原创 手机音频设计的演变 Part 2 ( English )

An outgrowth of analog dis-integration from the Communications and Applications processors is the need to communicate with and manage multiple audio devices.

Because of the number of audio sources and devices are increasing in the cellular handset, designers have begun using single integrated circuits called audio subsystems which group together the various audio inputs and outputs with the necessary data converters and amplifiers.

Figure 3 shows an Audio Subsystem applied to a Communications Analog Baseband solution.

These audio subsystems are designed with mixed signal data converters, and either (1) all analog, (2) all digital, or (3) a mixture of analog and digital circuitry for the amplifiers, and interfaces to the processors.

The advantage of an audio subsystem is that the designer can deal with audio as a single block within the architecture of the phone and the results are very often dramatic reductions in component count, board area, and system costs compared to any other way to accomplish the same thing.

The various functional blocks of an audio subsystem must generally meet performance specifications equivalent to an individual IC performing the same function.

Microphones
The communications microphone is mounted within the case of the handset and is typically an Electret microphone in a miniature canister. Many of them include a J-FET amplifier.

Additionally, there is another input for an analog Electret microphone in a hands-free kit such as the Mono Headset with Microphone accessory. Typical microphones for cell phone use have sensitivity values on the order of -45dB to -55dB referred to 94dBPa SPL. The speech output levels for the microphones are typically less than 100mV, too small to be used directly. Therefore, an additional amplifier with gain values from +20dB to +36dB is used on the PC board to get the signal level to a value that the baseband processor can use with its internal analog to digital converter.

To improve the immunity of the microphone signal path to noise pick up from other circuitry on the PCB, the additional gain block is usually placed very close to the microphone.

A major trend is to incorporate the gain block inside the microphone canister (“amp in a mic”), thereby providing a much higher analog output signal that can be routed across the phone PCB. In this case, any noise interference will have a lesser effect than if the amplifier were mounted on the main PCB, and the SNR will be higher which allows better conversion in the baseband A-D.

An additional trend is toward using digital microphones in the handset because they provide better signal-to-noise ratio as well as better immunity to RF and EMI interference than the analog microphone. Also, digital microphone data can be more directly manipulated in the cell phone CPU in order to provide voice processing functions such as echo cancellation.

The digital microphone could be made with MEMS technology or conventional electret condenser microphones, followed by A-D converter circuitry to produce a bit stream output, or audio samples at a given sample rate output to a particular standard audio buss, such as I2S.

There is also a strong trend toward using two or more microphones in the handset case for the purpose of creating microphone arrays to provide far field (user surrounding acoustic environment) noise suppression of 20dB or better while maintaining very good response for speech very close (within 70mm distance) to the microphone array. Far field acoustic noise suppression can greatly enhance the intelligibility of speech through the microphone signal path.

To enhance speakerphone and PTT performance, multiple-microphone arrays will also provide focus on the talker by using beam forming techniques.

Speakers
Over the years the speakers used in cellular handsets have changed dramatically. The major change has been in the size, having gone from roughly 26 millimeter in diameter to as low as 11 millimeters in diameter. As a result, the frequency response and sound pressure level quality over the audio band (20Hz – 22KHz) has also diminished. Handset makers attempt to make up for the quality deficiency either by (1) increasing power applied to the speaker, by (2) signal processing which compensates for the poor frequency response (example: Bass Boost), or (3) combining both methods.

Increased output power places additional mechanical stress on the speaker coil and diaphragm. The industry has begun asking for both power limiting as well as automatic gain control (AGC) in order to manage the speaker performance. Some investigations have been done regarding speaker feedback mechanisms which would allow dynamic control of the speaker characteristics via software in the handset.

Basically, the goal is to achieve higher performance from cheaper speakers. The goal would be better served by using better speakers. Traditional speakers are the dynamic moving coil type with typically 8 impedance. Earpiece speakers are predominantly 32 (although some are 16 ), while ringer and speakerphone speakers and car-kit speakers currently remain at 8 . If 4 speakers were used, larger wiring size would be needed to handle the extra current.

New types of speakers for cell phones are coming onto the market based on piezoelectric principles, and are commonly called ceramic or piezoelectric speakers. The piezoelectric ceramic speaker should not be confused with a moving coil speaker that uses ceramic instead of iron magnetic core material that is often mistakenly called a ceramic speaker. Piezoelectric speakers can be made very thin – less than 1mm thickness – which is helpful to handset makers designing thinner phones.

Since piezoelectric speakers are essentially capacitive (20nF to as high as 3000nF depending upon the particular design), they have far less dissipation losses than resistive type moving coil speakers and, therefore, promise improvements in system power efficiency. As expected, the effective impedance also varies with frequency.

The frequency response of piezoelectric speakers is similar to small geometry moving coil speakers up to ~10 KHz bandwidth. The 10 KHz bandwidth is quite sufficient for voice and ring sound music which have major frequency content below 10 KHz.

Currently however, the piezoelectric speakers require voltages higher than the handset battery voltage and some of the system efficiency is lost by boosting the battery voltage to the required amplifier operating levels. Typical driving voltage requirements are 10Vp-p to as much as 30Vp-p, depending upon the individual speaker design, achieving output sound pressure levels of 75db to 85dB. The audio amplifiers must be designed to drive the high capacitance loads, which, at the power on stage, are effectively short circuit loads.

There are some rapid changes occurring and future improvements are expected which will bring the operating voltage of piezoelectric speakers much lower, and possibly within the range of the cellular handset battery voltage, as well as improve frequency response and output sound pressure levels.

