Who the show is not reaching
Four access provisions, all of them engineering problems with numbers attached, all of them cheap at the plot stage and expensive afterwards. None of this is front of house’s department. Every one of them is decided by somebody in a technical role, usually without noticing they decided it.
Flash
Three per second, and why it is not a style choice
Photosensitive epilepsy affects a small proportion of the population and a large proportion of a stadium. The threshold is not a house rule — it is the same number in WCAG 2.3.1, in ITU-R BT.1702 and in the Ofcom broadcast guidance: no more than three flashes in any one-second period.
Two refinements matter on a show. Sensitivity peaks between roughly 15 and 20 Hz, which is exactly where a strobe lands when somebody sets it by feel against a track. And saturated deep red is judged more strictly than any other colour, because the response it provokes is not the same as an equivalent white flash. Static patterns count too: more than about five clearly discernible light-dark stripe pairs is its own hazard, which is why a fast-moving high-contrast gobo pattern can be a problem at no flash rate at all.
The useful reframing is musical rather than numeric, because nobody types a frequency into a console. A strobe on every beat of a 128 BPM track is 2.13 flashes a second and fine. On every eighth note it is 4.27 and it is not. The threshold sits between two ordinary programming decisions, which is exactly why it gets crossed by accident.
- Signage instead of design. A notice at the door discharges a duty and does nothing for somebody who is already inside when the cue fires. It is the last line of defence, not the first.
- Effects that stack. Two fixtures each at 2 Hz, offset, are 4 Hz in the room. The threshold is about what the eye receives, not what any one fixture is set to.
- Video content nobody checked. Supplied content, VJ loops and camera cuts all flash. Lighting gets scrutinised for this and screens routinely do not.
- Cameras in the audience. A broadcast or IMAG feed carries the flash to everyone watching a screen, where the assessment standards are stricter and a Harding test is the normal instrument.
Hearing
Three ways to get sound to somebody, and they fail differently
An assistive listening system takes a clean feed of the show and delivers it to a listener without the room in the way. That last part is the point: for most people with hearing loss the problem is not level, it is separating a voice from reverberation and crowd noise, and turning the PA up makes it worse rather than better.
There are three delivery technologies and they are not interchangeable. An induction loop drives a wire around the seating area and couples magnetically to the telecoil built into most hearing aids — nobody has to collect anything, which is the single biggest determinant of whether a system gets used at all. Infrared needs line of sight, which means under a balcony is dead, but it does not leak out of the room, which is why it is used where confidentiality matters. RF covers the building comfortably and is therefore one more set of frequencies to coordinate against your radio mics.
- A system nobody can switch on. The commonest failure by a distance. The receivers are in a cupboard, the batteries are flat, and the duty manager has never been shown the panel. This is a technical handover problem, not a policy one.
- Feeding it the wrong mix. The main PA mix is not the right source: it is voiced for a room. A dedicated mix, dry and speech-forward, is what makes the system worth having.
- Headphones counted as hearing-aid compatible. They are not. Compatibility means a neckloop or similar that couples to a telecoil, and it is a separate count in the table.
- RF assistive listening left out of the coordination. It is another transmitter in a band you are already fighting over. Coordinate it with the mics, not after them.
Words
Captioning is a latency budget
Captioning and surtitling are the same technical problem with different politics: get accurate text on a surface, in time. The engineering constraint is latency, and it is unforgiving in a way pre-recorded captioning is not. A caption that lands two seconds after the line is tolerable. Four seconds and the audience is reading the previous joke while the room laughs at this one, which is worse than no captions at all because it actively removes them from the shared moment.
That budget is spent in three places: the human or machine producing the text, the transport, and the display. A live captioner working from speech is the largest and least compressible share. Automatic speech recognition is faster but wrong in a characteristic way — it fails on proper nouns, technical vocabulary and crosstalk, which on a show is most of what matters. The transport is where a technical team can actually win time, and it is usually the part nobody looks at.
Two delivery choices follow. Open captions are burned into a surface everyone sees, cost nothing per person, and cannot be turned off. Closed captions go to a personal device or a seat-back unit, which serves people who want them without changing the show for anyone else — at the cost of a distribution system, a network, and something for the audience to collect and return.
| Provision | What it actually costs technically | Where it usually fails |
|---|---|---|
| Open captions | A surface, a feed, and a sightline study — the caption screen is scenery and has to be in the plot | Placed where half the house cannot see it, or too far from the action to watch both |
| Closed captions | A network, devices, charging, handout and return | Wi-Fi that was specified for the crew, not for 300 audience devices |
| Audio description | Another wireless channel, a describer position, and a mix that ducks | Coordinated last, so it lands on top of a radio mic |
| Assistive listening | A dedicated speech mix and a distribution system | Fed the main mix, or nobody on duty knows where the receivers are |
Describing
Audio description is a mix problem before it is a script problem
Audio description narrates what is happening visually, on a separate channel, for blind and partially sighted audience members. Technically it is a small live radio station running in parallel with the show: a describer in a position with a clear view, a microphone, a mix, and a distribution path to a receiver in somebody’s hand.
