🔭 astrophysics

A High-Bandwidth Backplane for Wideband Radio Interferometers and Integration with the CHORD Telescope Correlators

This paper presents a high-bandwidth, scalable backplane architecture for the CHORD telescope's CRS FPGA platform that enables robust, tightly synchronized, and high-throughput signal processing for wideband radio interferometers through a validated multi-crate system supporting 128 digitized inputs per crate.

Original authors: Wellington Avelino, Joshua Montgomery, Jean-Francois Cliche, Graeme Smecher, Matt Dobbs

Published 2026-06-03

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Wellington Avelino, Joshua Montgomery, Jean-Francois Cliche, Graeme Smecher, Matt Dobbs

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to organize a massive, high-speed conversation between 1,024 people (the antennas) who are all shouting different things at once. In the world of radio astronomy, these people are listening to the universe, and they need to share their "shouts" (data) instantly to figure out what's happening out there.

The problem is that if you just plug 1,024 microphones into a standard computer, the wires get tangled, the signals get messy, and the timing gets off by a tiny fraction of a second. In radio astronomy, being off by even a tiny fraction of a second is like trying to take a photo of a hummingbird with a camera that has a blurry shutter; the picture comes out useless.

This paper describes a custom-built "super-connector" (called a backplane) designed to solve this exact problem for a new telescope called CHORD. Here is how it works, broken down into simple concepts:

1. The Problem: The "Traffic Jam" of Data

Older systems used standard computer buses (like VME or PCI) which are like narrow, single-lane country roads. They are too slow and don't keep everyone's clocks synchronized. If one person speaks a millisecond later than the others, the whole group gets confused.

For the CHORD telescope, the data traffic is so heavy (about 10 Terabits per second—that's like downloading the entire internet in a few seconds) that standard roads would collapse. They needed a new kind of highway system.

2. The Solution: A Custom "Data Super-Highway"

The team built a custom backplane, which is essentially a giant circuit board that acts as the central hub for the telescope's computer brains.

The "Crate" Concept: Think of the system as a large shipping container (a "crate"). Inside this crate, they can stack up to 16 different computer boards (called CRS boards).
The "Backplane" as the Glue: The backplane is the wall inside the crate that all 16 boards plug into. Instead of using messy cables, the boards plug directly into this wall.
The Speed: This wall has 25 "lanes" of traffic for every single connection, running at 25 Gigabits per second. It's like turning a dirt path into a 25-lane superhighway where every car can drive at the speed of light.

3. Keeping Time: The "Conductor"

The most critical part of this system isn't just speed; it's timing. In an orchestra, if the violinist plays a split-second before the drummer, the music sounds terrible. In a radio telescope, if the antennas aren't perfectly synchronized, the "image" of the universe blurs.

The Metaphor: The backplane acts like a strict conductor. It sends a single, perfect "beat" (a 10 MHz clock signal) to all 16 boards at the exact same time.
The Result: The paper proves that this system keeps the timing so precise that the difference between any two boards is only a few hundred femtoseconds (that's a millionth of a billionth of a second). It's so stable that even if the room gets hot or cold, the "beat" doesn't wobble.

4. Testing the System: The "Stress Test"

The team didn't just build it; they put it through the wringer to make sure it works for the CHORD telescope.

The "Eye" Test: They sent massive amounts of data through the system and looked at the signal quality (like looking at an "eye" diagram). The "eyes" stayed wide open, meaning the data arrived perfectly clean with zero errors, even when all lanes were running at full speed simultaneously.
The Heat Test: They ran the system for 24 hours straight with all 16 boards working hard. The system stayed cool enough, and the power supply remained steady, proving it won't overheat or crash during long observations.
The Thermal Drift Test: They checked if the heat from one board would mess up the timing of its neighbor. They found that while heat does cause tiny shifts, the system is predictable and stable enough that the telescope's software can easily correct for it.

5. Why This Matters

This backplane is a "plug-and-play" solution. You can start with one crate (16 boards) and add more crates later without having to rewrite the computer code. It provides a solid, reliable foundation that allows the CHORD telescope to listen to the universe with extreme precision.

In short: The paper presents a custom-built, high-speed, perfectly synchronized "data highway" inside a computer crate. It ensures that thousands of radio antennas can talk to each other instantly and perfectly in time, allowing the CHORD telescope to take sharp, clear pictures of the universe without the data getting lost or garbled.

