Reference

Four different circuits, one connector

A three-pin XLR tells you a cable has a screen and a pair. It tells you nothing about what's on them, and the four things that commonly are have amplitudes three orders of magnitude apart, bandwidths five orders apart, and two of them put DC on the signal pins.

They have very little in common. A microphone output is ten millivolts; a two-wire intercom line is thirty volts of DC. AES3 is a 6 MHz square wave. DMX512 is an EIA-485 bus that the standard says shouldn't be on this connector at all. They all ended up on the same connector for historical reasons, none of them electrical.

Below, each circuit gets its numbers and its source, and then the page computes what happens at every one of the twenty-five combinations. The interesting cells are the ones where the destination pushes something back up the cable at a device that wasn't built to receive it.

The four circuits

Circuit	Connector	Amplitude	Bandwidth	Impedance	Cable	Governed by
MicrophoneAnalogue, balanced	XLR3	0.01 V p-p	20 Hz – 20 kHz	≈150 Ω source into a bridging input of 1.5 kΩ or more	Screened twisted pair; capacitance is the only real limit	IEC 60268-12 / AES14
Line levelAnalogue, balanced	XLR3	3.5 V p-p	20 Hz – 20 kHz	≈100 Ω source into 10 kΩ or more	Screened twisted pair	IEC 60268-12 / AES14
AES3AES/EBU digital audio	XLR3	5 V p-p	DC – 6.1 MHz at 48 kHz frame rate	110 Ω source, 110 Ω terminated	110 Ω ±20% screened twisted pair	AES3 / EBU Tech 3250 / IEC 60958-4
DMX512EIA-485 lighting control	XLR5 / XLR3XLR3 is not permitted	5 V p-p + 2.5 V DC	250 kbit/s, 8N2	120 Ω terminated, 32 unit loads maximum	100–120 Ω, ≤65 pF/m; 120 Ω preferred	ANSI E1.11-2024
IntercomTwo-wire party line	XLR3	1 V p-p + 30 V DC	300 Hz – 5 kHz	200 Ω or 30 kΩ line, depending on the system	Screened twisted pair	Manufacturer convention

Microphone — The only port on this page that deliberately puts 48 V back down the cable at whatever is plugged into it.
Line level — Nominal +4 dBu, clipping around +24 dBu. About 350 times what a microphone sends, and twelve times the most a mic input takes.
AES3 — Biphase-mark coded, which makes it immune to polarity — the one signal here where a reversed pair costs nothing.
DMX512 — An EIA-485 transceiver. Its common-mode range is −7 V to +12 V, and beyond that it is a semiconductor with a rating.
Intercom — Carries its own DC power on pin 2, unbalanced against pin 1. No standard governs it; Clear-Com and RTS differ.

The three-pin pinout

Pin	Analogue audio	AES3
1	Screen / chassis	Cable screen or signal earth
2	Hot (+, in phase)	Signal
3	Cold (−, inverted)	Signal

Pin 2 is hot. AES14-1992, and IEC 60268-12 behind it, settled this: a positive-going pressure at the microphone produces a positive-going voltage at pin 2. Equipment from before the convention took hold, and a certain amount of equipment made since, does the opposite. The failure it causes is the quiet one: a single inverted channel sounds entirely normal until it's summed against a correctly polarised one.

AES3 declines to name either pin hot. EBU Tech 3250 §6.4 lists pin 2 and pin 3 identically as "Signal", because the interface is biphase-mark coded and a polarity reversal is decoded transparently. It's the one signal on this page you can wire backwards for free.

Pin 1 goes to the chassis, not to the audio ground. That's AES48, and it needed a standard because doing it the other way turns every piece of equipment in a rack into a node in a ground loop (the "pin 1 problem"). It's a design rule for the box, but it's why a cable with the screen lifted at one end sometimes fixes a hum that nothing else touches.

AES3, in numbers

Everything here is from EBU Tech 3250, the freely published twin of AES3. It's worth reading the figures next to the analogue ones above: the same cable is being asked to carry a signal with about three hundred times the bandwidth.

