Almost all music we listen to these days is transported in digital form. But we hear analogue. Stephen Dawson looks at the all-important device that converts digital to analogue.

DAC is a slippery term at the consumer level. What we expect from a DAC has evolved over the years, although the letters represent a simple concept: Digital to Analogue Converter.

Back in the early to mid-1980s the DAC simply referred to a part of a CD player that converted the digital audio on a disc to analogue audio, ready to be fed to an analogue amplifier.

But people were sometimes unsatisfied with the DACs in their CD players and used third party DACs. So for many audiophiles the CD player was relegated to the role of CD transport, merely spinning the disc and extracting the digital data. That was decoded by an external DAC.

Since then, DACs have gone high resolution – supported higher standards than that of the CD. And more recently they’ve added the ability to decode audio from computers. Indeed, most DACs come equipped not only with S/PDIF inputs, but also with a USB socket so that they can become audio output devices for computers.

DIGITAL TO ANALOGUE
Analogue signals vary continuously. A microphone responds to continuously varying pressure waves in the air by producing a continuously varying voltage in a cable. An LP carries the music in the form of a continuously varying groove on its stylus. Your loudspeakers convert a continuously varying voltage into pressure waves in the air.

Digital slices up that continuously varying signal into a series of measurements at regular intervals. Or, at least, the digital scheme that is almost universally used in the consumer arena: PCM (for Pulse Code Modulation). DTS and Dolby and MP3 and almost all the others are simply methods of compressing the amount of data in PCM. When decoded, they all produce PCM.

The one exception is Direct Stream Digital, used on the Sony/Philips Super Audio CD, and is now becoming somewhat popular among some audiophiles as a download format. This is a Pulse-Density Modulation system, and works very differently. Here I’ll be sticking with PCM alone.

So instead of continuously varying like analogue, PCM merely ‘samples’ the signal periodically and keeps a record of those samples. How accurately it records the signal depends on two things: how often it samples the signal, and how precise it is in recording the value of each signal.

The first of these is called the sampling frequency. For a CD (and most MP3s) that’s 44,100 times per second, or hertz. For movies that’s usually 48,000Hz. For audiophile recordings, it can be 88,200 or 96,000, or even higher.

Nyquist and others determined that to record a signal to allow it to be perfectly reconstructed, the sampling frequency must be twice the maximum frequency of the signal you’re capturing. If we say that people can hear sounds up to 20,000Hz, then a 40,000Hz signal should suffice.

But in the real world there are complications. If a 44,100 sampling rate is used and the signal being captured has a component at, say, 25,000Hz, then the digital samples will be corrupted with incorrect values. Later when the PCM is converted to analogue, these will appear as audible artefacts within our hearing range. So a filter has to be applied to the signal before it is converted to digital to eliminate such content.

And when the signal is being decoded to analogue, the sampling frequency itself would form part of the signal, producing a powerful high-pitched squeal. For CDs, this would be at 22,050Hz and would likely damage loudspeakers (and infuriate dogs). So the signal has to be filtered again. Ideally 22,050Hz would be completely eliminated without having any effect on frequencies of 20,000Hz and below.

The other parameter is precision: how accurately the value of each sample is recorded. That depends on the number of bits used. An 8 bit system has to roughly manhandle the original value to the closest of just 256 levels. With 16 bits – the system used in CDs – there are 65,536 levels available. And for 24 bits, that goes up to 16,777,216.

Everyone agrees that the 8 bit, 11,025Hz sampling used on some early computer systems is way too rough to provide anything like truly accurate sound. Just about everyone agrees that 24 bit, 96,000Hz sampling captures the audio virtually perfectly (although some hold out for 192,000Hz).

Most consumer music, though, exists in 16 bit, 44,1000Hz format. And it’s about this that there’s the most disagreement.

IMPROVEMENTS
The very first CD player DACs were compromises. The technology available in the early 1980s when the CD first became available was expensive. The first Philips CD player got by with a 14 bit DAC. It just ignored the two least significant bits. The first Sony CD player used a 16 bit DAC, but only one of them so it had to switch between the left and right channels, delaying one of the channels slightly. That annoyed some purists, but it amounted to nothing more than one speaker seeming to be a quarter of an inch closer to you than the other.

