Multifunction Codecs Simplify Wireless Telephones
Jamal Sarma and Scott Gardner
This article was developed specifically for publication in Wireless Design & Development magazine.

A s new air-interface standards enable mobilization, the wireless industry is undergoing dramatic changes. Cordless phones continue to be a key component in mobilization. In 1995, residential and small office/home office (SOHO) users purchased over 10 million cordless phones.

To improve clarity while maximizing battery life and security, cordless handset designers are now turning to digital 900-MHz architectures. However, this movement imposes significant new constraints on baseband hardware. Simultaneously, with the unfolding revolution in digital communications, designers can benefit from the emergence of highly integrated baseband ICs.

Multifunction devices can reduce cost significantly by integrating many functions onto a single chip. Increased integration also reduces power dissipation, an important consideration in portable devices. Furthermore, designers look for increased channel capacity through improved companding methods, as well as higher signal/noise ratio. Voice detection (VOX) is a valuable way to help isolate the signal and improve noise immunity.

Such requirements most directly influenced by the selected Codec for the system. OKI has recently introduced a new Codec, the MSM7708 , which could prove an ideal solution for many designs. Companding methods help solve the major issue of overcrowding of the 900 MHz ISM band by increasing channel capacity. Moreover, the MSM7708 operates with a single 3-V power supply. With battery life becoming a paramount selling feature of cordless handsets today, the implementation of 3-V IC technology and greater integration keeps power consumption to a minimum.

As shown in Figure 1, various baseband functions are all folded into this single device. Packaged in a small, 64-pin TQFP package, this Codec frees up board space for other feature enhancing concepts. Included functions are:

The block diagram, shown in Figure 2, illustrates the included functions. When transmitting, the MSM7708 receives an analog input from a microphone, with operation optimized for a standard configuration with 16-ohm impedance. The signal is amplified through a low-noise, gain-adjustable amplifier. The amplifier output is fed into an active 3-pole anti-aliasing filter. Once the signal is conditioned, an over-sampling delta-sigma analog-to-digital converter provides accuracy and low power consumption. The signal then passes through a band-pass filter stage, to notch the frequency spectrum, and is subsequently companded via either µ-law or A-law PCM schemes. The resulting 64 kbps PCM signal is routed to the Adaptive Differential Pulse Code Modulation (ADPCM) transcoder block via a parallel-to-serial converter, before the data stream finally makes it off chip to the IF/RF section of the system.

On the receive side, the reverse operation occurs as the ADPCM transcoder pipes a PCM signal into the expander for conversion into a linear signal. Once decoded, equalization is applied to the signal using DSP circuitry to perform low-pass filtering and noise correction. A 10-bit over-sampling DAC performs quantization, sending an analog signal to an RC low-pass filter. The filtered output is fed into a power amplifier that is capable of driving 350-ohm speaker loads.

In total, the integrated Codec can provide a full-duplex 32 kbps data stream with so little distortion that it is virtually undetectable. Lower bit rates, providing further channel capacity, are possible but introduce noticeable distortion. Figure 3 illustrates the input and output for a sine wave at 1 kHz at 32 kbps and 16 kbps data rates. These waveforms were made by looping back the input to a 2-V peak-to-peak sine-wave signal and looping back the output to the input. At 32 kbps, the output waveform preserves the signal integrity, even with ADPCM compression. At 16 kbps, some signs of signal deterioration are visible. With the Codec's 2.048-MHz data rate, 64 signals can share one channel at 32 kHz, and 128 signals can share one channel at 16 kHz.

An external 8-bit microcontroller ("protocol MCU") is used to synchronize timing and control register operation throughout the data path channel, and additionally provides system control functions, such as redial memory, RF VCO frequency control, LCD display updating, keyboard scan, and so on.

Recording and Playback

In response to the need to integrate many functions onto one IC, OKI has also integrated into the device a complete interface for speech acquisition and playback. With this capability, users can record information, such as addresses and telephone numbers, by pressing a button on the handset. The voice data is stored in external serial RAM. The captured data can then be played back over the air and through the handset at a later time. Cordless phone users find this feature particularly helpful while driving or otherwise moving around, when taking written notes is often inconvenient.

The logic block provides serial speech data for recording and play back together with the control signals (DIO, SAD, SAS, TAS, RWCK, CS1, CS2) to the serial RAM/ROM memory. Up to four minutes of sound data can be stored and played back by the device, with each minute requiring 1 Mbit of RAM (for recording and playback) or ROM (for playback of fixed messages).

During transmit-side recording, data is piped into the serial register after ADPCM encoding at 32 kbps. During transmit-side playback, the ADPCM-coded signal is directly sent to the RF stage for transmission, as the signal is already compressed. During receive-side recording, the received signal is piped straight into the serial register before ADPCM decoding, and then piped through the ADPCM decoder during playback. By this scheme, data is always stored in a compressed format, and compression/decompression only occurs when necessary.

The protocol MCU controls all recording and playback operations by setting control bits in the Codec's registers. The designer can thus program custom functions, such as multiple recording/playback slots and the playing of fixed messages.

ADPCM Algorithm

The Codec uses ADPCM transcoding to reduce the required data rate while maintaining sound clarity. The standard 64-kpbs signals created by the PCM circuitry is compacted to fit into a reduced 32-kpbs ADPCM (G.721) signal or vice versa in digital transmission systems, effectively doubling the available channels. The MSM7708 Codec also includes provisions for 24 kbps (G.723) and 16 kbps (G.726) usage, which respectively triple and quadruple the number of effective channels.

In ADPCM technology, "differential" refers to the error (difference) between the expected value and the true PCM data which is sent down the line. 'Adaptive' (ADPCM) refers to the filter performing prediction by changing its transfer function based on the rate of change (slope) of the PCM input data. Thus, the statistical data produced from the error (differential) data is used by the adaptive filter to enhance fidelity in a halved data rate (G.721 - 32 kbps) environment. Going below the 32-kbps threshold, implementing G.723/726 standards, is unpopular in telephony applications (especially digital cordless applications) because a user can normally hear the difference.

Figure 4:
An illustration of 4-bit ADPCM, showing the polarity bit, quantization factor (), and the number of data sorts required to form each sample

In OKI's ADPCM implementation, the upper bit of each sample defines the sample polarity (increase or decrease), and the lower three bits define the multiplication factor applied to the fundamental quantizing width (). OKI has developed proprietary systems for optimal formation which provide higher tone quality than available from other sources. OKI's implementation maintains a record of the quantizing factor used over the last four seconds and dynamically recalculates the quantizing factor, based on this history. The 16 possible step values in any one sample (eight up and eight down) can consequently vary over a total range of more than 800 possible values, depending on the dynamically adaptive factor that has been calculated at any one point in time. Resolution, therefore, approaches that provided by a 10-bit ADC at a 32 kHz digitizing conversion rate, but the required data bandwidth is only 4 Kbytes/sec, compared to 40 Kbytes/sec for an uncompressed sample. Moreover, subjective "double-blind" tests have shown that listeners prefer speech recorded with OKI's ADPCM compression over uncompressed digitized recordings at even higher frequencies and resolution, as the distortions introduced by ADPCM are subjectively pleasing to the listener.

In total, designers will find that the MSM7708 is a powerful, cost-effective engineering solution that extends battery life of cordless handsets, while simultaneously providing many value-added features that enhance product differentiation.

Specifications on the MSM7708 and other OKI products can be obtained by contacting http://www.okisemi.com/us.