Send an Inquiry
Single line drawing of annunciator inputs fed by potential free contacts from two protection relays and one breaker auxiliary switch.

Annunciator Panel Guide: Sequences and Wiring

An annunciator maps each protection trip to a labelled window and one horn. See how ISA-18.1 sequences work, what the contact inputs need and when you need one.

What is an annunciator panel and what does device 30 do?

An annunciator panel gives every abnormal condition one labelled window and one shared horn. Each window is wired to its own field contact, usually a spare output from a protection relay, a breaker auxiliary switch or a thermostat, and the operator reads the label rather than the log. The device carries function number 30 in ANSI/IEEE C37.2-2022. A normal window is dark; a new alarm flashes and sounds the horn; silencing cuts the horn and leaves the window flashing; acknowledging stops the flash and leaves it steady. ANSI/ISA-18.1-1979 (R2004) fixes those four operating sequences, which suppliers label A, B, C and D though two suppliers quoting the same letter may not mean the same behaviour, so ask for the sequence drawing. Immunity requirements follow IEC 61000-4-2:2008, IEC 61000-4-4:2012 and IEC 61000-4-5:2014, with assemblies to IEC 61439-2:2020. A wall unit stays lit when the network is lost, so hardwired windows survive three generations of control-room software. Setting review and quotation within 12 hours, MOQ 1, 7-15 working days.

Annunciator panels give every abnormal condition one labelled window and one shared horn. Each window is wired to its own field contact, usually a spare output from a protection relay, a breaker auxiliary switch or a thermostat. The operator reads the label, not the log. The device itself carries function number 30 in ANSI/IEEE C37.2-2022.

At a glance

  • One window per alarm. Each window carries a separate field contact, and the wiring decides the label.
  • ANSI/ISA-18.1-1979 (R2004) fixes four operating sequences. Ask for the sequence drawing, not the letter.
  • Device number 30 marks the annunciator relay itself in drawings built to ANSI/IEEE C37.2-2022.
  • Inputs are normally potential free. Wet 24 V DC cards will not accept a 110 V DC station contact.
  • Silence stops the horn only. Acknowledge turns flashing into steady, and reset clears the window.
  • Lamp test proves the LEDs and the horn. It says nothing about the contact behind any of them.
  • First-out designs resolve which contact arrived first, often inside a 5 ms scan.
  • A window panel keeps working when the SCADA server does not. That is why it stays on the wall.

What an annunciator actually does

A protection relay decides. A breaker clears. Neither one tells the person standing in the switchroom what just happened, or which of the eleven things on that cubicle caused it. The panel hanging on the wall answers three questions, quickly and without a login: which equipment complained, which condition set it off, and whether anybody has admitted seeing it yet. The window label answers the first two. The difference between a flashing window and a steady one answers the third. None of that needs a processor. That is the whole reason the function has outlived three generations of control room software, and why the oldest units still earning their keep use filament lamps behind a printed legend.

The labels are the product. A window marked “Auxiliary supply fail” saves more minutes than any amount of logging nobody opens afterwards.

Two jobs sit on either side of the panel. Upstream, somebody decides which contacts exist and whether each one is spare, honest and dry. Downstream, somebody has to read the thing at night. Both decisions get made during detail design, usually by different people, and neither survives being deferred to site.

Why installers still fit annunciators next to a SCADA screen

Every new substation ships with an alarm page. Fewer of them have anybody looking at that page when a transformer trips during a night shift. A wall mounted unit beside the switchgear stays lit whether or not the server is running, whether the network switch got patched correctly, or whether the historian filled its disk last Tuesday. During a fault cascade, software will hand you several hundred lines in order of arrival. The panel hands you one label you can act on immediately.

Selecting a relay for a live project?

Send the system voltage, CT ratio and tripping scheme. Our engineers confirm the model and reply within 12 hours.

View relay modelsRequest a quote

The practical resolution most of our own build sheets use looks like this: keep somewhere between 24 and 32 of the highest consequence points hardwired to windows, and let everything else live in software. Breaker trip, protection relay fail, DC supply fail and emergency stop belong on the panel. Trending, thresholds and anything an operator can safely wait two minutes to read belong on the screen.

There is a second reason, less often written down. Regulators and insurers ask what happens if the network is lost. A hardwired window with its own battery supply answers that question in one sentence.

