Mouse Dead After Sleep: A Power Drop Windows Cannot See or Prevent
Device Manager said everything was healthy, the touchpad worked, and the mouse was completely dead. The receiver loses power during standby and its firmware stops listening for the mouse — and that happens below the Windows power-management layer, which can neither observe it nor stop it. Full forensics, the trigger design that actually works, hardware-level fixes, and the false lead I chased first.
My G304 wireless mouse went dead again after the laptop had been idle. The mouse had power, the receiver was plugged in, the touchpad worked fine — but the mouse produced no motion and no clicks.
I have now fixed this twice, and both times I was sure I had found the root cause. The first was wrong, and the second was only half right. This is the final version, including both of the false leads.
1. Why the symptom is so confusing
| What you see | What is actually happening |
|---|---|
Device Manager shows the mouse as OK, no warning icon | The device was never disconnected at the USB level — it just stops sending any HID report |
| The touchpad works, the rest of the system is fine | The USB controller and the OS are healthy |
| Replugging the receiver fixes it | A re-enumeration is a USB reset of the receiver — the firmware restarts |
| Rebooting fixes it | Same, plus it clears the suspend state |
| Nothing in the event log | No driver errored; it simply stopped reporting |
Status Class FriendlyName InstanceId
OK Mouse HID-compliant mouse HID\VID_046D&PID_C53F&MI_01&COL01\8&2C2C0656&0&0000
Problem = CM_PROB_NONE. The system believes the mouse is perfectly fine — which is exactly why “reboot / try another port” always works and nobody ever finds the real cause.
2. The first false lead: the logi_lamparray filter
The receiver carried Logitech’s USB lower filter driver:
HKLM\...\Enum\USB\VID_046D&PID_C53F\<inst>\LowerFilters = { vhf, logi_lamparray }
An official Logitech driver (logi_lamparray_usb.inf, WHCP-signed) whose job is to bridge Logitech RGB lighting into Windows 11 Dynamic Lighting. My reasoning:
The G304 is a single-colour DPI-indicator mouse with no RGB at all, so this driver is installed by mistake; it sits in the receiver’s USB I/O path, so if it wedges it swallows the HID reports.
There was corroboration: Microsoft’s own documentation instructs users to disable Logitech_Lamparray_Service; OpenLogi#1129 documents the same service overwriting HID++ writes; Valve’s Dota 2 #35162 shows a different machine with a different mouse (G Pro Wireless) and the identical filter stack producing main-thread freezes.
So I removed the driver package, stripped the filter and disabled the service. The mouse did come back.
But when the symptom recurred, the first thing I checked was the filter:
LowerFilters = [] <- the filter is not there any more
logi_lamparray driver package <- does not exist
logi_lamparray kernel driver <- disabled
The symptom happened anyway. Which means: the earlier “fix” did not work because the filter was removed. It worked because the procedure happened to re-enumerate the receiver — the same thing as replugging it.
When a fix “works”, first establish what it actually fixed. Verifying “the mouse came back after stripping the filter” is not enough; you also need the control case: “does re-enumerating alone — without touching the filter — fix it just as well?” That control is what separates a cause from a coincidence. Running it first would have saved a whole detour: “uninstall the package → a WinUSB package claims the composite device → the mouse’s HID node disappears entirely → another round to pull it back”.
(The filter is still a genuine defect worth reporting upstream, and the issue filed with Logitech has been corrected accordingly: https://github.com/Logitech/cpg-docs/issues/25)
3. The second attempt: right direction, only half right
The second time I turned to power, with this chain:
The receiver is USB-powered. During standby the USB hub enters selective suspend and cuts port power → the receiver’s firmware loses power → after the resume it no longer listens for its paired mouse.
