Digital Audio Interfaces

A digital audio interface allows equipment such as audio mixers, recording devices or audio processors to be interconnected without degrading the signal quality. Unfortunately, there are different types of interface, most of which are incompatible, although in some instances you can wire a device with an AES/EBU interface to a device fitted with a S/PDIF socket.

Most interfaces convey stereo (2-channel) audio over a single set of wires, although some also require a separate word clock (WC) circuit, which is often wired via a standard BNC connector.

Common interfaces are described in the following sections.

Sony/Philips Digital Interface (S/PDIF)

S/PDIF conveys both channels of a stereo signal, usually via an RCA phono (PIN) connector, although some devices use a TOSLink fibre-optic connector. This interface is commonly used for domestic equipment, such as CD/DVD players, Digital Audio Tape (DAT) recorders, and MiniDisc (MD) machines, conveying 16, 20 or 24-bit samples at rates of up to 96 kHz.

• An S/PDIF circuit can be used to convey data instead of normal audio, including AC-3 data for Dolby Digital 5.1 surround sound, or DTS surround sound data.

Electrical Hardware

S/PDIF in its ‘copper’ form uses a signal level of between 0.5 and 1 volts into a 75 Ω load over an unbalanced circuit. The similarity to a standard television signal makes it suitable for routing through equipment designed for video material. Since it’s meant to operate with 75 Ω cable it can also be used over long distances without equalisation, and in this regard is superior to the professional AES/EBU interface (see below).

• Some devices incorporate an output transformer for improved circuit isolation.
• Although a standard RCA phono (PIN) connector is normally used for ‘copper connections’, some professional devices and adaptors employ a BNC connector.
• You may encounter equipment that has a fibre-optic connector, usually of the TOSLink variety, using 1 mm plastic fibre-optic cable and visible red light. The high attenuation in this kind of fibre limits the maximum cable length to 10 metres or less. In addition, you may need an adaptor box to connect other S/PDIF or AES/EBU devices.
• PC-based hardware sometimes provides a digital output via a 2-pin header, also known as an HDR-2 interface. Although the data conforms to the S/PDIF standard, the 5 volt TTL signals are incompatible with S/PDIF.

Connecting to devices with an AES/EBU interface is possible, although the following points should be noted:-

1. A standard AES/EBU input accepts data with minimum eye height of 200 mV. This means that it should accept the S/PDIF minimum signal level of 500 mV peak-to-peak.
2. An S/PDIF input normally requires an attenuator to protect it from being damaged by the comparatively high level signal produced by an AES/EBU output.
3. The AES/EBU interface and S/PDIF convey different data flags. This means that you may find that a connection, although electrically satisfactory, simply refuses to work.

Further information about connecting different digital interfaces appears elsewhere in this article.

Encoding

S/PDIF is a domestic version of the AES/EBU interface (see below). The data is assembled in the same way, except for differences in the Channel Status bits (C-bits). It’s possible to interconnect AES/EBU and S/PDIF devices, either directly or via an adaptor box, the latter correcting differences in signal levels or modifying any offending C-bits.

• S/PDIF is sometimes called CD/DAT, IEC 958-11 or IEC 60958, although there are some differences in settings of C-bits between the S/PDIF and IEC 958-11 standards (see below).

• What appears to be an AES/EBU interface may actually be an SP/DIF interface with extra balancing circuitry, also known as the IEC 958-11 interface. Equipment with this kind of interface may not work due to incorrect C-bits, such as bit 0 which is different for domestic and professional audio material.

• More about encoding can be found in the section concerning the AES/EBU Interface.

Channel Status Bits

The C-bits for S/PDIF are organised as follows:-

The most important bits are usually set as follows:-

whilst bits 4-8 are normally at 0. Assuming an SP/DIF input ignores bit 0, bit 2 is 1 (Copy Protect set to off) and bit 3 is 0 (No audio emphasis) then transfer from an AES/EBU source is possible. Bit 2 is usually at 1 in the AES/EBU interface, allowing copying via this kind of connection, even though copying the same material via an S/PDIF circuit may be prohibited.

