Sample Rate Conversion

It is changing a recording from one snapshot speed to another, like 48,000 down to 44,100 per second.

48 kHz to 44.1 kHz: the SRC engine rebuilds the wave and recomputes new samples at tighter spacing, same pitch and length.

What it is

Resampling a recording from one sample rate to another (e.g. 48 kHz down to 44.1 kHz) by recalculating new sample values.

Key facts

Sample rate = snapshots captured per second, in Hertz (Hz) / kilohertz (kHz)
44.1 kHz = 44,100 samples/sec (CD standard); 48 kHz = 48,000 samples/sec (video/broadcast standard)
48k to 44.1k ratio = 147:160, so it is NOT a clean integer divide
Nyquist theorem: highest stored frequency = sample rate / 2; 44.1 kHz holds up to 22.05 kHz, 48 kHz up to 24 kHz
Human hearing tops out ~20 kHz, so both rates cover the full audible band
Other rates: 88.2k, 96k, 176.4k, 192k (high-res) and 32k (some broadcast)
Bit depth is SEPARATE: 16-bit = 96 dB dynamic range, 24-bit = 144 dB; SRC does not change it
Aliasing = frequencies above Nyquist folding back as fake lower tones; bad SRC adds this as harshness
Pro converters (SoX VHQ, r8brain) exceed 150 dB SNR; sloppy ones sit near 90 dB
SRC keeps pitch and length identical; it is NOT time-stretch or pitch-shift

How it works

Pick source rate and target rate (e.g. 48,000 to 44,100 samples/sec).
Compute the ratio between them (44100/48000 = 0.91875).
Mathematically rebuild the smooth original waveform from existing samples using sinc interpolation.
Apply a steep low-pass filter just under the new Nyquist to block aliasing.
Read off brand-new sample values at the new spacing in time.
Export the file at the new rate, same pitch and length as the original.

Real examples

Film/video edit delivered at 48 kHz, but the client wants a CD master at 44.1 kHz
Music master at 44.1 kHz dropping into a 48 kHz video timeline
Podcast recorded at 48 kHz exported to 44.1 kHz MP3 for older players
96 kHz sample library imported into a 48 kHz live-show session
DAW set to 44.1 kHz importing a 48 kHz field recording (auto-resamples on import)

How it helps in live sound

Lock ONE sample rate (48 kHz) across the whole rig: console, Dante/AVB, playback laptop, recorder, video feed.
Dante/AVB clock the network at one rate; mismatched rates cause clicks, dropouts or refused subscriptions.
If a backing track is 44.1 kHz and the show runs at 48 kHz, convert it offline BEFORE the gig, never live.
Set DAW 'SRC/resample quality' to its highest (Pro Tools Best, SoX VHQ) for exports.
QLab/playback: pre-convert all cues to the session rate so the engine is not resampling live under load.
Match camera/video audio at 48 kHz to avoid drift and lip-sync slip over a long show.

Everyday analogy

Like redrawing a flipbook at a new frame rate: you recompute the in-between drawings so the motion stays smooth instead of just dropping or duplicating pages.

Watch out

Myth: 'just change the rate field and it's done.' Truth: that's renaming, not converting; real SRC recalculates every sample with a filter, and skipping it shifts pitch/speed or adds aliasing harshness.

Fun fact

44.1 kHz exists because early digital audio was stored on video tape: it's the number that fit neatly into both PAL and NTSC video timing in the late 1970s.

Key takeaways

Sample rate = snapshots per second; 44.1 kHz (CD) vs 48 kHz (video) are the two big ones.
Nyquist = rate/2 sets the highest frequency you can store; downsampling lowers that ceiling.
Good SRC = rebuild waveform + steep low-pass filter; bad SRC = aliasing harshness.
SRC keeps pitch and length the same; it is not time-stretch or pitch-shift.
Live: lock the whole rig to one rate (48 kHz) and pre-convert any odd files offline.