Fullscreen HTML5Med Res (??MB)HTML4

Digital Amplifiers are here! Analog Audio could soon be dead! Audio recordings have been revolutionised by digital technology, which has allowed dramatic improvements in signal-to-noise ratio, distortion, wow/flutter and frequency response. Now the same kind of revolution is well under way in an area which many of us probably still see as the exclusive domain of analog circuitry: audio power amplifiers. By JIM ROWE 6  Silicon Chip W hen Compact Discs burst onto the audio scene in the early 1980s, they quickly changed the definition of ‘hifi’. Suddenly we had a recording and playback technology which could deliver a signal-to-noise ratio of 96dB, distortion levels below 0.01%, negligible wow and flutter and a frequency response which was near enough to ‘ruler flat’ over the complete audible spectrum. CDs delivered these benefits mainly because they took advantage of digital technology. Instead of trying to record audio waveforms faithfully in the grooves cut into vinyl records, they ‘sampled’ the waveform 44,100 times per second and turned it into a stream of binary numbers – ones and zeros – which could be recorded and played back much more faithfully. This made it possible to reconstruct a much more accurate replica of the original, when the digital samples were converted back into analog form. The same sort of benefits came when digital technology was applied to tape recording to provide us with DAT (digital audio tape). And the improvements continue, with the new enhanced digital recording techniques such as HDCD (High Definition CD), SACD (Super Audio CD) and DVD-Audio – which are just coming onto the market. But until very recently, the high quality audio available from these digital technologies still had to be converted back into analog for the last crucial step in the audio chain: power amplification to drive the speakers. That’s because up till now, analog circuitry has provided the only way to achieve a high-quality audio power amplifier. Sony’s Playstation 2, announced and released with such fanfare a few months back, contains a digital audio amplifier courtesy of Tripath. Sony’s new VAIO (Video Audio Integrated Operation) notebook computers also contain a similar digital audio amplifier. Even today, some audiophiles will tell you that the only kind of power amplifier worth listening to is one with a class A (or at worst, class AB) push-pull output stage, with a whopping great power supply, plenty of heatsinking and loads of negative feedback. And there’s the rub: traditional high quality analog power amps and their power supplies are big, heavy, expensive and wasteful of power. These disadvantages have been becoming more and more of a problem as manufacturers were pushed to make their products smaller, lighter, more efficient and of course, cheaper. Many products do need to incorporate audio power amps – and in some cases more of them than ever before, like home theatre systems and computer sound systems. So there’s been a huge amount of R&D effort invested in digital technology to achieve the kind of improvements with audio amps that it’s already provided in areas like recording and playback. Lofi digital The first digital amplifiers to come out of this R&D were pretty terrible and anything but ‘hifi’. Sure they were efficient but the signal-to-noise ratio was poor and most of them radiated so much switching noise that you couldn’t even bring a distortion meter near them, let alone measure their distortion level! In the last couple of years, though, that R&D has really begun to bear fruit. Many of the latest stereo TVs and home theatre systems from firms like Sony, Sharp and Hitachi now have true dig- This Evo 200-2 “bel canto” digital audio amplifier has been receiving outstanding reviews around the world – but then with its $US2395 price tag it would want to! Specs are 1Hz-80kHz, 240W RMS per channel. Unfortunately, though, one reviewer was a little over the top in saying it was at least as good as the best tube (valve) amplifier he had ever heard. . . July 2001  7 Tripath’s TA-3020 (right) and TA-2022 (below) digital amplifier driver chips. They’re already used in a range of consumer electronics equipment and they’ll be found more and more in the future. ital audio amps, as does Sony’s new Playstation 2 and its VAIO handheld computers. Apple Computer’s latest Power