• When copying between DAT machines the interface conveys indexing information, also known as start IDs. You should therefore turn off auto-indexing on the receiving machine.

Copy Protection (CP)

Apart from the Copy Protect bit (see above), you may not be able to copy material via an S/PDIF when:-

1. Using an early DAT recorder

   Older DAT machines prohibit recording at 44.1 kHz to stop piracy of CDs, although recording at 48 kHz is possible. Such machines can sometimes be modified to record at all rates.

2. Using a DAT recorder fitted with SCMS

   Modern domestic DAT machines use the Serial Copy Management System (SCMS), which allows the recorder to make digital copies from some sources, such as an original CD, but prevents you making a second-generation DAT recording from such a copy. Hence a recording’s owner can copy material but endless duplication is prevented. Analogue copying is unaffected but ‘second generation’ copying via S/PDIF is again prohibited.

• SCMS is part of the S/PDIF standard and is not used for data transferred via an AES/EBU interface. Some professional machines with S/PDIF connections aren’t fitted with SCMS.
• Enabling Copy Protect whilst making an original recording prevents any digital copying.

SCMS works by recording two bits of data, collectively known as ID6, which are buried within the data recorded onto DAT. These bits control Copy Protection (CP), working as follows:-

The system works in conjunction with the S/PDIF Category Codes (see above). The following table shows how this operates in an SCMS-equipped DAT recorder:-

• Digital copying from an SCMS DAT machine to a pre-SCMS machine is only possible if ID6 is set to 00. This means that even ‘one-copy-allowed’ tapes will be blocked.
• Some professional machines allow you to manually set the condition of ID6.

AES/EBU Interface

This interface, created by the Audio Engineering Society and European Broadcasting Union (AES/EBU) is really a professional version of S/PDIF. It’s connected using a 3-pole XLR connector or, when fitted on a digital audio card, via a standard quarter-inch 3-pole jack.

The signal voltages used by the AES/EBU interface are much higher then those used by S/PDIF and can therefore damage the latter if a direct connection is made. However, the actual data conveyed by the interfaces are so similar that they can sometimes be connected directly or via an adaptor box.

Unfortunately, some equipment responds to the consumer/professional flag within the data stream, refusing any material from the wrong category of equipment. There isn’t any easy way of getting around this problem, apart from buying a specialist device that can modify the flags in the data.

16 or 20-bit samples at 48 or 32 kHz are preferred, although 44.1 kHz can also be used. Unlike S/PDIF, this interface can’t normally be used for 24-bit audio, since the necessary bits are required for other purposes.

• Some of the information in this section also applies to S/PDIF.
• An AES/EBU circuit can convey data instead of audio, such as AC-3 data for Dolby Digital 5.1 surround sound, or DTS surround sound. Alternatively, it carry Dolby E data, accommodating eight channels of broadcast-quality digitally compressed audio, which can be transferred to a pair of audio tracks on a video tape recorder (VTR).

Electrical Hardware

The stereo data is carried on a balanced RS-422 circuit with alternating current (AC) coupling, using a signal level of 5 to 10 volts peak-to-peak, working into a 110 Ω load: a transformer is used inside some equipment. The signal can travel over 350 metres of cable without any problems and for even longer wiring you can install equalisation circuits or a repeater device.

• The permissible cable length can also be further extended by using transformers at each end of the circuit. These normally take the form of an inline device, fitted with an XLR plug or socket at one end and a BNC connector at the other. The signal conveyed via the latter is at a lower impedance of 75 Ω, requiring the use of matching coaxial cable.

As mentioned above, 3-pole XLR connectors are used. The connections are often marked as Digital In (DI) and Digital Out (DO) to avoid confusion with analogue circuits that are also on XLRs.

