We are using Digi XBee 3 Global LTE Cat 1 modules as Modbus TCP servers in Transparent Mode.
Configuration:
- Module: XBee 3 Global LTE Cat 1 (
XB3-C-G1-UT-001) - Firmware:
11521 - Transparent TCP Server, listening on port
502 - XBee server timeout
TS = 0x4B(7.5 seconds) - Cellular private network / VPN with carrier-side NAT
- Windows Modbus TCP polling client
- Poll interval approximately 2.5 seconds, request timeout 1.5 seconds
- The UART-attached Modbus MCU has no unsolicited traffic; it only replies to polls.
Observed failure sequence:
- The Windows client has an established TCP/Modbus connection to the XBee.
- A cellular-path disruption occurs, such as a mobility event, PDP-context update, or carrier/NAT routing change. There is no observed FIN or RST for the original TCP session.
- The Windows client times out, closes its local socket, and later attempts a fresh connection from a new ephemeral source port.
- In the failed state, the client sends valid TCP SYN packets to
XBee_IP:502, but receives no SYN-ACK and no RST. - The XBee remains unreachable on port 502 for a prolonged period. We have observed this for many hours.
- A module/network reset or LTE detach restores normal operation.
The problem appears consistent with the XBee retaining a stale internal TCP-server connection after the former cellular route has silently disappeared. Since Transparent TCP Server mode supports a single active connection, the listener may be occupied by that stale connection and ignore replacement SYNs.
We cannot use API-mode socket control in this product at present, and the remote MCU cannot initiate a keepalive because it has no unsolicited UART traffic.
Questions:
- Is this a known issue in XBee 3 Global LTE Cat 1 firmware 11521, particularly for Transparent TCP Server mode after cellular route/PDP/NAT changes?
- Is there a documented way in Transparent Mode to detect or clear a stale TCP server connection without resetting the module?
- Is there a recommended server timeout, connection-close delay, reconnect backoff, or watchdog/reset strategy for this deployment?
- Are there newer firmware versions that specifically improve stale TCP socket/listener recovery?
We can provide a filtered PCAP showing the old connection, the new SYN attempts, and the absence of any XBee reply.