Ultrawideband (UWB) communications systems offer great promise of fast data rates and video transmissions without RF carriers. Based on short, time-sequenced pulses, these systems exhibit wide bandwidths, but considerable engineering challenges for reliable spectrum shaping. Engineering challenges also exist in testing UWB devices, since traditional test instruments are based on evaluating narrowband signals. Fortunately, with advances in RF signal generators, analyzers, and digital oscilloscopes, engineers currently have the tools they need to test UWB transmitters.

In February 2004, the United States Federal Communications Commission (FCC) opened 7500 MHz of bandwidth for UWB applications, from 3.1 to 10.6 GHz. The FCC regulation calls an UWB transmitter an intentional radiator with a fractional occupied bandwidth of greater than 20 percent or an instantaneous bandwidth of greater than 500 MHz.1 The FCC limits its equivalent isotropically radiated power (EIRP) to −41.3 dBm from 3.1 to 10.6 GHz for both indoor and handheld devices.2 Additional restrictions are placed on the radiated power outside of this band to minimize interference with Global Positioning Systems (GPS), aviation systems, and other communication systems (see table). The EIRP limits are measured using a spectrum analyzer with a 1-MHz resolution bandwidth (RBW) and average detection or root-mean-square (RMS) detection. In addition, the regulations limit the average radiated power over the 1164-to-1240-MHz and 1559-to-1610-MHz bands to −85.3 dBm when measured using a 1-kHz RBW.

The FCC does not specify the RF physical layer interface in the regulations but rather dictates the spectrum content to balance the needs of the many licensed and unlicensed users. The RF physical layer interface is then defined through the standards community as in the IEEE 802.15.3a standard for UWB communications, and industry groups like the Multi-Band OFDM Alliance (MBOA).3

The candidate proposals for UWB wireless personal area network systems can be divided into two groups: single-carrier and multiband devices. Single-carrier devices use a pulse-modulated RF carrier and a direct sequence (DS) spread-spectrum technique to realize the wideband requirements.4 Multiband devices, based on the MBOA proposal, use 528-MHz-wide orthogonal-frequency-division-multiplexing (OFDM) modulation to meet the 500-MHz minimum bandwidth requirements. The multiband RF carrier is then frequency hopped within a specified band group to spread the signal over a much larger operating bandwidth. The MBOA proposal specifies that the 3.1-to-10.6-GHz frequency range is divided into 14 channels or bands. These bands are grouped into five Band Groups (1-5) consisting of four groups of three bands each and one group of two bands.

At introduction, the first MB-OFDM devices will operate in Band Group 1, which specifies three usable RF carriers of 3432, 3960, and 4488 MHz. Assigning frequency-hop patterns or time-frequency codes (TFC) to each radio system will provide multiple user access and allow piconets to be formed. By implementing switching of the OFDM carrier, the system can use the full frequency band and the electronics need only to operate at speeds relative to the 528-MHz modulation bandwidth. By comparison, the single-carrier DS-UWB system would require the electronics to operate at a much higher chip rate in order to instantaneously cover the wide frequency band.

The MB-OFDM switches frequency for every OFDM symbol, at the very high rate of 312.5 ns. The frequency must settle within 9.5 ns, which makes using a tunable phase-locked oscillator impractical. An alternative is to generate the carrier frequency from a single phase-locked oscillator and a single-sideband (SSB) beat product from another frequency derived from that same oscillator (Fig. 1).5 In this configuration, if the oscillator frequency is set to 4224 MHz, then any one of the three carrier frequencies, 3432, 3960, and 4488 MHz can be rapidly selected.

Due to the wideband properties and rapid frequency switching of MB-OFDM signals, it is important to understand the types of measurements available to accurately characterize MB-OFDM device performance. The rest of this article will focus on several measurements useful when characterizing MB-OFDM transmitters such as average and peak power, error vector magnitude, and power density measurements.