Further information about connecting different digital interfaces appears elsewhere in this article.

Encoding

Bi-phase mark encoding is used, eliminating any direct current (DC) component, which means that reversing the wires in the balanced circuit has no effect. This form of encoding also ensures that many of the transitions in the data match the timing of the bit clock, as shown below:-

This arrangement, known as self-clocking, allows the clock signal to be easily extracted at the receiving device. The bit rate is determined by the sampling frequency, as shown below.

Organisation of Data

Data frames are sent out at the same rate as the sampling frequency. Each frame contains two sub-frames, the first containing audio for channel 1 (left), the second for channel 2 (right). The interface also carries non-audio data bits, spread over a number of frames and grouped into blocks. Each block consists of 192 frames, from 0 to 191. Hence the start of frame 0 is called the start of block.

Synchronisation

At the beginning of each sub-frame there’s a preamble, which is used by the receiver to extract the sample clock, ensuring that devices at each end of the digital link are synchronised. As shown in the above diagram, there’s normally a transition in the data for each pulse of the bit clock. However, the preamble is made unique by breaking this rule. This coding violation allows the preamble to be easily identified, without confusing it with other data. Different preambles are used to identify each channel and the start of a data block, the latter always being in channel 1. These are illustrated below:-

Sub-frame Structure

The sub-frame is made up of the following bits:-

The V, U, C and P bits constitute 4 bits in every sub-frame; in other words, 8 in every frame. Each of these bits appears in every one of the 192 frames contained in each block.

The V and P bits are ‘tied’ to the associated audio sample, whereas the U and C bits, each with a total capacity of 192 bits or 24 bytes, can carry information as required. Note that the data in subsequent blocks needn’t be identical. It may be updated at the interface’s frame rate; around 200 times a second.

Channel Status Bits

Equipment responds to C-bits in various ways. Some devices ignore certain data bits, allowing the audio data to be used, whilst others ‘lock out’ the material if the C-bits aren’t as expected.

The C-bits in the AES/EBU interface operate as follows:-

The usual states for these bits are as follows:-

Bits 5-7 are often set to 0 whilst many devices simply ignore bits 8 to 127. When copying between DAT machines the interface doesn’t convey indexing information, also known as start IDs.

Synchronisation Techniques

In a complex studio system there can be problems with timing between the different digital devices. There are two basic methods of synchronising the equipment in such a system:-

Time Delays

Signals are synchronous if the start of the preamble is within a set margin of the reference clock. Outputs should be within ¹⁄20 of the sampling frequency and inputs within ¼. It may be necessary to align the timing of clock inputs by adjustments within the equipment, but once set, it shouldn’t need to be changed again, since drift isn’t usually a problem.

• Most timing delays are caused by the effects of different lengths of cable.

• Any jitter in the digital waveform can cause serious problems. This blurs the edges of the data causing an increase in noise on the audio output.

Unsynchronised Devices

Some sources can’t be locked to the studio system, such as a CD player that provides a nominal output of 44.1 kHz but really operates at 44.098 kHz. Unfortunately, such devices often lack a clock input, whilst others run at the wrong rate, such as a CD player supplying a 44.1 kHz signal to a studio working at 48 kHz. There are three solutions to such problems:-

1. Use Analogue Circuits

   This can be a very cost-effective remedy, causing little degradation in quality.

2. Use a Sample Rate Convertor (SRC)

   An expensive option, often best avoided, except when converting between frequencies. The device adds or deletes samples as necessary, sometimes causing a small amount of distortion.

3. Use a Synchroniser

   The ideal solution where rates are nominally the same. A buffer within the device fills or empties over time, depending on which frequency is highest. The buffer is reset at its mid-point, when a 10 ms cross-fade is made between the current sample and the mid-point data.

