Window length (sec) ≈ 1 ÷ frequency resolution (Hz). 10 Hz resolution NEEDS a 100 ms window. No escape.
Transients (kick, clap, click) are SHORT in time = WIDE in frequency. A click is broadband by nature.
A pure steady tone is the opposite: NARROW in frequency = must last a LONG time to measure precisely.
Speed of sound ~343 m/s at 20 degC; +6 dB = double pressure; -3 dB = half power; double distance = -6 dB SPL
1/3-octave RTA = 31 bands (fast/coarse) vs high-res FFT = thousands of bins (fine/laggy): same uncertainty wall
How it works
Pick a window: grab a slice of the incoming audio (e.g. 1024 or 4096 samples).
Short slice = sharp timing, fuzzy pitch. You see WHEN but not WHICH note.
Long slice = sharp pitch, smeared timing. You see WHICH note but not exactly WHEN.
Multiply window seconds by bin hertz: the product can never drop below ~0.5. That is the wall.
Choose your poison per job: transients want short windows, tuning/feedback want long windows.
An analyser just shows you the result of this choice as 'fast vs detailed'.
Real examples
Tuning a bass guitar: use a LONG window (slow, high pitch resolution) so 41 Hz reads as a stable number not a blur.
Hunting a snare 'crack' or timing a delay: use a SHORT window (fast) so you catch the exact moment.
Ringing out monitors for feedback: long FFT pinpoints the exact ring frequency to notch (e.g. 2.4 kHz).
An analyser feels 'snappy but coarse' or 'detailed but laggy' depending on the FFT size you picked.
A 1/3-octave RTA on a graphic EQ: 31 fast bands, great for fast moves, useless for finding one narrow ring.
How it helps in live sound
Feedback hunting: set analyser to a LARGE FFT (8192-16384) for fine Hz so you notch the exact ring, not a guess.
Catching transients / setting gates: switch to a SMALL FFT (512-1024) for fast time response.
Use a 'multi-resolution' / FDW mode if you have it: long windows for bass, short for treble.
Don't trust an RTA reading on a kick drum's sub: short events smear low frequencies, the number lies.
Room tuning: use slow averaging (long windows) for steady response; go fast for FOH troubleshooting.
If your EQ resolution looks razor sharp, your analyser is slow to react. Physics, not a fault.
Everyday analogy
Like a camera shutter: fast shutter freezes the moment but the picture is dark and grainy, slow shutter is bright and detailed but motion blurs into a smear, and you can only ever pick one.
Watch out
Myth: 'just buy a higher-resolution analyser and you get sharp pitch AND sharp timing.' Wrong, no analyser, FFT size, or money beats Δt × Δf ≥ ~0.5; better gear only lets you choose the trade-off, never escape it.
Fun fact
It is the exact same maths as Heisenberg's quantum uncertainty principle, just with time-and-frequency instead of position-and-momentum, which is why physicists and sound engineers fight the identical wall.
Key takeaways
You can nail the timing OR the pitch, never both perfectly.
The hard limit is Δt × Δf ≥ about 0.5. No gear beats it.
Short window = sharp time, blurry pitch. Long window = sharp pitch, blurry time.