Mac G4 computers also use them, as do Altec Lansing’s latest PC speaker systems. Car audio firms like Alpine and Blaupunkt are using them in their latest models, and some of the new personal audio players and mobile phones are said to use them as well. Manufacturers like Sharp have even released true audiophile-level digital amplifiers, with specs that more than compare with the best traditional analog designs. The end of analog? In short, the writing is on the wall: the future of high quality audio amplifiers is eventually going to be digital, probably much sooner than most of us expected. Is this just because designers have found ways to get acceptable performance out of digital amplifiers or because they can get away with smaller heatsinks and power supplies? No, although those factors obviously help, it’s because digital amplifiers actually make more sense, now that so much audio material is being recorded and transmitted in digital form. With digital amplifiers the audio can be kept in this form right up to the high power level, ready to drive the speakers – which not only makes the whole system more efficient but provides the potential for even higher quality of reproduction (because there’s less signal processing). For example, audio is stored on the HD layer of the new SACDs in ‘Direct Stream Digital’ form, which is already 8  Silicon Chip how designers have been able to come up with digital amps which achieve true hifi performance, as well as very high efficiency. Let’s see how they’ve done it. How they work suitable for passing directly through many of the new digital amps. In fact, this is the best way of achieving the full enhanced audio performance from SACDs, rather than converting the signal back into analog form and feeding it through a traditional analog amp. The same applies to normal CDs and the new DVD-Audio discs, even though their LPCM (linear pulse-code modulation) audio has to be passed through additional upsampling and other digital processing circuitry. Even compressed multichannel digital audio like Dolby Digital, MPEG2 and DTS bitstreams can give better results, as the decoding and other processing can again all be done in the digital domain where they introduce far less noise and distortion. So there are compelling reasons for digital amplifiers. In a couple of years, just about all new consumer equipment is likely to have digital amps. Probably diehard audiophiles will keep using traditional analog amps (hey, some are even still using valve analog amps!) but eventually.... By now you’re probably wondering Digital amplifiers don’t handle multi-bit LPCM digital audio – the kind of 16-bit samples that come from a CD or the higher resolution 24-bit samples from DVD-Audio. Not directly, anyway. They handle a ‘bit-stream’ or single-bit digital signal. Single-bit signals have been used in recent years to make many of the highest quality digital master recordings. Well known examples are Philips’ Bitstream, Matsushita’s MASH and Sony’s Super Bit Mapping (SBM). Another general name for this format is Sigma-Delta Modulation (SDM), and it’s used for making master recordings because it’s actually capable of much higher fidelity than LPCM. The resolution can be equivalent to 24 bits or better, with a frequency response of DC to 100kHz and a dynamic range of over 120dB, as a result of noise-shaping techniques. To produce normal CDs, master recordings made using SDM are converted into PCM by digital processes known as decimation and brick-wall filtering. This inevitably causes some degradation, which is why Philips and Sony decided that to ensure higher fidelity on SACDs they would directly record the SDM bitstream – renamed Direct Stream Digital. Part of the reason why SDM can deliver much higher fidelity is that it’s a lot simpler to convert analog audio into SDM, than in a conventional multi-bit analog-to-digital converter. This is done by a Sigma-Delta modulator. As you can see from Fig.1, Sampling Clock (fc = 64 x fs) MIXER ANALOG INPUT + NOISE SHAPER & INTEGRATOR QUANTISER PDM OUT – (NEGATIVE FEEDBACK) Fig.1: A basic sigma-delta modulator, which converts analog audio into a PDM digital bitstream. They’ve been used for years as A/D converters for making digital master recordings. ANALOG INPUT Similarly, any analog audio signals must first be converted into PDM using sigma-delta modulators, before