Accelerating ZigBee and 802.15.4 Module Testing with LabVIEW and an NI RF Vector Signal Analyzer

Jeremy Willden, MaxStream, Inc.

"For all measurements in the MaxStream test fixture, the PXI-5660 outperformed the Agilent 8594E by a factor of more than 10."

- Jeremy Willden, MaxStream, Inc.

The Challenge:

Improving throughput, repeatability, and accuracy in the current production test environment at MaxStream.

The Solution:

Using the National Instruments PXI-5660 vector signal analyzer and NI LabVIEW for ZigBee/802.15.4 and proprietary radio RF production test.

 

ZigBee is a wireless networking technology targeted at consumer, commercial, government, and industrial applications for low- to medium-data rate communication. The specification includes mesh networking for high reliability, sleep modes for low power consumption, and low implementation cost. The physical over-the-air transmissions are governed by the IEEE 802.15.4 specification, which provides 250 kbps in the 2.4 GHz band. Other frequency bands work at lower data rates, but worldwide acceptance affords the greatest popularity to products based on 2.4 GHz technology. ZigBee offers up to 16 channels using direct sequence spread spectrum, OR DSSS.

 

MaxStream is a leading provider of embedded wireless solutions, both standards-based and proprietary. By adding the NI PXI-5660 vector signal analyzer to our standard production test fixture, we improved throughput, repeatability, and accuracy. The expandability and flexibility of PXI helped us to evolve our existing production test system to keep pace with our very high growth rate.

 

The NI LabVIEW graphical development platform has long been a key component in the MaxStream production test environment. We use it for a wide range of functional, RF, programming, and operational tests in a critical manufacturing setting. The rack-mounted test fixture, or RMTF, performs basic electrical testing, microcontroller programming, parametric RF testing and calibration, and module customization.

 

Each test fixture in our original production test system included an RF spectrum analyzer, an RF signal generator, an RF power meter, a DC power supply, an antenna for over-the-air tests, an off-the-shelf PC running the NI LabVIEW run-time engine, two National Instruments PCI-6025E data acquisition (DAQ) boards, and one PCI-GPIB interface to control the rack-and-stack instruments.

 

System benchmarking revealed that the RF tests – particularly the spectrum analyzer measurements – were limiting system throughput. The full battery of spectrum analyzer measurements required approximately 10 seconds. An analysis of execution time revealed two bottlenecks: The sweep time of the spectrum analyzer consumed a substantial percentage of the measurement time, followed closely by the setup and data transfer time across GPIB.

 

Changes in a production environment are serious due to the high potential impact on the company’s core business. To minimize any potential risk, we borrowed an NI PXI-5660 vector signal analyzer from National Instruments to benchmark the possible speed improvement. We also performed a full set of tests to identify any significant risks.

 

Within the first few hours, the speed advantage was clear. We placed the NI vector signal analyzer side-by-side with the Agilent 8594E. We performed several representative measurements at matching resolution bandwidths, data array sizes, reference levels, and spans. For all measurements in the MaxStream test fixture, the PXI-5660 outperformed the Agilent 8594E by a factor of more than 10.1

 

The spectrum analyzer measurements, which required approximately 10 seconds with a rack-and-stack instrument across GPIB, now complete in less than one second. The high IF digitizer bandwidth combined with the low latency and high-bandwidth PCI backplane in the PXI chassis offer much faster measurements and higher production throughput.

 

Virtual instrumentation provided significant signal generator improvements even though we retained the system’s original signal generator. In the original system, a proprietary piece of hardware provided the modulation signal to the RF signal generator. As the number of products and the data rates increased, the proprietary hardware was unable to meet the required performance – specifically the data rates required for ZigBee and 802.15.4. During the development process, we used a LabVIEW VI and an NI PCI-6250 M Series DAQ board to produce the modulating waveforms required for 802.15.4. During the transition to production, we enhanced this code to support all MaxStream radio types, replacing the proprietary hardware and improving performance.

 

The purpose of a production test system is to determine device assembly correctness in as short a time as possible. Implicit in this purpose is the desire to minimize test time while maximizing test accuracy. Some production test philosophies require little more than a typical use verification of operation. At the other end of the spectrum lies full rigorous validation of every specification across temperature, while subject to high vibration, temperature swings, and voltage variations. Obviously, test time and equipment requirements strongly influence test costs, which vary widely across the test completeness spectrum. Most test systems lie somewhere between the two extremes on this spectrum.

 

One benefit of using a standards-based chipset – as most or all ZigBee solutions do – is that some level of testing occurs before the chips are placed on the board. This reduces the cost of production test, and allows the test system to spend less time validating the chip and more time testing the system. A ZigBee module requires fewer tests than a proprietary transceiver, saving time otherwise required for detailed mask and modulation tests and allowing simpler, faster RF calibration, mask testing, and modulation analysis.

 

In production, the most critical measurements involve transmit power, frequency, and spectral masks. The PXI-5660, with a high 20 MHz IF bandwidth, offers much faster measurements of these parameters than most rack-and-stack instruments in the same cost range, which typically support a maximum of 3 to 5 MHz. The 802.15.4 spectrum exceeds the IF bandwidth of many of these instruments, requiring longer sweep or acquisition times to obtain the same data.

 

One of the features of the PXI-5660 is a digital downconverter with the ability to demodulate a transmitted signal in hardware, presenting the baseband data to LabVIEW and allowing analysis of the transmitted RF spectrum with far greater precision than a spectral mask.

 

While production testing focuses on maximizing throughput in a minimal amount of time, the goal of development testing, in contrast, is to maximize coverage and examination of all possible corner cases. Again, the PXI-5660 excels in this area. The NI Spectral Measurements Toolkit for LabVIEW includes VIs for many measurements applicable to ZigBee and 802.15.4, including in-band power, adjacent channel power, occupied bandwidth, and spectrum peak search. The NI Modulation Toolkit for LabVIEW provides more advanced signal generation and analysis useful in 802.15.4 testing, reducing the effort required to create data packet generators and receivers for bench testing and validation.

 

NI PXI instruments provide substantial performance improvements at cost parity or better compared to the rack-and-stack equipment that we replaced.

 

Future plans include the migrating all of the PCI boards in the test fixture PC to the PXI equivalents. National Instruments software and hardware make this transition seamless, with no code changes required. 

 

1There was only one exception – the NI PXI-5660 was faster by about a factor of two for a full 2.7 GHz spectrum scan, due to the large size of the data array returned for this measurement, while the 8594E always returns the same array size for any array measurement. This gives the 8594E a speed advantage, but with a substantially reduced number of measurement points. It is possible to configure the PXI-5660 to reduce the returned waveform array in the same way, but it was preferred to examine the larger data set rather than shorten the measurement time, which is a very small percentage of the overall test time.

 

Author Information:

Jeremy Willden
MaxStream, Inc.
Tel: 801-765-9885
jeremyw@maxstream.net

Figure 1. The illustration above shows the final production test system.