Preventing Pops and Glitches in AES3 digital audio for HD Radio and DTV
Digital Audio Testing vs Analogue Audio Testing
One of the key differences between analogue and digital audio transmission is the need to characterise the digital audio carrier (or bearer) signal separately from the audio content. Unlike analogue systems, digital transmission systems do not degrade the audio content gracefully or progressively. As the carrier degrades, audio transmission will suddenly fail. Symptoms can manifest themselves as audible pops and glitches, or sometimes even complete dropouts, a problem frequently encountered by broadcasters maintaining large systems with long and complex signal chains.
Audio Pops and Glitches at Cox Radio
Paul K. Reynolds, Chief Engineer of the Cox Radio Company in the greater San Antonio, TX area, was only too aware of the kind of problems that digital audio can bring to the broadcast engineer; "After installing a wireless Ethernet bridge between our studio and our transmitter we were hit with audible pops and glitches that would happen live on air once every few days - or worse, every few minutes. Debating how to make a test that would track down the cause became a hot topic in our engineering department. Then we heard about the Wednesday Webinar series hosted by Prism Sound, which had a session covering this very thing: Test strategies for reliable, high quality AES3 digital audio in HD Radio and DTV".
The Challenge Detecting Transient Audio Problems
The problem faced by Reynolds and his engineering team was that conventional audio analysis techniques assume steady state conditions. Typically one measures parameters such as frequency response, noise and distortion etc at a single moment in time, and assumes these values to hold true indefinitely. But this simply isn't good enough in the world of digital audio, where a system close to the edge may be working (and measuring) absolutely fine when probed at one moment in time, and then without warning can drop out completely, live on air.
Figure 1.1: Eye Diagram - short cable
Figure 1.2: Eye Diagram - long cable
Figure 2: dScope Screen (click for larger image)
Figure 3: Data Jitter
Figure 4: AES3 frame structure (click for larger image)
One Possible Factor Digital Audio Carrier Degradation
So how is this possible? Figure 1 shows the 'eye diagram', or the electrical carrier display of an AES3 digital audio signal in two different scenarios; eye diagram at the top in fig 1.1 shows an image of a signal at the receiving end of a short cable. Below it in Fig 1.2 the eye diagram shows the image of the same signal as received at the end of a long cable run, with the resistive and capacitive elements of the cable causing attenuation of the signal and degradation in the rise and fall times. There is also additional jitter added to emulate poor performance of the AES3 equipment at the source plus some electrical interference added to emulate nearby sources of noise encountered in a real world signal chain. As a result, the 'eye' has closed significantly, and is now close to breaching the (red) limits of receiver operation defined in the AES3 specification. Just a small amount of further degradation (maybe switching-in a slightly longer cable via a router, or a small increase in interference) could be enough for the receiver to interpret an incoming audio data bit incorrectly, potentially causing a pop or a glitch. Or worse still, the receiver could loose lock altogether, causing a complete drop out in the audio stream until the situation improves. And when communication resumes, everything then looks OK again from the point of view of conventional audio analysis tools.
The Solution - Prism Sounds dScope Series III
During Prism Sound's webinar, Reynolds and his team at Cox Radio learned about proven test techniques for the task of tracking down troublesome components and interfaces in broadcast audio. Paul realised that; "The dScope Series III Digital/Analogue audio test instrument designed and manufactured by Prism Sound was precisely the tool we needed to help find what was causing these elusive pops and glitches. Although the instrument can measure both analogue and digital performance, we were interested primarily in its digital analysis capabilities. We wanted to use the instrument's digital measurement tools both on-air and off to help us get to the bottom of what was plaguing our broadcast."
Reynolds adds; "After we received the dScope Series III instrument we immediately put it to work probing our signal chain. We were quickly able to check the AES3 Channel Status, frame rate, and the rest, and we soon discovered that the dScope Series III had some powerful and unique capabilities. This included a Watchdog function providing us with a comprehensive log of failures over an indefinite amount of time. This log consists of state changes such as biphase violations, loss of lock, eye narrowing, and the like, as well as numerical results measuring our digital audio transmission path. The log also included any limits that we defined that were breached, along with a stamp of the date and the time of occurrence. Now we had a way of putting those pops and glitches together with any digital audio technical hiccups occurring in our broadcast chain - and we could correlate the time stamps on the log file with events that happened in our facility that may have been aggravating factors."
