by Pat Brown
Pat Brown will help you understand the pro and cons of analog vs. digital interfaces.
Are you new to digital audio? Are you confused by the landscape of formats that have emerged in the marketplace? Following is a short expose that may shine some light on this potentially confusing topic.
Inputs and Outputs
It’s all about inputs and outputs (I/O). How do I get an audio signal from one to the other? The ongoing evolution of professional audio has produced a number of viable digital interfaces to complement legacy analog I/O practices. The choices may seem confusing at first, but when you break them down the strengths and weakness of each become apparent.
In this overview, I will start with analog since it is familiar to most readers and serves as a reference for the discussion of digital formats. I will focus on professional interfaces only. While similar in many ways to consumer I/O, professional are more robust against electromagnetic interference (EMI) and allow much longer cables – both requisites for large sound systems.
Figure 1 – An analog interface is point-to-point, with no exact requirements with regard to cable type, connector type, and cable.
A professional analog interface is a point-to-point connection between an output and an input (Fig. 1). It is electrically balanced, which provides strong immunity to EMI. Cabling is shielded twisted-pair. The interface is impedance mis-matched, where a low output impedance (typically < 200 ohms, or Ω) drives a high input impedance (typically > 2000 Ω). The impedance mismatch simplifies the interface by (usually) eliminating the need to consider specific impedance values. For example, a 100 ohm output would produce the exact same signal level into any high impedance input (10kΩ, 20kΩ, 30kΩ, etc.). An “output” simply connects to an “input” – end of story. It is often permissible to passively split an analog output to drive several inputs. It is not permissible to “Y” multiple outputs together.
The signal is in the form of a time-varying analog voltage that can span a level range of over 100 dB. Signals are classified by the magnitude of this voltage (e.g. mic level, line level, loudspeaker level). The signal flows in one direction only – from output to input. Cable lengths are limited by cable capacitance, and can approach 300 m (1000 ft) in some applications.
Analog connectors are classified by the number of electrical contacts. The impedance of cables and connectors is not a consideration, since analog audio signals are low in frequency in terms of the electromagnetic spectrum (Fig. 2).
Figure 2 – Analog audio (sometimes called baseband audio) is below 100 kHz in terms of spectral content. This is “low” frequency in terms of the entire electromagnetic spectrum.
- Simple and mature.
- Easy to troubleshoot.
- Does not require impedance-matching between output and input.
- Signal propagation through a device or down a cable is instantaneous, with no practical delay or “latency” that must be considered.
- The electrical properties of the cable can degrade the signal.
- Multiple audio connections, along with electrical “grounds” can produce ground loops, which may in turn cause hum and buzz problems.
- Multi-pair cables are heavy and expensive.
Digital I/O – Audio Industry-Specific Formats
The professional audio industry has standardized several point-to-point digital audio formats. These are designed to largely insulate the user from the complex inner workings of the interface and the complexity of the data. The most common are AES3 (USA) and AES-EBU (Europe). I’ll use AESx to refer to both. They are mostly identical except for a few details.
Figure 3 – A digital audio interface has more exact requirements than an analog interface. This includes an (approximate) impedance match, and there are less connector options.
Like analog, AESx is a one-way connection between an output and an input. AESx was designed to allow the use of balanced analog cabling and connectors, most often an XLR male output driving and XLR female input via a shielded twisted-pair cable. For short runs (< 30 m, or 100 ft), one may use the same cable as for a balanced analog interface. Low-capacitance cable designed specifically for digital I/O can extend the cable length to 100 m (300 ft). As always with cabling, your mileage may vary depending on the specifics of your application. Other cabling options exist including coax (AES3id – essentially a video interface) and category cabling (e.g. CAT5e). AESx can also be deployed using space-saving DB25 connectors on multi-pair cable (up to 16 channels). The exact implementation is product specific, and there are passive methods for converting between each. The manufacturer determines the specifics based on the target market for the product.
An important difference between analog and AESx is that an AESx connection carries two audio channels over a single twisted-pair. There is nothing special about the XLR connectors, and some products allow the user to select between analog and digital using the same input or output connector, saving space on the chassis (Fig. 4). The high compatibility with analog I/O cabling and connectors is a major strength of AESx.
Figure 4 – The input connectors for a four-input AESx processor. Note that for XLRs are required for analog, but only two are required for four channels of AESx (two channels on each connector). There is a switch to select between analog and AES (courtesy Marani).
