EM9305: The Swatch Group's BLE Chip That Beats InPlay on Battery Life
A 2 µA delta across 10 million smart labels works out to years of field life. That’s the entire margin between a product that lasts the warranty period and one that triggers a recall. And yet most BLE silicon comparisons get the average current wrong by more than 2 µA before the board even ships.
The reason is boring: vendor datasheets advertise peak TX current under conditions nobody runs in production. Supply voltage, DC/DC mode, advertising interval, payload, and SDK build flags each move the number more than the silicon does. A 3.0 V DC/DC EM9305 figure compared against a 1.8 V LDO nRF52 figure tells you nothing.
This article gives you a repeatable benchmarking method, a worked example on the EM9305 from EM Microelectronic, and honest comparison anchors against the Nordic nRF52 family and InPlay’s IN100 / IN612. Numbers below are datasheet-derived, not measured by us. Run your own bench before committing to a socket.
Currency note: figures referenced here track datasheets as of November 2025. Verify against current silicon revisions and SDK releases before quoting them in a design review.
Meet the EM9305 (and Why It Keeps Winning Smart Label Sockets)
EM Microelectronic is the semiconductor arm of the Swatch Group, headquartered in Marin, Switzerland. They’ve been shipping ultra-low-power silicon into watches and metering for decades. Their design culture chases nA-class sleep budgets, not throughput.
The EM9305 sits in their EM Bleu BLE family. Headline specs:
- BLE 5.3 controller, LE 1M / 2M / Coded PHYs
- Cortex-M0+ application core, 64 KB RAM, 512 KB flash
- Integrated buck DC/DC, 1.5 V to 3.6 V supply range
- ~3×3 mm WLCSP, often cited as one of the smallest BLE chips with a full application core
- Datasheet sleep current in the few-hundred-nA range with RAM retention
The claim worth investigating: lowest average current for a connectable BLE advertiser at 0 dBm. Where it’s actually shipping today is mostly ePaper smart labels, medical adhesive patches, and asset tags where the budget is a coin cell or a printed battery.
The Six Variables That Break BLE Power Comparisons
Before any numbers mean anything, you have to pin down six things. Change one, and the comparison’s void.
- Supply voltage. 1.8 V vs 3.0 V vs 3.3 V. DC/DC efficiency curves and LDO drop both bend here.
- TX power setting. 0 dBm vs +4 dBm vs -20 dBm. Peak current scales roughly linearly with output power; average current scales with TX duty cycle.
- DC/DC vs LDO mode. DC/DC typically saves 30 to 50% on radio current at 3.0 V. It also adds an external inductor and a BOM line.
- Advertising or connection interval. A 1 s interval and a 100 ms interval are different products.
- Payload size and PHY. 31 B legacy adv on 1M PHY is the lingua franca. Extended adv on 2M PHY isn’t comparable.
- SDK / stack version and compiler flags. Power numbers regress between SDK releases. We’ve watched a Nordic minor version bump add 8% to average current in advertising.
Pin down all six. Document them. Treat any vendor benchmark that omits one as marketing.
A Reproducible Benchmark Setup
Here’s the bench we’d recommend. It’s the same setup that holds up in a design review.
+-----------+ +----------+ +-----------+
| SMU |--Vdd-->| DUT | RF -->| Spectrum |
| (3.000 V) |<--I----| Board | | Analyzer |
+-----------+ +----------+ +-----------+
| |
v v
[ I(t) log ] [ GPIO trigger ]
\ /
\ /
v v
+----------------------+
| Host PC / Otii / PPK |
+----------------------+Steps:
- Fix supply at exactly 3.000 V from an SMU (Keysight N6705C, Otii Arc Pro, or Nordic PPK2 if you don’t need full SMU range). Don’t power from a dev kit USB rail.
- Configure an identical workload across all DUTs: connectable undirected advertising, 1 s interval, 31 B payload, LE 1M PHY, 0 dBm TX power.
- Drive a GPIO high at the start of each advertising event so the current trace can be aligned. This matters more than people think when you’re integrating per-event charge.
- Capture a 60 s window. Compute average current and per-event charge in µC. The per-event number is what you scale into battery life math.
- Repeat with DC/DC enabled and disabled. The delta tells you what your BOM is buying you.
- Log SDK version, silicon revision (read from the chip, don’t trust the reel label), and ambient temperature. Power scales with temperature, especially leakage.
A single advertising event looks roughly like this on the current trace:
I(t)
mA | ___
5 | | |
4 | | |__
3 | __| |
2 | | |__
1 |___| |________
0 +--------------------------> t
pre TX RX postThe “pre” ramp and “post” tail are where silicon vendors actually compete. Peak TX current is mostly determined by physics and the PA topology. Wakeup time, scan window, and how fast the chip falls back to sleep is where EM9305 tends to come out ahead.
EM9305 vs nRF52: Worked Example Numbers
The table below is datasheet-derived, not measured. Treat it as a starting point for your own bench, not a citation.
