Hello,
In our product we used four independent LTC4279 controllers in a multi-port PoE PSE design. Schematic attached.
We have observed an issue related to hot-plug operation that affects for some units in small batch production (about half of 2 different batches (produced by different factories) have these issues).
Our multi-port PoE PSE works with PD devices: Silvertel AG59624, AG5800, AG5424
Observed behavior:
- If all PDs are connected before power is applied, the system starts and operates normally. All 4 channels works fine with load 20-30W/channel
- If the PSE is powered first and PDs are connected one by one (hot-plug), one of the channels may fail to start. The affected channel repeatedly attempts to turn on and then immediately turns off. This cycle repeats approximately once per second.
The affected channel is not fixed. Depending on the connection sequence, different channels may exhibit the behavior. For example, if channels 2 and 4 are already active, the issue may occur on channel 1 or channel 3.
During the retry cycle, the GATE signal periodically turns on and off.
The voltage across the 0.1Ω sense resistor remains very low during these attempts and does not reach the normal operating level.
In normal operation with load connected, the sense resistor voltage is approximately 50mV (~0.5A @ 56V). - The issue only occurs when other channels are already supplying loaded PDs (about 20-30W/channels). If the existing PDs are connected but unloaded (1-2W/channel), the problem does not occur. Also our other devices (single channel) with totally same schematic for one channel works fine.
- Change supply voltage from 48V to 57V. did not change the behavior.
- The issue not depend on length of Ethernet cables. Were tested various cables from 1m to 100m.Same behavior.
- For tests was added extra 470µF bulk capacitance to the power input. But this did not resolve the issue.
- Switching between UltraPWR and standard LTPoE++ modes did not change the behavior.
- Changing prtection diode to Ulta fast UF4007 did not change the behavior.
- We used a bit different design for LED indication, but even if additional mosfets that control LEDs removed, this did not resolve the issue.
Additionally we have also observed at similar situations (all 4 channels works with load 20-30W) that after hot-unplugging a PD, the LTC4279 occasionally does not detect the disconnect event and keeps the port powered. If the same PD is reconnected, it powers up immediately without an apparent new detection/classification cycle.(voltage on sense resisotr = 0mV with disconnected PD but line keep powered)
Are there any known errata, application notes, recommended measurements, or layout considerations that could explain this behavior?
Any guidance would be greatly appreciated.
Best regards,
