Bluetooth RF Testing - Testing for Proper Radio Design

JasonYoo

Bluetooth RF Testing - Testing for Proper Radio Design [Copy link]

Today’s Bluetooth radio designs use a number of system architectures, ranging from traditional IF systems using analog modulation to digital IQ modulator/demodulator configurations. Regardless of the design configuration, several issues must be addressed during product development:

? Global regulatory requirements
? Bluetooth technology certification
? Easy, high-yield manufacturing and testing
? Perfect interoperability with other vendors’ designs, some of which may only barely meet the Bluetooth specification In

the following pages, we will examine some of the different aspects of the design, the implications of R&D testing, and the tools that can make the development work easier. Then, we will explain how to perform these measurements and discuss some of the measurement values that can be expected.

Bluetooth RF Technology – Overview

Bluetooth devices operate in the ISM band from 2.402 to 2.48 GHz, typically on 79 channels. They communicate with each other using a digital frequency modulation technique called 0.5BT GFSK (Gaussian Frequency Shift Keying). This means that the carrier is shifted up 157 kHz to represent a '1' or down to represent a '0' at a rate of 1 million symbols (or bits) per second. The '0.5' limits the –3 dB data filter bandwidth to 500 kHz, thereby limiting the occupied RF spectrum.

Communication between the two devices is time division duplex (TDD), meaning that the transmitter and receiver transmit alternately in different time slots. In addition, ultra-fast frequency hopping patterns of up to 1600 hops/s are used to increase the reliability of the link in a band that can be crowded. Reliability is important if the use of the band is expected to increase almost certainly as a result of recent US FCC regulations.
　

Figure 1. Bluetooth RF power envelope and VCO frequency timing.

　

Figure 1 shows possible timing for transmitting and receiving a 366 μs DH1 packet in a 625 μs time slot . The settling time interval can be seen in the lower trace. During this interval, the device must jump to the next channel frequency and the voltage controlled oscillator (VCO) must settle in time to transmit or receive the packet data. Note that the beginning of the packet is not directly related to the rising edge of the RF burst, as can be seen from the dotted lines representing possible alternative rising edges. The rising edge of the burst is also independent of the beginning of the time slot. The design may reduce power immediately after all the packet data has been transmitted, or wait until near the end of the time slot to reduce power.


　

Figure 2. Direct frequency modulation VCO, analog frequency detector

　

The receiver layout in the Bluetooth example shown in Figure 2 uses only a downconverter. (Gray blocks are parts that are omitted or swapped in different designs). A design like this would use only a local oscillator. The output frequency is doubled and switched between receive and transmit functions. Simple direct modulation of the VCO is possible using FSK. The baseband data is passed through a Gaussian filter and characterized with fixed timing delay and no overshoot. The pulse is applied only to the transmitter. Using a sample-and-hold circuit or phase modulator prevents the phase-locked loop (PLL) from removing the phase modulation within the bandwidth. The intermediate frequency is usually very high, so the physical size of the filter components can be limited and the IF frequency can be kept far enough from the LO frequency to achieve satisfactory image rejection. Antenna swapping

can be used when the level is high enough to overload the receiver input.

A power-T- output amplifier is an option that can be used to increase the power required for the Class 1 (+20 dBm) output version. Level accuracy specifications are not critical, but care must be taken to avoid excessive power output and to ensure that the battery is not drained unnecessarily.

Whether the design provides +20 dBm or less, the receiver must be prepared to provide received signal strength indicator (RSSI) information so that devices of different power levels can interoperate. Power up and down switching in the design, such as this, can be easily achieved by controlling the bias current of the amplifier.

Unlike TDMA systems such as DECT or GSM, Bluetooth spectrum testing is not gated to isolate power control and modulation errors. The measurement interval must be long enough to capture the effects of up and down switching and modulation. In practice, this may not cause certification problems, but time-gated measurements may be important because of the ability to quickly find defects.

As shown in Figure 3, some designs use unspecified cycles to prepare the receiver before modulation begins. In this example, neither 1 nor 0 is transmitted.

　

Figure 3. Applied power before FM

　

Frequency Error – All frequency measurements in the Bluetooth specification rely on short gate cycles of 4 μs or 10 μs , which can cause differences in results that can be understood in several ways. First, the narrower time window means that the cutoff frequency of the measurement bandwidth is higher, so various noise structures are included in the measurement. The second way is to consider that error structures, such as quantization errors or oscillator sideband noise in the measurement device, will be more proportional in short cycles than in longer measurement intervals, because in the latter case these errors tend to be averaged out. This fact must be considered in the design constraints in addition to the static errors caused by the crystal reference.