Cable impedance110 Ω nominalFrom 0.1 to 128× the frame rate. §6.1

Driver output110 Ω ±20%Balanced. §6.2.1

Signal amplitude2 – 7 V p-pMeasured across 110 Ω. §6.2.2

Receiver termination110 Ω ±20%Substantially resistive. §6.3.1

Receiver eye200 mV minimumAt half the nominal symbol period. §6.3.3

Reach without EQ100 mEqualisation extends it. §6.3.4

The 100 m is a threshold rather than a cliff edge. §6.3.4 says receiver equalisation exists "to enable interconnecting cable longer than 100 m to be used", which is a statement about where unequalised transmission stops being guaranteed. §6.1 is blunter about the family: V.11 balanced circuits allow "signal transmission over distances of up to a few hundred metres".
Microphone cable is the wrong impedance and works anyway, for a while. Typical analogue mic cable is 40–90 Ω against a required 110, so every junction reflects. Over a few metres the reflections arrive inside the eye and nothing happens. Over a long run they arrive where the receiver is trying to sample, and it doesn't fade out gracefully: it's lock, then intermittent lock, then nothing.
AES3id is the same data on 75 Ω coax, at 1 V p-p on BNC, specified for around 1000 m. If a run is long enough to be an argument, this is usually the answer, and the conversion is a transformer and a couple of resistors.

DMX512, from the standard

ANSI E1.11-2024 is published free by ESTA, which makes it one of the few standards on this site I can quote rather than paraphrase. Two of its clauses are more restrictive than common practice, and one is more permissive than most people expect.

The 5-pin XLR — Table 3

Pin	Function	Link
1	Data Link Common	common
2	Data 1−	primary
3	Data 1+	primary
4	Data 2−	secondary
5	Data 2+	secondary

The 8P8C option — Table 4

Pin	T568B	Function
1		Data 1+
2		Data 1−
3		Data 2+
4		Not assigned
5		Not assigned
6		Data 2−
7		Common for Data 1
8		Common for Data 2

Permitted only in fixed installations that are "not normally accessible except to qualified, authorized users" (§7.3). Each data link lands on a true twisted pair, which this page checks at build time. See the Ethernet page for why that matters.

Bit rate250 kbit/s245–255 permitted. Table 6

Slot format8N21 start, 8 data, 2 stop. Table 5

Break92 µs min176 µs typical. Table 6

Mark after break12 µs minReceivers accept from 8 µs. Tables 6, 7

Full-frame refresh44 Hz max22.7 ms for 513 slots. Table 6

Unit loads32Per EIA-485. §5.9

Termination120 Ω +5/−10%Across Data+ and Data−. §4.9

Cable impedance100–120 Ω120 preferred. E1.27-1 §4.4

No DMX standard states a maximum cable length

The figure everyone quotes (a thousand metres, or three hundred, depending on who's quoting) is in neither document. E1.11 §1.1 puts cable outside its scope entirely: "Cable requirements and premises wiring are not within the scope of this Standard." The cable standard that does exist is just as explicit the other way: E1.27-1 §4.2, headed "Maximum and minimum cable lengths", says they are "specifically omitted", because signal quality, device capacitance and the installation environment decide it.

What the standards do give you are the parameters that set the limit, and they're worth having instead: 32 unit loads, 120 Ω terminated, and a cable capacitance ceiling of 65 pF/m between conductors (E1.27-1 §4.5). A long run is a budget across those three.

DMX on a three-pin XLR

It's everywhere. A large share of the moving lights, dimmers, hazers, LED pars and DMX-controlled effects in circulation have a three-pin XLR on the back, and in practice a great deal of DMX is sent down three-pin cable and works.

It isn't to standard, and the standard is unusually direct about it. E1.11 §7.1.1 requires the 5-pin XLR. The concession in §7.1.2 that allows a different connector, available "only when it is physically impossible to mount a 5-pin XLR", then closes the obvious door: the alternate connector "shall not be any type of XLR connector". So a 3-pin DMX port is excluded by name, and a compliant alternate connector has to be marked NCC DMX512-A and ship with an adapter to 5-pin.