UWB signals are meant to have noise-like properties in order to underlay narrowband radio systems with minimal interference. To accurately measure the time-averaged power using a spectrum analyzer, the instrument should be configured with an RMS detector. The time-averaged power spectral density (PSD) is the main regulatory test for an UWB transmitter. The FCC requires that the PSD be measured using a swept-tuned spectrum analyzer with a specified RBW and rms detector. The PSD is typically reported in dBm/Hz or normalized to a 1-Hz bandwidth. It is also acceptable to report PSD with an alternate RBW.

Using a swept-tuned spectrum analyzer with an RMS detector, it is possible to measure the average power over a selected frequency range and obtain the same reading as a traditional power meter. As an example, Fig. 2a shows the averaged transmitted power of a noise modulated, fixed-carrier signal as 0 dBm. This measurement was obtained using the band power function of the analyzer and gives the same result as a power meter. Figure 2a also shows that the maximum PSD of the signal is slightly above the −17 dBm/100 kHz reference line using the RMS detector. If the RF carrier of this signal is now rapidly switched equally in time between two frequencies, the total band power remains the same at 0 dBm but the PSD is reduced by 3 dB (Fig. 2b). This rapid switching allows the MB-OFDM device to reduce the PSD in order to meet regulatory requirements without sacrificing the total transmitted power. Recall, that the regulatory guidelines are put in place in order to allow UWB signals to coexist and underlay existing narrowband radio systems. It is often necessary to verify the proper operation of any narrowband communication system that would co-exist with the MB-OFDM system. Modern signal generators can be used to create UWB interference signals in order to determine the effects of UWB pulsing on any narrowband receiver.

Although an UWB signal has many noiselike properties, the transmitted signal can also contain unwanted spurious signals generated from local oscillators, and clock signals and various other components generated within the transceiver. Therefore, the FCC also places limits on the highest radiated emission at 0 dBm EIRP. These peak measurements are made using a swept tuned spectrum analyzer with a RBW setting of up to 50 MHz. The accuracy of the RBW filter generally degrades for wider bandwidths and the FCC guidelines allow for measurements with RBW settings between 1 and 50 MHz. Should the selected RBW be greater than 3 MHz, the FCC requires a detailed description of the test procedure, equipment calibration, and instrumentation employed when the application for certification is filed.

The following scaling factor is used to determine the peak emissions with RBW settings other than 50 MHz: 20log (RBW/50), where RBW is specified in MHz. As an example, the maximum peak emissions using a 1-MHz RBW would result in an EIRP of −33.9 dBm. The scaling factor is based on the way the peak detector responds to pulsed signals under various RBW settings.1 The 20log scaling will result in a higher, more conservative value than a 10log scaling, which would be used for pure noise signals.

As a measurement example, Fig. 3 shows a spectrum analyzer response to an OFDM signal using a peak detector (upper trace). As a comparison, the same signal is measured using an rms detector (lower trace). When the signal is noiselike, peak measurements are a very useful indication of spectrum occupancy, but generally not ideal for measuring an absolute power level. On the other hand, time-average measurements allow the total signal power to be easily determined using the band power function. This measurement was made on the Agilent PSA spectrum analyzer that has the capability of simultaneously displaying measurements from the two detector types on the same display.

When characterizing the MB-OFDM symbols and transient responses of the UWB signal, the swept-tuned spectrum analyzer does not have an adequate measurement bandwidth to fully capture all the information contained within the wideband signal. Real-time oscilloscopes with bandwidths to 13.5 GHz are the natural tools for this case. Wideband spectral analysis of the MB-OFDM signal can be obtained using an FFT-based measurement, such as found in the Agilent 89601A vector-signal-analysis (VSA) software.

