Unconventional Uses of FPGAs

Original Article ↗ Hacker News Discussion ↗

Evolutionary / unconventional uses of FPGAs

  • Multiple comments highlight evolutionary hardware, especially evolved FPGA circuits exploiting analog/physical quirks, as a canonical “unconventional” use.
  • Some practitioners describe being inspired by evolutionary algorithms but later burning out due to difficulty getting useful results and dependence on carefully crafted fitness functions.

Asynchronous logic and oscillator-based computing

  • Discussion covers asynchronous FPGAs and references work on reconfigurable asynchronous logic and “asynchronous systems on programmable logic.”
  • One personal project uses interacting ring oscillators on FPGAs to emulate Ising-like dynamics for graph problems.
  • Others argue such approaches cannot fundamentally speed up NP-complete problems beyond constant-factor gains; debate centers on parallelism vs oscillator interaction and continuous vs discrete-time formulations.

Physical effects, side channels, and accidental sensors

  • Ring oscillators acting as strain gauges fascinates many; compared to similar behaviors in microcontrollers and solid-state strain gauges.
  • Several war stories:
    • A game controller erroneously triggering inputs from PCB flexing.
    • A Raspberry Pi model resetting under camera flash due to light-sensitive silicon.
    • An FPGA image processor overheating only on certain vibrating helicopter setups because every pixel changed every frame.
  • Takeaway: vibration, strain, temperature, light, and humidity can all create unexpected behavior; many designs inadvertently become sensors.

Cost, overuse, and when FPGAs are appropriate

  • One view: FPGAs primarily make short-run/high-end systems more expensive, especially in government/military markets.
  • Counterview: compared to ASICs, FPGAs make low-volume, high-end designs feasible and cheaper.
  • Several comments complain about “this FPGA could have been a microcontroller / a few logic chips,” citing tool and ecosystem overhead and vendor lock-in.
  • Others note FPGAs’ dependence of power on design and switching activity, and the difficulty of predicting that accurately.

Security, military applications, and ASIC definitions

  • FPGAs with encrypted bitstreams are described as useful in munitions: logic is loaded at arming and vanishes on power loss, limiting reverse engineering.
  • Some point out FPGAs are still just ASICs in a broader sense and that similar “ephemeral IP” can be built with other architectures.
  • Debate over how secure FPGA bitstreams really are:
    • One side claims extracting meaningful design information from a running FPGA is extremely difficult, and power loss can irrevocably erase secrets.
    • Another side argues FPGAs are not inherently secure; sophisticated attackers may reconstruct netlists, so security must be systemic, not assumed.
  • There is also disagreement on terminology: whether ASIC should be reserved for single-product custom chips or is a broader category including processors, FPGAs, and converters.

Audio applications and measurement skepticism

  • A high-end FPGA-based headphone DAC/amp is praised subjectively as “ultimate” for some listeners.
  • Others reference measurement-focused reviews showing it is outperformed by cheaper, conventional DAC/amp designs.
  • General sentiment: FPGA-based audio can be technically interesting (e.g., very fast DACs, custom filters), but does not automatically yield better objective performance than modern delta-sigma DACs, and much of audiophile marketing is viewed skeptically.

Learning and higher-level HDLs

  • One commenter is learning Clash (a Haskell-to-FPGA compiler) and finds it feels like relearning programming, but rewarding; prior experience is mostly software and simple GPIO.
  • This is framed as an example of how accessible FPGA experimentation is becoming, even for people without formal hardware backgrounds.

On-chip sensing, thermal control, and randomness

  • Beyond strain gauges, FPGAs can act as:
    • Temperature sensors (timing ring oscillators against a reference).
    • Power-supply monitors (time-to-digital converters and oscillator frequency shifts vs VCC).
    • Self-heating “ovens” for stabilizing analog behavior via controlled switching and feedback to an on-chip temperature sensor.
  • These techniques are noted as both useful (e.g., for analog stability) and potential side channels in multi-tenant FPGAs.
  • A compact open-source TRNG design is praised; another project offers FPGA-based RNG IP without ring oscillators, emphasizing the need for extensive verification, especially in cloud FPGA contexts.

Other unconventional projects and questions

  • A “one-bit Bluetooth” project uses a SERDES at ~5 GHz on an FPGA plus a short wire and filter to communicate directly with phones, showing how far raw digital I/O can be pushed into RF territory.
  • Fault-injection / glitching attacks using FPGAs to extract secrets from embedded systems (e.g., game consoles) are mentioned as a common but “unconventional” use.
  • Someone asks about adapting incompatible engine sensors to an ECU; replies suggest FPGAs are overkill for simple analog level shifting, recommending analog gain/conditioning unless the interface is digital and complex.
  • A brief educational overview lists FPGA resource types (clocking, interconnect, sequential/combinational logic, specialized blocks) and notes that tools normally prevent undefined behavior, which unconventional uses sometimes exploit.
  • Precision reference designs are cited to show how PCB layout and packaging (plastic vs metal cans, humidity effects) are engineered to minimize strain and drift, reinforcing that mechanical and environmental factors tightly couple into silicon behavior.