Joining offices and PBXs over SIP

There are two cities, A and B. City A has a Panasonic KX-TDA200 PBX with its own extensions and external connectivity. City B is a branch with a Panasonic KX-TDA100 PBX, also with internal numbers and PSTN access. For a long time the link between the branches ran over E1 PRI, but the operator introduced TDMoP between the cities — and users at both offices began complaining about call quality. Management also would not mind cheaper calls to other cities and to mobiles. Alternatively, there is no E1 link between the offices at all and a solution is being chosen.

We use standard hardware-and-software tools that do not require large capital investment. Despite the specific TDA100/TDA200 example, the solution also fits uniting PBXs from other vendors compatible with ISDN PRI.

E1 and PRI signaling are circuit-switched, while the Internet is packet-switched. The designers of ISDN PRI (Q.931) never imagined anyone would swap proper links for a pseudo-E1 (pseudowire). Packet transport in TDMoP/CeS/TDMoE brings noticeable jitter, possible partial packet loss and clock-recovery problems — especially on high-latency links.

TDMoP equipment replaces lost packets with a duplicate of the last received E1 frame or with a filler frame (usually D5 or FF). Imagine what happens to an HDLC-based PRI signaling link: messages are lost, calls hang. It is all compounded by the incomplete PRI stack in modern PBXs and Asterisk. A misconfigured jitter buffer in the TDMoP tunnel affects all channels at once, wildly increasing delay or causing "gurgling" with a high chance of call drops.

The worst part — on the subscriber side it is almost transparent: unlike ordinary VoIP, you cannot "ping" it through ping or iperf, you cannot measure jitter and it is hard to estimate loss.

• Native SIP gateways (SIPGW). Not cheap: you must buy expensive licenses to activate SIP trunks (the vendor’s policy), and NAT-traversal problems force you to set up a VPN.
• H.323 connection. Again you need a SIPGW, the base configuration cannot connect to VoIP operators, and you must buy codec and echo-cancellation support. NAT falls on the users, debugging H.323 dialects is not for the faint-hearted, and with a different-vendor PBX (LG, Siemens) compatibility problems are inevitable.
• FXO/FXS gateways and an analog link. It will work, but quality is mediocre: you cannot tell who called, and multichannel and forwarding cannot be organized properly. It is partly patched with Busy disconnectors, and fax problems keep bothering you for a long time afterwards.

We propose replacing the inter-city E1 link with local E1 links to the Alvis-GW-2E1 SIP/E1 gateways, and carrying the inter-city link over SIP. This avoids the PRI ailments — there is no more TDMoP in the scheme.

• E1 stream synchronization is local or taken from the TDA — no delays in E1.
• Echo cancellation is handled by the built-in hardware LEC module, G.168+ compatible and supporting NEC (Networking Echo Canceller).
• You can use the full E1 stream, all slots; no extra licenses and no SIPGW needed. To save bandwidth there are G.729/G.723 codecs and AMR-NB support.
• Delays, packet loss and jitter are handled by standard SIP/RTP means.

• Both sides can be behind NAT — no need to complicate things with a VPN.
• The solution can be combined with connecting to VoIP operators to cut long-distance and international costs.
• With Asterisk you can define multiple call paths depending on various factors.
• Network problems are easily diagnosed with Wireshark/ping/iperf.
• Advanced redundancy schemes and FailOver solutions are possible.
• Not only point-to-point but also a distributed branch network with "each-to-each" connectivity is possible.

Centralized monitoring of the whole network is available via the web interface, the Alvis Distributed System tab. Plus SNMP support: besides CPU load and memory, you can monitor E1 Layer 1 and PRI signaling status, get real-time alarms on link problems, and manage the gateway.

A web-interface add-on for the Alvis-GW-2E1 is under development that automatically applies configuration sets for different PBXs — here, for connecting to the KX-TDA. No need to repeat others’ mistakes — just apply the standard settings; no Linux or Asterisk knowledge required.

Hackers scan and break into VoIP systems, especially SIP. We do not ship equipment on a "sold-and-forgotten" basis: we perform security audits and minimize the risk of huge phone bills for customers. A separate finding — some TDA firmwares have vulnerabilities over PRI too; our equipment can be used as a filter on PRI-PRI calls. We developed a specialized PRI filtering system for several clients and are exploring a public release.

For especially critical applications there are dedicated solutions: redundancy systems, Industrial builds with an extended temperature range, and specialized software — a built-in E1/T1 Analyzer and ISDN PRI Monitor. All our equipment includes hardware Watchdog timers that let the system recover after failures.