The Video Doorbell That Kept Dropping Offline

Why this matters

A video doorbell the dealer installed as part of a security package went offline most evenings and came back by morning. The customer missed two deliveries and a visitor, and the alarm app showed the doorbell as "offline" in red every night, which made the whole system look unreliable. Video doorbells fail for two main reasons, power and Wi-Fi, and both can produce the same red "offline". The tech had to tell them apart with readings, at the hour the fault happened, before buying anything.

The pattern

The doorbell was on a brick-fronted two-story house, mounted on the brick beside the front door, wired to the existing doorbell transformer and chime. The app's device history showed the same shape for nine of the previous fourteen days: connected all day, then offline in blocks starting somewhere between 19:00 and 20:30, reconnecting for a few minutes, offline again, and steady from about 23:00. Weekends were worse than weekdays.

The homeowner's router was the internet provider's combined modem and router, in an upstairs back bedroom used as an office, about as far from the front door as the house allowed.

Three candidates

Power. A video doorbell runs from the low-voltage doorbell transformer, typically through the chime, and needs a minimum voltage and enough capacity at its terminals, which its maker states. An undersized or failing transformer can let the voltage sag when the doorbell's heater, IR or radio load peaks, and the doorbell reboots. See related: Doorbell Transformers and Why Video Doorbells Fail.

Wi-Fi coverage. Brick and the distance to a back-of-house router weaken the signal at the door. A weak link is fragile, and something else happening in the evening can push it over.

The device itself. A faulty radio or a firmware fault in the doorbell.

The evening timing fitted both power and Wi-Fi. Evenings are when the outdoor temperature drops, when people stream video, and when neighbors are home using their own networks.

Readings in the afternoon

The tech arrived at 15:00, when the doorbell was online.

Power at the doorbell. The transformer was in the basement next to the panel, labeled 16 V 30 VA on its nameplate. Before opening anything he checked the doorbell maker's requirement, which for this model called for a voltage range and transformer capacity the nameplate met. He removed the doorbell from its bracket to reach the terminals, leaving the low-voltage wires connected, and measured AC voltage across the terminals.

• Idle: 17.8 VAC.
• With the button held, chime sounding: 16.9 VAC.
• With live view running from the app at the same time: 16.7 VAC.

All three readings sat inside the maker's range. The sag under load was under 1.2 V, which is normal for the resistance of the old doorbell wire plus the chime. The transformer's 120 V primary was hard-wired into a junction box; he did not open it, because nothing on the low-voltage side pointed at it, and work on the line side goes to an electrician where local rules require it.

The device's own uptime. The doorbell's diagnostic page in the app showed uptime since its last restart: 11 days. It had gone offline on several evenings in those 11 days without restarting once. A power sag low enough to drop a doorbell's radio usually resets the whole device and restarts that counter. This one had stayed powered through every offline episode. That moved power well down the list without condemning it entirely, and it put Wi-Fi at the top.

Wi-Fi at the door, measured where the device lives. The doorbell reported its own received signal strength (RSSI, measured in dBm, where values nearer zero are stronger) at -73 dBm. The tech confirmed it with a Wi-Fi analyzer on his phone held against the brick at the doorbell's position, not standing in the doorway or inside the house: -72 to -75 dBm on the household network, 2.4 GHz band. See related: Measuring Wi-Fi Signal Where the Equipment Lives.

That is weak. Many device makers give a minimum somewhere in the high minus sixties to low minus seventies for reliable video, and a doorbell pushing live video sits at the demanding end. Standing in the open doorway the same phone read -64 dBm, and inside the front hall -61. The brick wall and the steel-reinforced door frame were costing roughly 12 dB between the hall (-61) and the doorbell's position (-73).

Readings in the evening

The tech came back at 20:00 the same day, because a weak signal explains why the link is fragile and not why it fails at night.

 FRONT DOOR, DOORBELL POSITION, SAME DAY

                   15:00         20:00
 doorbell RSSI     -73 dBm       -74 dBm
 phone analyzer    -72 to -75    -73 to -76
 nearby networks
   on same channel 3 visible     7 visible
 ping to router,
   100 packets     0 lost        9 lost
 round trip,
   worst packet    38 ms         410 ms
 doorbell state    online        offline twice
                                 in 30 minutes

The signal level barely moved. -73 in the afternoon, -74 in the evening. A common assumption is that the household streaming "uses up the signal". It does not. Received signal strength is a property of the radio path between the router and the door, and that path had not changed.

