Knowledge · GNSS-denied navigation
GNSS-denied navigation: what to specify and check when selecting an IMU
QUICK ANSWER
When GNSS is jammed or spoofed, the IMU's datasheet errors integrate freely – a constant gyro bias produces an angle error that grows linearly with time, and the resulting tilt error drives position error roughly with the cube of time. Selecting an IMU for GNSS-denied operation therefore means two things: building an error budget from documented typical values, and checking that the part is specifiable – export documentation, lifecycle status and evaluation before design-in.
A recent Forbes article on secure supply chains for navigation in GNSS-denied environments has put the topic back on engineering agendas. This article stays on the engineering side: what goes into the specification and what to verify before a sensor is designed in. How long an IMU can bridge a single outage was covered in an earlier article; here the focus is selection.
01
Why does GNSS-denied operation need its own IMU specification?
Interference is no longer an edge case. According to IATA's Incident Data eXchange, cited in IATA's release of its 2025 Safety Report, reported GNSS jamming events in 2025 rose by 67 % compared with 2023, while reported GPS spoofing incidents rose by 193 %. In the Baltic region, Lithuania's regulator RRT identified 36 GPS spoofing antennas near Kaliningrad in May 2026, up from 3 in early 2025, with a combined estimated radius of about 450 km (estimated from ADS-B disturbances). A monitoring station on Bornholm logged 30 jamming/spoofing incidents so far in 2026, compared with 16 in 2025 and 11 in 2024. In civil aviation, Airbus has selected an inertial navigation system for the A220 that switches to inertial navigation when GNSS becomes unreliable, with entry into service expected around 2030.
For a system designer, this means the IMU must be specified for the phase in which it works alone – not only for its role as a smoothing aid in a GNSS fusion filter.
02
Which datasheet values drive the error budget?
Four parameters dominate open-loop error growth. The underlying relationships are described in Woodman's An introduction to inertial navigation (University of Cambridge, 2007):
- 01Gyro bias instability (°/h): a constant gyro bias ε produces an angle error θ = ε·t, i.e. linear growth.
- 02Angle random walk, ARW (°/√h): the angle error's standard deviation grows with √t; the ARW value equals σθ after 1 h.
- 03Accelerometer bias instability (µG): a constant acceleration bias ε produces a position error s = ε·t²/2.
- 04Tilt error and gravity: a tilt error ε leaks gravity into the horizontal axes as a residual acceleration g·sin(ε).
Combining the first and last relationship (our own derivation): a gyro bias makes the tilt error grow linearly, so the horizontal position error grows approximately as s ≈ (1/6)·g·ε·t³. This cubic term is why gyro bias instability usually dominates the budget after the first minute.
Velocity random walk (VRW) adds a further noise term on the velocity, which should be included in a full budget.
03
How do current Epson IMUs compare on these parameters?
All values are typical values from the Epson datasheets (Typ = average or 1σ at 25 °C, 3.3 V).
| Model | Gyro bias instability (°/h) | ARW (°/√h) | Accel bias instability (µG) | VRW ((m/s)/√h) | Gyro range (°/s) | Output rate (Sps, max) | Interface | Size (mm) / weight (g) | Operating temp. (°C) |
|---|---|---|---|---|---|---|---|---|---|
| M-G370PDG0 | 0.8 | 0.06 | 24 | 0.02 | ±450 | 2000 | SPI/UART | 24×24×10 / 10 | −40…+85 |
| M-G366PDG0 | 1.2 | 0.08 | 24 | 0.02 | ±450 | 2000 | SPI/UART | 24×24×10 / 10 | −40…+85 |
| M-G330PDG0 | 3 | 0.1 | 34 | 0.03 | ±400 | 2000 | SPI/UART | 24×24×10 / 10 | −40…+85 |
| M-G570PR20 | 0.5 | 0.04 | 14 | 0.012 | ±475 | 2000 | RS-422 | 65×60×30 / 150 | −30…+70 |
The PDG series shares one 24×24×10 mm, 10 g footprint and the same SPI/UART interface across three performance levels. This makes it possible to fix the mechanical and electrical design early and choose the performance grade once the error budget is known.
04
What does the budget look like in numbers?
The following open-loop values are an illustration, not a navigation performance claim. They treat the bias instability as a constant bias (an optimistic simplification), ignore Schuler effects, sensor fusion and temperature, and use g = 9.81 m/s² and 1 µG ≈ 9.81·10⁻⁶ m/s².
