The following is an abridged and edited version of a paper presented as part of the Broadcast Engineering Conference of the 2016 NAB Show.
Audible blending artifacts are the top complaint from auto manufacturers and consumers regarding the HD Radio experience. The majority of the problems can be traced back to individual broadcast stations. While our industry has long relied on manual management of diversity delay and time alignment, results in the field often have shown that stations set manually have been known to �drift� over time.
There are a myriad of potential problems that can cause drift, many of which are out of control of any individual device that has been traditionally installed in the broadcast chain. What is needed is a systems approach to this problem.
In this paper, we will discuss the implementation of various combinations of hardware to provide continuous monitoring and adjustment of diversity delay in real-time to reduce or eliminate the objectionable blending artifacts impacting listeners.
WHY DOES IT MATTER?
HD Radio is here to stay and growing rapidly. The biggest area of growth is with factory-installed radios in new automobiles. Ten years ago, the broadcast industry wondered when these digital radios would show up. They are here now, and listeners have them.
In 2015, DTS reported that over 35 percent of new automobiles came with HD Radio as standard equipment. In 2013, that number was about 33 percent. Millions of HD Radio receivers have been sold and are being used, and millions more are added each year. Should these trends continue for the rest of the decade, it will soon be hard to buy a new vehicle without HD Radio. The digital experience of a radio station is something broadcasters should be paying attention to today, because within a few years, whether listeners are aware of it or not, an HD Radio will become the primary radio.
Market-based research conducted by DTS shows that, in many of the largest metro areas, 15 percent or more of vehicles registered on the road are equipped with HD Radio, and in a few markets, that number is nearly 20 percent.
Let that number sink in: One-fifth of the audience could be listening to an HD1 signal instead of analog FM. Those listeners may not necessarily know that they have HD Radio, or be seeking out HD Radio content, but the acquisition of a vehicle installed with HD Radio means they will listen to this content.
If your station is not precisely aligned all of the time, the transition will sound bad, whether from analog to digital, or back from digital to analog when the HD Radio digital signal fails. Just how objectionable this is to listeners depends on how far off the time alignment is and what type of programming is currently broadcast on the station.
My experience, gained from listening to hundreds of stations, is that music formats with light audio processing and spoken-word formats seem to be more susceptible to noticeable audible problems, even with a small offset in samples.
On loud, densely processed music formats, the effects of being slightly out of spec can get buried in the program density to a point. But keep in mind, even these stations have periods of less density that are important, such as talk-intensive morning shows and commercials.
Many broadcasters may think, �Well, the radio blends once every 8 seconds after it tunes to the station, so it skips once � why does it matter that much?�
Perhaps they think of the blend as an initial acquisition; then you keep digital forever. But that is not always the case. Receivers of all types lose digital lock at some point.
The easiest way to think about this is in the car. The blend point has traditionally been thought of as occurring when someone is driving out of a station�s market. At some point on the highway, their receiver loses digital lock, it blends to analog, and eventually as they go out farther and farther, the signal is lost into noise.
Unfortunately, that is not the typical blending experience for listeners because most do not leave the market or the station�s signal coverage each day.
Let�s consider the station�s 60 dBu coverage contour, which is where many FM band HD Radio stations with digital sidebands at �20 dBc tend to start to lose digital lock.�� If a listener lives and works in that area they can experience the constant drifting in and out of digital lock, causing the radio to blend frequently.
Fig. 1: There are plenty of spots to expect blending, as shown in this set of field measurements.
Click to Enlarge Let me give you just one example of what I mean. I recorded measurements for WNIC, Ann Arbor, Mich., as seen in Fig. 1. (Green indicates good signal; red, bad.) If you lived and worked in the vicinity of the University of Michigan campus, you�d hear blending often as you drove from where you work to where you live to where you do your errands.
AUTO COMPANIES HAVE TAKEN NOTICE
Diversity delay blending issues are the number one HD Radio-related complaint from auto manufacturers.
Fig. 2: Blending artifacts at 1 sample offset.
Image Courtesy DTS I can independently vouch for this because I work closely with many automotive companies. I have had emails, phone calls and idle conversations over lunch about this problem with their engineers. I have been in vehicles with these same engineers when they hear stations skip, and often they ask me to help fix it.
General Motors temporarily removed HD Radio in some vehicle models to help fine tune their implementation to address consumer feedback about HD Radio blending. Their customers have been complaining to them about this for a long time. Instead of calling up the radio station to complain about it, the customer takes the car back to the dealer, thinking there is a problem with the radio.
Fig. 3: Blending artifacts at a 3 sample offset.
