Despite all of the talk about the benefits of a time-division-duplexing (TDD) scheme, the reality is that initial third-generation (3G) wideband CDMA (W-CDMA) systems will be developed around a frequency division duplex (FDD) approach. Why you ask? Simple. FDD provides a similar operating format for 3G operators and equipment developers (separate bands for transmit and receive operation) while TDD calls for carriers to abandon traditional approaches and use a single 5 MHz channel for transmit and receive.
But while the FDD format is similar in nature to existing cellular approaches, the complexity of an FDD signal format complying with the 3GPP specifications imposes new performance requirements for linearity and dynamic range of amplifiers and other RF components. Signal statistics, such as ratio of peak power to average power and complementary cumulative distribution function (CCDF), take on an increased importance. In addition, it can be quickly deduced that traditional two-tone tests are not sufficiently accurate to characterize the nonlinearities of an amplifier on such measurements as adjacent channel power ratio (ACPR).
To further complicate issues, 3GPP base station and handset compliance measurements require specific test signals (or test models) for each measurement, such as:
- ACPR or adjacent channel leakage ratio (ACLR)
- Error vector magnitude (EVM)
- Spectral emission mask
- Output power dynamics
From the perspective of RF/analog engineers, the task of designing to 3GPP FDD compliance testing becomes much more complicated if they have to construct a specific signal, even before an EVM measurement can be made. Although each of the above mentioned topics merits a paper of its own, this article below will give an overview of CCDF, and a solution that meets the needs of an RF/analog engineer in terms of making 3G compliance measurements with minimum effort.
Key Statistics
The nature of W-CDMA leads to the superposition of many single channel signals and, therefore, a higher crest factor than communication systems such as TDMA GSM. Crest factor or ratio of peak power to average power is one of the key parameters of a 3GPP signal.
Another key parameter is the statistical behavior of the signal, in particular, how often the peak values occur. A CCDF graph indicates the probability, or percentage, of a particular peak-to-average ratio value occurring, which is an important factor in designing high-power linear amplifiers. The statistical behavior of the 3GPP signals depends on:
- The number of code channels or simultaneously transmitted signals
- The selection of channelization codes. Unlike IS95 where the spreading factor is always 64, the 3GPP spreading factor can vary between four and 512, yielding a larger selection of channelization codes.
- The correlation of user data. An unlikely event
- The timing offset between code channels. The offset between a base station frame and a mobile frame. For a particular selection of channelization codes, choosing different timing offsets for each channel can reduce constructive interference and in turn minimize crest factor.
In Figure 1, the CCDF of a base station signal that is generating 64 channels simultaneously is compared to that of the CCDF of an additive white Gaussian noise (AWGN) source of equal average power. Each channel was arbitrarily assigned a different channelization code and timing offset.
Click here for Figure 1
Figure 1: Comparison of the CCDF of a base station signal that is generating 64 channels simultaneously and the CCDF of an AWGN source of equal average power.
In Figure 1, the dedicated physical channel (DPCH) and all of the common channels were transmitted along with the 64 channels. The first thing to note from this figure is that the maximum peak-to-average power ratios differ by approximately 2.3 dB (10.47 - 8.21). The second observation is that 1% of the time the AWGN signal will exceed a ratio of 6 dB while 2.5% of the time the base station signal will exceed the same ratio.
The implication derived from the CCDF graph shown in Figure 1 is that if the amplifier was designed using the AWGN source and an actual 3G signal were passed through it, it would be over driven. This would lead to an increase in ACPR.
Downlink Test Modules
Several downlink test models are specified in section 6.1.1 of "3GPP TS25.141", the document that covers 3GPP base station (node-b) conformance testing.1 According to what measurement is being made, RF/analog designers have to use the appropriate test model.
Constructing the necessary model (base-band signal) from scratch, by writing proprietary code, or from a series of primitive models, is an overwhelming task for engineers not involved in signal processing. The 3G specifications for the test models are complex and highly varied. For example, test model 4 (section 6.1.1.4) used for an EVM measurement is defined in Table 1.
Table 1: Downlink Test Model 4 used for EVM Measurements
Click here for Table 1
Then, from section 6.1.1.5, the structure of the DPCH and all the common channels (P-CCHP, CPICH, PICH, SCH...) are defined. A high-level block diagram of what is needed to construct the base station downlink DPCH is shown in Figure 2.
Click here for Figure 2
Figure 2: Components necessary to generate a 3GPP DPCH signal.
