Reference
Things worth writing down once
Formats, protocols and pinouts that took me a day to pin down, written up so they don't cost anyone else a second one. Each is cross-checked against primary documentation where any exists, and tells you when none does.
Two lines, two limits
LED processor and power lines
"How many panels per line?" is two questions sharing a word: one processor port, limited by bandwidth, and one power circuit, limited by current. Pixels per port computed with power-of-two containers rather than 3 x bit depth, which reproduces NovaStar's three published MX40 Pro figures to the pixel (the naive formula can't fit them with any one constant). Why LED refresh rate costs no bandwidth at all, panels per circuit from watts and volts, and what the one manufacturer table that publishes the power figure turns out to encode.
- Pixel containers
- NovaStar
- Brompton Tessera
- Port capacity
- Power distribution
- NEC 210.20(A)
What reaches the back row
Inverse square law and air absorption
A point source loses 6 dB per doubling of distance at every frequency, and a line array only 3, out to a border distance that depends on frequency and on the square of the array's height. Then the air takes its own frequency-dependent share on top, which at 8 kHz can exceed 100 dB per kilometre and peaks in dry air, not damp. Both are computed (ISO 9613-1 in full, checked against the published table) and combined in a calculator for your distance and your weather.
- Inverse square law
- Line arrays
- ISO 9613-1
- Air absorption
- Wavefront Sculpture Technology
- Delay towers
A sign, an angle, and milliseconds
Polarity, phase and time alignment
Three words people use as if they meant the same thing, for three different things. Polarity is a sign, phase is an angle at one frequency, and a time offset is a different angle at every frequency, which is why a delay produces a comb filter and a reversed cable doesn't. Speed of sound by temperature and what a cooling evening does to a delay tower, comb depth by level difference, subwoofer and fill alignment, and the precedence effect with Haas's own numbers rather than the folklore.
- Polarity
- Comb filtering
- Haas effect
- Precedence effect
- Speed of sound
- Delay alignment
ISO 226, and the level knob
Equal-loudness contours
The ear's sensitivity depends on frequency and on level, and you can't separate the two: a 40 phon contour needs 38 dB more at 50 Hz than at 1 kHz and a 90 phon contour needs 18, so the level a mix is played at is part of the mix. ISO 226 computed from the published formula rather than traced, marked where the standard's data stops and a PA keeps going, and A- and C-weighting checked against IEC 61672-1.
- ISO 226
- Fletcher–Munson
- Phon
- A-weighting
- C-weighting
- IEC 61672-1
- Contourtonist
Series, parallel, and 100 V lines
Loudspeaker impedance
An 8 Ω loudspeaker is allowed to measure 6.4, a multimeter reads lower still, and a daisy chain is a parallel circuit whatever the connector implies. Series and parallel loads with every total computed and checked, how unequal loads share power (opposite rules in series and in parallel), what halving the load does to amplifier current, and constant-voltage lines, including why a 70 V tap on a 100 V line draws exactly twice its rating.
- IEC 60268-5
- Series and parallel
- Daisy chains
- Minimum load
- 100 V line
- 70 V line
Length, conductor size and temperature
Cable voltage drop
One formula, ρ·L/A, reproduces the AWG table from its definition, IEC 60228's conductor resistances, and BS 7671's mV/A/m volt-drop column to within 3 %. What falls out on the way: flexible cable is 11–16 % worse than pure copper of its nominal size, a conductor at 70 °C is 20 % worse than at 20, and the regulation's figures assume both without saying so. Tables for mains and DC runs, and a calculator that also finds the size for 3 %.
- IEC 60028
- IEC 60228
- BS 7671
- AWG
- Volt drop
- Temperature coefficient
Categories, screening and pinouts
Ethernet cabling
A category is a bandwidth rating verified over 100 metres, not a speed, which is why 10GBASE-T runs 37 metres on Cat6 and Category 7 isn't a TIA category at all. The categories, the ISO/IEC screening grammar behind U/UTP and S/FTP, T568A and T568B, PoE, and a computed matrix of how far every BASE-T PHY runs over every cable.
- ANSI/TIA-568.2-D
- ISO/IEC 11801
- 10GBASE-T
- T568A / T568B
- Power over Ethernet
- IEEE 802.3bt
One connector, four circuits
XLR, AES3 and DMX
The same three-pin shell carries microphone level, line level, AES3 digital audio, two-wire intercom at thirty volts of DC and, against the standard's explicit wording, DMX512. Pinouts, cable specifications and levels for each, from ANSI E1.11-2024, ANSI E1.27-1 and EBU Tech 3250, with a computed matrix of what happens when you cross them.
- ANSI E1.11-2024
- DMX512-A
- AES3 / AES-EBU
- AES14
- Phantom power
- RDM
Party line, matrix and hybrids
Intercom: 2-wire and 4-wire
Both names are wrong as counts: a “2-wire” party line runs on three conductors and carries the DC that powers every pack, and “4-wire” is two pairs with no power at all. How a distributed-amplifier party line works, the three incompatible ways Clear-Com, Audiocom and RTS TW wire the same 3-pin XLR (any one of which powers up and passes some audio on either of the others, which is worse), hybrids and nulling, and computed power and cable-reach budgets.