The part that is genuinely an engineering decision is what the listener hears underneath. Description on its own leaves them without the show; the show at full level makes the description unintelligible. So the channel carries a mix — programme plus description, with the programme ducked under the voice — and how that duck is set is the difference between a usable service and an unusable one. It is the same problem as a broadcast voice-over and it wants the same care.
Two practical consequences for a technical team. The describer needs a position with sightlines, light to read by, and comms that will not appear on the channel. And the receivers are more transmitters and receivers in a band that is already contested, so audio description belongs in the frequency coordination from the start rather than being handed a leftover.
The rest of the room
Sightlines, seats and the performance that gets adjusted
Wheelchair spaces are not just floor area: the standards require them to be dispersed rather than clustered, with companion seating, and with sightlines comparable to the rest of the house. That last requirement is where technical departments quietly do damage. A front-of-house mix position, a followspot platform or a delay tower placed after the seating plan was signed off can eat an accessible route or block the only sightline a wheelchair space had, and it will not show up in anybody’s drawings as an access problem.
Relaxed and sensory-adjusted performances are the other common provision, and they are almost entirely a technical brief: a cap on peak SPL, house lights at a low level rather than out, strobes and haze removed or reduced, sudden transitions softened, and a quiet space with a video and audio feed so somebody can leave and still be at the show. None of that is difficult. All of it needs deciding in the plot rather than negotiated in the last hour.
- Technical positions placed last. Mix position, followspot platforms and delay towers get sited after the seating plan and take out accessible routes and sightlines. Put them in the same drawing.
- “Comparable sightlines” measured from an empty room. A standing audience in front of a wheelchair space removes the sightline that existed at the survey.
- Relaxed performances treated as a lighting change. The SPL cap, the haze, the transitions and the relay feed to the quiet space are all separate departments, and all of them need telling.
- Access provision with no rehearsal. The one run where nobody tests the loop, the captions and the description together is the run where they all get discovered at once, in front of an audience.
Where this goes next
The calculators on this page
Flash rate converts a tempo and a division into flashes per second and checks it against the guidance, and tells you the fastest division that still fits. Assistive listening receivers works Table 219.3 for a seat count, including the hearing-aid-compatible column and the induction-loop exception. The standards index has the documents themselves — and unlike most of this site, these are ones somebody may need to cite rather than merely apply.
4 questions on the mechanism rather than the trivia. Nothing is scored and you can answer again — the explanation under a wrong answer is the point of it.
-
1A track runs at 128 BPM and the strobe is programmed on every eighth note. Is that inside the photosensitivity guidance?
-
Colour changes the strictness — saturated red is judged harder — but every colour is subject to the three-per-second limit.
-
Intensity matters to the assessment, but the rate limit is a rate limit. 4.3 a second is over it regardless.
-
The tempo is moderate; the division is what decides it. Every eighth note at 128 BPM is 4.27 flashes a second.
-
Right, and this is the whole problem: on every beat it would be 2.13 and fine. The threshold sits between two completely ordinary programming decisions, which is why it gets crossed by accident.
-
-
2A 900-seat theatre is fitting an induction loop that covers every seat. What does that change about the receiver requirement?
-
The exception is about the hearing-aid compatible column specifically, not a reduction of the total.
-
The total count is, but the second column is not. There is a specific exception for full-coverage loops.
-
Right. A loop couples directly to the telecoil in a hearing aid, so the hearing aids in the room already are the receivers. The overall count still applies for people without a compatible aid.
-
Not everyone in the audience has a telecoil-equipped hearing aid. The main count in Table 219.3 still stands.
-
-
3Live captions on a show are landing about four seconds after the line is spoken. Why is that worse than it sounds?
-
There is no single universal figure of that kind, and the reason this matters is not compliance.
-
Right. The damage is not missing information, it is exclusion from the shared moment, which is the thing the audience actually came for. It is the same reason a few frames of lip-sync error is unbearable.
-
Accuracy and latency trade against each other in the other direction: more time usually buys more accuracy. The cost here is social, not textual.
-
That is an operational nuisance if both exist, but it is not why four seconds is bad for the person reading.
-
-
4Where does an assistive listening system most commonly fail in practice?
-
Right, and by a distance. The receivers are in a cupboard, the batteries are flat, and the duty manager was never shown the panel. That is a technical handover problem, and it is the one most worth fixing.
-
Loop wire is passive and durable. Installed systems rarely fail this way.
-
Level is adjustable and rarely the issue. Being fed the wrong mix — the reverberant PA mix instead of a dry speech mix — is a real problem, but it still comes second to nobody turning it on.
-
It happens, and it is why the system belongs in the frequency coordination. But it is not the commonest failure.
-
Answers are kept in this browser on this device. Nothing is sent anywhere.
- Is this strobe over the line?
- How many receivers do I need?
- Why is a late caption worse than none?
Where to go next
These pages are a chain rather than a library — the next one along usually answers the question this one raised.