Is the number of components needed to implement a complete Bluetooth communications link sabotaging the industry's goals to drastically lower the technology's costs? With the Bluetooth industry's ambitious cost goal of $5 to implement a complete receiver/transmitter, digital signal processing subsystem, and microcontroller with associated ROM and RAM, every effort must be made to eliminate unnecessary components, particularly those that carry any significant cost burdens.
Most Bluetooth suppliers have adopted a multichip approach to system design, employing CMOS devices for the baseband DSP and microcontroller and bipolar devices for the RF functions. While this approach undoubtedly helps simplify chip design, inherent disadvantages such as higher component count, inadequate board space, and other system integration issues, can lead to higher implementation costs. A typical implementation of a Bluetooth radio system, for example, involves a considerable number of relatively expensive RF and intermediate frequency (IF) filters.
Unlike these traditional solutions, the new single-chip Bluetooth solutions use a number of modern techniques to significantly reduce the external component count, thereby decreasing costs. Cambridge Silicon Radio's BlueCore product range incorporates this new design approach. Each product is a fully integrated Bluetooth solution fabricated entirely in CMOS and includes the baseband DSP, radio and a general-purpose 16-bit RISC processor in a single chip. This proprietary "RF-inside-the-chip" architecture offers considerable technical and commercial advantages that extend beyond a reduction in external components needed to implement a Bluetooth communications link. These advantages involve the voltage-controlled oscillator (VCO); the receiver IF; and the RF front end between the antenna and the transmitter/receiver.
The VCO circuits in most Bluetooth systems require a number of off-chip active and passive components, typically including an inductor, a high-tolerance ceramic resonator and one or two varicap diodes. Apart from their direct costs, these components consume circuit-board real estate and often require the use of RF screening cans. Setting up the VCO also demands some form of frequency measurement equipment and automated trimming system, presenting other hidden costs.
In contrast, BlueCore products feature a fully integrated synthesizer with all the VCO elements, including the resonator and tuning diodes, contained on-chip, and require no external trimming during production. Some observers may be concerned that the fully integrated VCO may be highly susceptible to noise effects from other circuits within the IC. However the VCO circuit uses patented techniques to reduce significantly the apparent VCO conversion gain-to-noise signals and thus avoid superimposing cross-coupled digital noise on its output signal.
In addition, each BlueCore chip features built-in self-test facilities, including an analog calibration, alignment and adjustment routine that is performed every time the chip powers up. This routine completely eliminates the need for RF tests and trimming during product assembly and automatically adjusts for drift during day-to-day operation in order to ensure optimum quality of the Bluetooth communications link.
Superheterodyne receivers work by mixing the received signal with the output from a local oscillator, the frequency of which is varied to tune the receiver. The receivers then feed the result through some form of bandpass or IF filter. For example, if the wanted incoming signal is centered on 100 MHz, mixing this with a local oscillator frequency of 110 MHz would yield a wanted difference output of 10 MHz (and an unwanted summation output of 210 MHz). The characteristic of the bandpass filter at the intermediate frequency of 10 MHz then defines the receiver's channel selectivity performance.
Unfortunately, an unwanted 120-MHz signal also produces a 10-MHz signal in the pass band that cannot easily be discriminated from the wanted signal. Consequently, if the 120-MHz signal is of any significant strength, it will severely impact the receive response of the receiver. The ability of a receiver to reject this unwanted signal is known as the image rejection.
Traditionally, designers choose to overcome these limitations by using a high first IF or by placing a highly selective tunable filter in front of the mixer stages. Designers of conventional Bluetooth systems may employ the first approach, using a relatively high first IF of 100 MHz; they are then forced to use an off-chip surface acoustic wave (SAW) device for channel filtering. To avoid the time and cost overhead of developing an application-specific device, designers invariably choose the type of SAW filter designed for DECT phones. These SAW filters are bulky devices that are difficult to design in, exhibit inferior adjacent channel rejection and offer no easy migration path toward future on-chip filtering. The filter also usually needs at least a couple of capacitors and a couple of inductors for matching purposes.
The second approach, using a highly selective front-end filter, is not really applicable for Bluetooth applications. Apart from the difficulties of making the filter tunable, it introduces losses that cannot be tolerated in high sensitivity receivers.
The single-chip solution instead employs a very low (near-zero) IF, which enables all channel filtering to be performed on-chip and obviates the need for any external components. The receiver also uses an all-digital demodulator, which furnishes much better co-channel rejection characteristics than its analog alternative. It also avoids the inductors normally associated with receivers that use standard quadrature demodulators. However, the near-zero IF approach moves the image response into the receiver's front-end passband-close to the wanted signal, which means that it has to be removed by the use of another technique known as an image rejection mixer.
In an image rejection mixer, the inherent qualities of the mixer circuit suppress the image response, which requires the phase and amplitude of the local oscillator signals feeding the mixer be carefully controlled. The BlueCore chip includes circuitry between the synthesizer driver and the image rejection mixer structure that automatically and continuously maximizes the image rejection performance.
