High-fidelity audio testing and the latest semiconductor solutions

If you’re designing the analog portion of an audio system, it helps to know two things: what the listener expects, and how you can objectively prove that you’ve achieved those expectations. But correlating engineering measurements with something as subjective as hearing is very difficult. In June of this year, it became clear to me that hearing is not a linear continuum.

National Semiconductor wanted to show me some products that have been optimized for the hard-core, two-speaker, hi-fi enthusiast who demands absolute perfection in their audio systems. These products include a series of high-voltage op amps with very low distortion and drivers for audio power amplifiers. Both are designed to replace the large number of discrete components in high-end hi-fi preamps and amplifiers. Engineers designing high-end audio systems are still clinging to discrete devices, but National Semiconductor has begun to change their thinking.

National Semiconductor has set up an acoustically ideal sound room, complete with a pair of Wilson Audio's professional Watt-Puppy speakers and a reference-design analog audio chain built around the new chip. It also has its own CD to showcase the features revered by the faithful, but I was invited to provide my own discs, too. So I pulled a disc of Beethoven piano sonatas from a shelf in the car's sun visor. A grand piano is a delicate instrument, and Beethoven piano sonatas are not likely to be subjected to much post-processing "enhancement."

My expectations were low, since my hearing has been diminishing with age, cars, planes, motorcycles. Knowing this, I walked into the listening room, where its $26,000 speakers hooked up to custom National Semiconductor electronics were overkill for my $50 ears. That is, I figured $50 was all I could spend on headphones or speakers for my personal use. The result? I was shocked by what I heard.

With the National Semiconductor demo, I could tell which high note was the longest. I could also hear the sound the singer made with his lips, the sound you make when you part your lips and there's some saliva on them. On my Beethoven CD, I could hear the pedal action not as a mechanical action, but as the artist's foot tapped with a different tone - it was right on the beat.

I was puzzled. The most I knew about the Golden Ear was that one could hear these sounds without headphones. How could I hear so many new sounds from recordings I had listened to hundreds of times? I posed this question to Mark Brasfield, senior audio applications engineer at National Semiconductor and the creator of the listening room (Figure 1).

Interestingly, Brasfield admits that he has hearing loss, but says that industry research shows that, at a certain point, people with reduced hearing acuity can actually hear more high-end audio than people with perfect hearing.

Professional Testing

I asked Bruce Hofer, founder and president of test equipment maker Audio Precision, perhaps the most visible maker of pure audio test equipment (Figure 2), what kind of tests I could use to compare audio systems (see "The Challenges Of Audio Testing" at www.electronicdesign.com). But if I wanted to make a definitive measurement, Hofer's words discouraged me.

"Thirty years on, people are realizing that the ear and brain system is not well understood," Hofer said. "At Audio Precision, we do objective measurements on audio equipment. But the world is all about perception and measurement, which, by its very nature, doesn't quite track with what you're actually trying to measure."

Hofer said Audio Precision does not have audio distortion analysis, which means it is currently just "big/good/bad." His test equipment simply provides measurements of quantity and performance, usually cutting along different axes, such as frequency response, distortion factor or noise. Hofer said that while these indicators are all valid, the behavior of the ear is much more subtle than this.

Many companies and researchers use humans in their evaluations, rather than high-precision test equipment. Hofer says the best examples include Fraunhofer and Dolby, which use humans to listen to real recordings to measure how well lossy compression algorithms capture real-world performance (see “Testing For Audio Transparency”).

Hofer explains that the analytical measurements that Audio Precision's analysts make are designed to avoid the failures of subjective measurements by describing highly mathematically defined signals. Describing the effects of distortion, hum, and noise relative to the original mathematically defined signal is a straightforward process.

"But even so, if you record a wide range of multi-frequency inputs digitally, measurements will show that some tones are completely lost, rather than attenuated," Hofer said. "Lossy compression algorithms are designed to not waste any bits on certain parts of the signal because those parts do not meet the threshold of audibility."

If the algorithm is truly optimized, the human ear and brain will not notice the missing information even with analog test equipment.

Other Views

So, if designers can't identify objective measurements that correlate to subjective experience, what should they do? Clearly, most companies focus on power efficiency and integration levels.

I have talked to some of National Semiconductor's competitors in the analog audio chip market, and none of them are pursuing the extreme hi-fi market. Instead, they focus on mobile phones, personal media players, and home theater. Most companies believe that direct comparison of listening experience is impossible.

When it comes to earbud/headphone oriented devices, such as MP3 and AAC players, such comparisons are especially impossible, they say, in part because of the contribution of the actual acoustic transducers connected to or in the ear, which are significantly higher than any electronic components in the signal chain.

This is also impossible, in part because the processed digital signal deceives the ear in various ways. For example, the digital processing tries to make you think that the sound source you hear is somewhere in front of you, rather than in the middle of your head, which is what you hear as a pure stereo signal. In fact, to a considerable extent, Fraunhofer and Dolby algorithms can simulate the full surround sound of a decent set of headphones or high-end earbuds.

With this in mind, other chip companies have recently announced their analog audio products (including Class D amplifiers as part of the "analog" category).

