Sample-Rate Conversion
Canonical Definition
Sample-rate conversion produces output samples on a different sampling grid from the input. Synchronous conversion uses rates with a defined clock relationship, while asynchronous sample-rate conversion estimates and tracks the relationship between independent clock domains. Both require interpolation and filtering that control imaging, aliasing, passband, stop-band, latency, and numerical error.
Synchronous Versus Asynchronous
| Type | Clock relationship | Typical use | Key checks |
|---|---|---|---|
| Synchronous SRC | known or derived rate ratio | format conversion inside one clocked system | ratio, filters, latency, word length |
| ASRC | independent clocks | joining an external digital source to a local DSP clock | tracking range, buffer behavior, lock, latency |
Rational Resampling Example
Converting 44.1 kHz to 96 kHz can be represented by the reduced ratio:
96,000 / 44,100 = 320 / 147A conceptual implementation interpolates, applies a low-pass filter, and selects output samples on the new grid. Practical polyphase implementations avoid calculating the many inserted zeros directly.
What Must Be Specified
- input and output rates and clock sources
- passband edge, ripple, stop-band attenuation, and transition width
- latency and buffer depth
- input and output word length, rounding, dither, and saturation
- lock, mute, and recovery behavior
- PCM versus non-PCM payload handling
Encoded Data Warning
An encoded IEC 61937 surround payload is not ordinary PCM audio. Passing it through a PCM SRC changes the payload bits and destroys the compressed frame. Detect non-PCM data and route it to a compatible decoder or bypass path.
Verification
- Measure frequency response, noise, distortion, and latency.
- Check source-rate changes and independent-clock drift.
- Interrupt the source and document mute and relock behavior.
- Verify pitch with a known tone.
- Confirm channel order and word format after conversion.