Workload: Connectable adv, 1s interval, 31B, 0 dBm, 3.0V, 25°C
Datasheet-derived, November 2025. Silicon rev / SDK: [verify at publish]
Chip Avg Current TX Peak RX Peak Sleep
-----------------------------------------------------------
EM9305 (DC/DC) ~3.0 µA ~4.5 mA ~3.0 mA ~400 nA
nRF52832 (DC/DC) ~5.5 µA ~5.3 mA ~5.4 mA ~1.5 µA
nRF52810 ~6.2 µA ~5.0 mA ~4.6 mA ~1.2 µA
InPlay IN100* ~2.5 µA† ~4.2 mA n/a ~700 nA
* IN100 is beacon-only (non-connectable). Not apples-to-apples.
† Beaconing workload, not connectable advertising.What the numbers say if you trust the datasheets:
- EM9305 wins on average current for connectable advertising at 0 dBm, mostly thanks to the integrated DC/DC and short radio ramp. The sleep current floor also pulls the 1 s interval average down.
- nRF52832 in DC/DC mode is the industry baseline. It pays roughly a 50 to 80% current overhead in this advertising-only workload. In high-throughput connections that gap closes, and the Nordic ecosystem advantage starts to dominate.
- IN100 looks lowest on paper, but it’s a configuration-only beacon. It can’t accept a connection. If your product needs GATT, IN100 is out of the running before you start.
The honest reading: EM9305 genuinely beats nRF52 in advertiser-heavy roles. It doesn’t beat nRF52 in throughput, ecosystem, or anything that depends on Zephyr maturity.
EM9305 vs nRF52 vs InPlay: Honest Trade-offs
EM9305. Best raw average current we’ve seen in connectable advertising at 0 dBm. The application core is a Cortex-M0+, which is fine for sensor pre-processing but not for heavy DSP. Tooling is functional but smaller than Nordic’s. FAE coverage is regional. If your design lives or dies on an extra year of coin cell life, this is the chip to bench.
Nordic nRF52 family. Best ecosystem, period. nRF Connect SDK and Zephyr support are mature, the community is large, and you can hire engineers who already know the part. Pays roughly a 50 to 80% average-current overhead vs EM9305 in pure advertising. For most products this is the right default unless power is the binding constraint. If you’re staying on Nordic, our terrestrial device SDK reference applications cover the Zephyr integration path.
For InPlay, the split matters: the IN100 (NanoBeacon) ships with no firmware. You configure it over a tool and it beacons, which is the fastest time to market for non-connectable tags. The IN612 is the connectable sibling and the fairer EM9305 rival. If you’re in the IN100 socket, also look at how the Hubble advertising packet format maps onto the chip’s payload constraints before locking the design.
Worth a glance during shortlisting: Dialog/Renesas DA1469x for ultra-low-power with a Cortex-M33 application core, and TI CC2640R2 / CC2652 if you’re already in the TI ecosystem. Neither tends to beat EM9305 on advertising current, but both have specific niches.
Third-party benchmark suites like BlueJoule are useful as a sanity check on your own bench. Don’t cite their numbers without re-verifying against your silicon rev and SDK.
A Design-In Checklist Before You Commit
Run this before you order the first board spin:
- Re-run the benchmark on the silicon revision your distributor will actually ship, not the rev on the eval kit.
- Confirm the DC/DC inductor is on your AVL. EM9305 specifies a particular external L value and saturation rating; a substitution will quietly cost you efficiency.
- Validate whether a pre-certified module is available. If you can’t afford full radio cert, this decides the chip.
- Get a written long-term supply commitment from EM Microelectronic for your program length. The Swatch Group is stable, but the EM9305 is one part in their catalog.
- Pull SDK release notes for the last 4 versions. Look for power regressions. They happen.
- If you’re integrating with a network back-end, confirm payload format and provisioning flow before the board freeze. Hubble’s provisioning at scale guide covers what tends to bite at the factory.
Before You Cite These Numbers in a Design Review
EM9305 beats nRF52 on average current for advertiser-heavy BLE designs, and beats InPlay IN100 anywhere a connection is required. Both claims hold only with an honest bench. Pin down supply voltage, DC/DC mode, interval, payload, PHY, and SDK version, and document them on the same page as the table. Run the bench yourself against the silicon rev and SDK release you’ll actually ship.
Editor flags:
- Insert verified silicon revision and SDK version in the table footnote and currency caveat before publish.
- All numerical figures are datasheet-derived; if measured data becomes available, swap in and update the “datasheet-derived” disclaimer.
- Confirm latest BlueJoule publication status before final pass; remove the reference if their dataset is stale.
- Verify EM9305 package dimensions and the “smallest BLE chip” framing against current EM Microelectronic marketing claims (other vendors contest this).
Hubble Network lets BLE devices connect directly to satellites at standard advertising power, so chips like the EM9305 can reach global coverage without redesigning the radio stack. See how it works →