The Addition Alarms That Did Not Talk to the House

Why this matters

A bedroom addition passed its final inspection with two new smoke alarms that worked perfectly, with each other. They had never been joined to the three alarms in the original house, and seven months later a homeowner pressing the hall test button found out the way nobody should: her daughter, asleep in the new room, heard nothing. The first fix was correct and the problem came back anyway, which is the part of this case worth reading closely. It also explains why five trades and an inspector all walked past it.

What the homeowner found

In November, on the weekend the clocks changed, the homeowner did what the fire service tells everyone to do and pressed the test button on the hallway alarm in the original house. The hall unit and both bedroom units in the original house sounded. Her daughter came out of the addition a minute later and asked what the noise was. In the addition, nothing had sounded.

She pressed the test button in the addition's bedroom. Both addition units sounded. Nothing sounded in the original house.

Two groups of alarms, each interconnected within itself, and not connected to each other. She called the general contractor who had built the addition.

The layout

 ORIGINAL HOUSE                  ADDITION (new)
 H1 hall --- B1 bed --- B2 bed   A1 bedroom --- A2 hall
   |                               |
   | hall box: interconnect        | addition circuit
   | conductor from addition       |
   | arrives here, capped  <-------+  run through attic
   |  (never landed)                  above the hall
 group 1: sounds together        group 2: sounds together
 nothing crosses between the groups

The addition was a bedroom and a short hall built off the original hallway, with its own branch circuit. Its two alarms had been wired with a three-conductor cable carrying power, neutral and the interconnect conductor, which is the extra wire that carries the "sound now" signal from one alarm to the rest.

The first diagnosis, and why it was right

The contractor sent the electrician who had wired the addition. He took down H1 and opened the hall box. The interconnect conductor from the addition's run was in the box, coiled behind the house wiring with a wire nut on its end. It had never been connected to anything.

He remembered why. At rough-in he had pulled the cable from A2 across the attic to the hall box, planning to land the interconnect at trim. On trim day he found the original house alarms were an older line from a different maker than the units the contractor had supplied for the addition. Mixing makers on a hardwired interconnect is something manufacturers generally tell you not to do, because the signal on that wire is not standardized between them. He capped the conductor, meant to raise it with the contractor, and moved on to the next job on his list. Nothing on paper recorded it.

So the diagnosis was simple and correct: the addition was not connected to the house because the wire that would connect it was capped in a box. He landed the interconnect conductor to the house's interconnect splice, put H1 back, and tested from the hall. All five units sounded. He tested from A1. All five sounded. He left.

It came back

Three weeks later the homeowner, now in the habit, tested again from the hall. H1, B1 and B2 sounded. A1 and A2 did not.

She tested from A1. All five sounded.

So the connection now worked in one direction and not the other, and not even reliably in the direction that had worked on the day it was fixed. The electrician's landing was not loose; he checked the splice and it was tight. The first diagnosis had been right: the wire was capped, and connecting it was necessary. It was not sufficient, because the problem he had recognized on trim day and then set aside was still there. The two groups were different makers' products on one interconnect line, and nothing guaranteed that one maker's alarm would recognize the other's signal.

The behavior is what the manufacturers' instructions warn about in general terms: interconnecting alarms outside the maker's compatibility list gives results nobody can predict, which is why the list exists. A test that passes on the day is not evidence of compatibility, because it only proves that one press, at that moment, worked.

The four days in between

The replacement units had to be ordered, and they took four days. The contractor did not leave the family to guess what that meant. He told them in writing that, until the new alarms were in, a fire in the original house might not sound in the addition and a fire in the addition might not sound in the house, and asked them to sleep with the addition's bedroom door open and the hall between the two parts of the house clear. He also left a battery-only smoke alarm on a shelf in the addition's short hall, tested, with a note that it was not connected to anything and was there only to give the new room one more sensor nearer the old house. That is a stopgap, not a repair, and he said so; it was removed when the new units went in.