The four ISA-18.1 sequences, described in behaviour rather than letters

Every standard panel walks through the same four states. A normal window is dark. A new alarm flashes and sounds the horn. Silencing cuts the horn and leaves the window flashing. Acknowledging stops the flash and leaves the window steady. Reset clears it once the field contact has gone back to normal. Suppliers package those four states into sequences and label them A, B, C and D, though the letters are vendor shorthand rather than clause numbers. Two suppliers quoting “sequence D” may not mean the same thing. All of it is latching logic in one form or another, because the point of the box is to remember what happened after the contact has stopped saying it.

SequenceWhere it is usedAfter silenceAfter acknowledgeClears when
Sequence AManual reset systems, the default choice for unsupervised plantWindow keeps flashing, horn stopsWindow goes steadyContact returns to normal and reset is pressed
Sequence BAutomatic reset systems on continuously staffed plantWindow keeps flashing, horn stopsNo separate acknowledge stepContact returns to normal on its own
Sequence CFirst-out designs where order of arrival mattersFlashing continues on the earliest windowEarliest window steady, later ones keep flashingEvery contact is normal and reset is pressed
Sequence DFirst-out designs fitted with ringback indicationAs sequence CAs sequence CRingback windows flash again briefly when the condition clears, then go dark

Table 1. The four sequences most suppliers build to ISA-18.1, written as behaviour rather than part numbers.

Ringback is worth asking for on any site where somebody clears an alarm by standing on the contact rather than fixing the cause. Without it, a window can go dark while the condition behind it is still wrong.

Single line drawing of annunciator inputs fed by potential free contacts from two protection relays and one breaker auxiliary switch.
Figure 1. One contact per window. The inputs sit on the same DC bus as the trip circuit supervision.

Annunciator panel hardware: windows, flasher and horn

Strip away the legend strips and there are four things in the box. A card per group of windows decides which contacts closed. A flasher drives every unacknowledged window at somewhere between 60 and 120 flashes per minute. A horn, usually quoted at 80 dB to 90 dB measured at 1 m, does the part the LEDs cannot. And four buttons, silence, acknowledge, reset and lamp test, give the operator the only things they need to touch.

Consumption is not trivial once the window count grows. A filament lamp sits around 1 W to 2 W per lit window, while an LED module runs nearer 20 mA. Across a 64 point faceplate with autumn rain keeping half the bay in alarm, the difference decides whether the 24 V DC supply has any headroom left for the trip circuits it is supposed to serve. A suppressor across the horn coil costs very little and spares the output contact the energy a DC coil dumps back into whatever opened it.

Faceplates come in two shapes that behave quite differently on site. Individual window tiles, commonly around 30 mm square, can be relabelled one at a time when a bay gets added. Compact touch screen units in a 96 mm by 96 mm cut-out hold far more points in far less metal, cost less to wire, and need somebody with a laptop and the configuration file before a single window labels change. Read the nameplate rather than the quotation: it is the figure stamped on the case that the commissioning engineer will hold you to.

State diagram of the four ISA-18.1 annunciator sequences from new alarm through silence, acknowledge and back to normal.
Figure 2. The four sequences differ only in how the window behaves after silence and acknowledge.

Input contacts: potential free, wet, NO and NC

Most protection relays and auxiliary switches hand over a potential free contact, and most cards are built to feed it from their own supply. The contacts themselves come from everywhere: a relay output, a breaker auxiliary switch, a contactor auxiliary, a thermostat or a pressure switch. The expectation matters more than the voltage. A card expecting to wet its own inputs at 24 V DC will not thank you for wiring a 110 V DC station contact into it, and a 250 V DC contact from a larger station will finish that conversation quickly. Input currents are usually small, around 5 mA once the contact closes. A solid-state card draws far less from the station battery than the relay based panels it replaces, which is sometimes the deciding argument on a retrofit.

Whether the field contact is normally open or normally closed is a supervision decision rather than a logic one. A normally closed contact that opens on alarm also covers a broken wire and a blown terminal fuse, so the window lights for a missing signal as well as a bad one. Operators complain about nuisance indications until the day the same arrangement catches a loom nobody had noticed was damaged.

