How warm are the microcontroller and LDO supposed to get?

[blu006]

Hello all,

I'm assuming that the esp32-c6 supermini boards aren't great for this application. (Unless something is wrong with the GPIO's) The "ceramic" antenna honestly looks like a PCB with castellated holes, and the LDO and ESP32 SoC get a bit warm for comfort.

Aliexpress Link to ESP32-C6 board

Front thermal image (a bit blurry) with the white cursor on the periphery of the SoC and the red cursor automatically pointing towards the LDO:

Rear thermal image, included to illustrate how warm the entire PCB gets:

Thanks,
blu006

[tostmann]

Thanks for bringing actual thermal data — that makes this answerable with numbers instead of hand-waving.

Why the board runs warm — this part is by design. A Zigbee coordinator can never sleep: the radio has to listen continuously for its network, so this firmware deliberately runs with no power management at all (no light sleep, no frequency scaling — CONFIG_PM_ENABLE is off). Steady state is therefore: receiver permanently on + CPU active, 24/7. The ESP32-C6 series datasheet (Table 5-9) specifies 802.15.4 RX at ~74 mA, plus the CPU on top — so the board continuously dissipates very roughly a quarter of a watt, around the clock. TX bursts draw more, but at coordinator duty cycles the RX listening dominates. There is no firmware knob to reduce this; any power-save mode would make the coordinator deaf to its own network. So "gets warm and stays warm" is the expected steady state here, not a symptom.

The LDO on boards like this typically takes the USB 5 V down to 3.3 V linearly, which means it dissipates the 1.7 V drop times the same current — on the order of 100–150 mW in a tiny package. That's inherently the hottest spot on most small ESP32 boards; again not a fault indicator by itself.

Your readings: ~58 °C max on the front hotspot with ~29 °C ambient, ~52 °C on the back. For the SoC that is comfortably within spec — the ESP32-C6 chip is rated for ambient temperatures up to 105 °C (series datasheet, Table 5-2; modules typically 85 °C). Whether ~58 °C is fine for the LDO on that particular board depends on which regulator part they used — that's board-designer territory, and we haven't characterized that board, so I can't vouch for it specifically.

On the "supermini boards aren't great for this" assumption: I can't confirm or refute that for this specific board — we simply haven't measured it, thermally or RF-wise. The warmth alone isn't the disqualifier. The metric that actually matters for a coordinator is link quality: Zigbee2MQTT shows LQI per device — if your devices hold stable LQI at realistic distances through your home, the antenna is doing its job, castellated-PCB look or not. If LQI is consistently poor at short range, that would be the reason to pick different hardware.

[blu006]

Oh yeah. The LQI is dismal. I need a little time to provide concrete real-world data. Here are the three devices in my test network, all within the same room.

[blu006]

Here is a table with how far apart these devices are from the coordinator:

Device	Distance (cm)
Thermometer	93
Shortcut Button	266
PIR	300

All devices are on channel 11 which seems to work best with this antenna...

[tostmann]

Sorry for the long silence.

We measured it in the meantime — not on your board, we don't have one, but on hardware we do. Same test device, unchanged position, factory reset and a fresh join per arm, 8 round-trips each:

Coordinator	LQI values	Median
C6 stick (busware form factor), same firmware you run	112 · 117 · 117 · 117 · 117 · 122 · 127 · 127	117
Espressif Thread-BR / Zigbee gateway board, H2 half	183 · 183 · 183 · 183 · 183 · 183 · 188 · 188	183

Limits: the two coordinators sat at different positions, so this doesn't separate antenna from placement, and it's one device, one position, no distance sweep. A data point, not a product claim.

Your thermometer at 93 cm reports 142 — above our C6 stick's 117 at comparable distance. LQI normally isn't comparable across setups, since 802.15.4 leaves the derivation to the implementation, but yours and ours come from the same firmware on the same SoC family, so the scale matches. At short range your board is not the outlier you suspected.

One thing about the number itself: the firmware passes the LQI straight out of the ZBOSS data indication to the host, so what Zigbee2MQTT shows is what the coordinator measured on the last frame it received. It describes the uplink at our end, nothing about the other direction.

What stands out is the drop, not the level — 142 at 93 cm down to 51 at 300 cm, in one room. Your overview shows why it has nowhere to hide: 3 end devices, 0 routers. Everything talks directly to the coordinator, with no mesh to route around anything. One mains-powered router — a plug, a bulb, anything not battery powered — changes that topology, and it's a lot cheaper than swapping coordinator hardware.

Before concluding anything about the board: are you actually losing messages? Missed PIR triggers, button presses that don't arrive, devices going unavailable. A low LQI on a link that never drops a frame is cosmetic. If things are genuinely dropping, that's the more interesting report.