A bigger concern was the difficulty of that final filtering stage – getting rid of 22,050Hz without hurting the high frequencies within the signal itself. This was hard to do with analogue filters. By hard, I mean impossible. A steep filter curve means introducing phase shift. A slow filter curve means too much HF gets through.

But ways around this were developed: principally techniques such as oversampling. The 44,100Hz sample rate was converted to a higher rate (typically eight times) and new samples interpolated. Then a gentle analogue filter could be used to eliminate high frequency artefacts while having just about no effect on the audible band. Timing is important in another way.

One possible problem with digital audio is ‘jitter’. This is where the sampling intervals – the time between each sample and the next – is not perfectly uniform. At the DAC stage this introduces distortion, just as though values of samples were being incorrectly calculated. The quality of the ‘clock’ in a DAC is therefore important. Good quality DACs rarely suffer from this.

CONNECTIONS
In the consumer field there are two ways of transporting a digital audio signal from a CD transport (or player) to a DAC, although they use the same protocol. That’s called S/PDIF, or Sony/Philips Digital Interchange Format. Sony and Philips jointly developed the CD, thus the use of their names. This is implemented by means of either electrical or optical connections. The first is generally called ‘coaxial digital audio’, while the other is called ‘optical digital audio’. The coaxial cables use a standard RCA plug. The optical ones use a Toslink connector.

Any decent external DAC will support both.

You might want rather more than CD standards of PCM. All decent DACs will support 96,000Hz, 24 bit audio, and most will go up to 192,000Hz. Recently Universal Music has taken to re-releasing classic albums in high resolution audio on Blu-ray, mostly at 96,000 sampling, but some are 192,000Hz. There are also various online stores for high resolution audio.

For many people, DAC doesn’t mean a gadget for converting CD digital audio to analogue, but for extracting high quality audio from a computer. These DACs will have a USB connection, typically a USB-B socket (the squarish type found on printers and other USB client devices).

In most cases you can simply plug this into a computer – Windows or Mac – and the computer will recognise the unit as an audio device, which it can use to output sound instead of its built in sound system. You may have to use Control Panel in Windows or its equivalent in a Mac to select it as your default device.

Recent versions of Windows support signals of up to 24 bits and 96,000Hz sampling with USB DACs. If the DAC supports it, they can go to 192,000Hz with an ASIO driver (Audio Stream Input/Output), but you will need playback software that supports ASIO. Foobar2000 can, with the plug-in, but Windows Media Player and iTunes don’t.

Up until Windows Vista, the default path for audio was through the Windows audio handling system. This included a mixer (to include system sounds) and would typically force all the output to a particular sampling frequency, resampling where required. That is not conducive to quality.

With Vista, and continuing on, a new audio handling system was introduced. WASAPI (for Windows Audio Session API) routes past the old routines, allowing your music player (again, I use Foobar2000 – you have to install an additional component for the purpose) to communicate directly with your hardware. That stops the mixer and the resampling, and it ensures that the most pure version of the music gets through to your external DAC.

The downside is that while in use you will get no system notification noises because the mixer is blocked out.

NOT QUITE THE END
An increasing number of DACs are beginning to appear which support Direct Stream Digital (as mentioned, the PDM system used in SACD), including a double data-rate version. DSD has proved to be a difficult digital format to work with when it comes to editing and mixing and such, so in professional circles it is converted to an ultra high resolution PCM format called DXD (Digital eXtreme Definition). This uses 24 bits of resolution and has a sampling rate of 352,800Hz, or eight times that used for the CD. That figure lends itself to relatively easy conversion to and from DSD, but it is even higher in resolution than DSD, so there is demand for the ability to use these files.

So, of course, already some DACs are appearing which support DXD.

All of which suggests that the development of DACs is a long way from complete. Who knows what new digital formats tomorrow might bring.