RF Channel Emulator
Put your radio through the propagation environment it will actually meet — on a bench, repeatably, before it ever leaves the lab. Every impairment is computed in FPGA fabric at the sample rate, with a fixed and published latency, so closed-loop testing produces results you can trust.
- < 1 µs
- Deterministic end-to-end processing latency
- 400 MHz
- Instantaneous emulation bandwidth per channel
- 8 × 8
- MIMO channel matrix with spatial correlation
- 100 %
- Sample-accurate — nothing dropped under any load
Field testing is expensive, and it is not repeatable
A radio that fails once in a hundred connections at the edge of coverage is almost impossible to debug outdoors: you cannot reproduce the conditions, and by the time you have instrumented the failure the weather has changed. A channel emulator turns that into a bench problem — same scenario, same seed, same result, as many times as it takes.
Reproducible failures
Every stochastic process is driven by a seeded hardware generator. Re-run a scenario with the same seed and the sample stream is bit-identical — so a failure found on Tuesday is still there on Friday while you bisect it.
Conditions you cannot arrange
A 900 km/h Doppler profile, a −8 dB Es/N0 satellite link in rain fade, or a barrage jammer at a known power ratio. Some of these are illegal to transmit and none of them are available on demand.
Regression, not one-off tests
Scenarios are scripted and run unattended. Link performance becomes a number your CI system reports on every firmware build, rather than something you discover at integration.
What the emulator does to your signal
Every effect below runs concurrently, per channel, at the full sample rate. They are not alternative modes — a realistic scenario needs most of them at once, and the cost of one more is fabric, not latency.
Propagation and multipath
- Path lossDistance-based free-space attenuation, static or driven from a trajectory, with a programmable exponent for non-free-space environments.
- Log-normal shadowingSlow fading from obstruction, with configurable standard deviation and spatial decorrelation distance.
- Multipath delay spreadTapped delay line, up to 24 independently configurable taps per path, each with its own delay, power and fading process.
- Rayleigh fadingNon-line-of-sight fading with selectable Doppler spectrum — classic Jakes, flat, rounded or bell-shaped.
- Rician fadingLine-of-sight fading with programmable K-factor per tap, including K-factor variation along a trajectory.
- Nakagami-m fadingGeneralised fading distribution for environments that fit neither Rayleigh nor Rician.
- Dynamic delayContinuously varying propagation delay from platform motion, interpolated to sub-sample resolution so range rate and Doppler stay physically consistent.
- Standard channel models3GPP TDL and CDL profiles, EPA/EVA/ETU, ITU pedestrian and vehicular, COST-207, and user-defined tap tables.
Doppler and geometry
- Doppler shiftCarrier frequency offset from relative velocity, static or computed continuously from a motion profile.
- Doppler spreadFading rate set independently of shift, so a scattering environment and a fast-moving platform can be emulated together.
- Doppler rateAcceleration-induced frequency drift — the term that breaks carrier tracking loops on fast platforms and satellite passes.
- Orbital Doppler profilesLEO and MEO pass profiles generated from orbital parameters, with the matching delay and path-loss variation applied together.
- Trajectory scriptingPosition and velocity supplied as a time series; the emulator derives delay, Doppler, path loss and K-factor from it consistently.
- MIMO spatial correlationCorrelated fading across an 8 × 8 channel matrix using a specified correlation matrix, including the 3GPP low/medium/high presets.
Noise, interference and jamming
- Additive white Gaussian noiseCalibrated injection with the level set directly as Es/N0, Eb/N0 or C/N, so a BER curve maps onto the axis you publish it against.
- Continuous-wave interferenceTone interferers at programmable offset and power, static or swept across the band.
- Narrowband and partial-band jammingShaped noise occupying a defined fraction of the band, with programmable duty cycle.
- Barrage noise jammingFull-band noise at a specified jammer-to-signal ratio, for anti-jam margin measurement.
- Pulsed and swept jammingTime-gated and frequency-agile interference, including follower-jammer behaviour against frequency-hopping links.
- Adjacent-channel interferenceA modulated interferer in the neighbouring channel at a programmable ratio, for selectivity and blocking tests.
- Co-channel interferenceA second modulated signal in-band, for capture-effect and multi-user testing.