Contact bounce and induced noise both get filtered out, typically over 10 ms to 20 ms. That filter also sets how far apart two events have to be before first-out logic can separate them, which is why a 5 ms resolution claim means very little unless the input filter is quoted alongside it.

Immunity deserves a level rather than a tick box. Cards destined for switchgear are usually asked to meet IEC 61000-4-4:2012 at 2 kV on the signal ports and IEC 61000-4-5:2014 at 2 kV to earth, with IEC 61000-4-2:2008 level 3 covering the electrostatics of somebody walking across a control room carpet. Breaker tripping operations inside the same cubicle produce the interference those input filters have to swallow, so ask what the card was tested against rather than whether it passed anything.

Ratings that decide which model turns up

Line on the datasheetWhat it actually tells youTypical catalogue valueWhat goes wrong if nobody checks
Auxiliary supply rangeHow far the station battery may sag before the logic restarts and forgets everything19 V DC to 60 V DC, or 88 V DC to 275 V DC depending on the cardA low battery alarm that also removes the alarm layer, hiding the event it should have reported
Inputs per cardHow many separate contacts one card accepts, and whether they share a common return8, 16 or 32 per cardShared commons put two unrelated bays on one return, and a single earth fault takes out both
Output contact ratingWhat the horn or the interposing relay is allowed to pullAround 5 A quoted against a utilization category such as AC-15, dropping well below that on DC-13 dutyContacts rated only for AC weld shut on DC duty and leave the horn running until somebody pulls the fuse
Operating temperature rangeBehaviour inside a kiosk that reaches 50 °C in direct summer sun-10 °C to +55 °C on most industrial cardsWindows drifting to look faintly lit, which reads as an alarm that nobody can clear
EMC immunityWhich levels were actually tested, not whether the brochure says compliantIEC 61000-4-4:2012 at 2 kV and IEC 61000-4-5:2014 at 2 kVWriting compliant instead of a level leaves the cheapest card on the table acceptable to the purchaser

Table 2. Five datasheet lines worth arguing about while the panel is still in detail design.

Wiring it to relays and switchgear without creating tomorrow’s fault

Field contacts rarely reach the panel directly. They land first on marshalling terminal blocks, get fused or linked there, and cross to the input card in 1.5 mm² stranded cable with the rest of the cubicle wiring. Route alarm cores separately from anything carrying CT secondaries or VT output. Cross talk between a 1 A CT circuit and a sensitive input filter is one of those faults that only appears at load, in the rain, twice a year.

One contact to one input. It is tempting to loop a single auxiliary contact across two destinations, a common alarm card and the protection relay input block, and save a spare contact. Those two circuits usually sit on different supplies, and the shared return they create will turn a local fault into a panel wide one. Where absolutely no spare contact exists, fit an interposing relay and give each destination its own dry output.

Every loop needs proving before the panel is energised, and every window label needs checking against the current single line rather than the one issued last year. Our own preference is to prove each point by opening the contact at the marshalling terminal rather than at the device, because that is the part of the circuit somebody re-terminated during the outage.

Serial links to SCADA, and what to keep hardwired

Most modern units will also publish their alarm states over Modbus RTU on an RS-485 trunk, over Modbus TCP, or over IEC 60870-5-103 and the IEC 61850 station bus where the substation already runs them. Doing this removes the cost of a second contact for every point and gives the SCADA team the timestamps they want for sequence of events work, typically quoted at 1 ms resolution rather than the 5 ms to 20 ms you get from an input card.

The trap is treating the serial route as a replacement. A gateway, one Ethernet switch and one software driver stand between the operator and that alarm. Keep the points that trip plant or endanger anybody hardwired to a window, whatever the communication philosophy says, and let the serial link carry the copy. The same contacts often feed the site PLC as well, so settle during design which system owns the wording of each alarm. Two systems describing one condition differently is its own kind of fault.

Generator annunciator: the windows every set needs

A genset controller will happily display sixty fault codes on a four line screen. Nobody reads sixty codes while standing next to a running engine with a fault. The set of windows worth paying for is much smaller, and it maps cleanly onto standard device numbers, which makes it easier to reconcile with the single line later.