Frequency Drift – Drift measurements combine short-term, 10-bit adjacent data groups with long-term cross-burst drift results. If a sample-and-hold design is used in the transmitter, the errors introduced by this design may be obvious. In other designs, unwanted modulation components or noise from 4 to 100 kHz can be seen as ripples in the graph. This proves to be another way to confirm that the power supply is adequately decoupled.

Modulation – In the transmitter path, the VCO shown in Figure 2 is directly modulated. To avoid the PLL removing the modulation component within the bandwidth, it can be turned on during transmission, or phase error correction (two-point modulation) can be used. Sample-and-hold techniques can be effective, but care must be taken to avoid frequency drift. Unless digital techniques are used to adjust the synthesizer division ratio, the phase modulator should be calibrated to avoid lack of flatness in the modulation response for different data patterns. Figure 4 shows a typical modulation pattern used for certification testing.

Figure 4a & 4b. Modulation patterns used for certification testing

The Bluetooth RF specification checks the difference in peak frequency between two different patterns, 11110000 and 10101010. The output of the GMSK modulation filter reaches its maximum after 2.5 bits, and this is checked with the first pattern. The cutoff point and shape of the GMSK filter can be checked with the second pattern.

Ideally, the peak difference for the 1010 pattern is 88% of that for the 11110000, although some designs show higher ratios because they do not use a 0.5BT Gaussian filter when transmitting. The highest fundamental modulation frequency is 500 kHz, even at a bit rate of 1 Msymbol/s. The light grey trace on the left hand side of the graph in Figure 4 shows the effects of I/Q imbalance. This can occur when a system with the block diagram shown in Figure 7 is not fully calibrated.

The intra-band spectrum - '-20dB' test confirms that the modulation and pulse signals fit within a 1 MHz bandwidth. The blocks in Figure 5 can be thought of as limiting time windows. The resolution bandwidth of 10 kHz is set for this purpose. Because of the amplitude pulse, this measurement must use 'peak hold'. This method takes into account the deviation of the waveform from the exact center frequency, so it becomes a frequency width mask rather than a fixed mask. If the signal is located in the middle of the mask, the result will be very similar. The dots in Figure 5 are caused by non-data 0 in the packet start code.
　

Figure 5. -20dB measurement

　

Adjacent channel measurements are specified as a series of random frequency measurements. Non-gated sweeps are a quick and easy way to check for these issues. Unlike other TDMA systems such as GSM, DECT, and PDC, gates may still be used even for a combined measurement.

Out-of-band spectrum – Frequency doubling techniques are often used to prevent RF coupling back into the VCO and causing center frequency drag. Subharmonics must be excluded from the RF output path, especially when they may affect the performance of co-sited functions such as GPS receivers (L2 frequency is 1222.7 MHz) or cellular radios.

Figure 6 shows the signal in a design that does not contain subharmonics but generates harmonics up to 9 GHz. This measurement can be performed using a standard spectrum analyzer. For research work, faster sweep times can be used, but they still take several seconds. If a longer sweep time is selected, newer spectrum analyzers with deep data acquisition buffers allow you to zoom in on specific sampling points after the sweep.

Figure 6. Broadband bypass

As shown in Figure 7, some designs use IQ mixing in the transmit and receive paths. The advantage is that it can increase the level of circuit integration and offload signal processing to digital signal processing rather than analog circuits. This figure describes a hybrid approach. Some designs add image rejection mixing in the front end. The higher level of silicon integration makes it more affordable.

The calibration of all these IQ stages must be carefully described. Published techniques for radar and cellular applications describe the sequences and signals that can be used. Applying IQ modulation directly to the RF output may have unexpected effects on the signal. However, errors in the calibration of the modulator will not have any effect on the frequency error because the frequency is just the rate of change of the phase. However, it may not be easy to identify the error in the spectrum.

　

Figure 7. IQ modulator, digital demodulator
　

Errors in IQ modulation indicate amplitude modulation. This can be detected using a power vs. time display, or a vector analyzer for more detailed investigation.
IQ modulators can also be used to create power up and down transitions, indicating the values that gated measurements may produce. Error measurements must be processed digitally in the receive chain. A DC block is found between the mixer output and the ADC input of the receiver to confirm a zero IF system. Imperfections such as LO-RF feedback create a DC component that varies with input frequency and must be properly processed. Near-Zero IF, which is usually set at half the RF channel bandwidth, is more likely to be detected early. Sideband suppression can therefore be an issue. A quick calculation of the sideband: a 0.1 dB gain error, or 1 degree phase error, will cause the sideband to drop by about 40 dB.

Analyzing IQ Waveforms – Vector analyzers can inherently demodulate a wide range of signals. While additional precision may not be warranted when only FSK is directly applied, this parameter will change during IQ design or when considering other formats such as Bluetooth 2, cellular or WLAN.

