01 to $0.03 per placement) and you’re trimming 6 placements from every board.
The layout benefits add up too. Crystals need keep-out zones, controlled trace routing, and ground isolation. Strip those out and your PCB designer gets breathing room, the kind that’s hard to quantify but easy to appreciate on a 6 mm diameter board.
For a disposable glucose patch or a single-use cold-chain tag, this math is compelling.
What You Actually Lose Without a Crystal
Five specific tradeoffs, quantified where possible.
Sleep Clock Accuracy and Drift
An external 32.768 kHz crystal holds ±20 to 50 ppm across temperature. An internal RC oscillator? Expect ±250 to 500 ppm, sometimes worse at the edges of the operating range.
The BLE stack uses sleep clock accuracy (SCA) to size the receive window around each connection event. At ±50 ppm, the RX window for a 1-second connection interval is roughly 100 µs. At ±500 ppm, that window balloons to ~1 ms or more. Wider RX windows mean the radio stays on longer, burning current for every single connection event.
Connection Stability and Supervision Timeouts
Clock drift accumulates between anchor points. With ±500 ppm and a 4-second connection interval, drift can reach ±4 ms per interval. If a few missed calibration cycles happen during a temperature swing, the peripheral’s timing can wander far enough to miss events entirely.
What this looks like in practice: you get intermittent disconnections that only show up in the field, specifically in environments with thermal fluctuations. Cold-chain logistics devices moving between a 35°C loading dock and a -18°C freezer are a classic victim. You’ll chase this bug for weeks because it never reproduces on your bench.
The Current Consumption Paradox
Going crystal-less is supposed to be about efficiency. But the internal RC oscillator needs periodic calibration against the HF source. Each calibration cycle wakes the CPU and runs the HF oscillator briefly.
Net result: some crystal-less BLE designs consume 10 to 30% more average current in connected states compared to their crystal-equipped equivalents. If you’re running a tight energy budget on a 40 mAh coin cell, that’s months of battery life evaporating.
HF Crystal Removal: Even Tighter Constraints
The BLE spec requires RF carrier accuracy within ±150 kHz. An internal HF oscillator can hit this fresh off calibration at 25°C. Drift it to -20°C or +60°C and you’re flirting with the boundary.
The failure scenario: your device passes FCC/CE testing at room temperature in the lab, then drifts out of spec on a rooftop in Phoenix or a warehouse in Minnesota. Packet loss climbs. Worst case, you’ve got a regulatory compliance issue in deployed product.
Advertising Interval Jitter
Higher clock variance makes advertising intervals less predictable. For a beacon that just needs to be “eventually” discovered, this is tolerable. For a device that needs fast, reliable scanner-side discovery (think: tap-to-pair consumer products), the added jitter can noticeably increase connection setup time.
When You CAN Skip the Crystal
Some applications genuinely don’t need one:
Advertise-only beacons with no connection requirement and modest duty cycles. If you’re broadcasting a payload every 2 seconds at room temperature, internal oscillator accuracy is fine.
Single-use disposable devices with 6 to 12 month lifetimes in thermally stable environments. A medical patch worn on skin (roughly 32 to 36°C, always) is about as benign a thermal environment as you’ll find.
Ultra-compact form factors where the crystal physically won’t fit. Ingestible sensors, subdermal implants, and some in-ear devices simply don’t have the board area. The crystal-less tradeoff isn’t optional; it’s the only path.
Devices with infrequent, short connections in controlled environments. A sensor that connects for 200 ms once per hour in an office building is unlikely to hit drift problems.
For asset tracking applications where devices operate in a wide range of environments, it’s worth reviewing how different use cases handle these constraints before committing to crystal-less.
When You CANNOT Skip the Crystal
Anything requiring sustained connected throughput. Audio streaming, OTA firmware updates, and bulk data transfers all need tight connection event timing. A 500 ppm sleep clock will cause retransmissions and throughput collapse.
Wide-temperature-range devices. If your operating spec spans -20°C to +60°C (industrial, outdoor, automotive-adjacent), the internal RC oscillator’s drift curve works against you hard.
Products that need a precise real-time clock. Timestamping, medication scheduling, and time-synchronized sensor logging all suffer from RC oscillator drift. You’ll accumulate minutes of error per day.
Multi-role designs running central and peripheral simultaneously. The timing coordination overhead is already tight; adding clock uncertainty makes it fragile.
RF-certification-critical products that must pass testing across the full operating temperature range. Keep the HF crystal.
Medical devices with reliable-delivery requirements. If a missed packet means a missed alarm, the crystal stays. Period.
Decision Checklist
Pin this to the wall during your architecture review.
CRYSTAL ELIMINATION DECISION CHECKLIST
───────────────────────────────────────
[ ] Connection intervals ≤ 1 second? → Keep crystal
[ ] Operating temp range > ±20°C from 25°C? → Keep crystal
[ ] Sustained connections > 30 seconds? → Keep crystal
[ ] OTA DFU required? → Keep crystal
[ ] Product lifetime > 1 year? → Keep crystal (LF)
[ ] RF certification at temp extremes needed? → Keep HF crystal
[ ] Advertise-only, room temp, short life? → Crystal-less viable
[ ] Board area budget < 15 mm² total? → Crystal-less may be
the only optionIf you checked any of the first six boxes, keep the corresponding crystal. The BOM savings aren’t worth the field failures.
Practical Tips If You Go Crystal-Less
You’ve run through the checklist and crystal-less makes sense. Here’s how to not regret it.
Tune the calibration loop. Every vendor ships a default calibration interval. It’s usually conservative. Profile your actual power budget and adjust; too frequent wastes power, too infrequent lets drift accumulate. The device SDK documentation for your platform will typically expose these parameters.
Characterize drift across your real temperature range during DVT. Don’t trust the datasheet’s “typical” number. Measure your specific silicon lot at your actual temperature corners.
Increase connection supervision timeouts and slave latency. Give yourself margin to absorb drift. A supervision timeout of 6 seconds instead of 2 seconds costs nothing if your application can tolerate the longer disconnect detection.
Budget the extra RX window power in your energy model. If you modeled your battery life using crystal-grade SCA values, redo the math with ±500 ppm. You might be surprised.
Test connection reliability at temperature corners. Not room temp. Not “briefly in a thermal chamber.” Run your full connection scenario for hours at your min and max operating temperatures.
Design crystal pads into the PCB even if you plan to DNP. Adding unpopulated footprints for both crystals and their load caps costs you a few mm² of copper and zero BOM dollars. If crystal-less operation fails qualification, you populate the pads and keep your board. Without them, you’re looking at a respin. A $0 insurance policy against a $20,000 to $50,000 respin is the easiest engineering decision you’ll make this quarter.
Making the Call
Crystal-less BLE is a real option with real savings: 4 to 10 mm² of board area, $0.17 to $0.60 per unit, and a simpler layout. For disposable, thermally stable, advertise-only devices, it’s often the right call.
But the tradeoffs are equally real. Wider RX windows, higher average current, connection instability at temperature extremes, and potential RF compliance headaches. These aren’t theoretical risks; they’re field failure modes that show up 6 months after launch in environments your bench setup never simulated.
Run the checklist. Characterize at temperature. And always design in those DNP crystal pads.
“Hubble Network connects BLE devices directly to satellites—no gateways, no infrastructure, no crystal oscillator debates. See how it works →”