Window labelWhere the contact comes fromANSI numberTrigger point most sets use
Generator breaker tripAuxiliary contact on the breaker, 52a52Any trip from protection or manual opening
High coolant temperatureThermostat or controller output on the jacket26Commonly 95 °C to 105 °C on water cooled units
Low lube oil pressurePressure switch on the engine gallery63Around 2.0 bar to 2.5 bar at running speed
OverspeedGovernor or tachometer relay output12Usually 110% of rated speed
Low fuel levelLevel switch in the day tank71Set by tank geometry, often 30% remaining
Emergency stop operatedAuxiliary contact on the stop relayn/aContact closure at any stop button
Battery charger failVolt free alarm contact on the chargern/aMonitored on both DC banks where fitted

Table 3. The seven generator points that earn a window, with the device numbers used in ANSI/IEEE C37.2-2022.

Whatever you fit a generator set with, keep those points off the starting circuit supply if there is a separate one. A window that goes dark while the engine is cranking tells you nothing at the exact moment you want to know something.

Nine lines that stop arguments later

  • Number of windows today, and again after the next bay or panel gets added.
  • Whether inputs take potential free contacts or wet ones, and at what voltage.
  • Which sequence you want, written out as behaviour instead of a letter.
  • Whether first-out is needed, and how many milliseconds it has to resolve.
  • How the horn is silenced, how long it stays quiet, and who is allowed to reset.
  • Whether the panel must survive a total loss of AC supply on the station battery.
  • What outputs are needed: common alarm relay, repeat contacts per window, serial map.
  • Label material, language and who signs off the legend drawing before manufacture.
  • The retest period after handover, agreed with whoever carries the maintenance contract.
Faceplate layout of a 16 window annunciator panel with the silence, acknowledge, reset and lamp test buttons along the bottom.
Figure 3. A 16 window faceplate. The bottom row holds the four buttons operators reach for first.

Selecting a model for this scheme? Send the voltage level, CT/VT ratios and your relay list. An engineer returns a model mapping and quotation within 12 hours →

Commissioning checks worth the extra day

  • Prove every window from the marshalling terminal, not from the device, so the cable run is included.
  • Sound the horn, then silence it, then acknowledge, then reset, in that order, on each group.
  • Close two contacts close together and confirm the first-out order comes out the way the drawing says.
  • Run lamp test and count what lights up. An unlit window is either unlabelled or failed.
  • Trip the input supply fuse on purpose and watch what the panel reports about itself.
  • Record the auxiliary current drawn with all windows lit, then again with one bank off.
  • Photograph the legend strip after approval, then compare the as built faceplate against it.

The checks are boring and take about eight hours on a 32 point panel. Every one of them has caught something on the jobs we have run, most often a label that describes the previous incarnation of the drawing.

Failure modes that wait until the first real trip

The dominant failure is not electrical. It is a legend that stopped matching reality after a cubicle was added, so the window that lights up points at equipment nobody has owned for six years. Everything else is ordinary: LEDs fade over about ten years of continuous duty; output contacts weld on DC duties they were never rated for; a card with poor input filtering produces a nuisance trip during every breaker operation on the adjacent panel; first-out logic with a slow scan confidently reports the second alarm as the first. Nobody discovers any of it until something genuinely wrong happens, which is a bad time to learn the panel was lying.

The defence is short and cheap. A weekly lamp test during rounds. An annual loop check with a shorting link at the terminal. A spare card on the shelf. It sounds like nothing, and it is the difference between a panel that reports and a panel that decorates.

Timing diagram of two closing contacts showing how a first-out annunciator locks the earliest alarm and holds the later one behind it.
Figure 4. First-out logic within one scan window. Both contacts closed, only one window claims first.

Where an annunciator panel is the wrong tool

  • You need sub-millisecond sequence of events records for a regulator or an insurer. Buy a dedicated SER device with its own time source instead.
  • The alarm count runs past roughly 64 windows. Past that size nobody scans a faceplate, and an HMI or SCADA list serves readers better.
  • Nobody staffs the panel location for most of the day. A wall mounted window in an unattended kiosk is a decoration, and remote signalling costs the same either way.
  • The site has no DC auxiliary supply and would need a new battery and charger purely to feed indication. Ask whether the AC fed unit with its own backup has already been considered.
  • The signals are really process setpoints that need trending, not discrete contacts. A logger or a protection and control panel with the right I/O will serve you better than a wall of windows.
  • Every point already exists in a certified IEC 61850 system with operator workstations in duplicate. Adding a second alarm layer then creates two sources of truth, and the second one will drift.