In order to understand the behavior of a component, it is important to analyze it from multiple directions. Figure 8 shows an example of viewing the same data in four different ways. Deviation viewing provides quick visual confirmation of the correct modulation mode. Eye diagrams and FSK errors can show the quality of the modulation. Demodulated data viewing allows the user to check the preamble, start code, sync text and payload data.
　

Figure 8. Various views of FSK

　

Design Simulation – A higher level of integration focuses on simulation tools. In addition to being able to quickly evaluate different circuit topologies, some of the more advanced tools can provide the receiver with more valid and imperfect signals.

Some of the biggest RF challenges for Bluetooth technology lie here. As the battery depletes, the effects of limited level compression performance can be tested, as well as phase noise, differential path loss, signal imperfections and interference – including the effects of nearby transmitters, which occur when the Bluetooth unit is coupled to a mobile phone.
There are two areas of recent product development that have made great strides. The first is the integration of digital signal generation and vector signal analysis blocks, which allows you to alternate between simulation and real-world testing. The link between software products and physical instruments allows you to quickly compare results from prototypes.

The second feature is the design guide, which automates the setup of the tool. It has led users to switch to design software that can evaluate real circuits instead of programming with basic configuration information related to a specific radio technology.

Receiver Testing – The frequency detector shown in Figure 2 is a frequency detector in the mixer/tuning circuit category. It looks simple, but it requires some verification. When characterizing a design, it is important to note that some results will not have a normal (Gaussian) distribution.

This is due to the circuit technology used and there is a limit based on the phase/frequency characteristics of the tuned circuit/mixer combination. Delay line discriminators are another option but require verification.

Front-end amplifier design and testing must focus on interference rather than the most likely noise figure or 1 dB compression characteristics. There are various techniques that can be used to dynamically change the gain of the receiver chain to optimize the rejection of unwanted signals. Using synchronized pulse amplitude modulation in the signal generator can be a worthwhile test of the burst-to-burst response of the AGC system, especially if software controlled. Figure 9 shows the measurement path for testing an isolated receiver.
　

Figure 9. Measurement path for testing an isolated receiver.

　

Testing Receiver Frequency Hopping - As mentioned earlier, all Bluetooth designs use a local oscillator. This has the side effect of causing it to spin around in less than 300 s of full tuning range. This can also occur when the device is operating in Bluetooth test mode. During
the transmit cycle, a frequency may be selected at the end of the ISM band opposite the receive test frequency, or at any other point. The VCO must switch back to the receiver frequency each time. See Figure 10.

Each burst can be used for data transmission, so a continuous sequence can be used. This eliminates the need to perform frequency hopping BER tests where the signal source must hop. However, the user must still arrange for synchronization of the signal generator with the device under test before link signaling occurs.

Once the bits are converted to digital format, the BER test can be performed. There are several ways to perform this, and Table 1 provides a summary of the various BER test techniques.

　

Figure 10. VCO swapped/fixed RX channel

　

Data recovery point	Note
IF	Use an eye diagram.
Demodulator Output	Defines an original PN sequence output and sends out BER measurement.
Baseband Output	Recover the clock and decode the payload data. Perform BER measurement.
Reflection	The complete installation requires the use of Bluetooth test mode; both baseband and link processing must be included. Some designs allow you to perform these tests with custom settings.

Table 1. Receiver bit error measurement methods

Conclusion

Bluetooth technology uses fast frequency hopping (up to 1600 hops/sec) and operates at 2.4 GHz. The use of GFSK modulation and relaxed receiver sensitivity requirements allow for simple radio designs. These features have led to some modules that use technology from previously designed systems, such as the European DECT standard.

However, the lower price target for Bluetooth devices has forced other designs to take a different approach with higher integration. The goals of system integration on a single chip, minimum power consumption, enhanced interference suppression, and better-than-spec sensitivity make its design as challenging as higher-performance radio designs. This article reviews the differences between direct FM VCO design and digital IQ techniques and how they affect measurements. It shows how Bluetooth modulation characterization measurements verify the quality of the signal produced by a Direct FM design and how carrier frequency drift and ICFT can make IQ modulation artifacts disappear. In addition

, the article shows the advantages of performing time-limited Bluetooth specification measurements. Obviously, radio designers must have access to complete simulation and measurement tools to complete a reliable Bluetooth design.

About the Author

Peter Cain obtained a degree in electronic engineering with honors from the University of Southampton in the UK and has worked in the electronic measurement industry for 20 years, 16 of which have been dedicated to Agilent Technologies [HP]. Peter has served in engineering, project and program management, and recently technical solution planning. Peter, who used to like diving, still loves to design things, and his young son gave him the best reason to do this.