Image Courtesy DTS This customer feedback is not unique to GM vehicles. Automotive designers are sensitive to this problem, and it is my experience that most of the problems come back to issues on the broadcast side.
The official specification from DTS is that the analog and digital signal time alignment should be 0 samples, plus or minus 3 samples. One sample refers to 1 out of 44,100 samples per second in the 44.1 kHz digital audio bit stream.
Fig. 4: Blending artifacts at 6 sample offset.
Image Courtesy DTS What exactly is objectionable for an average listener? DTS has performed research, and found there are various, objectionable blending effects, that depend upon the sample offset. Figs. 2, 3 and 4 show the effects of small sample offsets and the impact on the audio. The aural impact of not being within spec quickly gets out of hand, with only small sample errors. A very noticeable �comb-filter� effect is heard in the audio. Larger errors create more problems; for example, there are 23 deep nulls created by a 50-sample offset.
We�ve identified the problem and seen how it important it is to solve. Now we turn to real world examples that can create the issues.
EXPORTER LOCATION AND NETWORK DESIGN
FM broadcasters demanded flexibility in equipment placement, and thus the Exgine system was born.
Earlier implementations of HD had the entire system located at the station�s transmitter site; there was little to go wrong then. Those who have implemented and studied more than a few stations have always known that locating the Exporter at the transmitter site produces the most stable results; still, many in the industry pushed to move the Exporter back to the studio. (By necessity, the exporter must have a GPS reference.) However, when the Exporter is located at the studio and connected to the exciter across a LAN bridge, problems can be introduced, such as wide variations in the diversity delay shift over time. The installation of a GPS-disciplined 10 MHz reference source at the transmitter site, as a time-base for the Exciter, mitigates this problem, but does not completely eliminate HD Radio delay drift.
The variable latency on these Ethernet links generally creates jitter conditions far beyond the jitter design tolerances of the Exporter to Exgine (E2X) connection. Another related issue plaguing the industry is that many of these extended Exgine networks are not properly isolated; I have discovered networks where the Exporter and the exciter are on the same network as many other different devices.
Some networks I found even had automation systems and standard end-user workstations on the same subnet as the Exporter at the studio and the exciter at the transmitter site. A Wireshark analysis of these networks will show you that your E2X packets are competing with broadcast traffic that is inherent on large networks, such as the Address Resolution Protocol. I have seen some large networks attached to these bandwidth-limited LAN bridges that the remote connection is nearly saturated with broadcast traffic.
The Exgine network should be isolated on its own subnet, and there are several ways to accomplish this. One is by creating a physically different network for the E2X link that just has these two components on it. The more advanced method (and more commonly accepted practice with larger corporate network environments) would be to create a separate VLAN for this network and use a virtual router in a Layer 3 switch. If a VLAN trunk is established between the two ends, make sure the bandwidth needed for all traffic is adequate, and configure VLAN priority so that the VLAN accommodating the E2X packets has the highest priority.
The reality is, in present day third- and fourth-generation hardware, the exporter belongs at the transmitter site if you want to remain in 0 � 3 samples specification that DTS has defined as �perfect� alignment.
Exporter or exciter reboot. Third- and fourth-generation HD Radio transmission hardware is far more reliable than the earlier generations, so the need to reboot due to a hardware or software lockup is typically lower compared to the past.
Still, there are circumstances that can cause reboots, such as power interruptions to the exporter or exciter. These can be mitigated during power failures or switchover from utility power to generator and vice versa by installing an Uninterruptable Power Supply to increase reliability. The Exgine system in its current design does not have deterministic startup; so, at every reboot, the exporter can start at a different time reference with respect to the receiving Exgine interface in the exciter. The manufacturers have worked with DTS on this issue to some degree, but it is not perfect.
Improper clocking configuration. It would be impossible to cover every Exporter and exciter clocking configuration in the scope of this paper. Suffice it to say that you must follow the proper configuration for your exporter and exciter installation as recommended by your manufacturer. Make sure you are on current known-stable firmware and software builds from your manufacturers.
GPS synchronization. There is a common misconception or myth that exists among broadcast engineers that says if the Exporter is located at the transmitter site, GPS synchronization is optional. This couldn�t be further from the truth. GPS-disciplined oscillators are a requirement for long-term stability of the station to maintain the 0 � 3 sample spec. Additionally, many exciters use the GPS reference to make sure their oscillators, sampling and 10 MHz subsystems are disciplined and fall within acceptable operating tolerances.
Hardware failures or defects. Sometimes components in these systems fail or need to be recalibrated due to age. These can be tricky to troubleshoot in the field if the engineer has little experience with the hardware. Often times, loaner or spare Exporter or exciters can be used to help troubleshoot the issue; sometimes it is be best to send them back to the factory for analysis.
