PACERIFT
03 / THE SIGNAL PATH

Clean signal.
Honest numbers.

Phone sensors are imperfect. Pacerift processes GPS and inertial motion on-device to make the useful signal easier to read.

RAW SENSOR INPUT FILTERED MOTION TRACE
01 / INPUT

GPS + IMU

Location, acceleration and rotation arrive as time-stamped sensor samples.

02 / CLEAN

Reject spikes

Robust outlier checks help prevent isolated bumps from dominating a ride summary.

03 / ALIGN

Separate motion

Orientation-aware processing estimates useful linear motion in the phone reference frame.

04 / REVIEW

Read the ride

Inspect speed, route and force trends after the session, even when recording offline.

[ 🛠️ Technical Telemetry & Signal Engineering Specs ]EXPAND SPECS +

Engineering notes for riders who want to understand how GPS and inertial signals become a readable ride profile. These are processing concepts, not a claim that a phone replaces calibrated chassis-mounted motorsport instrumentation.

01 / LOW-PASS

Butterworth smoothing

A low-pass Butterworth response attenuates high-frequency noise while retaining the ride-scale motion that matters. For order n and cutoff fc, the squared magnitude response is:

|H(jω)|² = 1 / (1 + (ω / ωc)2n)

Cutoff selection trades noise rejection against transient response. Filtering should preserve acceleration and braking events rather than over-smooth them.

02 / ROBUST OUTLIERS

Hampel rejection

For samples in a local window, compute the median m and median absolute deviation (MAD). Flag an outlier when:

|xᵢ − m| > k · 1.4826 · MAD

The scale factor makes MAD comparable to standard deviation for normally distributed noise. A flagged sample can be replaced with the window median or marked for downstream handling.

03 / FRAME ALIGNMENT

Quaternion rotation

Orientation quaternions rotate measured acceleration from device coordinates into a chosen reference frame:

aworld = q ⊗ adevice ⊗ q⁻¹

Gravity compensation and axis alignment help estimate linear acceleration. Orientation drift and phone placement remain sources of uncertainty.

04 / VEHICLE DYNAMICS

Useful relationships

Longitudinal acceleration is estimated from velocity change over time. For a turn at speed v and radius r, lateral acceleration is:

alat = v² / r  G = a / 9.80665 m·s⁻²

An ideal coordinated-turn lean estimate follows tan(θ) = v²/(r·g). Phone-only estimates are affected by mounting, vibration, road camber, tire dynamics and sensor drift.

PHONE SENSORS SUPPORT RIDE ANALYSIS · THE OPTIONAL PACERIFT POD IS DESIGNED TO ADD DIRECT CHASSIS MEASUREMENT

READY WHEN YOU ARE

Make every ride
count.

FREE · OFFLINE READY · ANDROID