- Clear-Com
- RTS TW
- Audiocom
- Party line
- 4-wire hybrid
- Nulling
SD to 24G, and how far it goes
SDI over coax
No SMPTE standard states a maximum cable length. Reach is a receiver loss budget divided by cable attenuation at half the clock frequency, which is why the same coax gets published at 78 metres and at 155 metres for the same format, and both are right. The formats, the real serial rates behind the marketing names, and a computed reach table that reproduces both of those figures.
- SMPTE ST 2082
- 12G-SDI
- 3G-SDI Level A/B
- 75 Ω coax
- Belden 1694A
- Loss budget
Link rates and what they carry
HDMI, DisplayPort and DVI
HDMI 2.1 is 48 Gb/s and DisplayPort 2.1 is 80, and neither figure is video bandwidth, because the line code takes a fifth, a ninth or a thirty-third before a pixel gets carried. Every version from DVI to HDMI 2.2 and UHBR20, with a fit table computed from pixel clocks and coding efficiency, plus why a cable can't be “HDMI 2.1” and why none of these specifications states a length.
- HDMI 2.1 FRL
- DisplayPort UHBR20
- DVI dual link
- Display Stream Compression
- DP++ dual-mode
- HDCP
Genlock and reference signals
Black burst, tri-level and TTL sync
A genlock receiver has one job, which is deciding when an edge arrived. Noise moves the signal vertically, and the slew rate at the decision point sets what that costs in nanoseconds. That's the case for tri-level sync, worked through with numbers here rather than taken on trust. Levels and standards for bi-level, tri-level and TTL, and the termination rules that decide whether any of it locks.
- Tri-level sync
- Black burst
- SMPTE ST 318
- Genlock
- 75 Ω termination
- SMPTE ST 2059 / PTP
Socapex, and the audio multipins
Multipin looms
A 19-pin Socapex carries six three-conductor circuits: six 16 A mains feeds on a lighting truck, or a loudspeaker loom fanned to NL4s on an audio one, and the two mate perfectly with each other. The pinout, validated at build time because a wrong mains table is a hazard, computed volt-drop and damping-factor figures by loom length, and why EDAC, VEAM and Harting audio multipins get no pinout here at all.
- Socapex 19-pin
- Speaker looms
- EDAC / Elco
- Harting Han
- Volt drop
- Damping factor
Colour codes for lengths
Cable length marking
No standards body has ever published a scheme for marking the length of a production cable. Two conventions are worth knowing: the resistor colour code, which encodes the number so anyone technical can read it, and one band per length, which is faster but has to be memorised. Band tables generated and round-trip verified, plus the practical business of making a mark that survives a season.
- IEC 60062
- Colour codes
- Cable management
- Heat shrink
- Neutrik rings
FFGL, OpenFX and Pixera RenderBridge
Video plugin APIs
Three ways to put an effect inside somebody else's video application: the interface, the lifecycle, the parameter model and the bundle layout of each, written so someone who already knows C++ and GLSL can start on Monday. Plus the traps: the 16-character truncation nothing plugin-side notices, the clock unit no host declares, and the teardown crash that lands at PC=0 after a correct render.
- FFGL
- OpenFX
- Pixera RenderBridge
- Resolume
- DaVinci Resolve
- OpenGL
VST 3 and Audio Units
Audio plugin APIs
The two formats that carry pretty much all of it: Steinberg's VST 3 and Apple's Audio Units v2 and v3. Interfaces, lifecycle, the processor/controller split, the pull render model, parameter identity, bundle layout and where hosts actually scan. The realtime rules come first, because getting a plugin to load is a day's work, and getting it not to click under load is the job.
- VST 3
- Audio Units
- AUv3
- JUCE
- pluginval
- auval
OSC control for Analog Way switchers
The LivePremier OSC dictionary
Every OSC address for driving an Analog Way LivePremier (Aquilon) switcher: takes, memory recalls on every bank, and the whole mappable parameter surface of a layer. Generated from the implementation rather than typed out by hand, with the rules that decide what an argument means and why a recall never reaches air by accident.
- Open Sound Control
- Analog Way LivePremier
- Aquilon
- AWJ protocol
- TouchOSC
- QLab
Analogue audio over Cat5
CAT5-to-XLR pinouts
Four balanced channels, four twisted pairs, and nine mutually incompatible ways to map one onto the other. What's inside a Radial Catapult, a SoundTools CAT Box, a Whirlwind Catdusa, an sssnake Cat Snake and Neutrik's AES72 stagebox, with a computed matrix of what comes out when you mix them.
- AES72-2019
- Radial Catapult
- SoundTools CAT Box
- Whirlwind Catdusa
- Neutrik NA-4I4O-AES72