Bluetooth radio frequency technology and its test items

Bluetooth devices operate in the ISM band and use Gaussian frequency shift keying (GFSK) digital frequency modulation technology to achieve communication between each other. The devices use time division duplexing (TDD) and use an extremely fast frequency hopping scheme to improve link reliability in crowded bands. For Bluetooth devices, the RF part is one of the main test contents.

Bluetooth RF design uses a variety of system architectures, including traditional analog modulation based on intermediate frequency systems and digital IQ modulator/demodulator configuration systems. However, no matter which design configuration is used, the following issues must be addressed during product development:
Regulatory requirements around the world
Bluetooth certification
Simple and efficient manufacturing testing
Good compatibility with other manufacturers' products

Bluetooth radio frequency technology

Bluetooth devices operate in the ISM band, usually on 79 channels between 2.402GHz and 2.48GHz. It uses a digital frequency modulation technique called 0.5BT Gaussian Frequency Shift Keying (GFSK) to communicate with each other. That is to say, the carrier is shifted up 157kHz to represent "1" and down 157kHz to represent "0", with a rate of 1 million symbols (or bits)/second, and then the -3dB bandwidth of the data filter is set to 500kHz with "0.5", which can limit the spectrum occupied by the radio frequency.

The two devices communicate via time division duplex (TDD), with the transmitter and receiver transmitting alternately in intervals, one right after the other, and a very fast frequency hopping scheme (1,600 hops/second) to increase link reliability in a crowded band. The FCC expects the band to continue to increase in usage, so reliability is essential.

In the Bluetooth architecture shown in Figure 1, only one down-conversion is used at the receiver. This type of design uses a simple local oscillator, the output is multiplied and switched between the receiver and transmitter. FSK allows direct VCO modulation, the baseband data passes through a fixed time delay and no overshoot Gaussian filter, and pulse shaping is only used in the transmitter. The phase-locked loop (PLL) can use a sample-and-hold circuit or a phase modulator to remove the phase modulation in the baseband. Usually the IF is quite high to limit the physical size of the filter components and keep the IF away from the LO frequency to ensure sufficient image rejection. If the level is too high to overload the receiver input, an antenna switch should be used.

Test items

Here are some tests applicable to the RF portion of a Bluetooth device.

The power-output amplifier is an option that can undoubtedly increase the output power of the Class I (+20dBm) output amplifier. Although the level accuracy is not required, excessive power output should be avoided to avoid unnecessary battery consumption.

Whether the design provides +20dBm or less power, the receiver needs to have a received signal strength indication. RSSI information allows devices of different power levels to communicate with each other. The power slope in this type of design can be achieved by controlling the bias current of the amplifier.

Unlike other TDMA systems such as DECT or GSM, Bluetooth spectrum testing is not limited to separate power control and modulation error testing, and its measurement interval must be long enough to collect the effects of slope and modulation. In practice this does not affect certification, and time-gated measurements are of great value because they can quickly identify defects. Some designs use an unspecified period before the start of modulation, which is usually used for receiver preparation.

Frequency error - All frequency measurements in the Bluetooth specification use a shorter 4-microsecond or 10-microsecond gating period, which will cause uncertainty in the measurement results. This can be understood from different perspectives. First, the narrow time opening means that the measurement bandwidth cutoff frequency is higher, which will introduce various types of noise into the measurement; second, the error mechanism should be considered. For example, in short-interval measurements, the quantization noise or oscillator sideband noise from the measurement device will account for a larger percentage, while these noise effects will be averaged out in longer measurement intervals. Therefore, the design scope should take this factor into account, which should exceed the static error generated by the reference crystal.

Frequency Drift - The drift measurement combines short 10-bit adjacent data groups with longer drift results across the pulse. This error can occur if a sample-and-hold design is used in the transmitter design. For other types of designs, unwanted 4kHz to 100kHz modulation components or noise can be observed on the waveform as ripple, indicating it as another way to ensure good power supply decoupling.

Modulation - In the transmit path, the VCO in Figure 1 is modulated directly. To avoid the PLL stripping off the modulation components within the bandwidth, the transmission device can be left open or phase error correction (two-point modulation) can be used. Sample-and-hold techniques should be effective, but care should be taken to avoid frequency drift. Unless digital techniques are used to adjust the synthesizer's division ratio, the phase modulator should be calibrated to avoid problems with low response flatness for different data pattern modulations.

The Bluetooth RF specification checks the peak frequency deviation of two different code types, 11110000 and 10101010. The output of the GMSK modulation filter reaches its maximum value after 2.5 bits. The first code can check this point, and the cutoff point and shape of the GMSK filter are checked by the second code. Ideally, the peak deviation of the 1010 code is 88% of that of 11110000. Some designs do not apply 0.5BT Gaussian filtering and will show a higher ratio. The highest basic modulation frequency is 500kHz, and the bit rate is 1 million symbols/second.