Another type of piezoelectric speaker technology involves attaching a piezoelectric mechanical beam to a flat surface on the handset, such as the display, which transforms the flat surface into a speaker when the piezoelectric beam is driven with an audio signal. These devices have somewhat higher output voltage requirements than piezoelectric ceramic speakers, but are also expected to benefit from industry efforts to bring down driving voltage requirements.

All of the piezoelectric speaker alternatives currently cost more than a moving coil type speaker solution, but they are beginning to appear in commercially available cellular phones where their unique characteristics bring desired benefits.

Headphones
Headphones are used for both the hands-free mode of operation as well as listening to music. Headphones for cellular phones typically come in 2 styles: (1) traditional headband style headset, or (2) the smaller, in-ear, earbud type.

Most headphones sound terrible. This is because headphones are a different acoustic configuration than loudspeakers, and they create a different acoustical environment in the ears. Music that sounds great with loudspeakers doesn’t sound the same with headphones. And their frequency responses are generally different than speakers.

There many international standards and methodologies for measuring headphone characteristics which makes it difficult to really compare headphones. Also, most headphone makers don’t spend much money on doing it well.

The key headphone specifications are: isolation, frequency response, impedance, and sensitivity. Isolation is a measure of how well headphones block outside noise of the audible frequency range, and is highly dependent on their design and type. The best isolation is achieved by noise-canceling headphones and in-ear-canal headphones. Frequency Response of headphones varies quite a lot since the small transducers used cannot physically have the same mechanical responses throughout the audio frequency range. Also, due to the small transducer size there will be natural frequency resonant peaks.

Professionals often use equalizers with headphones to roll off the high frequencies and boost the low frequencies. The high frequency roll off compensates for the fact that the headphone bypasses the filtering caused by the head and ears on upper midrange and treble frequencies as would normally happen when listening to speakers or any external sound source. The low frequency boost compensates for the fact that with speakers, the listener perceives the low frequency with the whole body as well as the ears. With headphones, the whole body experience is missing.

A frequency response of around +4dB (relative to the 1KHz value) in the range of 30Hz to about 600Hz and a sloping fall off of 0dB to -6dB from 1KHz to about 20KHz has be found to be acceptable compensation for most listeners. However, be aware. Each listener is unique and their listening experience will be different due to differences in head and ear shapes. The differences in normal responses of individuals can vary as much as 5dB because the acoustic impedance of the ear canal is very different for each ear. Extreme differences between individuals can be as much as 10dB.

Impedance values for headphones can also vary over a wide range. For example, professional audio headsets offer impedance levels from as low as 16 to as high as 600 .

For the mono headset with microphone or stereo headsets used with cellular phones, the speaker has a typical impedance of 32 , but some designs also use 16 impedance. Impedances can also vary up to +/- 15%, and are only specified at 1 KHz frequency. In general, the lower the impedance, the easier it is to deliver the necessary power to drive the headphone transducer, and therefore the volume will also be higher.

Sensitivity measurements for headphones are similar to those for speakers but at quite different power levels. For speakers, the measurement is Sound Pressure Level (SPL) in dB at a distance of 1 meter with an applied power of 1 Watt. For headphones, the measurement is SPL in dB at the earpiece (using a dummy head with built-in microphones) with an applied power of 1 milliWatt. The effective SPL value is lower by a few dB at the ear drum. Headphone sensitivity specifications are commonly listed as dB, or dB/mW. The correct term should be dB SPL for 1 mW input.

The headphone speaker output Sound Pressure Levels are easily in the range of 100dB – 120dB SPL with 1mW of applied power. However, the loudness appears much higher in an earbud headphone compared to a headband type headphone because the distance to the ear drum is much shorter when using earbud headphones. Thus, you will need much less power to drive headphones as to drive a loudspeaker of similar sensitivity SPL rating. Achievable SPL with the same applied power varies widely for different headphone models and manufacturers, ranging from a low of 107 dB to as high as 146 dB! Therefore, the audio output power level required for comfortable and safe listening volume will vary according to the type of headset used, making it very difficult to properly size the headset audio amplifier for all possible headset types. Too much output power could damage the user’s hearing. The threshold of pain is around 140dB SPL, but sustained listening at 90dB SPL can permanently damage hearing.

There are now discussions in Europe of trying to standardize output power levels for headset accessories as well as recognize the headphone type in order to limit the audio SPL in the ear canal of the user. This would also require some standardization of the type of headset that would be allowed to be connected to cellular handsets, as well as the connector itself to allow for consistent sensing of the headset type. Currently there are no such standards.

The purpose of “recognizing” the type of headset connected is to configure the audio amplifiers accordingly to avoid hearing damage. Some of this type of capability is being built into cell phones now, but more work is necessary to properly sense and select all the possible usage scenarios.

Headphone amplifiers are available in Single Ended output type requiring output coupling capacitors, as well as capacitor-less output types.

There are two capacitor-less output configurations; (1) true center point Ground requiring a charge pump to generate + and – power supplies, and (2) Vdd/2 buffer amplifier output which substitutes for the Ground connection in a headphone jack. Both provide very good low frequency response since there are no coupling capacitors.

The true center point Ground headphone amplifier is preferred because of ESD concerns surrounding a Vdd/2 voltage appearing on the headphone jack, and the desire to use the headphone output jack as a connector into other audio equipment. However, the added power dissipation of the charge pump reduces desired playback time and battery life when the phone is used as a music player.

The Vdd/2 buffer amplifier method is easily converted to a single ended configuration either by grounding the buffer amplifier output or by a selection pin on the amplifier.