Formulas: heading error after 1 h = ε_gyro·1 h; accel term s = ½·ε_acc·t²; tilt term s ≈ (1/6)·g·ε_gyro·t³ (ε_gyro in rad/s).
| Model | Heading error from gyro bias after 1 h (°) | ARW after 1 h, 1σ (°) | Accel bias term after 60 s (m) | Accel bias term after 10 min (m) | Gyro-bias tilt term after 60 s (m) | Gyro-bias tilt term after 10 min (m) |
|---|---|---|---|---|---|---|
| M-G370PDG0 | 0.8 | 0.06 | ≈ 0.42 | ≈ 42 | ≈ 1.4 | ≈ 1,400 |
| M-G366PDG0 | 1.2 | 0.08 | ≈ 0.42 | ≈ 42 | ≈ 2.1 | ≈ 2,100 |
| M-G330PDG0 | 3 | 0.1 | ≈ 0.60 | ≈ 60 | ≈ 5.1 | ≈ 5,100 |
Two conclusions follow. First, in this simplified model the gyro bias term already exceeds the accelerometer term after 60 s (by a factor of about 3–9) and outweighs it by more than an order of magnitude after 10 minutes – gyro bias instability is the parameter to negotiate first. Second, open-loop position holding over many minutes is not realistic for any MEMS IMU; the specification must define how long the IMU has to bridge and what aiding (odometry, vision, magnetometer, motion constraints) is available. The IMU Selection Guide with drift calculator lets you run the same calculation for your own outage duration.
05
What makes an IMU specifiable from a sourcing perspective?
An IMU that meets the error budget but cannot be documented, exported or supplied over the product's life is not a viable choice. Four checks belong in the specification phase:
- 01Documented typical values. Base the budget on values stated in the current datasheet revision, and record the revision (e.g. M-G370PDG0 datasheet Rev1.2, 2025-03-03; M-G570PR20 Rev1.4, 2026-01-21). Note that Epson states typical values; if your budget needs guaranteed limits, clarify this with the supplier.
- 02Export-control documentation. The EU Dual-Use Regulation (EU) 2021/821, Annex I Category 7, covers accelerometers (7A001), gyros (7A002) and IMUs/inertial systems (7A003). The classification of a specific part must be confirmed with the supplier for your project. Epson's IMU lineup page marks the listed models "ITAR free" (Epson does not define the scope of this label), and Epson datasheets require compliance with Japan's Foreign Exchange and Foreign Trade Control Act and the US EAR. Note also that Epson datasheets prohibit use in military or weapons equipment.
- 03Availability and lifecycle status. Epson labels older models "not recommended for new design" on its lineup page. Check the lifecycle status of every candidate before design-in, and do not assume long-term availability without a written statement from the supplier. For reliability planning, Epson's FAQ states an MTTF of more than 50,000 h.
- 04Evaluation before design-in. Typical values are averages or 1σ figures at 25 °C. Measure bias behaviour over your temperature range and motion profile with real hardware before committing the layout.
06
Which checklist should the IMU specification include?
| Check | What to verify | Evidence / document |
|---|---|---|
| Gyro bias instability (°/h) | Meets budget for the required outage duration | Current datasheet revision, own error budget |
| ARW (°/√h) and VRW ((m/s)/√h) | Noise terms included in budget | Datasheet |
| Accel bias instability (µG) | Position term acceptable over outage | Datasheet, own calculation |
| Value type | Typ. vs. guaranteed; conditions (25 °C, 3.3 V) | Datasheet notes, supplier confirmation |
| Operating temperature (°C) | Covers application range | Datasheet; own temperature test |
| Interface and output rate (Sps) | Matches host and filter rate | Datasheet, drivers |
| Export classification | Confirmed for the specific part and destination | Supplier statement; EU 2021/821 Annex I Cat. 7 check |
| Lifecycle status | Not marked "not recommended for new design" | Manufacturer lineup page, supplier confirmation |
| Supply continuity | Written statement where required | Supplier |
| Evaluation | Bias/noise verified on own hardware and profile | Evaluation-kit test report |
QWhich IMU parameter matters most in GNSS-denied operation?
Over periods of minutes, gyro bias instability usually dominates, because the resulting tilt error couples gravity into the horizontal channels and position error grows approximately with t³.
QAre datasheet typical values sufficient for an error budget?
They are a sound starting point, but Epson's typical values are averages or 1σ at 25 °C and 3.3 V. Verify them over your temperature range and motion profile, and clarify with the supplier whether guaranteed limits are needed.
QAre Epson IMUs export-controlled?
EU Regulation 2021/821, Annex I Category 7, covers accelerometers, gyros and IMUs. The classification of a specific Epson model must be confirmed with the supplier for your project; Epson's lineup page marks the listed models "ITAR free".
QCan the PDG performance grade be changed late in the design?
M-G370PDG0, M-G366PDG0 and M-G330PDG0 share the same 24×24×10 mm package, 10 g weight and SPI/UART interface, so the mechanical and electrical design can stay the same while the performance grade is decided by the error budget.
EVALUATION KIT
How can you evaluate a PDG-series IMU for your project?
AMCORIS, the motion sensing brand of Channel Microelectronic, provides evaluation kits on free loan for qualified projects, so you can verify bias and noise behaviour on your own hardware before design-in. The M-G370PDG0, M-G366PDG0, M-G330PDG0 and M-G570PR20 are available on request; the M-A352AD10 is the model offered directly. Short datasheets, studies and GitHub drivers are available in the downloads section; full datasheets are available on request.