So I applied four layers:
# (1) Globally: USB selective suspend off for every power scheme
powercfg /setacvalueindex SCHEME_CURRENT 2a737441-1930-4402-8d77-b2bebba308a3 48e6b7a6-50f5-4782-a5d4-53bb8f07e226 0
powercfg /setdcvalueindex SCHEME_CURRENT 2a737441-1930-4402-8d77-b2bebba308a3 48e6b7a6-50f5-4782-a5d4-53bb8f07e226 0
powercfg /setactive SCHEME_CURRENT
# (2) The hub the receiver is on: never let it be powered down
$hub = (Get-PnpDeviceProperty -InstanceId $recv.InstanceId -KeyName 'DEVPKEY_Device_Parent').Data
$obj = Get-CimInstance -Namespace root\wmi -ClassName MSPower_DeviceEnable |
Where-Object { $_.InstanceName -eq "$hub`_0" }
$obj.Enable = $false
Set-CimInstance -InputObject $obj
# (3) The receiver and each of its interfaces: never power them down
Get-CimInstance -Namespace root\wmi -ClassName MSPower_DeviceEnable |
Where-Object { $_.InstanceName -match 'VID_046D' } | ForEach-Object {
$_.Enable = $false; Set-CimInstance -InputObject $_
}
# (4) Per device node: clear the device's own suspend and link power management flags
$dp = "HKLM:\SYSTEM\CurrentControlSet\Enum\$($recv.InstanceId)\Device Parameters"
Set-ItemProperty $dp -Name SelectiveSuspendEnabled -Value 0 -Type DWord
Set-ItemProperty $dp -Name EnhancedPowerManagementEnabled -Value 0 -Type DWord
The next standby, the symptom came straight back. And that produced the most valuable evidence of the whole investigation — querying the device’s state properties while the mouse was still dead:
DEVPKEY_Device_ConfigFlags = 0 <- no SUSPENDED flag
DEVPKEY_Device_ProblemCode = 0
DEVPKEY_Device_LastArrivalDate = 23:48:18 <- no new arrival during standby
DEVPKEY_Device_DeviceSelectiveSuspended = (property does not exist)
Device Parameters\SelectiveSuspendEnabled = 0 <- mine, and it is set
Device Parameters\EnhancedPowerManagementEnabled = 0 <- mine, and it is set
Together those say three things:
- Windows never re-enumerated the receiver during standby (the device instance id is unchanged and
LastArrivalDateis still 23:48). The previous guard detected this viaLastArrivalDate→ that signal never comes. - Windows has no idea the receiver lost power. No suspend flag, no error code, no problem status, and
DeviceSelectiveSuspendeddoes not even exist as a property. - Everything in (2)(3)(4) above took effect (the values are
0/False) and none of it mattered.
Which means:
This power drop happens below the layer Windows power management controls. In Modern Standby, port power gating (port power gating / D3cold) is decided by the EC/BIOS, and Windows’
MSPower_DeviceEnable, the selective-suspend switch, and the per-device parameters do not reach it.
So the “root fix” I claimed last time was wrong — it only suppressed the symptom.
4. What actually works: stop asking “is the mouse dead?” and ask “is the user back?”
Since the drop cannot be observed and cannot be prevented, the only remaining road is: when the user comes back, reset the receiver once.
Resetting the receiver is a software-level USB reset — one physical replug — and it always restores the mouse in 4–5 seconds.
4.1 The signal: system idle time
GetLastInputInfo’s idle timer keeps running through standby and drops to zero the moment the user returns. It is the only reliable, essentially free “the user is back” signal.
4.2 Four triggers
| Trigger | Condition | Notes |
|---|---|---|
| T1 | The receiver’s LastArrivalDate changes | Fires when the receiver moves to another port or a driver updates (it did not fire during standby, but it does on a port change) |
| T1b | The device instance id changes | The receiver was moved to another USB port |
| T2 | System idle drops from >4 minutes to <1 second | The user is back (they touched the keyboard or touchpad) → reset the receiver now |
| T5 | Every 2 minutes once idle > 5 minutes | Proactive: reset the receiver while the user is away, so it is already fresh when they come back |
| T6 | A manual request file | The desktop “one-click recovery” writes a request file and the guard does the work — no UAC prompt |
T2 does not care whether the pointer moved — that was the biggest lesson of the previous version. It required “the pointer stayed still for 4 seconds after the user returned”, so a user who reached for the touchpad silently defeated the detector, and it never fired once.
4.3 Why T5 has to be proactive
A dead mouse produces no input at all — neither movement nor clicks reach Windows, so the idle timer never resets and the guard cannot tell the user came back. T5 inverts that: reset the receiver periodically while the user is away.
Modern Standby keeps user-mode apps running, and a USB reset re-powers the port and restarts the receiver firmware — so by the time the user comes back the receiver is fresh, and sliding the mouse is enough; no need to touch the keyboard at all. T2 stays as the fallback for the case where resets do not work during standby.
Measured over a 9.5-minute absence:
[00:50:58] T5 idle-period receiver reset <- idle 307s
[00:53:01] T5 idle-period receiver reset <- idle 427s
[00:55:03] T5 idle-period receiver reset <- idle 547s
The interval held at two minutes, and the guard was still running and resetting during standby — which is the precondition the proactive design depends on.
4.4 T6: a manual repair that needs no administrator rights
Resetting the receiver does need administrator rights — measured, an unelevated Disable-PnpDevice / Enable-PnpDevice fails with a generic error. But the guard already runs as SYSTEM, so the desktop entry now only writes a request file:
>"E:\LogitechMouseFix\state\repair.request" echo request
The guard sees it and performs the reset. Measured end to end: 11 seconds, no UAC prompt at all. If the guard is not running, the batch file falls back to an elevated run.
4.5 The cost, and what it feels like
Coming back from a long idle, the mouse has a 4–5 second window where it does not respond. But that is exactly the moment the mouse would have been dead anyway, so what you perceive is:
“it was already dead → 4 seconds later it suddenly works”
rather than “it stays dead until I unplug the receiver”.
And T2 fires at the moment the user touches the touchpad or the keyboard — i.e. exactly when they have just noticed the mouse is dead and are reaching for something else. The recovery latency is effectively imperceptible.
4.6 One full day of field testing: not a single freeze
On the first complete working day after the fix went live (2026.10.22), the report was: “the mouse never stopped working today, not once.”