they can be amplified (see Fig.4). So this is the basic way digital amps work. But, as you might expect, the various manufacturers have made their own modifications to enhance the performance. Let’s look a little more into the technology of practical digital amplifier chips. Tripath’s DPP PDM BITSTREAM OUTPUT Fig.2: How the pulse density in a PDM bitstream corresponds to the amplitude of the original analog signal. this consists of an analog mixer, a noise-shaping filter/integrator and a quantiser. Think of the quantiser as essentially a comparator which is gated by a high frequency clock signal, and driving a storage flipflop. Typically the clock frequency is 2.8224MHz, or 64 times the 44.1kHz sampling rate used for CDs. Because of the negative feedback loop, the integrator’s output represents the difference between the input voltage and the digitised output from the previous sample, accumulated over the 354ns sampling period. So the quantiser’s next output will be a 1 if this accumulated difference is positive (ie, the input voltage rose slightly), or a 0 if it’s negative. The output from the modulator is therefore a single pulse train of bits at the 2.8224MHz clock rate, with their density representing the instantaneous amplitude of the original audio waveform. This is shown in Fig.2. Another name for this kind of A/D conversion is pulse density modulation or PDM. As well as being a very simple and direct method of A/D conversion, sigma-delta modulation has much higher inherent linearity than a multi-bit A/D converter. As a result of the very high sampling rate it also doesn’t need sharp-cutoff ‘brickwall’ filters to prevent aliasing. And although the basic modulator shown in Fig.1 does tend to have a poor signal-to-noise ratio, it turns out that this can be dramatically improved by using ‘higher order’ noise shaping circuitry. This involves additional feedback and integrators, with the effect of shifting most of the noise up and out of the audio range. The signalto-noise ratio in the audio range (even up to 100kHz) can thus be improved to 120dB or so. Finally, a PDM bitstream has another huge advantage over multi-bit LPCM: it’s much easier to perform digital-to-analog (D/A) conversion. In fact, with PDM you don’t need a complicated D/A converter at all, just a simple low-pass filter after the power amp’s output switching, as shown in Fig.3. In fact, Fig.3 is so simple that you can see why digital amplifiers are going to take over. PDM makes everything a lot easier – it’s still a true digital signal but one that’s very easy to convert back into analog. Virtually all the new digital amps handle the digital audio as this kind of single-bit stream, so apart from the DSD audio from SACDs, all other kinds of digital audio have to be converted into this form by over-sampling and other digital processing. +V One of the leading players in producing ‘hifi’ digital amplifier chips is Tripath Technology, a fairly small firm in Santa Clara (California) founded in 1995 by semiconductor industry veteran Dr Adya Tripathi. Tripath chips are being used in Sony’s DAV-S300 compact home theatre system, in the Playstation 2 video game box and VAIO computers, and also in Sony’s new plasma screen TVs. Apple Computer is using Tripath digital amp chips in its latest Power Mac G4 computers, as is Hitachi in its own new 81cm plasma screen TVs. Marantz and Carver are apparently using Tripath chips in some of their latest hifi amps, while they’re also being used in car audio systems from firms like Alpine and Blaupunkt. Tripath describes its digital amps as having a ‘Class T’ configuration, to distinguish them from early ‘Class D’ digital amps which used fairly basic pulse-width modulation (PWM) technology. They stress that a Class T amp is very different from a PWM amp, both because of its output configuration and because it pre-processes the incoming digital bitstream using their patented Digital Power Processing (DPP) technology. They’re rather coy about the exact PDM OUTPUT ANALOG OUTPUT Q1 PDM INPUT MOSFET DRIVER CIRCUIT L Q2 SPEAKER C PDM TO ANALOG CONVERTER! –V Fig.3: A major advantage of using PDM for digital amplification is that it can be converted back to analog simply by passing it through a simple low-pass filter. July 2001  9 BASIC DIGITAL AMPLIFIER +V PDM INPUT FROM SACD, ETC PDM OUTPUT ANALOG OUTPUT Q1 MULTI-BIT LPCM DIGITAL