An FFT-based measurement system can provide the most informative view of the UWB signal but the measured spectrum can look different than the response obtained using a swept-tuned spectrum analyzer, partly because it is giving much more information about the changes in spectrum with time. Several factors cause detailed differences in the displayed data such as the shape and bandwidth of the selected RBW filter, the record length of the time captured waveform, the point in time when the signal is sampled and the way the signal is detected. The FFT-based measurement can be made to look similar to the measurement on a swept-tuned analyzer by selecting a Gaussian filter, increasing the time averaging, using a "max hold" function and proper triggering of the scope in order to capture all the transient effects of the signal. In addition to spectral analysis, the wide measurement bandwidth of the real-time oscilloscope allows additional insight into the physical RF layer of the UWB signal such as time-varying characteristics of the OFDM symbols. The time-varying properties of the MB-OFDM physical layer will now be introduced and related to a series of measurements using a real-time (not under-sampled) measurement system.

Page Title

It is possible to measure the dynamic characteristics of the MB-OFDM symbol using the spectrogram display from an FFT-based measurement system. A spectrogram is comprised of hundreds of spectrum measurements taken over time. The amplitude of the signal is represented by color or shade depth. A typical spectrogram of a measured MB-OFDM signal is shown on Fig. 4a. Wideband modulation of the OFDM signal and the rapid switching of the RF carrier at the symbol rate create challenges to measuring the UWB signal, because the signal is very complex. The MB-OFDM spectrum consists of 128 individual carriers spaced 4.125 MHz for a total bandwidth of 528 MHz. In practice, only 122 carriers are used for data, pilot, and guard tones. As in other OFDM systems, the data is modulated and demodulated onto individual carriers using an IFFT function in the transmitter and FFT function in the receiver respectively. The IFFT period is 242.42 ns (1/4.125 MHz) long and used to create the data portion of the OFDM symbol. The complete OFDM symbol includes an additional zero-padded prefix at the beginning and a guard time at the end for a total symbol length of 312.5 ns.

The OFDM system provides robustness to multipath dispersion by the addition of the zero-padded (no RF transmission) prefix. The multi-path energy that is not captured by the prefix will result in intercarrier interference (ICI) in the OFDM system. Therefore, a trade-off was made to improve ICI performance while not increasing the prefix overhead and was determined to be 60.6 ns.6 The 9.5-ns guard time is inserted at the end of the symbol so there is sufficient time to switch the carrier frequency to the next channel. An MB-OFDM frame consists of different types of symbols containing synchronization (sync) packets, preamble, channel estimation, header, data rate, frame length, and variable payload. Each symbol is transmitted at a different carrier frequency following the assigned TFC pattern (Fig. 4b).

The various types of MB-OFDM symbols are displayed on the spectrogram as functions of time and frequency. The switching effects at symbol transitions are clearly shown as horizontal lines on the spectrogram. The rising and falling edges of the bursts would need to be controlled to reduce this effect. The power level in the adjacent channel can be measured by this same system using a time-gated measurement. In this case, band power markers are used to measure the adjacent-channel energy during the time when the RF carrier is operating in a nearby channel. This measured spectrogram was captured using the using the 89601A VSA software running on the real-time model 54855 oscilloscope from Agilent Technologies .

The very low transmitted power level and noiselike properties found in UWB signals present another measurement challenge. At times, the received signal will be below the noise floor of the measurement system, as in the case when measuring MB-OFDM signals over the air. The challenge is to provide a system that can correlate the measured frequency response with the original transmitted signal. The real-time oscilloscope and VSA software is capable of coherence measurements on two separate channels. As an example, the UWB test signal can be split into two channels using a broadband power splitter. One signal is measured as a reference by channel 1 of the oscilloscope. The other signal can be transmitted over a wireless channel and measured by channel 2 of the oscilloscope (Fig. 5). Depending on the path loss and antenna gain and efficiency, it may be necessary to amplify the test signal with a low noise amplifier (LNA). Figure 5 shows the measured frequency spectrum of reference and test signals using the VSA software.

The upper trace in Fig. 5 displays the three OFDM channels of band group 1. This UWB signal was generated using the Agilent PSA signal generator with wideband modulation. The middle trace shows the measured test signal that is buried under the noise floor of the receiver. The lower trace shows the coherence between channel 2 and channel 1. Even though the received signal is buried in the noise, the relative level to the reference channel can be measured. Figure 5 shows that the received signal is −40 dB lower than the reference. The vector signal analyzer software calculates the coherence from the two measurement channels as a function of the type of averaging used. If rms averaging is used then the displayed frequency response is the cross power spectrum divided by the power spectrum of channel 1. This technique is very useful when the original signal is available as a reference channel. The same configuration will be used next to measure the delta error vector magnitude of a UWB component.