The guard’s telemetry for that day:
T2 user returned from long idle (a real standby/wake cycle) : 6
T5 proactive receiver resets while idle : 314 (steady ~2 min apart)
T6 manual requests : 8
guard errors : 0
receiver state at end of day : OK / CM_PROB_NONE
The T2 count is the number that matters — six of them, each one a genuine standby-to-wake transition, and every single one recovered on its own. No manual replug, no UAC prompt, no user intervention.
That confirms both key design decisions:
- The proactive T5 reset really works. With only the T2 fallback, the first mouse movement after waking would have had to feel dead (T2 waits until the user touches the keyboard or touchpad). Waking up to a working mouse instead means the periodic resets during standby had already kept the receiver healthy.
- The T2 criterion does not depend on the pointer. With 314 receiver resets in a single day, the guard neither false-triggered nor missed anything — zero errors, zero spurious repairs.
(The raw log ships in the tutorial repo as evidence/field-test-20261022.log.)
5. The real fix is at the hardware layer
Software can only clean up afterwards. What actually fixes it for good is keeping the receiver powered through standby:
- Put the receiver on a self-powered USB hub (a few dollars). The hub has its own supply, so the laptop’s port power-down does not affect it — the receiver never loses power and the mouse never goes deaf. Cheapest and most effective option.
- Try another USB port. AMD platforms have several USB controllers (this machine has three root hubs) and their Modern Standby port gating behaviour may differ. Moving the receiver costs nothing.
- Look for an S3 option in the BIOS. Many AMD laptops let you switch between “S3 Sleep” and “Modern Standby”. S3 is real ACPI sleep (the whole system suspends) rather than “the system stays awake, only the screen turns off”. Switching to S3 may change the behaviour completely.
- Also look for “USB standby power” / “USB Selective Suspend” / “Modern Standby” entries in the same BIOS area.
6. The shortest path for anyone hitting this
# 1. Confirm it is software, not hardware
Get-PnpDevice -PresentOnly | Where-Object { $_.InstanceId -match 'VID_046D' } |
Select-Object Status, Problem, Class, InstanceId
# 2. Find the receiver
$recv = Get-PnpDevice -PresentOnly |
Where-Object { $_.InstanceId -like 'USB\VID_046D&PID_C53F\*' -and
$_.InstanceId -notmatch '&MI_|LAMPARRAY' } | Select-Object -First 1
# 3. Emergency: reset the receiver (= one replug, ~4-5 s, always works)
Disable-PnpDevice -InstanceId $recv.InstanceId -Confirm:$false; Start-Sleep -Milliseconds 700
Enable-PnpDevice -InstanceId $recv.InstanceId -Confirm:$false
Step 3 always brings it back immediately — so when this happens, do that first instead of reaching for a reboot.
To automate it, use the guard script from the next section (which asks “is the user back?” rather than “is the mouse dead?“).
7. Traps worth recording
pnputil output is localised
In an elevated process it prints 发布名称:; in an interactive session Published Name:. Parsing the elevated output with an English regex matches nothing and fails silently — my first run cheerfully logged “all cleared” while nothing had been deleted.
$pub = ([regex]::Match($b, '(?i)oem\d+\.inf')).Value # language-agnostic
Set-CimInstance has no -Filter / -ClassName
PowerShell 7 offers only -InputObject and -Query. Fetch the object, change the property, pass it back:
$obj.Enable = $false
Set-CimInstance -InputObject $obj
Removing a driver package can stop the device from starting at all
Strip the filter but leave the package and moving the receiver to another port makes the INF write the filter back, while the kernel driver is now disabled — the device then fails to start entirely. Both have to go together.
A WinUSB package can claim the whole composite device
After the Logitech package was removed, Windows re-matched the driver and picked a third-party LIGHTSPEED_Receiver.inf (libwdi / WinUSB), which matches the entire VID_046D&PID_C53F with no &MI_ qualifier. WinUSB exposes no HID interfaces, so the mouse’s HID node vanished entirely.
Deciding what to remove and what to keep comes straight from the hardware ids the INF declares:
LIGHTSPEED_Receiver.inf -> VID_046D&PID_C53F <- claims everything, remove
usb_receiver_(interface_0).inf -> VID_046D&PID_C53F&MI_00 <- HID++ interface only, keep
Raw Input is not always usable
Measured on this machine, inside an interactive scheduled task (SessionId=2, UserInteractive=True):
GetRawInputDeviceList return code: 0 (no devices)
cursor: from=(1640,766) to=(1416,1040) moved=True
WatchBoth result: G304=0 other=0
-> the cursor moved but raw input delivered no events; RAW INPUT UNAVAILABLE
Any detection scheme built on “tell the devices apart by input source” is blind here. Do not build on it unless you have verified it first.
8. Scripts and source
The complete repair package (PowerShell scripts, the resident guard, task installation, logs):
https://github.com/Kirito14IT/logitech-mouse-not-working-after-sleep
It also links the issue filed with Logitech, the OpenLogi pull request, and every trap mentioned above.