An analogue versus digital war is set to break out at Electronica over the design of Class-D audio amplifiers. Based on pulse-width modulation (PWM) techniques, Class-D amplifiers are widely seen as the next step for portable audio players and a new generation of small, power-efficient home amplifiers.
Cirrus Logic has deployed new digital logic-based techniques that the company reckons will slash emitted EMI and improve their signal-to-noise and distortion ratings over existing Class-D amplifier designs.
But UK-based Zetex, which is using a more analogue-oriented approach has claimed that it will be able to reduce noise and distortion more readily.
Mark Taylor, director of field applications engineering in Europe for Cirrus Logic, said the company's approach was a "true digital" implementation, based on the delta modulation part of a sigma-delta converter.
Cirrus' definition of 'digital' PWM for audio is based on the fact that the company does not use one of the older techniques used for Class-D amplifiers. This combines the audio signal with a triangle wave to produce a signal that can drive the PWM.
Skip Taylor, vice president of strategic marketing for PWM devices at Cirrus Logic, added: "We are taking a natural derivative of the delta-sigma converter. We are not doing a classic PCM-to-PWM conversion. The design lets us push the signal-to-noise ratio into the 110, 120dB range."
Zetex' marketing manager for media delivery products Colin Ward countered that traditional techniques are capable of meeting and exceeding Cirrus' claimed performance targets.
"Our noise floor is as good as theirs and we are achieving sub-0.1% THD [total harmonic distortion] across the board. We believe we can make amplifiers two orders of magnitude better than today.
"The key is the output stage. If you don't design the output stage correctly, you won't get the performance. Output magnetics give you low-cost filtering.
"We use a triangle wave with comparators to slice up the waveform but we use two and invert one of them to cancel the carrier," added Ward.
Cirrus' design is based on intellectual property from AudioLogic, which the company bought earlier in the year, with patents bought from UK-based B&W Loudspeakers. The IP has been combined with the sigma-delta technology developed by Crystal Semiconductor, which Cirrus bought in the mid-1990s.
"We are using patented feedback methods, to cope with imperfections in driver devices. That is the technology we bought with AudioLogic," said Mark Taylor.
Compared with a classic Class-D amplifier, Cirrus uses a higher frequency of between 350kHz and 500kHz.
"We have traded off about 1 to 2% efficiency for 30 to 40% lower EMC," said Mark Taylor.
Most existing Class-D audio amplifiers tend to switch in the 150kHz to 200kHz range because higher frequencies have tended to lead to reduced efficiencies and some high-frequency distortion. This is because of the dead time that needs to be inserted to prevent shortcircuits as one transistor stage is turned off and its complement turned on. Mark Taylor contended that the higher frequency made it easier to design a low-cost EMI filter that would allow the audio frequencies through.
Skip Taylor said: "Your filters become critical at lower frequencies. The filter components become more critical. But at high frequencies, you spend more time in the switching region, so efficiency goes down.
"To deal with that, we had to implement multiple feedback paths in the past. But that doesn't appear to be required in this modulator."
Ward countered: "The problem with the magnetics is not to do with rolling off the filter from 200kHz, it's the audio band going through it. Most magnetics saturate too early. We have developed a design that can achieve the performance of inductors that would otherwise be four to five times bigger.
"We believe we have the lowest RF noise of any system. It's all in the shape and choice of the materials."
To deal with the larger dead-time component, Cirrus has changed the way in which the power amplifiers are driven. Instead of driving them straight into the on and off zones, they are ramped more gently using current feedback.
Skip Taylor said: "We control the switching times of the different devices. We shift the times of the upper and lower devices to minimise dead time."
Ward said: "Dead-time correction tends to introduce crossover distortion, which is the worst form of distortion. You can use the magnetics to deal with conductor correction."
The B&W technology will be used in later devices to perform speaker and room compensation.
"Initial products will probably have a separate DSP for that but, as time goes on, we will have products aimed at speakers.
"For portables, we are looking at putting modulators on the back end of the decoder. Up to headphone power ratings you can drive direct out of CMOS," said Taylor. "Beyond 100mW, you are talking power FETs."
For the power devices, Cirrus has signed a deal with International Rectifier to make a set of specialised FETs for PWM-based audio amplifiers. To work with those devices, Cirrus has designed an on-chip driver scheme.
"Our output drivers can float with respect to the rest of the chip. It makes driving the power FETs directly possible," said Mark Taylor.