Wolfson sounds out digital audio design

An new philosophy for audio design incorporating ADCs, DACs, headphone drivers and speaker drivers will improve the energy efficiency in the audio signal chain, writes Ian Smith


An new philosophy for audio design incorporating ADCs, DACs, headphone drivers and speaker drivers will improve the energy efficiency in the audio signal chain, writes Ian Smith

Over the last decade or more the complexity of mixed-signal ICs for consumer audio applications has increased from a single stereo digital-to-analogue converter (DAC) IC to an IC with multiple DACs, analogue-to-digital converetres (ADCs) and audio digital signal processors (DSPs), to meet the ever-demanding needs of the consumer-electronics market.

This has been made possible by the availability of lower-cost finer-geometry silicon with higher-volume demand. This has required a new philosophy for ADCs, DACs, headphone drivers and speaker drivers, to improve the energy efficiency in the audio signal chain in consumer products. This article explains some of the techniques in each of these areas.

The desire has been to increase, or at least maintain, the analogue performance along the way, but on these smaller transistors, the maximum voltage allowed decreases too. If the voltage is halved, so too must the noise to maintain performance, which means the transistor current must quadruple and the power consumption double.

Furthermore, as you cascade analogue stages to meet the more-complex functional requirements of modern systems, the SNR is further degraded; for example, if you take two 100dB stages and cascade them, you end up with a path of 97dB. The analogue headache just got worse.

In the digital world, the smaller transistor, which operates at lower voltage, has smaller gate capacitance. As power consumption for logic gates is approximately proportional to the product of capacitance and voltage, the picture becomes much better overall.

The answer to improved energy efficiency and analogue performance is to use the digital part to augment the analogue performance. Detailed calculations and simulations can be done for each silicon process to see how the balance is best struck.

As we have moved from 0.5µm to 0.35µm to 0.18µm and below, it has become apparent that cascading stages is less favourable for power consumption and SNR. This means that the IC with multiple inputs to a stereo ADC with analogue mixing is no longer as efficient as an IC with multiple ADCs and digital mixing. The same applies to DACs too.



Figure 1 Old and new CODEC architectures

There are other advantages of having more digital logic:
• Cascading stages is far easier, as signal degradation is down to precise mathematical accuracy rather than imprecise analogue noise;
• The digital mixing and routing can be more flexible;
• Digital filtering is much easier for audio;
• Sections which are not used can be easily turned off for more power saving, and
• The manufacturing cost can be lower, as it can be more-cheaply tested with digital scan patterns rather than analogue performance measurements (it is difficult to speed-up audio tests, as you cannot change the frequency of the audio band).

On the down side, digital mixing of audio from different sources is not so straightforward when they are of differing sample rates, requiring a sample-rate converter; but the cost of this reduces as the geometry shrinks and transistors become cheaper.

Also, the whole digital solution is also far more complex and consequently takes longer to design and verify.

Another area where power can be wasted is when there is a residual dc voltage across a speaker or headphones.

Speaker drivers connect directly to the speaker in a BTL (bridge-tied load) configuration to maximise the power output from the low battery voltage. One terminal is driven in anti-phase to the other to quadruple the power output.

When idling, there can be a small offset of a few mV due to analogue circuitry variations, but this can amount to a significant amount of wasted power in to an 8ohm speaker if it is too big,  something, which can be minimised with careful design. Also, when the speaker driver is not required, the software can turn it off, and reduce the idle power from a few mW to under 1µW.

Years ago, a headphone driver output idled at half the battery supply rail voltage and produced an audio signal that went between ground and the supply rail. To remove the DC offset, a series capacitor was used. This component ended up being quite bulky with headphones, because the low impedance of the headphones demanded a large capacitance, so that the bass frequencies could be passed without attenuation – they formed a high-pass filter together.

An intermediate approach was to generate an artificial ground at half the supply voltage, to eliminate the need for the capacitor, but this needed another output, which wasted power.

Since then, a Class-W driver has been developed, which uses an intelligent charge pump to produce positive and negative supplies.

This removes the need for large capacitors, as the outputs can swing above and below ground, but can introduce a small dc offset and waste power. With the advantage of low-cost digital control now, there is an automatic calibration to minimise this offset.

Class W also includes a variable charge-pump voltage. Because the one audio IC generates the headphone amplifier supply and processes the audio signal, the charge pump can be adjusted continuously, so the headphone amplifier always has just sufficient supply to generate the audio output without unpleasant-sounding clipping of the transient audio signals.

This variable supply improves energy efficiency dramatically, because the majority of audio signals are well below the maximum output power of the amplifier. The same technique can be applied to an earpiece too.
The sp eaker driver in the past was also an inefficient Class-AB amplifier, but now is usually a Class-D amplifier, using a high-frequency switching waveform to drive the speaker. The speaker itself then filters the waveform to produce audio.

Because the amplifier’s output transistors are acting as switches and not resistors, efficiencies can exceed 90%. And when the amplifier is not needed, it is powered down and will draw under 1µW. Issues with Class D, such as EMI, are largely resolved as a matter of routine now.

As you can see, there are many techniques combined together in a single audio IC to maximise energy efficiency, which is vital for battery-powered devices, such as mobile phones.

Ian Smith is system architect at Wolfson Microelectronics

Leave a Reply

Your email address will not be published. Required fields are marked *


You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>