In-band spectrum - -20dB test can confirm that the modulation and pulse signal are indeed in a 1MHz wide band. The box in Figure 2 can be regarded as the limit range. This requirement can be achieved by setting a 10kHz resolution bandwidth. The peak hold method is used for measurement because the amplitude has pulse characteristics. By testing the frequency width instead of just the fixed template test, this method can deviate the waveform from the exact center frequency. The effect is very similar to the alignment within the signal template. The ridge in the figure is caused by the non-data whitening zero of the packet header.

Adjacent channel measurements are specified as one of a series of spot frequency measurements, and a non-gated scan is a quick and easy way to check. Gating is sometimes still used, although it is a combined measurement, which is different from other TDMA system measurements such as GSM, DECT and PDC.

Out-of-band spectrum – Frequency doubling is a technique commonly used to prevent RF from pulling the center frequency by coupling back into the VCO. Subharmonics need to be eliminated in the RF output path, especially when they could compromise associated stations, such as GPS receivers or cellular wireless devices that function at the L2 frequency of 1,222.7MHz.

Figure 3 shows a signal from a device that has no subharmonics but produces harmonics above 9 GHz, which is what a standard spectrum analyzer can measure. For research work, faster sweep times can be used, but they are still several seconds. If a long sweep time is chosen, a new spectrum analyzer with a deep data capture buffer is required, which can zoom in on a specific point of interest after the sweep.

Some designs have IQ mixers in both the transmit and receive paths instead. This approach can increase circuit integration, convert signal processing to digital signal processing, and remove analog circuits. Figure 4 shows some hybrid circuit methods. Some designs can add image rejection mixing in the front end. The higher integration of current silicon chip technology also makes this approach more economical.

All of these IQ level calibrations require careful consideration and many technical articles have been published on radar and cellular applications describing the sequences and signals used. Direct application of IQ modulation to the RF output can have significant effects on the signal, but modulator misalignment to frequency errors will not have an effect as frequency is simply the rate of change of phase, although it may be difficult to discern the error in the spectrum.

IQ modulation errors mean that amplitude modulation is present and can be detected using a power-versus-time display or investigated in detail using a vector analyzer. IQ modulators can also be used to shape power ramps, again illustrating the value of gated measurements. There is also some digital processing required to measure bit errors before all measurements are made in the receive chain. Another possible zero-IF system can be identified by looking for a DC block between the receiver mixer output and the ADC input. Non-ideal conditions such as LO-RF feedback will produce a DC component that varies with input frequency and need to be carefully dealt with. Sideband suppression is also an issue, here is a quick calculation that 0.1dB gain error or 1 degree phase error will reduce the sideband by about 40dB.

Analyzing IQ waveforms – Vector analyzers are inherently capable of demodulating a wide variety of signals, and while direct application of FSK may not cover more complex cases, other formats such as Bluetooth 2, cellular technology or LAN may need to be considered during the IQ design process.

In order to understand the performance of the device, it is important to have the ability to analyze from multiple angles, and Figure 5 shows the results of observing the same data in four ways. Deviation observation provides quick and intuitive confirmation of correct pattern modulation, eye diagrams and FSK errors show modulation quality, and demodulated data observation allows users to check the presence of preambles, headers, sync words, and payload data.

Design simulation - Higher levels of integration focus on simulation tools that not only allow for rapid evaluation of different circuit topologies, but also advanced tools for injecting a variety of valid and impaired signals into the receiver.

There are two recent developments that are very beneficial to product development. The first is the integration of digital signal generators and vector signal analysis blocks, which provides mutual exchange between simulation and actual testing. The link between software products and physical instruments can immediately compare results when prototypes are delivered. The second is the design guide that can automate tool settings, allowing users to better evaluate actual circuits with design software instead of having to write programs according to specific wireless technologies in basic configuration information.

Receiver Testing - The discriminator in Figure 1 is a mixer/tuning circuit that is a straight-through device but also requires calibration. During the design characterization process, it is important to be aware of the non-normal (Gaussian) distribution of some of the results.

This circuit is of limited value due to the phase/frequency characteristics of the tuned circuit/mixer. A delay line discriminator is another possible choice, but it also requires calibration.

Front-end amplifier design and testing focuses on interference rather than the best noise figure or 1dB compression characteristics. Many techniques have been published to dynamically change the gain through the receiver chain to optimize the rejection of unwanted signals. Synchronous pulse amplitude modulation can also be used with the signal generator. This test is useful for the pulse-to-pulse response of the AGC system, especially when the system is controlled by software.

Testing Receiver Frequency Hopping - As mentioned previously, the components used in all Bluetooth designs are simple local oscillators whose sideband effects cause less than 300 microseconds of skew across the entire tuning range, even when the device is operating in Bluetooth test mode.

During transmission, a frequency must be selected at the other end of the ISM band, centered at the receive test frequency or at some other arbitrary point, and the VCO makes the transition back to the receiver frequency each time. Each pulse can be used for data transmission, so a continuous sequence can be used, eliminating the need for frequency hopping BER testing when using a frequency hopping source. Although this can be done, the user must arrange for simultaneous control of the signal generator and the device under test before using the link signal. Once the bits are converted to digital format, BER testing can be performed, and Table 1 lists several possible test methods.