The fix that held

The fix was to make the house one system. The contractor had the electrician replace the three original units, all of which were within a year of the 10-year replacement date printed on their backs anyway, with the same maker and line as the addition's units, using the harness adapter that maker lists for its own older bases. All five units were now one family on one interconnect, which is the condition the instructions require, and the electrician confirmed in that maker's instructions that interconnected units on two branch circuits were permitted, which not every maker allows.

Then the walk-test, which is the thing that had been missing from every visit before it. Two people: one presses, the other confirms every other unit is sounding, starting with the one farthest from the unit being pressed. Every unit is pressed in turn, and every other unit is confirmed sounding.

INTERCONNECT WALK-TEST       date ______  by ______ + ______
PRESSED  H1   B1   B2   A1   A2   all sounded?
H1       x    Y    Y    Y    Y    Y
B1       Y    x    Y    Y    Y    Y
B2       Y    Y    x    Y    Y    Y
A1       Y    Y    Y    x    Y    Y
A2       Y    Y    Y    Y    x    Y
5 presses, 20 cross-checks, 20 Y

Five units pressed, four others checked on each press, twenty checks, twenty passes. The homeowner repeated the walk-test twice more over the next two months, pressing each unit in turn. Every other unit sounded on every press.

Why it slipped past everyone

This is the useful part, because no single person did something obviously wrong.

• The electrician saw the incompatibility and capped the wire instead of raising it. That was a reasonable instinct, not landing a connection he expected to misbehave, followed by the miss: no note on the work order, no call to the contractor, nothing anyone downstream could read.
• The general contractor bought the addition's alarms without checking what was in the house. A product decision made in the supply aisle set up a compatibility problem in the wall.
• The drywall and paint crews never touched the hall box. Nothing about their work involved the interconnect, and nothing told them to look.
• The inspector at final pressed A1's test button and heard A1 and A2 sound together. That is a real test of the addition's own interconnect, and it passed. Nobody pressed H1 while listening in the addition. The adopted code reached the original house too: under IRC R314.2.2, in the 2021 edition as adopted, an addition requiring a permit brings the existing dwelling's alarms into scope, in substance, including interconnection where it can be done without removing finishes, and here the attic over the hall made it easy. See related: Smoke and CO Alarm Requirements Your Permit Triggers.
• The first fix tested once in each direction, and both passed. A single press per direction on a mixed system proved the wire was landed and nothing more.

Every one of those visits included a test. None included the test that crosses between the old alarms and the new ones, in both directions, and that is the only one that answers the question an addition raises.

What the contractor does at final now

On every addition, and every remodel that adds or relocates an alarm, the contractor's final punch list carries the walk-test card above, filled in, before calling for inspection. Three rules sit on the card:

• List every alarm in the dwelling, not just the new ones. The test covers old and new together, because the question is whether the house is one system.
• Check the makers before rough-in, not at trim. If the existing alarms are a different maker or close to their replacement date, the contractor decides at the estimate whether to replace them, so the decision is priced and owned, not discovered on trim day.
• A capped interconnect conductor is a written finding. Any trade that leaves one capped writes it on the work order with the reason and calls the contractor the same day.

The card takes two people about ten minutes on a five-alarm house. CO alarms, where the house has interconnected ones, go on a second line of the same card. See related: How to Test a Home's Alarms Correctly.

References

• International Residential Code R314 (smoke alarms), including R314.2.2 for alterations, repairs and additions, 2021 edition as adopted by your jurisdiction.
• Smoke alarm manufacturer's installation instructions, for interconnect compatibility and the maximum number of interconnected units.
• See related: Hardwired, Interconnected and Battery Backup: What Each Word Means
• See related: Reading the Manufacture Date and the Ten-Year Replacement Rule
• See related: The Smoke and CO Alarm Disturb-and-Restore SOP