An alternative type of synchroniser, known as a sample-slip synchroniser, works by dropping or repeating samples during silent periods. Both types of synchroniser perform the special operation once every 20 minutes for a 10 parts per million (ppm) error in sample rate. Another option is a short-term SRC, which uses interpolation to fix the problem at a faster rate.

Other Interfacing Problems

Most of the following problems can be solved, albeit with expensive hardware:-

1. Differences in bits per sample

   Even when things seem normal there may be problems with signal levels or distortion when connecting devices that use a different bit counts, such as a 16 bit ADC connected to 20-bit device. It may be necessary to ‘re-dither’ the material or use a digital gain control box.

2. Clash of C-bits with S/PDIF

   See the section in this document that gives further details on S/PDIF.

3. Some C-bits giving incorrect flags

   One example of this is pre-emphasised audio data which actually lacks a no-emphasis flag.

4. Pre-emphasis

   Digital processing can introduce or remove emphasis, although not all devices can deal with emphasised data. CCIT J17 (NICAM 728) emphasis should only be used for broadcasting purposes.

5. DC offset

   This problem is usually caused by a badly aligned analogue-to-digital converter (ADC). If this happens, you may have to use a special device to apply digital high-pass filtering.

6. Asynchronous timecode

   In this situation the timecode contained in the AES/EBU data gets out of step with real time. A special device can be used to remove or replace the timecode in the data.

Sony Digital Interface (SDIF2)

SDIF2 is rarely encountered, although it’s commonly employed in early analogue-to-digital converters, as used to adapt a video recorder for digital sound recording.

Electrical Hardware

The interface uses separate 75 Ω BNC connectors for left (L), right (R) and word clock (WC) signals. Each unbalanced circuit uses 5 volt transistor-transistor logic (TTL) levels into a 75 Ω load with direct current (DC) coupling. In a multi-track system several SDIF2 signals, in the form of balanced RS-422 circuits, can be wired via a single 50-way D connector.

Encoding

SDIF2 conveys audio in 16-bit to 20-bit form, complete with Emphasis and Copy Protect flags. Unfortunately, some equipment ignores these flags, requiring the use of manual switching. Each 32-bit slot, which occupies one cycle of the word clock, is made up as follows:-

The most significant bit (MSB) of the audio data is sent first, irrespective of the audio word length. Two’s complement coding is used and unused bits are blanked to 0.

Synchronisation

Although a clock is coded into the data itself, it’s ignored by some equipment. This means you’ll have to connect a separate word clock (WC) circuit. If such a clock isn’t provided you’ll hear a cyclic hiss, typically at one or two Hz. The WC circuit carries a square wave signal at the sampling frequency, the rising edge of which is aligned to the start of each data slot.

• The WC output of older devices, such as the Sony PCM1610 and 1630 systems, which operate at 44.1 kHz, don’t provide an accurately phased reference and should not be used.
• WC outputs are often fitted to items of equipment that don’t have SDIF2 connections.
• A single 75 Ω terminator must be fitted at the end of a clock circuit. If a clock feeds several devices, usually via T-adaptors, it should be connected to the end of the chain.

Mitsubishi Interface

This interface, also known as Melco, is similar to SDIF2, but doesn’t contain a clock signal in the data. A balanced RS-422 connection is used for each channel and separate word clock (WC) and bit clock (BC) connections are required. It operates with 16-bit or 20-bit audio.

• The status information for a ProDigi multi-track tape machine is carried over an extra two channels, similar to two audio channels. The first is for tracks 1-16, the second for 17-32.

Yamaha Stereo Interface

This interface first appeared on Yamaha’s famous DMP7 MIDI-controlled mixer, as well as on its digital successor the DMP7D, allowing several mixers to be ‘cascaded’ together. The connection is made via an 8-pole DIN plug, as shown below:-

This connector conveys two RS-422 signals, one for the audio data and the other for the word clock, wired as follows:-

Both circuits use direct current (DC) coupling.

