Problem classes the practice has experience with.

A low-Earth-orbit satellite traverses the visible sky in minutes, and the carrier frequency a ground receiver sees shifts continuously across that pass — tens of kilohertz of Doppler at typical UHF and S-band frequencies, with slew rates that exhaust naive tracking loops. Time-of-arrival also varies as the path length changes, which matters for any system where downstream processing depends on consistent symbol timing or coherent integration.

The engineering shape of a solution involves predicting the Doppler curve from orbital elements, designing receiver loops that track the slew rate without losing lock, and decoupling the timing variance from the frequency variance so each can be corrected independently. The work lives in the gap between theoretical link budgets and operational reality — what physics says is possible, and what hardware actually does under noise, multipath, and pointing error. Predicting the curve is half the problem; designing the loops that follow it without breaking is the other half.

Underneath every modern radio link is an I/Q sample stream — the complex baseband representation of the signal — and most of the engineering substance lives there. Filtering for adjacent-channel rejection, root-raised-cosine matched filtering for symbol recovery, phasor analysis for constellation diagnostics, energy detection and classification for unknown signals, demodulation under noise and interference. The work is mathematical, but the math has to survive contact with real hardware: ADC quantization, oscillator phase noise, sample clock drift, and channel impairments all leave their fingerprints on the samples.

A waveform engineer is judged not on the elegance of an algorithm in MATLAB but on whether it still works at the bit error rate the requirements demand, on the hardware the budget allows, in the environment the system will deploy to. The right answer is usually not the most sophisticated algorithm. It’s the simplest one that meets the requirement under the actual operating conditions.

An HF radio that has been in service for decades still works, and the cost of replacing it is substantially higher than the cost of integrating it. But the original schematics are gone, the original developers are unreachable, and the radio’s control interface is whatever the manufacturer chose to expose — often an obscure serial protocol, a parallel handshake, or a panel-switch interface designed for human operators rather than computers.

The work involves signal-level analysis on the existing equipment to characterize what’s actually happening at the wires, reverse-engineering the framing and control sequences, and writing an adapter — sometimes called a "blister" — that exposes a clean modern interface (REST, MQTT, gRPC, whatever the integrating system speaks) without modifying the radio itself. The goal is to make the legacy hardware look modern from the outside while leaving its proven internals untouched. The radio doesn’t change. The interface to it does.

A utility, manufacturing plant, or pipeline operates on a SCADA system whose master station is reaching end of support — the OS is unsupported, the original integrator is gone, the vendor has discontinued the platform — but the field equipment still works perfectly. Replacing the entire system is a multi-million-dollar capital project that takes years and risks operational disruption. Modernizing only the master station while preserving the field equipment is faster, cheaper, and almost always the right answer.

The engineering shape is a phased migration: a new network architecture that supports both the legacy field protocols (Modbus, DNP3, proprietary serial) and modern IT integration; a new HMI and visualization layer designed for current operators rather than original ones; secure remote access; and a cutover plan that keeps the system operational throughout. The standard outcome is a system the operations team understands better than the one they had before, with documentation they can actually use, on infrastructure that’ll still be supported in a decade.

The same engineering problem appears in wildly different domains: detecting a signal of interest against a noisy background, then presenting the detection to a human operator in a way they can act on. Cellular phone activity inside a correctional facility where it shouldn’t exist. Distress signals or motion anomalies from RF tags on livestock spread across a large pasture. The fundamental work is the same — energy detection, classification under noise, false-alarm management, geolocation through TDOA or RSSI — but the deliverable is shaped by who has to use it.

Detection without a usable interface is half a system. The pin on the map is what an operator actually consumes — not the receiver math underneath it. Web-based and desktop GUIs that visualize detection events, plot locations geographically, log events for review, and surface alerts at the right priority level are part of the deliverable, not an afterthought. A practice that does both the signal work and the visualization work delivers complete systems instead of pieces that need a second team to integrate.

Rotating machinery — pumps, fans, compressors, generators — produces characteristic vibration signatures, and an imbalance shows up as a clear peak at the rotational frequency with specific phase relationships. Identifying the magnitude and phase of the imbalance, then translating that into actionable correction (where to add or remove mass, how much, at what angular position) is a signal-processing problem with a mechanical answer.

The work involves accelerometer placement and signal acquisition under real industrial conditions (noisy, electrically hostile, sometimes inaccessible), FFT analysis with the right windowing to separate imbalance from bearing wear or alignment problems, and translation of complex-valued vibration measurements into trim-weight recommendations that a maintenance technician can execute with a balance grade in mind. The math points at the problem. The engineering produces an answer the operations team can act on without becoming signal-processing experts themselves.