RF WCDMA Benchmark Comparisons Whitepaper

Publish Date: Apr 05, 2012 | 1 Ratings | 5.00 out of 5 |  PDF

Overview

Understand the speed improvements that software-defined PXI RF instruments deliver over traditional instruments in a benchmark comparison. As the WCDMA measurement results illustrate, inherently parallel LabVIEW measurement algorithms achieve significantly faster execution times than traditional instruments on multicore processors.

Table of Contents

  1. Introduction
  2. Introducing a New 6.6 GHz RF Test Platform
  3. AmFax Uses LabVIEW to Achieve Faster WCDMA Measurements
  4. Configuring RF Instruments
  5. Measurement Time Benchmark Results
  6. Measurement Averaging versus Repeatability
  7. Conclusion

1. Introduction

You awake at 7:00 a.m. to the sound of classic rock. The RDS receiver on your radio alarm clock reports that you are listening to Welcome to the Jungle by Guns N’ Roses. Later, as you sip your morning coffee, a WLAN transceiver lets you check your e-mail from the den. When you are ready for work, you walk out the front door and use a 315 MHz FSK transmitter to unlock your car doors. Backing out of your driveway in your car, you’re thankful that satellite radio provides commercial-free entertainment. Moments later, a Bluetooth transceiver in your ear communicates with your 3G cell phone. Within minutes, your GPS navigation system has acquired a 3D position fix and you’re on your way. The voice on your GPS receiver tells you to use the turnpike, where an RFID reader charges your car the appropriate toll.

RF technology is everywhere. It is easy to understand how this affects you as a consumer, but it has an even greater impact on you as an RF test engineer. The plunging costs of wireless devices have made time away from work easier, but it produces new challenges when designing next-generation RF automated test systems. More than ever, you are tasked with the goal of lowering cost of test. As a result, a significant concern for today’s automated test systems is focused on reducing overall test time.

Back to Top

2. Introducing a New 6.6 GHz RF Test Platform

To meet this need, NI has developed a 6.6 GHz high-speed RF measurement platform. These new products, the NI PXIe-5663 vector signal analyzer and the NI PXIe-5673 vector signal generator, offer a high-speed, flexible solution for automated RF measurements. The PXIe-5663 is capable of signal analysis from 10 MHz to 6.6 GHz with up to 50 MHz of instantaneous bandwidth. The NI PXIe-5673 offers signal generation from 85 MHz to 6.6 GHz and offers 100 MHz of instantaneous bandwidth. (See Figure 1.)

Figure 1. A PXI system with the new 6.6 GHz RF test platform.

The 6.6 GHz RF test platform is ideally suited for automated test applications. Using highly parallelized NI LabVIEW measurement algorithms, PXI modular instrumentation delivers measurement speeds that significantly outperform traditional instrumentation. To understand why PXI modular instrumentation is capable of faster measurement speeds than traditional instrumentation, consider the architectural differences between PXI modular instruments and traditional instruments. While both systems use many similar components, a core distinction is that PXI systems can use high-performance multicore central processing units (CPUs). For an illustration of this concept, see the block diagrams of both types of instrumentation in Figure 2.

Figure 2. A user-defined CPU is a critical component of PXI RF instruments.

Although PXI and traditional instruments share many similarities, the user-defined multicore CPUs available on PXI modular instruments enables much faster measurement times. In many cases, RF measurement algorithms are written in the inherently parallel LabVIEW programming language. As a result, overall measurement time can be reduced by upgrading to processors with more processing cores. As CPU clock speeds (or the number of cores) increase in accordance with Moore’s Law, today’s PXI RF testers can achieve even more remarkable measurement speeds. As you will observe in the benchmark data in this article, many PXI vector signal analyzers can execute processor-intensive RF measurement algorithms up to 30x faster than traditional benchtop vector signal analyzers. 

To understand the benefits of PXI instrumentation, consider applications such as high-volume wireless devices testing. In this scenario, test time is often a significant percentage of the products COGS (cost of goods sold). Moreover, for wireless communications protocols such as the 3G UMTS (WCDMA) standard, the processor-intensive measurement algorithms require a significant processing power. For this application, National Instruments Alliance Partner, AmFax, offers highly parallel measurement algorithms for WCDMA physical layer measurements. In this case, NI RF instrumentation and partner software combine for a low-cost, high-speed, and high-accuracy test platform.

 

Back to Top

3. AmFax Uses LabVIEW to Achieve Faster WCDMA Measurements

To illustrate the measurement speed and accuracy of the PXIe-5663 RF vector signal analyzer, consider a head-to-head comparison with leading traditional instruments (see Table 1). Both comparison instruments are relatively new RF vector signal analyzers (VSAs), and are considerably more expensive than a full PXIe-5663 RF measurement system.

