USB Devices That Disconnect

Suspect power before suspecting the device. Selective suspend, an unpowered hub and a marginal cable account for most intermittent USB disconnections, and all three cost nothing to rule out.

A note on this guide: this rests on how USB power delivery and power management are specified, rather than on faults reproduced here. See how we test.

The problem, as it appears

A device works, then stops, then works again. A dongle drops for a second mid-session. An external drive vanishes and reappears. A headset re-enumerates with the little sound the operating system makes.

Nothing has changed, and only one device is affected — which is exactly why people conclude that device is broken.

Why "the device is faulty" disappoints

Because devices differ in how much they draw and how gracefully they resume, so a shared problem surfaces on whichever one is least tolerant.

If a port is entering a low-power state, or supplying less than a device wants, the first thing to fail is the device closest to the margin. Everything else on the same machine behaves perfectly, which reads as proof that the machine is fine and the device is not.

It is the same reasoning error as blaming a component in a machine that will not POST — the visible symptom points at the part, and the cause is upstream of it.

What actually matters

Selective suspend, first

The operating system can put idle USB ports into a low-power state to save energy. On a laptop that is sensible. On a desktop it is the leading suspect for intermittent disconnections, because a device that resumes badly from that state presents exactly as a failing device.

Turning it off is free and reversible, and it is the single highest-value thing to try. On a desktop the power saved is negligible; on a laptop, decide deliberately.

Power budget, and what a hub actually does

A USB port supplies a defined amount of current. An unpowered hub divides that budget among everything connected to it — it does not create more.

So a hub that worked with three devices can misbehave with four, and the failure is intermittent because draw varies with what each device is doing. A powered hub brings its own supply and removes the constraint.

How far one hub stretches is covered in how many peripherals one USB hub handles.

This page contains affiliate links. If you buy through them, Ugly Fixes may earn a commission at no extra cost to you. Products are selected independently.

Cables are the cheapest thing to eliminate

Long, thin or damaged cables drop voltage along their length, and a device operating near its margin notices.

Swap for a short, known-good cable and retest. If the problem goes, you have the answer for the price of nothing — and cables are the component most often chosen on length rather than quality.

Move it to a rear port

Front-panel ports reach the board through an internal header and a length of cable, adding connections and distance compared with a port soldered directly to the board.

Testing the same device on a rear port separates a panel problem from a device problem in about ten seconds, and it is worth doing before anything is ordered.

Position matters for wireless receivers

If the device is a 2.4 GHz receiver rather than a wired peripheral, "disconnecting" may be interference rather than power — a receiver behind a metal case, near the machine's own electronics, drops packets.

An extension that brings it out to the desk fixes a surprising share of these, and the distinction is drawn in wired vs wireless peripheral latency.

External drives deserve more caution

A drive that disconnects mid-write risks the data, not just your patience.

Stop using it for anything that matters until the cause is identified, and rule out power and cabling before concluding the drive is failing — the symptoms overlap with a drive that has started clicking, and the responses are completely different.

The order to work through

1. Turn off selective suspend and retest. Free.
2. Swap the cable for a short known-good one. Free if you have one.
3. Move to a rear port, directly on the board. Free.
4. Remove the hub from the picture entirely, or replace it with a powered one.
5. For a receiver, get it out to the desk on an extension.
6. Then suspect the device — and by this point you have a much better case.

The mistake to avoid

Replacing the device first. It is the most obvious action, and a replacement plugged into the same port, on the same cable, through the same hub, has a good chance of behaving identically — at which point the new device gets blamed too. Every step above is cheaper and more informative.

The second mistake is leaving selective suspend enabled on a desktop out of caution. The energy saved on a machine that is plugged into the wall is negligible, and the setting is the single most common cause of the symptom this page describes. On a laptop the trade is real; on a desktop it is not much of a trade at all.