What changed was everything sharing the channel. At 20:00 three televisions in the house were streaming, and the analyzer showed seven other networks on the same 2.4 GHz channel compared with three at 15:00. Wi-Fi devices on one channel take turns to transmit. A device with a weak signal has to use the slowest, most robust data rates, so each of its transmissions occupies the channel far longer than the same data would from a strong device, and it needs more retries when the channel is busy. The doorbell's small but constant traffic, keepalives to its cloud service and video when anyone opened live view, started losing packets and timing out. The ping test from the doorbell's position showed it: no loss in the afternoon, 9 of 100 packets lost in the evening, and worst-case delay up from 38 ms to 410 ms. Enough of that and the doorbell's cloud service marks it offline until the link settles.

So the doorbell was not dropping because evening made the signal weaker. It was dropping because a link with no margin cannot absorb a busy channel, and evenings made the channel busy.

The device itself. The doorbell was current on firmware. The tech held his phone, connected to the same network, at the doorbell position through a full evening outage: the phone's own connection also stalled and recovered at the same moments the doorbell went offline. A faulty doorbell radio would not stall a phone held beside it.

Options and why one was chosen

Move the router. It belonged to the internet provider and sat where the service entered the house. Moving it meant a new cable run from the entry point, and it would only have moved the weak spot to the back rooms the family also used.

A range extender plugged in near the front door. It would raise the signal at the door, but a wireless extender repeats over the same weak path it is meant to fix, and it doubles the airtime each packet uses on the channel that was already the problem.

An access point near the door, wired back to the router. An access point with a cable back to the router delivers a strong signal at the door and carries the traffic home over the cable, not over the air. The house had a coat closet off the front hall and a basement directly beneath it, where the tech could run a cable up from the router's location along the basement ceiling.

The customer agreed to the access point. It was set up on the existing network name and password so nothing else in the house needed changing. Home network design beyond that one placement is its own subject and was not part of the visit.

The finish test

With the access point mounted high in the closet, about 12 ft from the doorbell through the front wall:

• Doorbell RSSI: -51 dBm, up from -73 dBm.
• Analyzer at the doorbell position: -49 to -53 dBm.
• Evening ping test from the doorbell position, 100 packets at 20:30 with the same streaming load: 0 lost, worst round trip 24 ms.
• Live view opened from the app ten times between 20:00 and 21:00: connected every time within a few seconds.

He left the evening readings and the after-readings on the ticket, and asked the office to pull the doorbell's connection history after two weeks. It showed no offline events in 14 days, against nine in the 14 days before the visit.

What made the difference in the diagnosis

The afternoon readings alone pointed at Wi-Fi but could not prove it, because the doorbell was online then and the signal level, though weak, was the same as it would be at night. The uptime counter made the power path unlikely without anyone needing to be present for a sag. The evening readings showed the actual mechanism: same signal, busier channel, lost packets. Had the tech replaced the transformer on the strength of the timing, the doorbell would have kept dropping off every evening, and the customer would have had a second visit and less reason to trust the third.

What the install visit could have caught

The doorbell went in on the last afternoon of a full alarm install, and nobody read the signal at the mounting position before drilling the brick. The dealer's install procedure now puts that reading first: hold a phone analyzer against the wall where the doorbell will sit, record the reading on the ticket, and stop to talk to the customer when it is weaker than the doorbell maker's stated minimum. The conversation then happens before the install, with a quoted access point or a different plan, rather than after a month of red "offline" in the app. See related: How to Install a Video Doorbell on a Security Job.

One reading at install time would not have shown the evening problem, but it would have shown a link with no margin, and a link with no margin is the one that fails at the busiest hour of the house.

References

• See related: Doorbell Transformers and Why Video Doorbells Fail
• See related: Measuring Wi-Fi Signal Where the Equipment Lives
• See related: How to Install a Video Doorbell on a Security Job
• Doorbell manufacturer installation guide for voltage, transformer capacity and minimum signal