Technical Summary: A High-Bandwidth Backplane for Wideband Radio Interferometers and Integration with the CHORD Telescope Correlators

Problem Statement
Modern radio astronomy is shifting from single-dish telescopes to large interferometric arrays (e.g., SKA, CHIME, HERA) that utilize hundreds to thousands of antenna elements. These systems require real-time digital spatial correlation (X-engine) across broad bandwidths, generating massive data traffic (e.g., CHORD requires ~10 Tb/s). The primary engineering challenge lies in sustaining high aggregate throughput while preserving stable timing alignment and phase coherence across a large number of distributed signal chains. Commercial off-the-shelf interconnects often fail to meet the stringent phase-stability and latency-control requirements of precision correlators without prohibitive integration complexity, cost, or power consumption. Existing backplane technologies (VME, CompactPCI, ATCA) have historically struggled with precise synchronization or lack the bandwidth density required for next-generation wideband instruments.

Methodology
The authors present a custom-designed backplane architecture developed for the Control Readout System (CRS) FPGA platform, specifically tailored to meet the requirements of the Canadian Hydrogen Observatory and Radio-transient Detector (CHORD) telescope. The methodology involves:

Architecture Design: A modular backplane system where each unit hosts up to four CRS boards. A single 6U, 19-inch crate integrates four interconnected backplanes to support 16 CRS modules (128 digitized inputs).
Interconnect Strategy: The system utilizes passive high-speed channels operating at 25 Gb/s per lane. Data exchange between boards within a crate occurs via zQSFP+ connectors, while inter-crate expansion uses QSFP28 interfaces. This passive mesh supports a fixed-topology "corner-turn" operation, allowing frequency-partitioned streams from all antennas to be shared before routing to centralized X-engine units.
Timing Distribution: A dedicated controller backplane distributes a 10 MHz reference and IRIG-B timing signals via Low-Voltage Differential Signaling (LVDS) to all slots. The design minimizes phase noise accumulation by using a low-additive-jitter fan-out buffer and avoids cascaded clock regeneration. An onboard Oven-Controlled Crystal Oscillator (OCXO) allows for standalone operation without external timing instruments.
Physical Implementation: The PCB uses a 14-layer low-loss substrate (Panasonic MEGTRON-4) with passive differential stripline routing. Solder-free connectors and backdrilling on vias were employed to minimize impedance discontinuities and reflections.
Validation: The system was rigorously tested under full-load conditions (16 CRS boards) to evaluate signal integrity, timing stability, power delivery, and thermal behavior.

Key Contributions

High-Density Passive Interconnect: The design achieves a 25 Gb/s per-lane data rate across a passive backplane mesh, enabling a total intra-crate throughput sufficient for 128 inputs without active switching logic on the backplane itself.
Integrated Timing Backbone: The architecture embeds a precise timing distribution network directly into the backplane, ensuring sub-nanosecond alignment stability across all modules.
Scalability without Firmware Modification: The system supports expansion from single-crate to multi-crate configurations without requiring changes to the FPGA firmware, offering a flexible path for next-generation interferometers.
Co-Design Approach: The backplane was co-designed with the CRS platform and CHORD requirements, prioritizing instrument-driven performance (stable timing, controlled latency) over simply adopting the highest available commercial data rates.

Experimental Results

Signal Integrity: Bit Error Rate (BER) testing using the IBERT framework demonstrated error-free operation over 10 Tbits per channel at 25 Gb/s. Eye diagrams confirmed clear timing and amplitude margins with no measurable degradation due to inter-channel crosstalk.
Timing Stability: Short-term synchronization analysis revealed inter-board delay stability with jitter in the range of 200–500 femtoseconds over minute-scale timescales.
Thermal and Power Performance: Under a 24-hour continuous full-load test (1152 W total draw), the system maintained stable operation. The maximum FPGA junction temperature reached 52°C in a 21°C ambient environment. Thermal gradients between slots were managed (14°C max difference), and power rail regulation remained within 1.5% with minimal ripple (50 mVpp).
Thermal Drift: Characterization of differential thermal sensitivity showed that timing drift is thermally driven and predictable, providing a quantitative basis for calibration and compensation in deployed systems.

Significance and Claims
The paper claims that this requirement-driven, passive backplane architecture provides a practical and flexible foundation for wideband digital correlators. By delivering observatory-scale scalability with low-jitter and phase-consistent data transport, the system addresses the critical bottleneck of maintaining coherence in large-N interferometers. The authors assert that the design is sufficient to support the phase precision required for CHORD-class instruments and can be generalized to other high-throughput, synchronized instrumentation (e.g., TES/MKID spectrometers, passive radiometry chains). The work validates that a custom, co-designed backplane can achieve the necessary stability and throughput without the prohibitive complexity of active switching fabrics or optical interconnects at this scale. The system has been successfully exercised in an 8-board full $n^2$ correlator configuration, with end-to-end observations planned for the CHORD telescope by the end of 2026.