Separate audio processors for FM and HD Radio. The once-embraced tactic of implementing a separate audio processor for analog FM and the HD1/MPS digital audio has caused blending headaches that this industry did not have imagine thirteen years ago when systems were first implemented.
The processing requirements for analog FM and HD Radio are radically different. Because of these differences, separate processing was at first encouraged as a best practice for stations adding HD Radio. It was not until about 2005 or 2006 that the audio processor manufacturers started offering devices with FM+HD processing integrated in to a single box. Use of these single box processors has become the accepted practice for the new installations of HD Radio.
When using separate processors, and even if a station achieves perfect alignment with no drift whatsoever, the blends between analog and digital often do not sound good. Ultimately, the industry may find that it must eliminate separate audio processors for analog and digital, and implement integrated processors to achieve blends that are not objectionable to listeners and automotive companies.
For years, broadcast engineers have been working with iBiquity (and now DTS) along with transmitter hardware manufacturers to resolve the alignment problems discussed in this paper. Progress has been made, and some of the issues that cause drift have been addressed. However, as I have transitioned from a traditional broadcast engineer to an HD Radio implementation specialist for iHeartMedia, I have discovered there are many variables beyond the control of all the individual parties involved; too many in fact. The set of issues I have mentioned isn�t even complete.
There needs to be an industrywide push for updated, precision and continuous diversity delay measurement systems and automatic corrective systems. iHeartMedia and other broadcasters have been asking for products and solutions in this area, and we have seen the industry respond in the last few years. At the 2015 NAB Show there were many solutions introduced to the marketplace. (Even more were present at NAB Show in 2016.)
The correction devices take one of two forms. The two-device system pairs a monitoring receiver that can measure the diversity delay offset, and subsequently sends a correction offset to a device that is active in the analog air chain (an audio processor, delay unit or Exporter). If the station has some of these products already, this could make a lot of sense. By upgrading firmware on existing devices, it may be possible to put together two pieces of hardware to come up with an automated hardware solution.
Some stations may choose to implement a single device system that measures and corrects the diversity delay in a single device. The single devices have an integrated receiver to make the measurements of delay offset, and can be inserted into either the analog or digital air chain to make delay adjustments. A simple antenna or RF sample (in FM+HD) with the appropriate amount of attenuation and is used to drive the device.
Many engineers with whom I�ve spoken are most comfortable with the device right before the exporter, so that any adjustments it makes to the delay are contained to just the digital audio, leaving the analog plant untouched. In that implementation, the existing diversity delay in the analog chain is increased to a larger number than previously needed; then, the device will makes up the difference in the digital chain. For example, if the analog processor is performing the delay at eight seconds, this could be increased to ten seconds. The automatic correction device in the digital chain would add two seconds of additional delay to the digital chain to match the larger delay in the analog chain. The advantage of going in this direction is that the devices can then correct the digital audio level and provide level alignment between analog and digital. This configuration will not work for any station that has to eliminate all delay for live programming and enter what is commonly referred to as �ballgame mode.� I have compiled a list of products that can help stations monitor and manage diversity delay. I have not personally tested each solution; the list should not be construed as a product review or endorsement. Also, this is an evolving space; check with each vendor about pricing and delivery schedules for the devices under your consideration.
Front panel of the Belar FMHD-1Belar FMHD-1. This broadcast monitor receiver makes a variety of HD Radio related measurements, including continuous measurement of diversity delay. The correction offset can be sent to a variety of devices, such as products from GatesAir, Nautel, Omnia, Orban, 25-Seven and Wheatstone. For devices which jump to the correction delay value, the FMHD-1 has an optional ramp function which sends the correction in user defined incremental steps. After a recent software release, the FMHD-1 can now scan up to six station presets and send correction codes to supported delay devices. The latest versions implement the most recent measurement filtration techniques developed in fall of 2015 by DTS. Belar is actively releasing new software for this device, so look for the latest version.
Delay offset corrections are made in the analog chain. For GatesAir or Nautel Exporters, the station may be configured in two different methods: You can employ the traditional Exporter delay model, configuration option �GatesAir FM� or �Nautel FM� meaning the analog air chain is looped through the exporter, providing the full analog diversity delay (7-10 seconds).
There is a newer configuration with options �GatesAir HD� and �Nautel HD� that includes a fixed analog delay in another device, such as an audio processor in the analog air chain at a high fixed value (i.e. 10 seconds). The engineer then loops the HD1/MPS digital input through the exporter�s delay to increase the HD1/MPS delay 1-3 seconds to match the HD1/MPS to 10 seconds on the analog.
