The HD Radio Network

February 1, 2007

The advent of HD Radio is requiring the broadcast industry to take a closer look at its station's networking system and infrastructure. Minimizing network-related issues that can induce dropouts must be a prime consideration in any successful HD Radio implementation. A network that performs quite well for day-to-day data traffic may be significantly challenged by the near-real-time, isochronous demands of the HD Radio stream. The key issues that appear to be causing station engineers and IT personnel the most difficulties are traffic management, bandwidth provisioning and reference timing synchronization between the various HD Radio components across the network.

There are two distinct physical configurations that the station may implement for deployment of AAS for multicasting on the HD Radio system:

  1. Importer to exciter (I2E)
  2. Exporter to exgine (E2X)

Figure 1 shows the studio importer connection over bi-directional (duplex) STL.

Fig 1 - Studio importer connection over bi-directional (duplex) STL.
Click image to enlarge.

The I2E configuration connects an importer to an exporter/exgine via a bidirectional Ethernet connection. Only the Advanced Applications Services (AAS), such as multicast programming and data services, are transported by this link, which is not concerned with the main program digital service. A bidirectional link is required to accommodate the command and response nature of the I2E configuration.

In this configuration, even with moderately bad network conditions (up to one percent packet loss and 100 millisecond latency) the system continues to perform well. The key is to provide adequate bandwidth overhead to allow the system to recover lost packets through TCP packet retransmission.

For a station running MP1 mode with 48kb/s of AAS, the average utilized bandwidth will be 54kb/s, requiring at least 90kb/s to be available through the STL. A 128kb/s LAN/WAN extender or two DS0s should provide sufficient bandwidth for any MP1 configuration. For the maximum MP3 extended hybrid configuration of one SPS at 48kb/s and a second SPS at 24kb/s, the minimum bandwidth required of the STL/WAN link is 156kb/s, requiring three DS0s for 196kb/s.

Exporter to exgine (E2X)

Figure 2 shows the exporter to exciter configuration.

The importer-to-exporter-to-exgine (E2X) configuration is the most bandwidth-efficient method of deploying an HD Radio multicasting data network. With this implementation, a single data stream may be conveyed to the transmitter site over the STL/WAN link, which contains all of the MPS information as well as the Advanced Applications Services from the importer, such as SPS and associated data.

Fig 2 - Importer to exporter to exgine configuration.
Click image to enlarge.

Studio-to-transmitter transport of the E2X data stream is currently supported only as simplex (one-way) UDP and can operate over most unidirectional STL systems of sufficient bandwidth and robustness. With UDP transmission, the loss of a single packet results in the loss of the entire audio frame of which it is a part. The resulting outage will last for the duration of that single audio frame: 1.48 seconds.

A 128kb/s LAN/WAN extender or two DS0s will provide sufficient bandwidth for any MP1 configuration. For MP3, 256kb/s or four DS0s should be considered. Packet loss across the link becomes a critical factor and must be kept below 10-5 for successful operation due to a lack of error recovery inherent with UDP. It is not uncommon for wide-area networks and STL systems considered healthy, to deliver only 10-3 performance or one dropped packet in every 1,000, which will result in poor HD Radio system performance when running E2X.

Managing HD Radio network traffic

Fig 3 - Recommended network deployment.
Click image to enlarge.

Because the STL system is usually the tightest bandwidth bottleneck in the HD Radio network, it is imperative that broadcast, multicast and other extraneous traffic be kept off the network path to the transmitter site. All HD Radio devices — importer, exporter and exciter — should use statically assigned IP addresses within their own subnet. This subnet must be separate from the rest of the facility through the use of VLANs or physically separated networks. The only way to be sure that no extraneous traffic is traversing the STL link is to place the entire HD Radio system on its own IP subnet. Figure 3 shows a recommended network deployment of subnetting using VLANs.

The exciter should always be on the WAN subnet, which it may share with the exporter and importer, or the importer may be placed on program automation subnet. Except for equipment that may be necessary to build the infrastructure — that is, routers and switches — no other station equipment should be on the WAN link subnet.

The implementation of VLANs or connection of devices through a dedicated physical network will substantially reduce packet loss and data collisions. Monitoring the traffic across the WAN with a network protocol analyzer or packet sniffer such as Ethereal is essential if a problem is suspected.

Provisioning the STL/WAN link

For a TCP data stream to function properly under adverse conditions, the link that carries it must have reserve bandwidth above and beyond the data rate of the stream. For TCP, the STL/WAN link must have a minimum of 40 percent reserve bandwidth. This is necessary to accommodate the temporarily higher data rates that occur when the stream recovers from packet loss. If a TCP WAN link is provisioned such that the aggregate data stream, including VNC, utilities and other extraneous traffic, occupies no more than 60 percent of the WAN link's available bandwidth, then the installation should be successful under all but the most adverse network conditions. For UDP, the total traffic across the link should be no more than 75 percent of the provisioned bandwidth to allow for network contention.

Additional bandwidth beyond these guidelines allows operation under poorer conditions, but with diminishing returns. In general, bandwidth should not be used to adjust for a poor network.

If other traffic is going through the WAN, the link should have class of service, QOS or other prioritization techniques employed to ensure that the HD Radio traffic has the necessary bandwidth.

Reference timing synchronization

While not specifically a networking issue, reference timing between the importer, exporter and exgine are not maintained across the network infrastructure. The use of GPS as a timing reference for the importer, exporter and exgine to precisely lock their respective clocks in step eliminates the phase and frequency issues and is highly recommended. Without GPS, or some other method of providing absolute frequency lock between the exporter and exgine, buffer underflow in the exgine or data overflow of the exporter's audio cards will eventually occur resulting in data frame misalignment, eventual audio dropout and significant diversity delay slippage on the main HD program channel. Without the use of GPS as a 44.1kHz timing reference for the station's AES audio chain or at least to the importer's audio cards, any difference in the importer's audio clock frequency and the exporter's 10MHz reference will result in the eventual underflow or overflow of the importer audio cards, which will result in occasional audio dropout of the SPS channels. The frequency of these dropouts will be directly proportional to the frequency disparity of the two references.

For more information on HD Radio and networking implementation, several white papers are available on the Ibiquity website at

Anderson is president of TBA Communications and a contract engineer for Ibiquity Digital.