System and component designers often contend with specifying a component under a variety of stimuli. Different radio formats can create variations in the distortion products when these signals are applied to a nonlinear device under test (DUT), such as an amplifier or modulator. The signal distortion can be characterized and modeled using actual measurements made on the DUT. One such measurement is called error vector magnitude (EVM). EVM is a radio format-specific figure of merit of the modulation quality in a demodulated radio signal. This measurement requires the test system to be capable of demodulating the signal based on a radio standard.7 The EVM measurement can also be made by importing a time-captured waveform into a software simulation tool, such as the Advanced Design System (ADS) software suite from Agilent Technologies, and allowing the software to demodulate the signal and calculate the EVM. This is now possible for MB-OFDM, using a new Design Exploration Library in ADS.

An alternative approach to the EVM measurement, called delta-EVM, does not require demodulation of the radio signal but rather uses the input signal to the DUT as a reference. The delta-EVM measurement provides the level of deviation of the test signal relative to reference in terms of a percentage. Any distortion introduced by the DUT will result in a larger percentage of delta-EVM. Figure 6 shows the configuration for measuring the distortion parameters of an UWB amplifier. Here, the delta-EVM is calculated using the Agilent 89604A Distortion Suite loaded into the model 54855 real-time oscilloscope. The software can also determine the gain compression and AM to PM conversion of the amplifier resulting in best-fit curves using fifth-order polynomials. The upper two traces of Fig. 6 show the gain compression and AM/PM performance of the DUT and the associated polynomial curve fit. The resulting coefficients can be used by simulation software, such as ADS, in order to improve the model and optimization for UWB system simulations.

The OFDM signal has a high peak-to-average ratio. It is very useful to know the probability that the signal will exceed a certain value when designing or troubleshooting an UWB radio. Clipping due to limitations in the digital-to-analog converter (DAC) resolution or improperly setting the power-amplifier bias will degrade the radio-link margin. The probability-density function (PDF) and the related complementary-cumulative-distribution function (CCDF) show the percentage of time that the peak signal varies about the mean and the probability that the power level will exceed a specific amplitude, respectively.

The PDF and CCDF are statistical measurements; test time must be suitably long to gather enough samples to accurately represent true signal characteristics. Also, the instrument bandwidth must be wide enough to capture any transient responses. The CCDF curve is useful when specifying the saturation point and bias current for a nonlinear device. The VSA software can calculate the PDF and CCDF curves for the OFDM signal before or after the DUT. The middle curves of Fig. 6 show the PDF and CCDF of the signal after the OFDM signal passes through an amplifier under test. Curves for a Gaussian distributed noise source are shown for comparison.

REFERENCES

  1. "Ultra Wideband Communication RF Measurements," Application Note 1488, Agilent Technologies, Palo Alto, CA, May 2004.
  2. US 47 CFR Part 15 Technical Requirements for Indoor UWB Systems, 15.517.
  3. MBOA SIG website, http://www.multibandofdm.org.
  4. Peter Cain, "Direct-sequence UWB signal generation and measurement," RF Design, November 2004.
  5. A. Batra, "Multi-band OFDM: A New Approach for UWB," Circuits and Systems, 2004. ISCAS '04. Proceedings of the 2004 International Symposium, Volume 5, May 2004.
  6. J. Balakrishnan, et al., " A Multiband OFDM System for UWB Communication," Proceedings of the IEEE Conference on Ultra Wideband Systems and Technologies, November 2003.
  7. "Using Error Vector Magnitude Measurements to Analyze and Troubleshoot Vector-Modulated Signals," Product Note 89400-14, Agilent Technologies, Palo Alto, CA.