BBC Micro Performance Wiki

scarybeastsecurity — Sampled sound, 1980s style, from the SN76489

6 min read

Sampled sound, 1980s style, from the SN76489

Chris Evans’s June 2020 blog post analysing how the SN76489 — a 1980 square-wave-only sound chip — was made to play digital audio samples on machines like the BBC Micro, Sega Master System, and the Coleco/TI 99. The author developed sampstream, a BBC Micro tape/disc-streamed sample player that demonstrates the technique on real hardware.

This is a primary-source analytical write-up: oscilloscope traces, signal-path tracing, and direct hardware experimentation. Not a how-to tutorial — there’s no code listing — but the technique explanation is rigorous and includes downstream analog-chain detail that’s missing from most BBC sound documentation.

Bibliographic

Filed into

  • Created sampled-sound — wiki page documenting the carrier-modulation technique end-to-end with BBC-specific implementation notes.
  • Refined sn76489 — added the 4 MHz clock input fact (with reference to Master ARM Ch 1 as primary source); added the analog-chain downstream stages (LM324N, LM386N-1, normalising network); flagged the ~8 kHz LM324N low-pass as a possible explanation for high-frequency-tone distortion seen by the author.

Key technical claims

The technique itself

“The usual basis of sampled sound playback is driving the SN76489 as a PCM device, i.e. a sequence of sample amplitude values.”

“The PCM stream is modulated simply by rapidly altering the volume of the 125kHz wave”

The mechanism: set one tone channel to period = 1 (so the chip emits a 125 kHz square wave), then write a stream of 4-bit volume values to that channel’s attenuator. Each different volume produces a different DC offset in the 125 kHz signal. The downstream analog chain (LM324N op-amp + LM386N-1 + speaker normalisation) low-pass-filters away the 125 kHz carrier and leaves the PCM envelope on the speaker.

“The key for sample playback is likely that the mid-point voltage differs for each volume.”

So the volume DAC is the sample output. There’s no special sample register on the chip — you abuse the attenuator.

Clock input (disputed in comments — Master ARM is authoritative)

The blog says:

“In the BBC Micro, the clock input to the sound chip is 2MHz and it has an internal divide-by-8, making it a 250kHz device.”

This is wrong. master-arm Ch 1 (page 16): “It receives a reference clock of 4MHz from central timing.” The SN76489 internal divider is ÷16, not ÷8 — so 4 MHz ÷ 16 = 250 kHz reference. Both descriptions land at the same 250 kHz internal rate, but the input clock is 4 MHz. A commenter on the post raised this point against the Master schematic; the author did not amend the main text.

Either way: period = N gives reference ÷ 2 ÷ N Hz output. Period 1 = 125 kHz. Period 0 is undefined (some clones treat as 1024).

Downstream analog chain (the bit usually missing from BBC docs)

“There’s some extra ciruitry between the LM386N-1 and the speaker, to keep the speaker happy. It normalizes the oscillation around the positive voltage point”

Per the author’s signal tracing, the SN76489 output passes through:

  1. LM324N op-amp (quad). Likely contains an ~8 kHz low-pass.
  2. LM386N-1 audio power amplifier.
  3. Post-amp DC-normalising network for the speaker.

Evidence for the LM324N low-pass: an 8 kHz square wave from the chip emerges as nearly a sine wave, not a square — characteristic of a filter rolling off in that band.

Signal levels

  • 125 kHz square wave from the chip: 720 mV peak-to-peak, centred ~3.3 V (not symmetric around 0).
  • The asymmetric centring is what enables PCM modulation via volume changes — each attenuator level shifts the centre.

Why sample playback is comparatively quiet

“This is likely one reason sampled sound playback is relatively quiet.”

The PCM signal is the difference between different DC offsets — a fraction of the full square-wave amplitude. After low-pass filtering, only that smaller fraction reaches the speaker.

What the author’s sampstream demo achieves

  • Custom disc loader that auto-detects 8271 (Model B) vs WD1770 (wd1770) controller.
  • Real-time 50 Hz oscilloscope view on screen.
  • “Near instant start and uninterrupted 400 KB of sample playback” (i.e. continuous streaming from disc).
  • Acknowledged not using best-in-class pre-processing or compression.

Prior art mentioned

  • Spy Hunter (arcade) — early example of sampled speech on the SN76489.
  • Galaforce and Icarus (BBC Micro) — creative tone-channel use.
  • pcmenc project — state-of-the-art today; uses multi-channel balancing and mid-point correction to fight 4-bit resolution limits.
  • SMS Power — community resource on SN76489 deep details.
  • MAME SN76489 driver source — for chip variants.

Compatibility commentary

Some references claim the SN76489 can’t hold a steady square wave for long periods. The author refutes this with oscilloscope evidence:

“Some references state that the SN76489 cannot hold a steady square wave output, particularly for longer periods. However, this is clearly not true — at least for this variant.”

Quality / quote on emulation accuracy

“the sample playback quality on a 1980 chip is obviously not too badly mangled because YouTube content ID recognizes emulator recordings”

Real hardware: “less harsh and louder” than emulators (per the author).

What’s missing from the article

  • No code listing. The author doesn’t walk through the assembly of his player. The wiki’s sampled-sound page reconstructs the implementation pattern from the sn76489 write-protocol and system-via slow-bus dance.
  • No sample rate quoted. The 400 KB / disc-stream figure implies multi-minute playback at modest rates, but neither the achievable sample rate nor the bus-bandwidth bottleneck is computed in the post.
  • No T1-timing recipe. via-timers T1 free-run is the natural way to drive sample-rate-locked writes; the author doesn’t discuss it.
  • No comparison vs DAC-cartridge approaches (Music 5000, Acornsoft speech). Those bypass the SN76489 entirely.

See also

  • sn76489 — chip-level reference, including 4 MHz clock input + slow-bus write protocol.
  • sampled-sound — implementation pattern derived from this source + the SN76489 page.
  • stardot-sn76489-sampled — 2025 follow-on Stardot threads: the persistent-/WE 62.5 kHz write trick + scarybeasts’ multi-channel sampled-song demos building on this 2020 piece.
  • system-via — slow data bus, the bottleneck for SN76489 writes.
  • via-timers — T1 free-run as the natural sample-rate clock.
  • reference_emulator_accuracy (memory) — author Chris Evans is the jsbeeb / beebjit author.

This wiki is curated by Claude following the LLM-Wiki methodology — a human curates source documents, the LLM compiles structured cross-linked markdown. Content may contain errors, omissions, or stale claims. For authoritative information refer to the original source documents in the bbc-documents GitHub archive.

Graph View

Backlinks