Sample rate is the number of samples per second, rather than a note's frequency. 48 kHz means 48,000 samples per second. The Nyquist frequency is half the sample rate for an ideal sampling system; real filters need a transition region. 96 kHz is not automatically twice as good as 48 kHz. Bit depth relates to amplitude resolution and quantization noise, rather than pitch.
One buffer's duration = sample count / sample rate × 1,000 ms.
| Buffer | One buffer at 48 kHz |
|---|---|
| 64 | 1.33 ms |
| 128 | 2.67 ms |
| 256 | 5.33 ms |
| 512 | 10.67 ms |
This is not the complete guitar-to-ear round-trip latency. The total can include input/output buffers, driver or safety buffers, ADC/DAC, routing/resampling, and effects' intrinsic delay. Processing delays add along a serial path; parallel branches may need alignment. A Delay pedal's musical repeats do not necessarily delay the first arrival of its dry path.
A higher CPU load does not automatically add another buffer for every pedal. If processing finishes before the deadline, latency can stay the same; a missed deadline can cause dropouts or crackles. Begin evaluating at 48 kHz/128 samples, try 64 if stable, and increase the buffer when needed. Test the heaviest preset and preset changes. Driver figures are not proof of a complete round-trip measurement.
Concept reference: Ableton — How Latency Works. A buffer calculation is not a measurement of total interface latency.
Concept references
Buffer and signal-path calculator
How long is one buffer, and what is missing?
- One buffer
- 2.667 ms
- Two buffers
- 5.333 ms
- Hypothetical total
- 5.333 ms
N / sample rate × 1,000 gives the duration of one buffer. Measure device round-trip latency for the real result; these figures do not guarantee that a particular buffer is stable.
An estimate using two buffers plus the time you enter. This does not measure device latency. Stability, safety buffers, and driver paths need testing; block widths in this diagram do not represent duration.