Z-Transform

Numbered samples and their z^-1 delays become poles and zeros on the z-plane; the unit circle is your frequency axis.

What it is

The Z-Transform is the digital cousin of the Laplace Transform: the maths that designs every filter, delay and reverb inside your plugins and digital mixer.

Key facts

• Z-Transform: X(z) = sum of x[n] times z^(-n) over all samples. x[n] = the sample at step n; z = a complex number; z^(-n) = delay by n samples
• z^-1 means 'delay by exactly ONE sample' - the core building block of every digital filter
• z is complex: z = r times e^(j*omega). r = magnitude (growth/decay); omega = angle = frequency in radians/sample
• The unit circle (r = 1) on the z-plane IS the frequency axis: walking 0 to pi rad = DC up to Nyquist (half the sample rate)
• Link to Laplace: z = e^(s*T), where s = the Laplace variable and T = 1/sample-rate = seconds per sample
• Zeros (circles) pull the response DOWN to silence; poles (crosses) push it UP toward a peak/resonance
• STABILITY: a filter is stable ONLY if every pole sits INSIDE the unit circle (|pole| < 1). On or outside = rings forever or blows up
• At 44.1 kHz one z^-1 delay = 22.7 us; at 48 kHz one sample = 20.8 us. Nyquist = rate/2, so 48 kHz tops out at 24 kHz (omega = pi)
• +6 dB = double the amplitude; -3 dB = half the power (the standard filter cutoff point); -6 dB = half the amplitude
• FIR filters have zeros only (always stable, linear phase); IIR filters add poles (efficient, but unstable if a pole escapes the circle)

How it works

1. Chop the continuous signal into numbered samples x[0], x[1], x[2]... taken every T seconds
2. Write the effect as a recipe mixing the current sample with delayed past samples using z^-1 delay blocks
3. Convert that recipe into a transfer function H(z) - a ratio of polynomials in z
4. Find the zeros (top = nulls) and poles (bottom = peaks) and plot them on the z-plane
5. Check every pole is inside the unit circle - if yes, the filter is stable and safe to run
6. Read the frequency response by tracing the unit circle: distance to poles/zeros sets the gain at each frequency

Real examples

• Your digital console's 31-band graphic EQ: each band is poles and zeros placed on the z-plane
• A plugin delay set to 22.7 us = exactly one z^-1 step at 44.1 kHz
• A high-pass filter to kill stage rumble = a zero parked at z = 1 (DC) pulling the lows to silence
• An IIR low-pass smoothing a noisy signal = one pole near the unit circle creating gentle roll-off
• Reverb tails = feedback loops of z^-1 delays; if a pole creeps outside the circle the tail screams and never stops

How it helps in live sound

• Set plugin delay times in samples when phase-aligning: 1 sample = 20.8 us at 48 kHz, 22.7 us at 44.1 kHz
• Run the whole rig at ONE sample rate (48 kHz is the live standard) so every z^-1 delay matches across the system
• If a digital EQ or reverb self-oscillates or rings forever, a pole has gone unstable - pull the resonance/feedback down
• Steep FIR linear-phase EQs add latency (longer delay lines) - keep total system latency under ~10 ms for live
• Treat -3 dB as your true filter corner when reading plugin cutoffs, not where the curve visually starts to bend
• More poles/zeros = steeper, more surgical filters but more CPU and more phase shift - use the minimum that does the job