Conclusion

Annunciator panels survive every upgrade cycle because they are the one alarm layer that needs neither a server nor a network nor a login. Keep the high consequence points hardwired, write the sequence down as behaviour rather than a letter, and prove every loop from the terminal once a year. The decision that matters is not which unit you buy. It is choosing, during detail design, which twenty or thirty conditions deserve their own window, and being able to explain why the others did not.

Frequently asked questions

What is an annunciator in plain terms?

A panel of labelled windows with a shared horn. Each window is wired to one field contact from protection or plant, so when something abnormal happens the matching window flashes, the horn sounds, and the label tells you which condition complained.

What does an annunciator do that a SCADA alarm list does not?

It works without a server, a network or a screen. That independence is why alarm points stay hardwired in substations that also run full SCADA, and why most designs keep somewhere between 24 and 32 critical points on windows rather than in software.

Why does the window keep flashing after the horn is silenced?

Silence only stops the audible alarm. The flashing state means nobody has acknowledged the condition yet, which is deliberate: it separates an operator who heard the noise from one who has accepted the alarm. Acknowledging turns the window steady.

What voltage do the inputs use?

Most take potential free contacts and supply their own wetting voltage, commonly 24 V DC or 48 V DC. Cards fed from a station battery accept 110 V DC or 220 V DC instead. Expect around 5 mA per contact once closed.

How many windows can one panel carry?

Modular cards usually come in 8, 16 or 32 inputs, and a faceplate can be built out to 64 points or more. Beyond roughly 64 windows the faceplate stops being scannable, which is where an HMI or SCADA list starts serving the operator better.

What does first-out mean on an annunciator?

First-out logic decides which of several near simultaneous contacts arrived first and flags that window. It matters during cascading faults, where the first event identifies the cause. Ask for the input scan time, because a 5 ms scan will merge two contacts that a 20 ms filter separates.

How often should the panel be tested?

Lamp test weekly during rounds, which proves lamps, LEDs and the horn. Full loop proving with a link at the marshalling terminal annually, including the first-out order and what the panel reports about its own supply.

Sources and standards

Five published references sit behind the figures quoted above, along with ANSI/ISA-18.1 for the sequences themselves. Editions are listed because immunity levels and requirements move between them.

SourceWhat it coversTypeEdition
IEC 60255-1:2022Measuring relays and protection equipment, common requirementsStandard2022
ANSI/IEEE C37.2-2022Device numbers and function numbers used in diagrams, including device 30Standard2022
IEC 61439-2:2020Power switchgear and controlgear assemblies built from factory tested partsStandard2020
IEC 61000-4-2:2008Electrostatic discharge immunity testing and levelsTest method2008
IEC 61000-4-4:2012Electrical fast transient burst immunity for signal and supply portsTest method2012
IEC 61000-4-5:2014Surge immunity testing for lines exposed to switching and lightningTest method2014
ANSI/ISA-18.1-1979 (R2004)Annunciator sequences and the behaviour each window is expected to showPractice1979 (R2004)

Table 4. References used above, with the edition each figure was read from.

Window counts, supply ranges and cycle timings quoted in this guide are the ranges our own build sheets ask for on industrial and utility work. They are not limits set by any standard, and the datasheet governing your own equipment always wins.

Related guides: Generator interlock · Electromechanical relay · DCS control system

Tell us the voltage class and the duty, and the reply comes back with a model and a lead time.

Working on a similar scheme? Send the single-line diagram, relay list and CT/VT data. Quote in 12 hours →

About Author
Leon Zhang
Hello, I'm Leon Zhang. I have 15 years of experience in the power relay protection industry with extensive pre-sales and after-sales project experience. Our company specializes in various complete sets of relay protection and automation equipment. I can assist customers in solving all practical on-site project challenges and provide optimal integrated solutions.
Tell Us Your Requirement
Send an Inquiry

High Quality

Stable performance, reliable design, ensuring safe operation for power system protection and grid stability.

Fast Delivery

Timely delivery to support your urgent orders and project schedules efficiently and professionally at any time.

Best Warranty

Professional Warranty: Reliable after-sales support for stable relay protection and long-term customer satisfaction.