FAQ

Can I use it without QPT nodes?

Yes. QPT only powers the per-zone view. Without it you still get:

• CPU cluster frequencies and per-core utilisation
• GPU frequency / busy and DDR frequency
• Four thermal zones (CPU / DDR / GPU / battery)
• Wall power (via power_now or voltage × current)
• FPS and foreground-app recording

Only the zone bar chart and zone curves disappear.

su -c 'ls /sys/class/powercap/qpt'

If the directory is missing, there are no QPT nodes; if it exists but is empty, the vendor did not enable them.

Wall power is stuck at 0.00 W

Not a bug. Many BMS drivers only refresh power_now while charging and report a literal 0 on discharge.

WattScope Next therefore:

1. Probes /sys/class/power_supply/{battery, Battery, bms, hisi_bk_battery}
2. Computes current_now × voltage_now, adapting units to the physical magnitude
3. Falls back to power_now only as a last resort

If all three fail you see --.

Dual-cell devices read about half the real power

Some dual-cell BMS drivers report only one cell's current. Enable Settings → Sampling → Dual-cell device to sum both paths — live preview, OSD and stored data all follow.

FPS shows 0 in games

The primary source is SurfaceFlinger presentation timestamps; dumpsys gfxinfo is only a fallback. A few games use compositing paths that neither source can see, so FPS reads 0 while everything else works. Version 1.0 relied on gfxinfo first, which made this far more common.

How much does recording itself cost?

At 5 Hz the remaining cost comes from the shell sample, persistence and preview refresh. After the rewrite: zero forks per sample (1.0 forked ~32 times), batched writes (~1 transaction/s), and FPS/foreground detection decoupled from the sampling tick.

Can it coexist with the original WattScope?

Yes. Different package IDs (io.github.rusheng920.wattscope vs com.qpt.powermonitor), fully independent data. They do not share history either — switching means starting fresh.

Where is my data stored?

In a local Room database inside the app's private directory. Exports (JSON / CSV) are started manually by you. Nothing is uploaded.

Which sample interval should I use?

Use case	Suggestion
Casual monitoring, frame-rate drops	200 ms (default)
Long runs (1 h+)	500 ms – 1 s: lighter, still dense enough
Catching transient spikes	100 ms, at the cost of the tool's own footprint becoming visible

Everything shares one rate

Since v1.2.3, CPU frequencies, per-core utilisation, wall power, GPU, thermals and FPS are no longer sampled on a slow 1 s cadence — they all follow your interval and land in the same tick as SoC power, so every chart aligns point by point.

Why Android 13+ only?

minSdk is 33. Older versions are neither adapted nor verified; permission and API differences would produce unpredictable behaviour, so no support is claimed.

Why aren't the overlays frosted glass?

Per-frame blur keeps the GPU busy and pollutes the measurement. The overlays use a static liquid-glass layer; only the in-app nav bar uses live blur — and even that degrades while recording.

Non-Qualcomm platforms?

Per-zone power needs Qualcomm QPT. MediaTek and Samsung devices have no such node and only get the secondary metrics. Node names differ across vendor ROMs; missing charts hide themselves automatically.