Thus, the complexity of constructing just one test model from basic building blocks, in this case of test model 4, for EVM measurement is quite time consuming.
Pre-Configured Test Models Help
In Figure 3, a pre-configured test model is used with a system simulator for conformance testing to the 3GPP TS25.141 EVM measurements. The simulation tool that was used for this example also supports test models 1, 2, and 3 for base station conformance test as well as the models according to conformance testing for handsets.
Click here for Figure 3
Figure 3: 3GPP FDD test bench developed using a conformance test suite and a system simulator.
Through the models provided in the simulation tool, designers can concentrate on simulation work rather than reading and learning the details of the standard each time. Generic base station and handset models are also provided for the system engineer, or any engineer, who wishes to do trade-off studies on the impact of multiple users and channelization code selection on CCDF.
The system shown in Figure 3 was used to perform simulations of EVM measurement as defined in 3GPP TS25.141 section 6.7.1. From section 6.7.1, EVM is defined as a measure of the difference between the reference waveform and the measured waveform. This difference is called the error vector.
Under 6.7.1, both waveforms must pass through a matched root-raised cosine filter with bandwidth 3.84 MHz and roll-off α=0.22. Both waveforms are then further modified by selecting the frequency, absolute phase, absolute amplitude and chip clock timing to minimize the error vector. The EVM result is defined as the square root of the ratio of the mean error vector power to the mean reference power expressed as a percentage.
In this simulation, test model 4 is used to generate a base station downlink signal that is passed through a power amplifier design to an ideal receiver. In order to reduce the EVM to an absolute minimum, an identical root-raised cosine (RRC) filter characteristic is used in both the transmitter and receiver.
The receiver is designed to measure the timing offset and adjusts its chip timing-offset to within 1/8th chip of the ideal sampling point. The receiver's maximal ratio combiner (MRC) output is used to recover the sampled chips (prior to de-scrambling and de-spreading) and these samples are passed to the EVM measurement block. The EVM measurement sub-circuit performs an automatic gain control (AGC) function to normalize the amplitude of the received signal with respect to the reference signal prior to the EVM calculation.
All that is required from the RF/analog engineer is the amplifier design. However, an engineer can have the first cut of the design by making use of a behavioral amplifier that is governed by values entered for 1-dB compression point, third-order intercept (TOI) point, second-order intercept point, gain, and noise figure. As the design progresses, the user can replace the behavioral amplifier with a circuit design. The nonlinearities of the circuit design are automatically brought up to the system level. At this point, conformal measurements can be started using the actual amplifier designed at the transistor level.
In Figure 4, a graph displaying the AM-to-AM characteristic of the amplifier under test is shown along with the instantaneous power of the transmitted signal and the average signal power. As the average power of the base station signal is increased or decreased, the use of such a curve enables the engineer to monitor when the amplifier is going out of and into compression. The scatter plot or I/Q plot is shown for different power levels. As expected, as the amplifier is driven into saturation the measured EVM value increases substantially.
Click here for Figure 4
Figure 4: Diagram of the AM-to-AM characteristic of an amplifier under test.
With such a system, the RF/analog engineer can better understand the nonlinearities of a base station power amplifier on EVM.
Taking it One Step Further
The simulation shown in Figure 4 can be taken one step further by enabling the engineer to understand the impact of AM-to-AM distortion independent of AM-to-PM and vice-versa. Such visualization tools and pre-configured test models not only give the RF/analog engineer greater insight but they also enable the entire engineering team to do trade-off studies concurrently between system level and circuit level specifications.
The 3GPP test blocks provided in the simulation solution detailed above are compliant with the 3GPP FDD UMTS specifications (Rev 3.9).2 IC and system-level simulation solutions can significantly accelerate 3G design by incorporating transistor level effects into high-level simulations.
References
- "3GPP TS25.141" 3GPP FDD Base Station Conformance Testing; Latest specifications are available at: ftp://ftp.3gpp.org/specs/latest/
- 3GPP FDD UMTS Specifications; Latest specifications are available at: ftp://ftp.3gpp.org/specs/latest/
About the Authors
Joel Kirshman is a systems engineer at Applied Wave Research. He has a BA in mathematics from Occidental College and an MS in electrical engineering, California State University, Northridge. Joel can be reached at joel@mwoffice.com.
Kurt Matis is the director of systems research at Applied Wave Research. He holds a BS in mathematics from Empire State College as well as an MS in electrical engineering and Ph.D. from Rensselaer Polytechnic Institute. Kurt can be reached at kurt@mwoffice.com.