Tracking Down the Causes of Pops and Glitches
Figure 2 shows how Prism Sound's dScope Series III addresses detection of pops and glitches. There are a number of unique tools and techniques that can be used to track down these kinds of problems: The digital carrier can be analyzed to check for amplitude and jitter problems, and can be monitored and logged over time, allowing both immediate and post-mortem analysis. Unrelated to the integrity of the electrical carrier signal, the AES3 format includes embedded channel status metadata in addition to the audio data, and this metadata can be decoded by the instrument and checked for consistency relative to the audio data. Results can be logged and alarms can be set, allowing for a number of modus operandi including notification via email to the appropriate technical staff or department at the time of failure.
Catching and Logging Data Errors (Pops and Glitches)
And equally importantly, the audio data itself should be analyzed for errors. The precise method for audio data analysis will depend on the nature of the signal chain involved - for transparent digital audio signal paths such as routers, cabling, patchbays etc, a pre-defined data sequence in conjunction with direct inspection of the digital audio data can be used to monitor bit transparency. For non-transparent signal paths such as sample rate converters, signal processing, mixers etc, a low distortion sinusoid in conjunction with a peak THD+N measurement can be employed to pick up even the smallest degradations in the audio signal. The use of peak detection in the THD+N measurement is crucial; for example, an error in the least significant data bit may not cause a sufficiently large contribution to the RMS distortion to reliably detect a fault, but this error will cause a significant glitch in a peak measurement.
The dScope analysis panel in Figure 2 shows a number of measurements on a digital audio signal (for a non-transparent path), which includes a barely-audible or visible data error in the time domain trace of the sine wave (green). This glitch manifests itself as a very large peak in the residual THD+N trace (yellow), and is flagged as an error in the numerical distortion analyzer result (the reading panel flashing red). At the same time, details of the error are written to a log file, and inspection of the digital audio carrier shows us that severe carrier degradation is the likely culprit for causing the data error. We also see that the Channel A 'Invalid' flag is set in the AES3 stream, and there are warnings about the measured digital audio frame rate and wordlength being inconsistent with the channel status metadata. These issues could also lead to problems with glitches and mutes. This complement of measurement parameters, logging and alarms is not found in any other instrument of its kind in the world.
Digital Audio Carrier Analysis Jitter
If a jitter problem is detected in the digital carrier analysis, as in the example here, it is important for the broadcast engineer to determine whether the problem is due to the source equipment, the interface cabling, or potentially both. Even in an ideal world, where source clocks are jitter free, the mere use of cables to carry biphase mark encoded data (as used in the AES3 digital audio interface standard) actually causes jitter. Figure 3 illustrates this phenomenon: In the AES3 interface format, a 'zero' in the digital audio data sequence is represented by a single pulse with a width of two Unit Intervals, or UI (a UI simply represents the quantum pulse duration in the AES3 data stream, regardless of the sampling rate). A 'one' is represented by a pair of pulses of opposing polarity, each with a width of a single UI. These varying pulse widths result in varying pulse heights at the receiver due to the charge / discharge effect of the capacitive interface cable. This translates into timing variations in the zero-crossings of the received data bits, which by definition is jitter.
Isolating Fs Jitter and Data Jitter
By analysing the digital audio carrier at different points in the AES frame (Figure 4), we can determine how much jitter is present in the 4 bits of static preamble, which is directly attributable to source clock timing ('fs Jitter'), and we can compare that to the jitter measured in the 16-24 dynamic data bits of the AES frame ('data jitter', or 'intersymbol interference'). This enables us to isolate source-clock jitter problems from cable-related jitter problems.
Reynolds explains how he used this technique to help him: "With the dScope Series III we were able to determine numerically whether the jitter we were measuring in the chain was clock jitter stemming from a specific piece of equipment, or intersymbol interference induced by our cable runs. This proved invaluable in determining and isolating the degree to which each piece of equipment and its associated cable runs contributed to any jitter in our chain, and enabled us to confidently say whether a given component or interface was causing a problem.
The Portable Solution Prism Sounds DSA-1
"Out at the transmitter site we used Prism Sound's DSA-1 portable analyzer. This is a hand-held battery operated instrument that makes getting around tight rack spaces easy. Like the dScope Series III, the DSA-1 analyzes and measures digital performance and has built in automated tests for trouble shooting the integrity of digital audio transmission lines and equipment."
Conclusion Problem Solved?
So after all of this digital audio probing, are the pops and glitches now eliminated? Reynolds concludes: "The Prism Sound dScope Series III and DSA-1 were invaluable in helping us narrow down the culprit causing our pops and glitches. We successfully debugged all of our audio hardware and AES3 interfaces, and we now believe that there may also be an issue with the way clock/timing packets are handled across the ethernet link. We are working with the manufacturer to help solve this problem now that we have confidence in the rest of the system."