INPUT OVERSAMPLING DIGITAL FILTER ANALOG AUDIO INPUT SIGMA-DELTA (S-D) MODULATOR MOSFET DRIVER CIRCUIT L Q2 SPEAKER C –V Fig.4: A very basic digital amplifier configuration. The amplifier can accept the PDM signal from a SACD player directly, but multi-bit digital audio and analog audio signals require some input processing. details but it appears that their Class T power amp is actually a high power sigma-delta modulator stage with feedback right around the power switching stage and high-order noise shaping. The DPP block which drives it appears to use a technique of digitally modulating the basic DPM bit-stream. Its clock signal frequency and exact timing are varied (from 200kHz up to 1.5MHz) with the sensed analog signal level, to pre-compensate for the switching limitations of the power Mosfets. So the power amp switches at around 1.5MHz for low output levels but slides down to 200kHz for full analog output. It seems to be a bit like spread-spectrum technology but the end result is very linear and ‘clean’ analog output after the final low-pass filter. Currently, Tripath makes three different integrated digital stereo amp chips, all with impressive specs. The lowest power TA1101B device is a 30-pin PSOP package measuring only 11 x 16 x 3.4mm but it gives 2 x 10W RMS into 4Ω loads at 0.04% THD+N (distortion and noise) running from 12V. The efficiency is about 80%, so it needs only minimal heatsinking – an array of vias under the package, to conduct heat from its heat slug through to a copper pattern about 10 x 32mm on the other side of the PC board. At the other end of the range is the TA2022, which is a 32-pin SSIP package and delivers 2 x 90W RMS into 4Ω loads at 0.1% THD+N, running from ±35V and with almost 85% efficiency. For even higher power applications Tripath also make digital amp 10  Silicon Chip drivers, which provide everything for a complete Class-T stereo power amp apart from the power switching FETs. There are four of these amp drivers, one of which (the TA0104A) teams up with suitable power Mosfets to deliver up to 2 x 500W RMS into 4Ω loads (or 1000W in bridge-mode mono), at 0.05% THD+N and running at ±92V. You can find more information about Tripath’s digital amplifier devices at their website; www.tripath.com They have data sheets and application notes you can download as PDF files, along with white papers. Their patents are also available from the US Patent & Trademark Office website, at www.uspto.gov/patft/ Apogee’s DDX Another leading player in the digital amp field is Apogee Technology, of Norwood in Massachusetts. Apogee’s amplifier chips are used in Altec Lansing’s latest PC speaker systems and the firm also has a strategic partnership with European chipmaker STMicroelectronics to develop their technology jointly. Apogee’s chips are based on their patented Direct Digital Amplification or ‘DDX’ technology, which as the name suggests is designed to take a direct digital audio input and carry out all necessary processing and amplification in the digital domain. A feature of DDX is that it carries out special processing to convert the incoming digital audio into a pair of modified PDM/PWM bitstreams. These are then used to drive the output switching Mosfets in a special way known as ‘damped ternary’, to give high efficiency and improved audio performance. Instead of the two levels (1 and 0) in a normal binary waveform, the pair of digital signals produced by the DDX processor drive the Mosfets to produce a signal with three levels: +1, 0 and -1. (Hence the name ternary, meaning ‘of three parts’.) This output signal has much less switching noise and sampling clock content than a standard PDM/PWM signal, especially at low signal levels. Not only that, the speaker load ends up being much better damped as well, giving better transient performance. Fig.5 shows how with a normal PDM AudioSource’s “Amp Seven”, a 200W/8Ω (500W mono bridged) digital amplifier which uses Tripath’s Class T technology. Distortion is less than 0.01% and it is stable into 2Ω loads. It accepts standard (ie analog) input signals but processes digitally. (www.audiosource.net). +V Q1 DIGITAL OUTPUT HAS FULL RAIL-TO-RAIL AMPLITUDE EVEN FOR ZERO ANALOG OUTPUT (0V) (0V) PDM OR PWM INPUT MOSFET DRIVER CIRCUIT L Q2 SPEAKER C –V Fig.5: With standard PDM or PWM, the amplifier’s digital output signal has a very high ‘RF’ content even