home/ atoms/ delay-line-interpolation

Fractional delay times require interpolation, which trades tuning accuracy for frequency-dependent attenuation

To achieve non-integer delay times (needed for precise pitch tuning), digital delay lines must interpolate between stored samples. Interpolation is a form of filtering — like Gaussian blur on an image, it attenuates high frequencies. In a single pass the effect is barely noticeable, but in a recursive feedback loop with many passes, the cumulative high-frequency attenuation is significant. This creates an inherent trade-off: higher delay-time resolution (more accurate pitch) means more interpolation per cycle and faster high-frequency decay. The Nord Modular’s delay lines use fractional sample interpolation at estimated resolution of 5.05 nsec (about 2^19 steps in 64 control units), which provides good pitch accuracy at the cost of some extra high-frequency roll-off.

Examples

If interpolation cuts high frequencies by 0.1dB per pass and your pitch requires 100 passes per second, highs decay ~10dB/sec faster than lows — changing the timbre.

Assessment

Explain why simple integer-sample delay lines would not need interpolation. Describe the trade-off between tuning resolution and timbral accuracy in waveguide models.

“In any delay line, there is always this trade-off between accurate delay time (giving acurate pitch in a physical modelling waveguide configuration) and accurate frequency response.”
corpus · physical-modeling-karplus-strong-waveguides-strings-and-wood · chunk 4