showstack

The reframe

Specifications are measurements of people

It is easy to read a spec sheet as though the numbers came from physics. Almost none of them did. They came from experiments on human beings, run to find the point at which a difference stops being detectable — and then a margin was added.

That has a practical consequence. When a number in this industry seems arbitrary, the useful question is not "what is the standard?" but "what could a person tell, and under what conditions?" — because the conditions are usually where the number breaks. A flicker rate that is invisible to a still eye is glaring to a moving one. A delay that is inaudible on speech is obvious on a snare. The threshold was never a single number; it was a number for one experiment.

Light

Fusion, phantom arrays, and why the eye is not a camera

Present a light that switches on and off fast enough and it stops looking like flashing and starts looking steady. That point — flicker fusion — is not fixed. It rises with brightness and contrast, and it is markedly higher in your peripheral vision than in the centre of your gaze, which is why a flickering fixture is so often noticed out of the corner of an eye and disappears when you look straight at it.

The bigger trap is movement. Fusion assumes the image stays on the same photoreceptors. Move your eye across a PWM-driven light, or move the fixture, and each flash lands somewhere different on the retina — so instead of a smear you see a dotted line, a phantom array. Flicker and its health effects are the subject of IEEE 1789. This is why a moving-light manufacturer quoting a PWM frequency in the low hundreds of hertz is not making a claim about your eye being slow; it is making a claim you can disprove by turning your head.

And a camera is a different observer again. It samples with a shutter, so it has its own beat frequencies with the light — the subject of genlock and LED walls — and it will photograph flicker that nobody in the room can see.

Sound

Why we localise to the first arrival, and what that buys you

When the same sound reaches you twice within roughly the first 35 milliseconds, you do not hear two sounds. You hear one, and you hear it as coming from wherever the first arrival came from. The second is folded in — it adds loudness and body, and it does not add a direction.

This is the precedence effect, and Helmut Haas's work showed how strong it is: the later arrival can be substantially louder than the first and still not take the localisation. It is not a small perceptual bias, it is a hard preference built into how we survive rooms full of reflections.

It is also the single most useful fact in sound reinforcement, because it is what makes a delay tower possible. Delay the tower slightly more than the physical distance requires and the audience under it hears plenty of level from it while still locating the sound on stage. Get it wrong and the same speaker becomes an echo that pulls every head in the wrong direction. The sound page has the arithmetic; this is the reason the arithmetic matters.

Both at once

The window in which sight and sound are the same event

Your brain is willing to call a sound and a picture the same event even when they do not arrive together — and the window is lopsided. Sound arriving after the picture is tolerated far more generously than sound arriving before it, because in the physical world late sound is what always happens and early sound never does.

Broadcast practice puts detectability somewhere around 45 ms for audio ahead of video and around 125 ms for audio behind it, with acceptability further out again. On a live show the equivalent question is a room: at 30 metres from the stage the sound is already about 90 ms behind the light, and nobody minds, because that is what distance sounds like.

What people do mind is inconsistency — a video wall with a frame of processing delay against a PA that is aligned to the stage puts a mouth and a voice on different sides of the window.

Colour

Two lights that match, and a camera that disagrees

Human colour vision has three channels. That means a very large number of physically different spectra map onto the same three responses — and any two of them will look identical to you. This is metamerism, and it is not a defect; it is what makes colour reproduction possible at all.

It is also a trap, because a camera also has three channels and they are not the same three. An LED fixture tuned by eye to match a tungsten source can photograph noticeably different — often greener, or with skin tones that will not correct cleanly. Nothing has failed. The two observers simply disagree, and the one holding the camera is the one the audience at home is using.

CRI and the newer TM-30 measures exist to put a number on this, and both are summaries: a single figure standing in for a whole spectrum. Colour spaces such as BT.709 and BT.2020 are the other half of the agreement — what a camera and a display consider a colour to be. Use them to reject bad sources, not to promise a match.

One more asymmetry worth knowing: at low light levels sensitivity shifts toward the blue end as rod vision takes over — the Purkinje shift. A deep blue night state reads brighter, and a deep red one darker, than a meter says. Designers have exploited that for a century without needing the name for it.

Darkness

What a blackout actually is

Dark adaptation is slow and it is in two stages: the cones give up their improvement within a few minutes, and then the rods keep going for another twenty to thirty. Full sensitivity is half an hour away, and any bright cue takes a large part of it back instantly.

So the audience's "dark" changes across an act. A blackout early in the show is much brighter to them than the identical state an hour in. And the running lights, the exit signs and the phone that comes out three rows back are all doing their work against whatever adaptation the audience has managed to accumulate.

Attention

The audience does not see what you did not point at

Attention is a spotlight, it is narrow, and it is driven by contrast — a change in brightness, an onset of movement, the start of a sound. Those pull it before conscious thought is involved.

The practical consequence is that a scene change can happen in full view of a thousand people and go unnoticed, provided the change is not where attention is. This is not a trick, it is inattentional blindness: outside the spotlight, large changes routinely go unregistered.

It cuts both ways, and this is the part worth internalising. An unmotivated movement in the wings, a technician crossing upstage, an LED that fades a beat late, a laptop screen in a box — each of those is a contrast event, and each will grab the spotlight away from whatever the show was pointing at. Most of what a good crew does in the dark is avoiding accidental salience.

The one people ask about

Why music gives you goosebumps

Musical chills — frisson — are a real, measurable autonomic event: piloerection, a change in skin conductance, a shift in heart rate. It is the same machinery that raises hair when you are cold or frightened, recruited by emotional arousal rather than temperature.

The imaging work on it, most influentially Salimpoor and colleagues in 2011, found something more specific than "music feels nice". Dopamine release was found in two places at two times: in the caudate during the build-up, while the listener anticipated the moment, and in the nucleus accumbens at the peak itself. The anticipation is not a lead-in to the reward. Neurochemically, it is part of it.

That explains the structure of the triggers, which are remarkably consistent: a resolution delayed longer than expected, an unexpected harmonic turn, a sudden dynamic swell, a new voice or texture entering, a sudden expansion of stereo width or of the room. All of them are expectations set up and then violated or withheld. A passage that does exactly what it promised, on time, does not produce chills — however beautiful it is.

Not everyone experiences it, and the difference is not musical training. It correlates with the personality trait of Openness to Experience, and with how strongly someone absorbs into what they are listening to.

Bringing it back

Why this page is the floor of the site

Every other page here is about getting a signal accurately from one place to another. This one is about the only reason that matters.

The beam angles are shaped by what an eye reads as an edge. The delay alignment exists because of the precedence effect. The QoS config protects a latency figure that was chosen because of a binding window. The shared clock exists because we notice when things disagree. The drone show reads as a picture because of how we group moving points.

Get the engineering right and nobody notices anything, which is the job. But knowing which human limit each number is protecting tells you the thing a spec sheet cannot: where you have room, and where you have none.