for zero analog output. This calls for large filter components to get the electromagnetic interference down to acceptable levels. or PWM binary drive signal, a binary 1 switches on one output Mosfet (s3ay Q1) to apply the full positive supply voltage to the speaker (via the low pass filter), while a binary 0 level switches on the other Mosfet Q2 to apply the full negative supply voltage. These are the only two output levels available, but the amplifier switches back and forth between them at a very high rate to provide an average level which varies with the audio output waveform –after it’s filtered, of course. Note that for all audio signal levels between peak positive and peak negative, the output Mosfets have to be switched on in turn during each clock period (or in alternate clock periods, with PDM) in the time ratio necessary to give the right instantaneous average level for the audio waveform. So for zero output voltage, they must each be switched on for 50% of the time – giving a very low audio voltage after the output filter. However, at the same time there will be a very high level of sampling clock signal (and its harmonics) at the input to the filter. DDX neatly avoids this problem when it converts the PDM signal into the drive signals for its ternary form. The processor works out the degree to which the two output levels would cancel each other at the output when they were averaged, and then it removes this part of the digital signal in advance. So only one or the other of the output Mosfets is turned on during that sampling period, for a proportion of time which gives the right average value. Or if the audio waveform should have a value of zero at that instant, neither main Mosfet is turned on at all. Fig.6 shows the idea, using an a 4Ω load, again with less than 1% THD+N and running from 28V. To make a complete digital stereo amplifier, a DDX-2060 is driven from the DDX-2000 controller chip. This is a 44-pin quad plastic flat package measuring only 10.5 x 9.5 x 2mm. It provides all of the processing to convert a serial multi-bit digital audio signal (from say an S/PDIF receiver) into a stereo pair of Apogee’s ternary drive signals – to drive each channel of the 2060. The DDX-2000 also provides a digital volume control function, managed by an external controller via an I2C bus. Apogee provides details of a complete 5.1-channel surround sound amplifier system which uses three DDX-2000 controller chips to drive four DDX-2060 power amp chips. With the controller chips running from 3.3V and the power chips from 28V, this configuration can take the serial digital signals from a Dolby Digital decoder chip and provide 4 x 35W for the FR, FL, SR and SL speakers, 35W for the FC (centre) speaker and 70W for the LFE subwoofer. A single volume microcontroller can adjust both master volume and channel balancing. The second of Apogee’s DDX controller chips is the DDX-4100, which provides all of the processing to provide DDX drive signals for five separate DDX power amp channels (ie, 4.1 channels of surround sound audio), from two different types of digital input signal: either S/PDIF (Sony/ Philips Digital Interface) stereo, or I2S/AC’97 (a Microsoft development) four-channel inputs. This means that one DDX-4100 can drive three DDX-2060 power amp chips to provide an all-digital 4.1-channel surround sound amplifier for PCs, delivering 4 x 35W plus 1 x H-bridge output switching circuit. As you can imagine, this reduces the level of clock signal ‘RF’ at the amplifier’s output dramatically, and makes it possible to use significantly smaller values of L and C in the filter circuit. In addition, though, the DDX power switching driver circuit itself pulls a neat trick. When neither of the ternary drive signals is at the 1 level, it turns on the two lower output Mosfets (ie, Q2 and Q4). So instead of getting any output energy, the speaker voice coil has a low resistance connected across it: the ‘on’ resistance of the two lower Mosfets (typically a fraction of an ohm) plus the DC resistance of the filter inductors. This applies heavy damping to the voice coil, which improves its transient behaviour. (Editor’s note: we don’t quite believe all that but a simple damping test would show if it was true.) The net result of the DDX technology is a digital amp which has very good signal-to-noise performance, low RF radiation and good