Exceeding the standard for wireless audio

There are 2.4 GHz RF alternatives to Bluetooth for portable audio links that offer the dual advantages of CD quality and long battery life.

While Bluetooth is undoubtedly a solution, the compromises introduced to ensure universal interoperability mean it is a less-than-perfect solution for particular applications.

For example, the Bluetooth 1.2-equipped stereo headphones on the market have overall received generally poor reviews for audio quality and battery life.

Fortunately, there are alternatives for the RF engineer looking to add design wireless stereo headphones. One example is an RF chipset specifically designed for audio streaming and developed by my company, Nordic Semiconductor (distributed in Australia by Avnet).

Dubbed the nRF24Z1, it is rated at 4 Mbps yet consumes power at half the rate of a comparable Bluetooth 1.2 chipset in the same application.

Let's consider the advantages offered by the nRF24Z1 by reference to a wireless headphones/MP3 player application.

An MP3 player/wired headphones combination is shown in the schematic illustrated in Figure 1. The audio source - typically flash memory or mini-hard disk - outputs digital audio, via an MP3 decoder in either I2S or S/PDIF format.

A microcontroller (MCU) supervises the audio source and controls playback characteristics such as volume control or bass boost delivered from a combined DAC/amplifier.

Adding an RF link between the audio source and DAC/amp dispenses with the need for a wired connection between the MP3 player and headphones (see Figure 2). This means that one side of the RF link resides in the player and the other in the headphones. Unlike a wired system's fixed connection between the MCU and DAC/amp, the wireless system requires an additional control data channel alongside the audio channel (otherwise volume control is restricted to the headphones only, with other control buttons remaining on the player). Both MP3 player and headset need batteries.

There are two methods for wirelessly relaying streamed audio content from a portable music player to headphones. One method is to simply relay the compressed MP3 data across the link.

The de facto 'good quality' compression standard for MP3 is 192 Kbps, well within the capabilities of Bluetooth 1.2's nominal 1 Mbps bandwidth. Even the 'lossless' proprietary standards, such as Apple's own, require only 320 Kbps; again no problem for Bluetooth.

However, there are downsides to this technique. Firstly, the headset will require the necessary electronics (DSP, DAC/amp and batteries) to decompress the MP3 stream; this adds weight, bulk, complexity and cost to the headphones.

Secondly, the quality of the reproduced sound will be a function of the headset, no matter how good the player happens to be.

Finally, the MP3 player couldn't then be used with conventional wired headphones.

The second technique is to decompress the MP3 data in the player and stream the uncompressed audio information to the receiver in the headphones. This mimics the process in a conventional wired MP3 player/headphone combination and is perhaps the most practical configuration - reducing the complexity, weight and power consumption of the headphones but demanding greater bandwidth.

Audio quality is a key differentiator for portable products in the ferociously competitive consumer market. CD digital audio samples the original analog music signal at 44.1 kHz with 16-bit resolution for each channel.

This sampling rate and resolution generates a data stream of 1.41 Mbps. CD audio is generally agreed to be an acceptable benchmark for hi-fidelity.

Bluetooth has to maintain synchronisation (a legacy of the technology's requirement to support up to seven slaves) to avoid re-linking delays and does this by sending a 160-bit packet every 625 µS (1600 packets/s, or a net data rate of 256 Kbps) to maintain the link, whether it's in use or not.

Bluetooth 1.2 features a 1 Mbps nominal data rate that runs at around 720 Kbps in practical circumstances (providing there are few simultaneously transmitting 2.4 GHz sources demanding frequency hopping and subsequent reduction in bandwidth).

In contrast, Nordic's transceiver has a nominal bandwidth of 4 Mbps, sufficient to transfer 16-bit stereo at 48 kHz - a total data rate of 1.54 Mbps - providing CD quality.

The radio has a nominal transfer rate of 4 Mbps. This provides ample overhead for retransmission of lost packets, acknowledgement of received packets, user interruptions (for example, buttons being pressed), device addressing and time division multiplexing.

(Figure 3 shows the nRF24Z1-based reference design for wireless headphones.)

While audio streaming at 44.1 kHz, the nRF24Z1 transceiver remains at a given carrier frequency for 2.9 ms.

During this time interval audio and control information is sent to the receiving end of the link (audio receiver - ARX), any lost audio content is retransmitted and acknowledgement and control information is received from the ARX. The system then hops to a different frequency and repeats the process.

When there is no content to be streamed, the chip can enter various sleep modes. In the 'deep sleep' mode the radio is shut down apart from a small 5 µA current to retain memory content.

In a 'lighter' sleep mode the radio is woken at regular intervals to look for a counterpart. When the system is in sleep mode, the power consumption of any converters and microcontrollers in the system must also be taken into account.

When the transmitting end of the link (audio transmitter - ATX) and ARX are turned on, the devices are able to locate one another (typically within 10 ms) by means of an on-chip frequency-scanning algorithm.

A transmitting or receiving Bluetooth 1.2 audio chip runs at around 60 mA current consumption. (Note: This is an average figure, some chipsets are better and can achieve down to around 50 mA. This variation is because power consumption is primarily a function of the design of the silicon radio.) Consequently, operating at 2 V, the device draws 120 mW.

Assuming the power source is a Li-Ion battery operating at 3.7 V via a 90% efficient DC-to-DC converter, the power draw from the battery is 133 mW.

The headphone-mounted DAC/amp draws around 4 mA in operation. Assuming the DAC/amp runs directly from the converter output of 3.7 V, it will draw 14.8 mW.

A typical 3.7 V Li-Ion battery has a capacity of 900 mAh, supplying 3330 mWh. With a total power consumption of 147.8 mW during playback, the user can expect 3330 mWh/147.8 mW = 22.5 hours of battery life.

The device's average ARX current is 22.9 mA (see Figure 4) whilst average ATX current is 17.8 mA (Figure 5). Clever silicon design has ensured the proprietary solution is an 'ultra' low power device.

Note that these figures apply to transmitting and receiving a 44.1 kHz sampled, 16-bit audio stream without compression with a good radio link. The bottom line current values - points 1-6 (Figure 4) and 1-7 (Figure 5) - represent the average value over the whole interval.

The length of the intervals 4-5 (Figure 4) and 5-6 (Figure 5) depend on the radio link quality. These would obviously extend under bad link conditions due to the number of retransmissions required.

Running at 2 V (the same as the Bluetooth device), the Nordic device draws 45.8 mW from the DC-to-DC converter, requiring 50.9 mW from the battery. Adding the DAC/amp consumption yields 65.7 mW.

Using the 3.7-V Li-Ion battery, the user now gets 3330 mWh/65.7 mW = 50.7 hours of battery lifetime, more than double the 22.5 hours experienced with the Bluetooth chip*.

Table 1 summarises the results and, for comparison, includes figures for a power supply using two AAA batteries of 1.5 V connected in series, with a capacity of 900 mAh, providing 2 x 1.5 V x 900 mAh = 2700 mWh and without DC-to-DC conversion.

*These calculations assume a moderate rate of battery discharge and linear reduction in capacity over time. In practice, battery life times are likely to be shorter for both technologies (Bluetooth and proprietary) as self-discharge and other effects take their toll.

In addition, the calculation doesn't take into account factors such as the power needed to mechanically vibrate the headphone membrane, and linear regulation cutting the battery supply once the voltage falls below a set level, but before it is exhausted.

Børge Strand is a field application engineer with Nordic Semiconductor in Oslo, Norway.