Multi-channel Interfaces

A multi-channel interface allows multi-track digital audio equipment to be connected over a single electrical circuit. The most common systems are described below. Unfortunately, some these interfaces are proprietary designs that are incompatible with other systems.

A-DAT Interface

This proprietary interface provides connections to an Alesis Digital Audio Tape (A-DAT) machine, which employs a standard S-VHS video tape to record multi-track digital sound. An optical connector conveys up to eight channels of audio, sampled at 44.1 or 48 kHz. This means that two connectors are needed for an eight-track machine; one for the eight inputs and another for the outputs. In the same way, a 16-track machine requires a total of four A-DAT connectors.

Digital audio cards that accommodate interface usually have a separate 9-pin synchronisation connector, which lets your computer control the transport mechanism of an associated A-DAT machine.

Multi-channel Audio Digital Interface (MADI)

MADI conveys multiple digital audio channels over a single circuit, each channel conforming to the AES/EBU standard. The 56 AES/EBU sub-frames are carried over a 75 Ω coaxial cable, up to 50 metres long, and fitted with BNC connectors. A separate synchronising clock cable is required.

The interface operates at a fixed data rate of 125 Mbit/s, whatever sample rate is used, although the rate at the cable is reduced to 100 Mbit/s by using 4 to 5 bit encoding. This process breaks up each 32 bit sub-frame into 4-bit words, encoded as 5-bit words by means of a look-up table. This form of encoding reduces the direct current (DC) content of the signal.

Synchronisation blocks are inserted at least once per frame. If the link isn’t used to it’s full capacity, extra synchronisation blocks are inserted to ‘fill’ the space on the bus. This is done using a device known as a Transparent Asynchronous Xmitter and Receiver Interface (TAXI).

The AES/EBU sub-frames are as normal, except that the preamble bits 0-3 are replaced by:-

Tascam Digital Interface (TDIF)

This interface is used for connecting Tascam multi-channel digital tape machines. It employs the same optical connector as the A-DAT interface, conveying the same number of channels, although it uses an entirely different data format.

Yamaha 8 channel Interface

This special interface consists of eight sets of RS-422 mono audio data and a clock signal, all wired via a 25-way D connector. The data itself can be in Yamaha, SDIF2 or Mitsubishi format.

When used in Yamaha format this works in the same way as Yamaha’s stereo interface described above. However, if you feed stereo data into one channel of a multi-channel interface only the left-hand information is received. Also, there’s no direct method available for connecting two channels from a multi-channel device to an input that has a stereo interface.

Digital Audio Connectors

Most connectors for digital audio are in coaxial form, as used in radio frequency (RF) and video systems. They have a central signal pin, surrounded by a cylinder that provides screening. An appropriate coaxial cable should always be used.

BNC

This twist-and-lock coaxial connector is used for SDIF2 and various kinds of video interfaces. It’s also sometimes used for S/PDIF connections in professional systems. The 75 Ω type of connector is most common. Unfortunately, it’s rather too easy to plug this into a similar 50 Ω socket, which often results in jammed or damaged connectors.

Phono Plug

This popular connector, also known as an RCA or PIN plug, is used for S/PDIF connections, as well as for video and audio connections in domestic equipment. The older ‘long’ variety of plug can cause problems with sockets that are designed for the modern ‘short’ style of plug.

Recent connectors of this type are gold-plated and highly reliable, although older and cheaper versions are often shoddy. Those that come already moulded-on to a cable are surprisingly good. However, you should ensure that digital audio wiring is always made of real coaxial cable, since conventional audio cables often aren’t suitable for use at high frequencies.

TOSLink

This form of fibre-optic connector, which is often used for S/PDIF circuits, employs a 1 mm plastic fibre-optic cable and visible red light. The high attenuation introduced by this form of fibre limits the maximum cable length to 10 metres or less.