speaker damping, while still offering very high electrical efficiency even at low signal levels. Apogee offers a single DDX stereo power amplifier device, which can be driven by either of two DDX processor/controller chips. The DDX-2060 power amp is a 36-pin package measuring 16 x 11 x 3.6mm. It can deliver 2 x 35W RMS into 8Ω loads with less than 1% Another AudioSource model, the “Amp THD+N, running from 28V. Six” modular dual-channel (stereo or Alternatively, it can deliver bridged mono) digital amplifiers again using 70W in mono mode into Tripath’s Class-T technology. July 2001  11 +V Q1 DDX 'DAMPED TERNARY' DIGITAL OUTPUT HAS NO PULSES FOR (0V) ZERO ANALOG OUTPUT (PDM) L1 Q2 PDM INPUT SPEAKER (LPCM INPUT) DDX PROCESSOR DDX DAMPED TERNARY DRIVE SIGNALS DDX POWER SWITCHING CIRCUIT C +V Q3 (ANALOG AUDIO INPUT) L2 Fig.6: Apogee’s DDX technology pre-processes the PDM/PWM bitstream to remove the ‘cancelling’ components, so the output switches produce a ‘damped ternary’ pulse waveform with much lower RF content. The lower Mosfets (Q2, Q4) are also turned on to heavily damp the speaker between the pulses, giving better transient performance. 70W of high quality audio at over 85% efficiency. The DDX-4100 also provides sample rate conversion, digital bass, treble and volume controls, bass management for the LFE channel and parametric equalisation for all five channels. All this comes in a 44-pin TQFP package measuring 10mm square! Apogee’s website at www.apo-geeddx.com provides full data on their chips and a white paper on DDX technology. Incidentally, you might be wondering what the difference is between "4.1 channel" surround sound and "5.1 channel" surround sound. 4.1 channel sound is basically what you get from matrix-type ’analog’ surround decoders like Pro-Logic: front left and right, plus centre front (basically L+R) and a single surround channel (basically L-R), with the ‘0.1’ channel carrying the bass from the centre front channel, LP filtered to run a subwoofer. On the other hand 5.1 channel sound is usually that from a digital decoder, with discrete FR, FL and FC front channels, two surround channels (SR and SL), and a discrete ‘low frequency effects’ or LFE channel for the subwoofer. The term ".1" is used to indicate 12  Silicon Chip Q4 the fact that the frequency response of that channel is deliberately limited to cover bass frequencies only; this channel is invariably used by the subwoofer. Only the start... Tripath and Apogee are not the ony firms working on all-digital amplifier chips. Cirrus Logic has just released a new Crystal ‘TrueDigital’ PWM amplifier controller chip, the CS44210, which is a complete digital stereo amplifier apart from the output switching Mosfets and their driver ICs. It provides all processing for up-sampling and sample-rate conversion from up to 24-bit digital inputs, digital volume, bass and treble controls, muting and de-emphasis, and even low power digital outputs to drive stereo headphones – all in a 24-pin TSSOP package measuring just 7.8 x 4.4 x 1.1mm. Together with a set of driver chips and power Mosfets, the CS44210 forms a digital stereo amp with 2 x 50W output into 8Ω loads, with an output dynamic range of 100dB and an energy efficiency of 90%. You’ll find full data on the Cirrus chip at www.cirrus.com Texas Instruments has released a LP FILTER (DAC) low-power PWM amplifier chip, the TPA2000D2 (2 x 2W RMS into 4Ω at 5V), with higher rated versions due shortly. Both Motorola and National Semiconductor have announced similar products, while Philips plans to have digital amp chips available later in the year. Other firms are already active at the equipment level. Last year, for example, Sharp Corporation released its SM-SX100 single bit 2 x 100W stereo amp offering true audiophile quality with a breathtaking $25,000 price tag. This year they’re releasing two mini stereo systems using the same all-digital technology and outputs of 20W/channel and 25W/channel respectively, with price tags below $3000. It’s very likely that most of the main consumer audio manufacturers will announce digital amps and systems before long, so prices will soon plummet. The digital amplifier era has definitely begun. Before the year is out you could be to be listening to one in your car, lounge room or computer room – or in your Walkman or mobile phone. Welcome to the all-digital audio future! SC