 

Instrument A1

Instrument B2

PXIe-5663

Instrument Type

Traditional RF VSA

Traditional RF VSA

PXI Express RF VSA

Frequency Range

9 kHz to 8 GHz

1 MHz to 8 GHz

10 MHz to 6.6 GHz

1Instrument A is a Rhode and Schwartz FSG

2Instrument B is a Rhode and Schwartz FSQ

Table 1. Comparison of PXI and Traditional Instruments

To provide the most realistic benchmarking data, the PXI and traditional instruments can be timed in a series of standard-specific measurements. For WCDMA applications, consider instrument performance for a wide variety of measurements. PHY layer measurements such as complementary cumulative distribution function (CCDF) require long acquisition times, and are less affected by processor speed. On the other hand, measurements that require demodulation, such as error vector magnitude (EVM), require substantial signal processing. Finally, frequency domain measurements such as adjacent channel leakage power ratio (ACLR) and occupied bandwidth (OBW) should also be tested, as they require discrete Fourier transform (DFT) computation.

With a common test executive architecture such as NI TestStand software, you can quickly configure a sequence of automated measurements. Not only does NI TestStand software provide a built-in framework for sequencing measurements, but it can also be configured to report measurement times for each test. A screenshot of NI TestStand reporting measurement times for an automated test sequence is shown in Figure 3.

Figure 3. NI TestStand automates measurements in a production test environment.

In Figure 3, observe the nested for loops surrounding the EVM measurement steps (“NI Configure EVM” and “NI Measure EVM”). The outer for loop determines the number of averages for a given measurement, and the inner for loop sequences through multiple instances of the same measurement settings. By taking multiple measurements with the same configuration settings, you can perform statistical analysis of the measurement data to determine the result mean and standard deviation.

Back to Top

4. Configuring RF Instruments

When performing instrument benchmarking, it is crucial that you configure each instrument to yield the fastest measurement performance. For traditional instruments, you achieve the best performance by using the instrument’s onboard averaging functions instead of averaging each trace manually. In addition, the front panel display should be turned off while the measurement is running. Finally, it is important to choose the most efficient instrument control bus. Because the data size for these measurements is small, the instrument control bus chosen should be optimized for latency. Thus, choose the GPIB bus over LAN for best latency results. As a general rule, latency more significantly affects measurements when little or no averaging is employed. 

To benchmark an RF vector signal analyzer’s measurement speed, it should be configured in a loopback mode with an RF vector signal generator. To evaluate the PXIe-5663 VSA, the new PXIe-5673 6.6 GHz RF vector signal generator can be configured to source the test stimulus. Consistent with the WCDMA standard, this stimulus should be at center frequency of 1.95 GHz. Configure an RF output power of -10 dBm, and cable the generator directly to the vector signal analyzer. Figure 4,  illustrates the hardware configuration.

Figure 4. Connect VSA directly to VSG.

While you should use an actual device under test (DUT) to determine characteristics such as measurement repeatability, one benefit served by the loopback approach is that you can use it to illustrate the measurement performance of the instruments.

Back to Top

5. Measurement Time Benchmark Results

Using the setting described above, observe the measurement times (in seconds) for each of the following measurements. Note that the number of averages used in Table 2 was chosen based on values that are often used in design validation applications. In a subsequent section in this article, you will learn more details on the relationship between the number of averages and measurement repeatability.

Typical Measurement Times for Various Measurements

All times in seconds # of Avgs Instrument A Instrument B NI PXI-5663 w/NI PXIe-8130 NI PXIe-5663 w/NI PXIe-8106 NI PXI-5663 w/NI 8353
CCDF  1M 0.505 0.510 0.488 0.330 0.384
EVM time 20 3.142 3.130 0.822 0.577 0.519
ACLR time 20 3.070 3.100 0.200 0.174 0.168
OBW time 20 4.554 4.540 0.217 0.188 0.167
Total time   11.270 11.280 1.727 1.269 1.070
Speedup versus Instrument A   1x 1x 6.56x
Faster
8.88x
Faster
10.53x
Faster

Table 2. WCDMA measurement times for traditional and PXI instruments.

As Table 2 illustrates, the PXIe-5663 RF vector signal analyzer with either the embedded or rackmount controller delivers superior measurement speed over traditional instrumentation. In addition, observe the effect of processor speed on overall measurement time. With the controllers shown, the NI PXIe-8130 embedded controller uses an AMD Turon X2 2.3 GHz CPU and the NI PXIe-8106 uses a 2.16 GHz Intel Core 2 Duo CPU. A quad-core processor, the NI 8353 1U rackmount controller, uses dual 2.4 GHz Core 2 Duo CPUs. Because CPU performance directly determines measurement speed, the quad-core controller is able to deliver even faster measurement times than the fastest dual-core embedded controllers. See Figure 5 for a chart illustration of the overall measurement time reduction as percentage of traditional instrument time. 