Digital Audio Quality
The objective of digitization of the analog signal is lossless encoding. This requires a sample rate greater than 2 times the analog signal bandwidth, along with a dynamic range that approaches 100 dB. So, the main attributes of the data are the bit depth and sample rate. Most devices default to “24/48” which means 24 bit words flowing at a 48 kHz sample rate. Greater sample rates are possible, and are sometimes used for special applications.
The sample rate, bit depth, and number of channels can be multiplied to yield the “data rate.” This gives us a simpler, one number way to describe the digital audio resolution. Figure 5 shows the minimum data rate for one channel of full-range audio. A strength of digital audio in general is that the data rate can be reduced using lossy or lossless compression schemes. AESx signals are usually not compressed to reduce the data rate, since the minimum requirements for full resolution are easily met by the current technology.
Figure 5 – Sample rate and bit depth can be expressed as a data rate for the digital bit stream, a sort of conveyor belt for the one’s and zero’s that make up the digital signal. AESx digital audio consists of two channels, with a data rate of about 6 Mbps.
The bit stream contains the audio samples (or payload) along with metadata that carries information about the signal required for decoding. This “protocol” must be adhered to by both the output and input circuits, or no audio will flow.
Since the data rate approaches 10 MHz (a 1.5 MHz fundamental plus odd harmonics), the interface must be impedance-matched (110 Ω) to prevent degradation of the signal traveling down the cable by reflections and standing waves. Like all impedance-matched topologies, AESx is a one-to-one connection – one output drives one input.
Multi-channel versions include AES10 (MADI) and AES50 (HRMAI). While based on AESx, the details regarding clocking are quite different. These multi-channel interfaces are popular for connecting digital mixers to their respective stage boxes.
It’s About Time
All digital signals require a clock signal to keep multiple components in synchronization. AESx has the clock signal embedded in the data stream, so a dedicate word clock connection is often not needed for simple systems. AESx components often have word clock I/O for more complex systems.
The output of a digital component is always latent relative to the input. There is an unavoidable delay. Latency is cumulative, and system designers have a “latency budget” that must be observed to avoid timing issues.
- “Analog-like” regarding I/O.
- Two channels on one cable.
- High immunity to electro-magnetic interference and ground loops.
- 24/48 resolution is lossless with regard to analog I/O.
- Potential clocking issues.
- Requires special instrumentation to troubleshoot.
- Due to the high frequency nature of the signal, an impedance match is required between and output and an input.
- Output is always latent – the only variable is “how much?”
Digital I/O – Data Networks
Once the analog audio signal is digitized, it’s data. Technology has provided some very efficient means of data transport between computers, the most widespread of which is the Ethernet network. Audio-over-Ethernet (AoE) exploits the low cost and ubiquity of data networks to transport audio data. While AESx and its siblings are audio industry-specific, AoE utilizes technology from the Information Technology (IT) industry as a means of audio transport.
Figure 6 – AoE requires the connection of each component to a network switch. The I/O connections are made using a configuration program (e.g. Dante Controller™ or Cobranet Disco™).
The analog waveform is first sampled, and then “packet-ized.” A packet consists of a few audio samples and some additional “meta” data necessary to traverse the data network. The AoE protocol (e.g. Cobranet, Dante, Q-Sys, Ravenna – there are others) assures that the data gets “on” and “off” of the network to produce a contiguous waveform at the receiving end. High speed networks (e.g. 1000BaseT or “Gigabit”) can transport hundreds of channels. Unlike the “one-way street” AESx formats, a single cable can carry signals in both directions.
It’s Not Audio. It’s Data!
Regarding I/O, it’s a data network and the rules are rigid and established by the IT industry. The user is largely insulated from the complex inner workings of the network, and deployment in smaller systems is relatively plug-and-play.
Audio practitioners must acquire IT skills to deploy large AoE networks. Ideally, an audio-only data network is established using dedicated network switches. Increasingly, AoE may reside on the venue’s backbone with other data traffic such as email, web browsing, point-of-sale, etc. – the so-called converged network. On a converged network, the audio packets must have higher priority than other data traffic. This is called Quality of Service (QoS). A chunk of a large network can be reserved for audio data through use of a Virtual Local Area Network (VLAN). A VLAN requires a more sophisticated “managed” switch than a simple, dedicated network.