- Impulsive noiseBurst noise with configurable rate, duration and amplitude distribution — ignition, switching and industrial environments.
Transceiver and hardware impairments
- Carrier frequency offsetStatic or drifting frequency error independent of Doppler, for oscillator-tolerance testing.
- Phase noiseInjected against a programmable phase-noise mask, so an oscillator specification can be tested before the oscillator is chosen.
- IQ imbalanceGain and phase mismatch between the I and Q branches, frequency-flat or frequency-selective.
- DC offset and LO leakageCarrier feed-through and DC bias, the two impairments that dominate direct-conversion receiver performance.
- Sampling clock offsetSample-rate error between transmitter and receiver, driving the timing loop the way a real crystal mismatch does.
- Amplifier non-linearityAM/AM and AM/PM compression from a measured or modelled characteristic, including memory effects.
- Amplitude and group-delay rippleFrequency-selective response from filters and cabling, applied as a user-supplied mask.
- Spurious and quantisation effectsDiscrete spurs at defined offsets and configurable effective converter resolution.
Why sub-microsecond latency is the whole point
Many channel emulators are software running on a workstation with an RF front-end attached. That is adequate for open-loop measurement and useless the moment the device under test reacts to what it receives.
- Closed-loop protocols HARQ, link adaptation, ARQ and TDD turnaround all have timing budgets defined in microseconds. An emulator that adds milliseconds of its own latency pushes the device outside its protocol timing and you end up testing the emulator instead of the radio.
- Ranging and positioning When propagation delay is the measurement, the instrument's own delay has to be a known constant that you can calibrate out. Ours is fixed, published and calibrated at manufacture.
- Emulator delay is subtractable Because the pipeline is fixed, the instrument's own latency is a constant offset rather than an error term. The minimum emulated path delay can be set to zero and the residual accounted for exactly.
- Jitter is zero, not small There is no scheduler, no interrupt and no buffer that can drain. The sample-to-sample delay through the instrument does not vary — which is a different claim from a low average latency, and the one that matters.
ALF-CE series
Three platform sizes. Bandwidth, channel count and RF coverage are the variables; the impairment set and the latency behaviour are identical across all three.
| Model | Emulation bandwidth | Channels | Taps per path | Typical use |
|---|---|---|---|---|
ALF-CE-1000 |
Up to 100 MHz | 1 × 1 bidirectional | 12 | Single-link development and regression |
ALF-CE-2000 |
Up to 200 MHz | 2 × 2 MIMO | 18 | MIMO radios, dual-band and relay testing |
ALF-CE-4000 |
Up to 400 MHz | Up to 8 × 8 MIMO | 24 | Massive-MIMO, phased array and multi-node scenarios |
System specifications
| RF frequency range | 70 MHz – 6 GHz (extendable on request) |
|---|---|
| Digital IQ interface | JESD204C, eCPRI or 10/25 GbE streaming |
| Processing latency | < 1 µs digital, < 3 µs RF to RF |
| Latency jitter | None — fixed pipeline depth |
| Path delay range | 0 – 2 ms, sub-sample interpolated |
| Doppler range | ±2 MHz shift, ±10 kHz/s rate |
| Dynamic range | > 80 dB instantaneous |
| AWGN accuracy | ±0.2 dB calibrated Es/N0 |
| Scenario update rate | Up to 100 kHz channel state update |
| Repeatability | Bit-identical for a given seed |
Control and integration
| Local control | Front-panel touchscreen, standalone operation |
|---|---|
| Remote control | SCPI over LAN, REST API, Python bindings |
| Automation | Scenario scripting with unattended batch execution |
| Scenario import | Tap tables, trajectories and impairment masks as CSV |
| Logging | Per-run channel state capture for post-analysis |
| Synchronisation | 10 MHz / PPS reference in and out, IEEE 1588 option |
| Form factor | 2U or 4U rack mount depending on model |
| Calibration | Factory calibrated, annual recalibration available |
| Warranty | 12 months, extendable |
What is inside the fabric
The emulator is an assembly of cores from the licensable portfolio. If you would rather build the instrument yourself, you can license the pieces.
Send us your test case.
The standard, the bandwidth, whether the loop is closed and which impairments matter. We will tell you which model fits and what it will and will not do.