XLR Connector

An exceptionally robust connector, also known as a Cannon plug, since this company was one of the first to make this product. The full ‘Cannon’ range of professional connectors come in various types with differing numbers of pins. The 3-pole XLR version can be used for an AES/EBU digital signal or a mono analogue signal over a balanced circuit. Normally the connector is wired in ‘XLR’ order, as shown below. The different terminologies for the wiring are shown for reference:-

Multiway Connectors

The shape of a D connector, often used for multi-track audio data, is similar to an elongated ‘D’. It comes in 9, 15, 25 and 35-way form, as well as high density 15 and 50-way versions. The locking screws, often in UNF form, are essential, although metric screws are usually required for attaching connectors to Japanese hardware.

An Amphenol connector is similar to a D connector, but has 14, 28, 36 or 50 plug contacts spread over a central projection. The latches are awkward but are absolutely essential.

Digital Audio Interface Solutions

Variations in the AES/EBU interface and S/PDIF can make it difficult to connect some devices. The following information is suitable for an electronics hobbyist who wants to create the necessary hardware. Of course, these only work if the actual data is suitable for the receiving device.

AES/EBU to S/PDIF

The signal level used by the AES/EBU interface is meant to be between 5 and 10 volts peak-to-peak whilst that for S/PDIF is normally in the range of 0.5 to 1 volt. To convert from one to another you can use a simple attenuator, as shown below.

The following circuit is even simpler. However, this variation may upset AES/EBU interfaces that employ electronic balanced circuits in place of a standard transformer.

A transformer can also be added to the output of such an attenuator to ensure that the S/PDIF signal is properly isolated from the ground circuit. This should be in the form of a standard pulse transformer with a winding ratio of 1:1, of the kind commonly fitted in computer network cards.

The less active may consider a ready-made attenuator, usually in the form of an inline device with an XLR socket at one end and a BNC connector at the other. This may also include a transformer.

SP/DIF to AES/EBU

This is slightly trickier, since you’ll need to increase the signal level. This circuit, derived from articles on Usenet, employs a logic chip containing six inverters, such as the 74HC04 or 4049:-

You may have noticed that this circuit doesn’t have an output transformer, although this can be added to make the whole thing comply exactly with AES/EBU specifications.

TTL to S/PDIF

As already mentioned, some PC-based hardware provides a digital output via a 2-pin header, also known as an HDR-2 interface. This signal is at Transistor Transistor Logic (TTL) level, giving a signal of 5 volts, which is incompatible with S/PDIF circuits.

The kind of device used for the TTL output may or may not be capable of driving the 75 Ω load presented by an S/PDIF circuit. If it can deliver 12 mA into a 450 Ω load you can employ a simple capacitor-linked attenuator, as shown here:-

Once again, a transformer can be added to the output of this circuit so as to provide ground isolation in strict conformity to the S/PDIF standard (see above).

If your device doesn’t have enough current capacity you can use almost any 5 volt driver chip with an attenuator similar to that shown above, with or without a transformer. This example uses two inverters from a 74HC04 logic chip:-

whilst the circuit below uses the 7HCU04 device from Philips and a custom transformer.

S/PDIF to TTL

Under some circumstances you may want to convert an S/PDIF signal into a form that can be fed into a standard computer circuit at TTL levels. Here’s a simple solution, again using the 7HCU04 chip:-

whilst the simple circuit shown below includes an adjustment for DC offset, with the voltage on the ‘traveller’ of the preset normally set to around 2.6 volts.

References

Digital Audio Problem Solvers, Francis Rumsey, Studio Sound, July 1991

Elektor Electronics magazine, Jul/Aug, 1995

Interfacing, Synchronisation and Communication, Francis Rumsey, Digital Information Exchange 1989

Interface and Control for Digital Recorders, Phil Wilton, Broadcast Systems Engineering, January 1987

The Truth about SCMS, Francis Rumsey, Studio Sound, May 1991

©Ray White 2004.