Also known as: direct RF sampling, direct-sampling mode, RF sampling
Direct sampling feeds the radio-frequency signal from the antenna straight into an analog-to-digital converter with no mixer and no down-conversion — the ADC digitises the RF itself.1 Everything after the converter, including tuning and channel selection, happens in software. It is the purest form of software-defined radio: move the ADC as close to the antenna as the hardware allows and let DSP do the rest.
How it works
An ordinary receiver mixes RF down to a lower frequency the ADC can handle. A direct-sampling receiver skips that step and clocks the ADC fast enough to capture the RF band directly. To recover a signal at frequency f without a mixer, the converter must sample fast enough to satisfy the Nyquist theorem for the band of interest. Two regimes exist:
- Baseband (Nyquist) sampling captures everything from DC up to half the sample rate. To reach, say, 30 MHz directly the ADC runs at 60+ MS/s, and an anti-alias low-pass filter blocks anything above the first Nyquist zone.
- Bandpass (under)sampling deliberately samples slower than the signal frequency, relying on controlled aliasing to fold a higher Nyquist zone down to baseband. A tight bandpass filter must then isolate the one zone you want.
Either way, the whole captured span is available in software at once — you tune by selecting and down-converting a channel digitally rather than by retuning hardware.
Variants
RTL-SDR direct-sampling mode. The RTL-SDR is normally a quadrature-sampling receiver: its R820T tuner down-converts RF before the RTL2832U digitises it, and it cannot tune below about 24 MHz. Feeding an HF antenna to one of the RTL2832U’s ADC inputs — the “Q-branch” or “direct-sampling” hack — bypasses the tuner and samples RF directly at 28.8 MS/s. That opens up shortwave and HF reception below the tuner’s floor, at the cost of poor filtering and images from the first Nyquist zone. A cleaner alternative is an upconverter that shifts HF up into the tuner’s normal range.
Dedicated RF-sampling SDRs. Higher-end designs use ADCs fast enough to sample the whole HF (or VHF) spectrum properly, with real anti-alias filtering, so no mixer is needed at all.
Relevance to SDR
Direct sampling is attractive because removing the mixer removes a whole class of impairments — no image frequencies, no local-oscillator leakage, no mixer intermodulation. Its limits are the ADC’s clock rate (which caps the top usable frequency) and its bit depth (which caps dynamic range across a wide capture). For very high frequencies, superheterodyne or direct-conversion front ends remain necessary because no affordable ADC samples fast enough.
GopherTrunk is a software decoder that works on whatever IQ stream its source device delivers, so it treats a direct-sampling capture the same as any other. The trunking systems it targets (P25, DMR, TETRA) live in the VHF/UHF bands above the reach of the cheap RTL-SDR direct-sampling hack, so those are normally received with the tuner, but the direct-sampling concept is fundamental to how RF-sampling SDRs present their data.
Sources
-
Software-defined radio — Wikipedia, on placing the ADC at or near the antenna and doing tuning in software. ↩