Two practical consequences follow, and they're the reason to care beyond pedantry.

You lose the secondary data link. Pins 4 and 5 carry Data 2: a second, independent EIA-485 link in the same cable. Most modern equipment leaves them unused and sends additional universes on separate runs, but some older consoles put a second universe there, Avolites' AVOS-era desks being the example practitioners cite most often. Current Avolites Titan documentation lists pins 4 and 5 as "not used", so treat this as legacy behaviour to check for on old kit rather than something to design around today.
RDM depends on the link being bidirectional, and three-pin infrastructure often isn't. RDM (ANSI E1.20) talks back on the primary pair, so a three-pin cable carries it perfectly well. The problem is everything else on the chain: 3-pin adapters, splitters and buffers are frequently unidirectional, and gear built to a connector standard it's already ignoring isn't reliably built to the transceiver requirements either. If RDM discovery finds nothing through a 3-pin patch, it's usually the splitter rather than the fixture.

I wouldn't re-terminate a working rig over any of this. But it's worth knowing which of your cables are 120 Ω data cable and which are microphone cable, so a long 3-pin DMX run that behaves oddly doesn't come as a surprise.

Check a DMX run against what the standards specify

Since no standard gives you a maximum length, here are the parameters that do exist. Enter your device count and your cable's published figures: the ones on its datasheet, not the ones on the drum.

This calculator needs JavaScript. The specification figures are in the tiles above.

There's no length field, because there's no length to check against: E1.11 §1.1 puts cable outside its scope and E1.27-1 §4.2 leaves run lengths out on purpose. Meeting all three of these doesn't promise a working 300 m run, and failing them doesn't mean a short one won't work. They're the parameters the standards chose to constrain instead.

What happens when you cross them

Row is the output, column is the input. Every cell is computed from the amplitudes, DC levels and input windows declared at the top of this page, including what the input asserts back down the cable, which is where the damage is. Cells on a grey ground can't be made by accident, because the shells don't mate. The cells marked in red beneath the verdict are the ones that can, and only because a three-pin DMX port put a lighting circuit on an audio shell.

output ↓ input →	Microphone	Line level	AES3	DMX512	Intercom
Microphone	Works	Works, with a caveat 0.01 V p-p is well below what this input is scaled for — expect to run out of gain	No useful signal 0.01 V p-p is below this input's 0.2 V p-p floor an analogue audio signal carries no meaning to an AES3 input	No useful signalonly via a 3-pin DMX port, which E1.11 does not permit 0.01 V p-p is below this input's 0.2 V p-p floor an analogue audio signal carries no meaning to a DMX512 input	No useful signal an analogue audio signal carries no meaning to an intercom input
Line level	Overload phantom present — harmless to a balanced analogue output, but switch it off anyway 3.5 V p-p into an input that tops out at 0.3 V p-p	Works	No useful signal an analogue audio signal carries no meaning to an AES3 input	No useful signalonly via a 3-pin DMX port, which E1.11 does not permit an analogue audio signal carries no meaning to a DMX512 input	No useful signal an analogue audio signal carries no meaning to an intercom input
AES3	Risk of damage 48 V phantom arrives at an AES3 output that does not expect it 5 V p-p into an input that tops out at 0.3 V p-p an AES3 signal carries no meaning to an analogue audio input	No useful signal an AES3 signal carries no meaning to an analogue audio input	Works	No useful signalonly via a 3-pin DMX port, which E1.11 does not permit an AES3 signal carries no meaning to a DMX512 input	No useful signal an AES3 signal carries no meaning to an intercom input
DMX512	Risk of damageonly via a 3-pin DMX port, which E1.11 does not permit 48 V phantom arrives at a DMX512 output that does not expect it 5 V p-p into an input that tops out at 0.3 V p-p a DMX512 signal carries no meaning to an analogue audio input	No useful signalonly via a 3-pin DMX port, which E1.11 does not permit a DMX512 signal carries no meaning to an analogue audio input	No useful signalonly via a 3-pin DMX port, which E1.11 does not permit a DMX512 signal carries no meaning to an AES3 input	Works	No useful signalonly via a 3-pin DMX port, which E1.11 does not permit a DMX512 signal carries no meaning to an intercom input
Intercom	Risk of damage 48 V phantom arrives at an intercom output that does not expect it 1 V p-p into an input that tops out at 0.3 V p-p an intercom signal carries no meaning to an analogue audio input	No useful signal an intercom signal carries no meaning to an analogue audio input	Risk of damage 30 V DC on the signal pins, into an input rated for 10 V an intercom signal carries no meaning to an AES3 input	Risk of damageonly via a 3-pin DMX port, which E1.11 does not permit 30 V DC on the signal pins, into an input rated for 12 V an intercom signal carries no meaning to a DMX512 input	Works

