Device / model: Xiaomi Redmi Note 12 Pro 5G (22101316G, global)
Board number: 13510M16M0A
Symptoms:
- Slow to find service, sometimes none at all.
- Mobile data dead even though it is switched on.
- IMEI ×2 and baseband version present, and the SIM reads.
Step 1 — compare with a reference phone before any board work
4636 showed −105 to −115 dBm, which looks like weak reception. Before that, the radio chip (MT6190MV) had been reballed and the dock board and antenna replaced, with no change.
The decisive test was simple: same spot, same network, a known-good iPhone.
| Phone | Signal |
|---|---|
| Redmi | −102 … −110 dBm |
| iPhone (working) | −109 dBm |
Reception is fine. The antennas, coax and receive path are all OK; the weak dBm is just the building. (If the gap had been 10 dB or more, the loss would have been inside the phone.)
Step 2 — it hears the network but never gets an answer
- Ruled out: SIM swap with the iPhone (no difference), IMEI matches the box, factory reset and update done. SIM, IMEI and software are fine.
- With the SIM moved to the right slot, the phone showed “Out of service” while reading −106 dBm: it receives the cell but can’t register.
- The modem log over USB told the rest:
- Registration only ever showed “searching”, never “denied”. No network reject, so it isn’t a blocked IMEI or the subscription.
- The connection report was switched on, and in 30+ minutes it never once reported a connection. It searched band 20, 8, 3, 1 and 7 (every LTE band in use) without ever connecting.
Receives on every band, never connects on any band = the transmit side is dead.
Step 3 — the split that pinpoints it
- After returning to automatic network selection it did register: on 2G (GSM 900) with full bars.
- On 2G it can make and answer calls, but there is no data, and LTE still never connects.
- So 2G transmit works, LTE transmit doesn’t. The main antenna’s low-band path is proven good by the GSM 900 transmission.
- This also explains the customer’s “random service”: the phone falls back to 2G, and data is dead because there’s no LTE.
Where the LTE PA supply goes
- Every LTE band’s power amplifier gets its supply, VPA_0, from one tracker chip: U0900 = MT6308MP.
- VPA_0 feeds three parts: the low-band PAMiD U0600, the mid/high-band PAMiD U0700 and U1900.
- 2G runs straight off the battery, which is why it still worked.
- The 5G n77/n78 amp has a separate supply (VPA_PMU from the main PMIC).
Root cause: one shorted 100 pF capacitor
Inside the low-band PAMiD U0600, VPA_0 feeds two supply pins:
- pin 28 (3G4G_VCC1) directly, with its own decoupling caps;
- pin 30 (3G4G_VCC2) through R0602, which is a 0 Ω link. C0631 (100 pF, 50 V, 0201) decouples pin 30 to ground.
C0631 was shorted. Because R0602 is 0 Ω, that put VPA_0 straight to ground. Every LTE PA lost its supply at once, while 2G kept working off the battery. That’s exactly the “hears every band, never connects on any, 2G fine” pattern from the logs.
The schematic lists the low-band PAMiD as a QM77042, but this board has a 77043 fitted, an alternate part on the same footprint. Read the real marking before ordering a PA.
How it was found
- VPA_0 read shorted to ground. So the fault was on the rail, not inside the MT6308MP.
- Voltage injection + thermal camera: a low, current-limited voltage from a bench supply into VPA_0, with the board watched on a thermal camera.
- C0631 at U0600 lit up. It sits behind R0602, but that’s a 0 Ω link, which doesn’t isolate anything, so the short shows straight through it.
Fix: C0631 replaced with a new 100 pF 0201, and VPA_0 checked for no short. LTE registers and mobile data works again.
Lessons
- Do the reference-phone compare before any board work on “no service / weak signal”. Here it saved a wrong antenna or reception chase.
- “Good signal, no service” is a transmit problem, not reception.
- 2G works but 4G never connects → look at the LTE PA supply first.
- Short-check the whole PA rail before replacing the tracker or a PA. A single tiny decoupling cap took the rail down, and so can one sitting behind a 0 Ω link.
Was this your fault too? Post your readings below — and if it fixed it, mark the solution so the next person finds it.