Why 56k is the fastest dial up modem speed

Original Article ↗ Hacker News Discussion ↗

Engineers and enthusiasts revisit why consumer dial‑up modems topped out around 56 kbps, tracing the limit to the public switched telephone network’s original design for voice: ~3–4 kHz bandwidth, 8 kHz sampling, 64 kbps channels, and practices like robbed‑bit signaling that left only about 56 kbps usable for data. Commenters contrast analog last‑mile constraints with digital T1/ISDN/DSL links, share real‑world quirks like pair‑gain and loop carriers that capped speeds far below 56k, and reflect on how advances in digital signal processing and new access technologies—from DSL to Starlink—finally broke past those historical telephony limits.

Historical telecom technology

  • Early phone networks mixed analog long-distance (e.g., microwave relay) with emerging digital links (T‑1), with only the local loop staying analog in modern systems.
  • Microwave links were widely used due to the cost and difficulty of trenching long cables; later replaced mostly by fiber.
  • Discussion clarifies that “digital vs analog” refers to the information carried, not the physical layer, which is always analog (e.g., AMI on T‑1, analog or digital payloads over microwave).
  • Telecommunications drove much of computing: relays, vacuum tubes, early error correction, Shannon’s work, and later transistors and digital switching all originated in telco needs.

Why 56k and not more

  • 56k exploits the 64 kbps G.711 digital voice channel (8 kHz × 8 bits), with about 7/8 of bits usable due to robbed‑bit signaling, giving ~56 kbps downstream.
  • Key point: only one A/D conversion (at the subscriber end); ISP‑side is digital (T1/PRI, access concentrators).
  • Station‑to‑station analog connections top out around 33.6 kbps because the signal undergoes multiple unsynchronized A/D and D/A conversions.
  • Some posters argue robbed‑bit signaling is secondary; the real limit is POTS channel bandwidth (~3–4 kHz) and line quality per Shannon–Hartley.
  • Questions about “just using 16‑bit ADCs” are answered: you still hit analog bandwidth/SNR limits unless you effectively reinvent DSL.

Real‑world line issues and workarounds

  • Many users never saw full 56k; pair gain systems, subscriber loop carriers, extra conversions, and shared copper limited speeds to ~24–33.6 kbps.
  • Some used dual/bonded modems (Multilink PPP, vendor “shotgun” solutions) or ISDN to get ~112–128 kbps and better latency.
  • ISDN, T1, and variants (IDSL, HDSL) are discussed, including billing pain (per‑minute charges) and configuration details.

DSL vs circuit‑switched modems

  • One view: 56k hit a wall because circuit‑switching is architecturally limiting; DSL wins by only using the local loop then packet‑switching.
  • Counterview: faster circuit‑switched systems existed; the real enabler for DSL was advances in VLSI/DSP and using higher frequencies than POTS, not packet vs circuit per se.

Culture, nostalgia, and side topics

  • Strong nostalgia for dial‑up, ISDN gaming, and old exchanges; recommendations for telecom history resources and museums.
  • Notes on the decline of “soup‑to‑nuts” domestic telecom hardware engineering.
  • Brief digressions on inequality, tech unevenly distributed, and modern systems like Starlink as rural successors to old long‑haul infrastructure.