The combinations that actually damage something

These are the cells the engine flags as damaging and that share a connector shell, so nothing mechanical stops them happening:

AES3 output into a microphone input. 48 V phantom arrives at an AES3 output that does not expect it; 5 V p-p into an input that tops out at 0.3 V p-p; an AES3 signal carries no meaning to an analogue audio input.
DMX512 output into a microphone input. 48 V phantom arrives at a DMX512 output that does not expect it; 5 V p-p into an input that tops out at 0.3 V p-p; a DMX512 signal carries no meaning to an analogue audio input.
Intercom output into a microphone input. 48 V phantom arrives at an intercom output that does not expect it; 1 V p-p into an input that tops out at 0.3 V p-p; an intercom signal carries no meaning to an analogue audio input.
Intercom output into an AES3 input. 30 V DC on the signal pins, into an input rated for 10 V; an intercom signal carries no meaning to an AES3 input.
Intercom output into a DMX512 input. 30 V DC on the signal pins, into an input rated for 12 V; an intercom signal carries no meaning to a DMX512 input.

Most of these are harmless in one direction and expensive in the other, and the dangerous direction is usually the one where the input is a microphone preamp with its phantom switch left on from the last job.

Sources

DMX512 — connectors, timing, unit loads, termination — ANSI E1.11-2024, Entertainment Technology — USITT DMX512-A, published free by ESTA and read directly. Clauses cited inline.
DMX512 cable — impedance, capacitance, and the omitted length — ANSI E1.27-1-2006, Standard for Portable Control Cables. The copy read was the R2016 reaffirmation; ESTA currently lists it as 2006 (R2021), a reaffirmation without substantive changes.
AES3 electrical specification — EBU Tech 3250, Specification of the digital audio interface, the freely published twin of AES3. Section numbers cited inline. AES3 itself is paywalled.
Pin 2 hot — AES14-1992 (stabilised 2019), which codifies IEC 60268-12 for XLR polarity and gender. Pin 1 to chassis is AES48.
Phantom power — IEC 61938 defines P12, P24 and P48. The 6.81 kΩ feed resistors and the pin-1 return are the P48 case.
RDM — ANSI E1.20; ESTA currently lists the 2025 revision. The claim that AVOS-era Avolites consoles output a second universe on pins 4 and 5 is practitioner report, not documentation. Current Titan documentation lists those pins as not used, and I couldn't find a manual confirming the older behaviour.

Assembled 7 September 2026 with AI assistance, from the standards linked above. The pin tables are transcribed from E1.11-2024 and checked at build time against the twisted-pair invariant; the compatibility matrix is computed from declared electrical figures rather than typed. Those figures are nominal working values, not absolute maximum ratings. The matrix is a guide to which combinations are wrong by an order of magnitude, and nothing on this page has been tested by deliberately mis-plugging hardware. Before trusting a cell, read the datasheet for the device in front of you.

Companion pages: Ethernet cabling for the twisted-pair and screening background behind the 8P8C option, and what is inside a four-channel CAT5 stagebox for analogue audio on the same eight conductors.