Figure 5. NI 8353 1U controller reduces test time by 83 percent over traditional instruments.

For most WCDMA PHY layer measurements, processing time has the biggest influence on overall measurement time. For these measurements, overall time is often proportional to the number of averages used. One exception is measurements that require particularly long acquisition sizes such as CCDF. In this case, the processor has a smaller influence on overall measurement time. Observe in Figure 6 that for CCDF measurements, PXI measurement systems are slightly faster than traditional instrumentation. 

Figure 6. Number of averages has little affect on CCDF measurement time.

To determine the exact performance improvements that can be observed with PXI instrumentation, these measurements should be performed over several trials. All data shown below is the mean of 10 trials for each configuration. In Figure 6, CCDF measurement time can be reduced by 33 percent when using a PXI-based measurement system instead of traditional instrumentation. Here, you can observe that the NI 8353 quad-core rackmount controller yields the fastest measurement time.

For processor-intensive PHY layer measurements, the choice of processor greatly affects the overall measurement time. In Figures 7 through 9, observe the relationship between measurement time and number of averages for both traditional and PXI instrumentation. 

 Figure 7. PXI instruments yield biggest improvements in processor intensive measurements.

For measurements such as EVM, which are processor-intensive, the choice of controller can significantly affect measurement time. For example, while an EVM measurement with five averages might take 342 milliseconds with the NI PXIe-8130 embedded dual-core controller, it can be reduced by 33 percent to 228 milliseconds with NI 8353 quad-core controller. A similar result is observed with the adjacent channel leakage ratio (ACLR) measurement, shown in Figure 8.

 

Figure 8. Measurement time versus number of averages for ACLR

The measurement time for ACLR can be more than 16 times faster when using PXI RF measurement systems.   Note that the typical test time for a single ACLR measurement (not counting configure time) was less than 8 ms, a time that is significantly faster than even time-domain ACLR measurement implementations.  A final measurement result, occupied bandwidth, is shown in Figure 9.

Figure 9. Occupied bandwidth measurements can be performed up to 30X faster with PXI instruments.

In Figure 9, you can see that for some measurements, PXI RF instrumentation yields the same results up to 30x faster than traditional instruments. In addition, the absolute measurement time improvement is more pronounced in situations where significant averaging is required.

Back to Top

6. Measurement Averaging versus Repeatability

While averaging can significantly increase overall test times, it is often required to achieve more repeatable measurements through averaging. Because measurement averaging actually increases measurement time, it is important to understand the relationship between repeatability and number of averages. Because the noise can be mitigated through averaging, you will observe that run-to-run repeatability improves as the number of averages increases. In Figure 10, the EVM standard deviation is compared to the number of averages configured for each measurement.

Figure 10. EVM standard deviation versus number of averages

For the results shown in Figure 10, all EVM measurements were made over a period of 1 WCDMA frame, which is equivalent to 2600 chips. Observe the relationship between measurement repeatability and number of averages. While only 1000 trials were used to calculate the data set shown in Figure 10, many production test applications require substantially larger data tests. In fact, many test runs using multiple instruments should be used to produce a more accurate model of measurement repeatability.

Using the PXIe-5663 RF vector signal generator configured in a loopback mode with the PXIe-5673 RF vector signal generator, you can easily achieve better than 1 percent EVM measurements. Table 3 illustrates the mean and standard deviation for various configurations.

 

Trials

# Averages

Mean EVM Value

STD Deviation

1000

1

0.82343%

0.01276%

1000

5

0.82171%

0.00398%

1000

10

0.82076%

0.00225%

1000

25

0.82055%

0.00143%

1000

50

0.82056%

0.00098%

1000

100

0.82063%

0.00066%

Table 3. EVM and standard deviation as a function of number of averages

The NI 5663 RF vector signal analyzer and PXIe-5673 RF vector signal generator produce both accurate and repeatable EVM measurements for the WCDMA standard.

Back to Top

7. Conclusion

As the adoption of wireless technology continues to accelerate, even greater pressure will be placed on the requirement to reduce measurement time. Fortunately, PXI measurement systems are able to harness advances in the computing industry. In fact, as the data in this article describes, parallel measurement algorithms executing on multicore PXI processors are significantly faster than similar algorithms on traditional instrumentation. Thus, with the new PXI 6.6 GHz RF measurement platform from National Instruments, you address the increasing need to lower the cost of RF test. For more information on this platform, visit ni.com/rf/platform.htm or the RF Homepage.

 

Back to Top

Bookmark & Share

Ratings

Rate this document

Answered Your Question?
Yes No

Submit