Due to the simple wiring, quantity of channels, and low-cost network hardware, AoE has become a very popular digital audio transport. A major difference between AoE and AESx (and its siblings) is that AESx, like analog, is point-to-point, requiring a direct connection between the output and input. AoE is a network, so signals can be routed between any network devices, regardless of where they are connected to the network. This provides extreme versatility regarding signal routing for larger venues, campuses, studios, etc.
The audio practitioner must learn the specifics of the AoE “flavor” that they are using. This includes running a control application on a PC to route audio signals, and configuring network switches for the required QoS.
Which AoE “Flavor?”
There are competing AoE formats. As of this writing, Audinate’s Dante™ enjoys the most widespread use. This can (and probably will) change with time, and competing formats such as Audio Video Bridging (AVB) (now referred to as Time Sensitive Networking (TSN)), will emerge as alternatives. AES67 is an initiative to produce inter-operability between the various AoE formats, so it may eventually be possible intermix AoE formats.
It is important to point out that AoE is just a transport. The quality of the digital audio is determined by the sample rate and bit depth. As such, all AoE formats should sound the same.
- 24/48 digital audio.
- Hundreds of channels on a single CATx cable.
- Ubiquitous cabling and connectors.
- Low-cost hardware infrastructure.
- Data networks can get complicated.
- Problems can be difficult to troubleshoot.
- Converged networks require cooperation with the IT department.
- No intuitive relationship to analog audio.
- QoS issues.
- Multiple AoE protocols exist, and no manufacturer supports all of them.
- IT is a different skill set, and requires specialized training.
Spectral Content and the Need for “Exactness”
One way to frame the differences between the formats is in regard to the requirement for “exactness” of the interface variables. This, in turn, is based on where the signal spectrum falls in the overall electromagnetic spectrum. As with most things audio, it is a gradient, with no clear “right” or “wrong” answer. Since audio by definition is low in frequency, analog audio interfaces are fairly permissive with regard to the interface details. One can sort of “get away with murder” and still get audio. Many wire and connector types will “work” but it is good practice to use twisted-pair cabling, ideally with a shield. Cable testing is basically a simple continuity test. The signal pair can be “bridged” with a butt set to detect the signal (Figure 7).
Figure 7 – A simple “butt set” is all that is required for analog troubleshooting.
Digital I/O moves further up the spectrum. The interface details become more black and white. For example, analog audio simply requires a low impedance output be connected to a high impedance input. All digital I/O formats require an impedance match, but AESx is more permissive than AoE. This explains why we don’t need “special” XLR connectors for AESx signals. It’s also why Euroblock and DB25 connectors can work. Even analog cabling can work at short distances. Cable testing is often performed by the audio gear itself, and whether the AESx input can “lock” to the AESx output. Special testers are available that allow the lock signal to be detected, and the audio decoded and listened to (Fig. 8).
At the highest frequencies (AoE on a gigabit network) the interface details are so stringent that they can’t be left to chance. The interface, cables, and connectors have an exact specification and no decisions are left to the end user. While continuity testers can work at the most basic level, the testing can (and should) be much more sophisticated. Along with cable testing there is a need to verify the system bandwidth, component addressing, Power-over-Ethernet (PoE) requirements, and more. Instrumentation for troubleshooting can be PC-based or stand-alone (Fig. 9).
Figure 9 – A stand-alone tester for network troubleshooting (courtesy Fluke).
Figure 10 – The requirement for precision in interfacing depends on the spectral content of the signal. The higher the data rate, the more the details matter.
The Inevitable Metaphor
If you read my stuff then you knew this was coming. Here’s a way to describe the various interface methods to the lay person.
Analog is like driving a jeep on a one-way dirt road. In spite of the bumps, the vehicle can weave all over the place and still get to its destination intact. AESx is like a Formula One car on an oval track. The higher speed requires a smoother surface and some boundaries. AoE is like a bullet train. There is an exact path and the margin for error is razor thin. At those speeds, any small discontinuity can have disastrous consequences.
This also explains why there are so many poorly performing analog systems in use. Impaired analog interfaces can still pass audio and no one may even realize they are impaired unless there is a direct comparison to a more ideal interface. So, as we move up in the spectrum from analog, to AESx digital, to AoE the interfacing details become more exact and there is less room to bend the rules. What we get in exchange for increased complexity are more channels and greater versatility – valuable benefits for many system types.
Digital audio has not obsolesced analog, and AoE has not obsolesced AESx and its siblings. All have their place. Analog still remains the standard by which to judge the fidelity of a digital system. Sometimes the simplest approach is the best, and sometimes it isn’t. pb