Texas Instruments has introduced many new chips, many of which are Class D power amplifiers. For example, TI currently ships its DRV600 stereo 600Ω line driver, which does not require coupling capacitors at the input and output.

For driving speakers, instead of 600ω lines, TI's TAS5162 stereo digital amplifier power stage can drive 6Ω bridge-tied load (BTL) up to 210W per channel with only 10% total harmonic distortion. Its efficiency is greater than 90%. For maximum dynamic range, it can operate at 12V and 50V supplies respectively. For low output level requirements, TI's TAS5176 can drive six channels at 15W per channel or three channels at 30W per channel. Both chips only require simple LC output filters to eliminate Class D pulse modulation signals.

TI's TAS5414 and TAS5424 four-channel digital audio amplifiers are targeted for use in automotive headlights and external amplifier modules. These two amplifiers provide four continuous 23W channels into 4Ω or 30W into a 2Ω speaker at less than 1% THD+N. The difference is that the TAS5414 has a single-ended input, while the TAS5424 is a differential input.

To further enhance the signal chain, integration is critical. TI's TLV320AIC3101 for digital cameras is a low-power stereo audio codec with a stereo headphone amplifier, digitally controlled stereo microphone preamplifier, and automatic gain control amplifier (AGC) with mixing/multiplexing capabilities between multiple analog inputs.

The chip's programmable filters can eliminate zoom motor noise. The playback path includes mixing/multiplexing functions from the stereo digital-to-analog converter (DAC) and selected inputs to different outputs through a programmable volume control.

Analog Devices (ADI) has also been developing Class D amplifiers. The ADAU1590 and ADAU1592 are dual-channel BTL power amplifiers with sigma-delta modulators to drive pulse modulation. This allows the microcontroller to interface with control reset, mute, programmable gain amplifier (PGA) gain, and fault reporting output signals. For using separate modulators, the ADAU1513 is the basic dual-channel power stage.

Maxim Integrated Products offers a speaker driver entry, but it is not a Class D device. Instead, Maxim announced two Class G speaker amplifiers, the MAX9730 and MAX9788. The former is general purpose; the latter is optimized for driving ceramic speakers.

A Class G amplifier has a push-pull stage similar to a Class AB, but adds a second, higher-voltage supply that is removed only when signal peaks exceed a preset level. In the case of the MAX9730, this will drive a 2.4W 8ω load from a 3.3V supply. Piezo speakers are different because they require large voltage swings to achieve enough deflection to move enough air to produce a loud noise. The chip's charge pump can deliver more than 700mA peak output current at 5.5V DC, which guarantees output to a 14V peak-to-peak piezo speaker.

For digital camera audio recording, Maxim announced the MAX9814 microphone amplifier with automatic gain control (AGC) low-noise microphone bias. Integrated AGC allows designers to optimize signal levels before digital signal processing. The device also integrates a low-noise preamplifier, variable gain amplifier (VGA), output amplifier, and an internal low-noise electret microphone bias generator.

Behind the Hi-Fi

National Semiconductor's presentation was centered around two products. The first was a pair of audio op amps with a typical THD+N of 0.00003% (guaranteed to be 0.00009% at maximum). Other performance indicators included 2.7nV/√Hz input noise density, 60Hz 1/f noise corner, 20V/μs slew rate, and 55MHz gain bandwidth.

The 44V LME49860 dual op amp is available in two different packages, and the 34V LME4978x0 has single, quad and dual op amp versions. The ±22V rated LME49860 is unity gain stable down to ±2.5V. Over this supply range, the amplifier maintains a common mode rejection ratio (CMRR) and power supply rejection ratio (PSRR) better than 120dB and a typical input bias current of 10nA.

At the input, the LME49860 can drive a 2kΩ load within 1V per supply swing or a 600Ω load within 1.5V. The LME49710, LME49720, and LME49740 amplifiers have lower operating voltages and similar specifications.

Another new product, the LME49810, is a monolithic 200V audio power amplifier driver with an integrated Baker clamp. Similar to an operational amplifier, each LME49810 can replace dozens of hand-selected and paired discrete devices in a high-end hi-fi audio system. The function of the LME49810 is to drive high-power discrete output transistors up to 50mA, providing up to 3kW of power to the system. When a complete power amplifier design is implemented, the typical THD+N is 0.0007%. Other indicators include a slew rate of 50V/μs and a PSRR of 110dB.

The Baker clamp handles input signal peaks. It is implemented by connecting a diode array between the base and collector of the transistor. It also avoids saturation of the collector-emitter junction and clips less noticeable signals by eliminating high-frequency glitches that occur when the transistor recovers from the saturation region.

In the National Semiconductor listening room, all electronics except the CD player are custom made. The op amps are used in the DAC signal path (four per stereo channel), as well as the power regulators. Obviously, the audio power amplifier drivers are in the power amplifiers. What would it cost to produce commercially?

Brasfield estimates that National Semiconductor's current prototypes sell for about $300 each, with a few thousand dollars in signal chain electronics. That doesn't include the speakers. Unfortunately, Brasfield says Wilson Audio Specialties has informed him that these $26,000-a-pair Watt Puppies are obsolete and that their replacement will cost more overall.