Satellite monitoring enables continuous observation of satellite communications and navigation signals to ensure link quality, detect interference, and verify compliance with spectrum regulations. These systems provide global and consistent visibility of uplink and downlink behaviour, which is critical for applications ranging from GNSS integrity and spectrum surveillance to interference hunting and system validation.
The primary objective of satellite monitoring is to maintain the integrity of space-to-ground and ground-to- space links. This includes verifying uplink and downlink quality, detecting unintentional or malicious interference, and supporting regulatory enforcement. Architecturally, monitoring systems are typically structured around three elements: the space segment, consisting of satellites with transponders and antennas operating in defined frequency bands; the ground segment, which includes monitoring stations equipped with large antennas, RF front ends, and digitisers; and the user segment, where specialised software and hardware analyse and visualise the captured data.
Satellite services operate across designated radio-frequency bands, each divided into uplink and downlink sub-bands in bidirectional systems in order to minimise mutual interference. Downlinks are usually allocated to the lower portion of a band due to lower atmospheric attenuation, while uplinks occupy higher frequencies to support higher data rates. Sub-band definitions vary by system; for example, Galileo uses “E” designations within the L-band rather than the “L” nomenclature used by other GNSS constellations.
From a monitoring perspective, this diversity makes frequency planning and sampling strategy critical. The sampling rate must ensure that the signal of interest occupies a single Nyquist zone, with out-of-band components suppressed through analog filtering. For direct sampling, this typically translates to minimum rates of approximately 2 GSPS for L-band, 4 GSPS for S-band, and 8 GSPS for C-band, assuming
appropriate bandpass filters are applied.
Modern monitoring stations rely on wideband digitisers to convert analog RF signals into digital data streams. Devices such as the ADQ35-WB from Teledyne SP Devices support direct sampling of L- and S-band signals without frequency mixers, reducing system complexity and calibration effort. With 12-bit resolution and up to 9 GHz usable input bandwidth, such digitisers enable flexible deployment across multiple satellite bands. External low-noise amplifiers and anti-alias filters remain essential to preserve signal fidelity and prevent spectral folding during analog-to-digital conversion.
Sampling-rate selection directly impacts both data integrity and downstream processing efficiency. For example, sampling the L-band at 5 GSPS places the signal entirely within the first Nyquist zone, while S- band undersampling at 4 GSPS confines the signal to the second Nyquist zone with sufficient guard bands. In contrast, poorly chosen rates can split the signal across Nyquist boundaries, introducing unavoidable aliasing.
Raw data rates from wideband digitisers can exceed practical transfer and storage limits. At 10 billion samples per second and two bytes per sample, a single channel generates around 20 GB/s. To manage this volume, onboard FPGA processing is used to reduce data rates before transfer over PCIe links.
Two approaches are particularly relevant for satellite monitoring. Bit compression reduces the number of bits per sample, enabling continuous streaming within PCIe bandwidth constraints while preserving full-band information. Digital down conversion, implemented through FPGA-based numerically controlled oscillators, filters, and decimation stages, translates selected RF channels to baseband or intermediate frequencies. This not only reduces data rates but also improves signal-to-noise ratio through filtering and coherent processing.
For real-time and near-real-time analysis, PCIe-based architectures are preferred. Peer-to-peer data transfer allows digitisers to stream data directly to GPUs using DMA, bypassing host CPU and system memory. This minimises latency and enables aggregate throughputs approaching the limits of PCIe Gen5, supporting simultaneous streams from multiple digitisers.
GPUs complement FPGA processing by handling computationally intensive but less latency-critical tasks, such as channelisation, demodulation, and long-term statistical analysis. For example, extracting individual Galileo sub-bands from a wideband L-band capture can reduce data rates from hundreds of megahertz of spectrum to a few gigabytes per second, well within modern GPU capabilities.
When long-duration recording is required, storage bandwidth can become a limiting factor. RAID configurations based on NVMe SSDs, connected via PCIe carrier boards, allow parallel writes across multiple drives. Enterprise-grade SSDs maintain sustained write speeds over long periods, enabling aggregate recording rates of tens of gigabytes per second and total capacities reaching the petabyte scale per slot. Consumer-grade drives remain suitable for shorter captures but exhibit throughput degradation once internal SLC caches are exhausted.
By combining wideband digitisation, FPGA-based pre-processing, GPU acceleration, and scalable PCIe storage, modern satellite monitoring systems provide a cost-efficient and flexible foundation for RF intelligence. This architecture supports evolving requirements such as multi-band monitoring, real-time interference detection, and large-scale data capture, making it suitable for both operational monitoring networks and research-oriented measurement campaigns.
spdevices.com