The synthesizer in the BlueCore chip generates FM by using a quadrature modulator. Unlike designs that use varactor methods, it stays fully locked during the transmit burst. As a result, the transmission is inherently stable, optimizing the frequency synthesizer in reducing wideband noise. This approach, again, helps to prevent any cross-coupled digital noise appearing at the transmitter output.
In most conventional Bluetooth systems on the market, the radio-frequency circuits between the antenna and the transmitter/receiver (including all the front-end RF filters) are external to the chip (or chips). Since many of the prime Bluetooth applications provide wireless radio links between GSM cellular phones and ancillary devices such as personal digital assistants (PDAs), computers or headsets, there generally are two front-end filters to consider. One protects the Bluetooth receiver's sensitive input stages against overload and guards against desensitization caused by the GSM phone's transmitter signals. The other front-end filter removes any harmonics-and possibly out-of-band noise-from the Bluetooth transmitter's output signal.
For their RF input stages, most traditional Bluetooth system designs employ high-Q ceramic or SAW filters to reject blocking signals from the GSM transmitter. Some designs also use these types of filters on the Bluetooth transmitter output. Because BlueCore products are fabricated entirely in CMOS, however, the receiver has a much higher linearity than comparable bipolar designs. This difference means that the RF filters can have a more-relaxed specification, resulting in a less complex design that is easier to implement. In fact, using small printed components to produce both filters makes them extremely cheap to fabricate.
To combine the two balanced transmitter outputs into a single-ended output, a balun is required. An impedance matching network transforms the complex output impedance of the transmitter outputs to the balun input, then from the balun to the input impedance of the power amplifier (PA). The network must also match the output of the PA with the (real) input impedance of the filter.
Ordinarily, a lossless matching network could transform the transmitter complex output impedance into R ohms, where R can be 50 ohm. A line of 50 ohm and 1/2 wavelength long could be used as balun. The PA could be matched to R/2, or 25 ohm, and a matching network could transform the filter port impedance to 50 ohm. However, this approach would require numerous matching network components, and does not provide the most effective solution.
Engineers of the single-chip solution have evaluated extensively these printed components to ensure maximum Q and repeatability. Cambridge Silicon Radio publishes a free reference design for a complete Bluetooth system that includes all the printed components. This reference design is based entirely on the use of industry-standard FR4 circuit board material-one of the cheapest substrates available-and accommodates material with permittivity values ranging from 4.2 to 4.8. The printed components are specifically designed to tolerate material variables such as thickness, dielectric constant and loss and do not require excessively tight manufacturing tolerances. To keep line widths to at least 0.1 mm, circuit losses and circuit Q as low as possible and the circuit insensitive to component tolerances, the design is treated as a combination of both distributed elements and lumped elements.
A practical circuit is derived from the basic design by adding a small-value capacitor at each end of the balun and an L-network, consisting of a series transmission line and a parallel capacitor, at the output of the balun. The junction between the capacitor and the transmission line forms the output that connects to the PA input. Also, the balun is shortened and is divided into two lines. By observing proper geometry, the frequency response of one of these printed RF filters, together with that of the printed transmitter balun transformer, peaks in the operating range of the Bluetooth transmitter (i.e. at 2.4 GHz).
Bluetooth systems with bipolar receiver input stages also require an external transmitter/receiver switch, which is normally constructed fromSOT23 devices containing GaAs field-effect transistors (FET) or positive-intrinsic negative (PIN) diodes. These devices typically cost about 50 cents each in production quantities and can therefore account for 10 percent of the total system cost. As a lower cost alternative, Cambridge Silicon Radio's products have a built-in transmit/receive (Tx/Rx) switch that is more than adequate for many applications. If the application demands an extra 3dB link budget, however, an external Tx/Rx switch can be used. The reference design includes an external PA, low-/noise amplifier (LNA), RF switch and detachable antenna to ensure that the first BlueCore Evaluation Systems have sufficient range. The second pass of BlueCore01 (BC01b) will not require external amplifiers for a class 2 Bluetooth system.
Traditional Bluetooth systems require that screening cans be placed around some or all of their component parts. These are relatively expensive to manufacture and can be difficult to accommodate in many typical Bluetooth applications where space is frequently at a premium. To compound the problem, screening cans also act as antennae, requiring extra screening elsewhere in the system. No Blue-Core product needs any form of screening can.
Engineers of the BlueCore solution have tried to keep the external component count to an all-time low. The net effect: a complete Bluetooth system built around one chip with added external surface-mount components-a crystal, four low-value decoupling capacitors, four RF blocking/RF matching capacitors and a single 2.2 microfarad or 4.7 microfarad miniature tantalum capacitor. All these components are small, low-cost devices and require no RF inductors-wirewound or multilayer.
An alternative to the printed filter is a ceramic filter; these are becoming available for the Bluetooth frequencies in small packages and with low insertion loss from manufacturers like AVX Corp. Similarly, transmission lines can be shortened, within limits, by replacing them with fixed inductors.
Simulation can perfect the reference design for specific users, and can be accomplished by breaking the complete design-balun, amplifiers, filter-into smaller sections. Simulating a complete circuit normally is not a good idea; it tends to be very slow, especially when the modeling is done in detail.
See Related Chart