Broadcast Electronics XPi10esp Exporter. This Exporter has a built-in diversity delay alignment feature. The Exporter can be provided with audio of the analog and digital from a receiver running in split mode, and configured to measure and maintain diversity delay. Note that the station must be configured to have the analog diversity delay fully provided by the Exporter for this feature to work.
DaySequerra M4DDC. This is a single-box solution, inserted in the AES path for either the analog or digital air chain perform, where it makes the corrections necessary for time alignment, level alignment and phase reversal correction. It has a web GUI and can send alarms via GPIO, SMTP (email alarms), or via SNMP. The web GUI can show you how well the device is working over a period of time in a graphical fashion. The latest version implements the most recent measurement filtration techniques developed in fall of 2015 by DTS. DaySequerra is actively releasing new software for this device, so be on the lookout for the latest version. This is an FM-only device; an AM version with slight hardware variations is expected in the future.
DaySequerra M4.2 TimeLock. This broadcast monitor receiver makes a variety of HD Radio-related measurements. It continually measures delay diversity, which is helpful for aligning stations manually. In addition, it supports automatic diversity delay correction with many of the Nautel, Omnia and Wheatstone audio processing products listed below. Support for delay controls in the GatesAir HDE-200 Exporter, Nautel Exporter Plus, and Omnia 11 audio processor is planned soon. The latest versions implement the most recent measurement filtration techniques developed in fall of 2015 by DTS. DaySequerra is actively releasing new software for this device, so keep an eye out for the latest version.
GatesAir HDE-200 Exporter. This device can receive correction information from the Belar FMHD-1. Future support for the DaySequerra M4.2 TimeLock is planned. With Belar�s latest software, the exporter�s delay corrections can be placed in either the analog or digital chain.
Online view of Justin 808 measurement dashboardInovonics Justin 808. This is a single-box solution that is inserted in to either (AES) analog or digital airchain and makes the corrections necessary for time alignment, level alignment and phase reversal. It has a web GUI and can send alarms via GPIO, SMTP (email alarms), or via SNMP. The web GUI can show graphics on how the device is working over a certain time period. The latest versions implement the most recent measurement filtration techniques developed in fall of 2015 by DTS. As of this writing, development of the Inovonics firmware is stable, and new features are being considered. Check for the latest firmware before proceeding. This is an FM-only device; there are no current plans to make an AM version.
Nautel Exporter Plus. This device can receive correction information from the Belar FMHD-1. Future support for the DaySequerra M4.2 TimeLock is planned. With Belar�s latest software, the exporter�s delay corrections can be placed in either the analog or digital chain.
Omnia .7, .9, .11. Any of these audio processors can work in conjunction with the Belar FMHD-1 to automatically adjust the analog diversity delay. The Omnia.11 requires version 1.6 firmware or higher. At the time of this writing, the DaySequerra M4.2 TimeLock integration was available with Omnia.7 and Omnia.9 processors, with support for the Omnia.11 planned for the near future.
Orban 8600, 8600S, 8500, 8500S, 5700 and 5500. Each of these audio processors can work in conjunction with either the Belar FMHD-1 or the DaySequerra M4.2. Orban notes that even the non-HD versions of these processors make diversity delay available for stations running separate analog FM and HD Radio digital audio processors.
25-Seven Precision Delay. This device can be inserted in the analog air chain to achieve diversity delay and supports automatic correction adjustments from either the Belar FMHD-1 or the WorldCast/Audemat Golden Eagle HD.
Wheatstone AirAura X3, FM531HD, FM-55, AM-55 and VP8ip. Each of these audio processors can work in conjunction with the Belar FMHD-1 or the DaySequerra M4.2 TimeLock.
WorldCast/Audemat Golden Eagle HD. Golden Eagle HD is a broadcast monitor receiver that makes a variety of FM and HD Radio related measurements, including diversity delay. It can work with the 25-Seven Precision Delay for automatic diversity delay correction. Also, it can be configured to monitor diversity delay alignment and send alerts if alignment (or other parameters) out of specification. All alignment measurements are available via SNMP for use by third party equipment.
The experience of over 10 years of HD Radio transmission has shown that the majority of the problems with blending artifacts can be traced back to individual broadcast stations. That�s in no way meant to demean the efforts of engineers out there; even the stations with the most diligent adjustments will still drift, for reasons not in control of station personnel. A systems approach, with continuous monitoring and real-time adjustment of diversity delay will reduce or eliminate the objectionable blending artifacts impacting listeners.