Optimise Your Test System Smartly

0
118

As higher targets are set to increase profits and decrease operating costs, test engineers are subjected to the ever-increasing stress of improving the efficiency of their testing operations and ultimately getting higher throughput for the same number of work hours. Read on to find out some simple ways to enhance your testing productivity

DILIN ANAND


D28_TmJUNE 2012: Chips are becoming faster and more efficient keeping with Moore’s Law. This is pushing consumer electronics technologies to new heights while making them all the more complex. Smarter test equipment are required to keep up with the new technology. Also, test engineers need to improvise techniques in order to enhance the productivity of their operations. Here is how you can improve your test performance.

Upgrade embedded controller, lower measurement time
As Intel and AMD come up with newer microarchitectures on their processors, there is a definite impact on the system controller and, in turn, the testing performance and productivity. The just released Ivy Bridge microarchitecture—a 22nm die-shrink of the previous Sandy Bridge microarchitecture—brings in a plethora of new features that enhance computing performance.

If you go for an embedded controller that features a processor with a large cache, the controller wouldn’t have to access the comparatively slower DDR3 RAM for that data. The lower latency of the cache will, in turn, improve processor performance, delivering faster measurements and results. This is especially true for operations that require intensive signal and data processing.

Explore Circuits and Projects Explore Videos and Tutorials

“Especially for RF measurements and RF protocol testing, CPU performance is often the single most significant factor preventing faster measurement performance. While actual system performance can depend on a variety of factors such as memory available and other applications running in the background, a strong correlation exists between the CPU performance and measurement time for automated test systems,” explains David Hall, product manager for RF and Communications at the National Instruments Dev Zone.

Fig. 1 shows how the PXIe-8106 with 4MB cache and 2.16GHz clock is able to outpace the PXIe-8130 with 1MB cache and 2.3GHz clock.
Fig. 1 shows how the PXIe-8106 with 4MB cache and 2.16GHz clock is able to outpace the PXIe-8130 with 1MB cache and 2.3GHz clock.

If your system has a latest processor, you can further enhance the productivity of your test system. By smartly utilising multi-core processors in the test system, the testing time can be reduced by running multiple threads for each task. Tasks such as frequency analysis are heavily dependent on the processor’s computing performance for faster results. By optimising the algorithm to allocate each channel to multiple cores on the processor, you can improve the processor throughput. One of the reasons for the lower computing time is that the processor is able to execute Fast Fourier Transform (FFT) frequency analysis simultaneously on all cores.

READ
Make Your PCB Design Ideas Fly High With EAGLE

Implement pipelining, improve throughput
Pipelining is a method used in a variety of fields to improve the throughput of an assembly-line-like system. It does not reduce the time to execute a test, but rather increases the number of tests that can be done simultaneously. More importantly, in pipelining, every test cycle takes as long to complete as the slowest cycle. But in most cases, the gain from implementing pipelining far outweighs the minor losses. Pipelining works best when each test takes almost the same time, as it minimises waste of time.

Pipelining works by delaying the start of testing each unit by one cycle. This delay allows later test sequences to run concurrently. Table I shows improvements in performance due to pipelining.

Re-order and schedule, maximise instrument utilisation
Once you have a faster processing and pipelining set-up, the next step is to improve instrument utilisation using test automation software. Even when you optimise your processor to improve upon the compute-intensive tasks, there is always a chance that your instruments will lie idle. To prevent such wastage of resources, auto-schedule tasks to run whenever the instrument is detected to be idling.

3F4_table-1

Not every set of tasks can be reordered. Move the tasks that can be reordered to fill in the idle instrument cycles. If you compare Table II and Fig. 2, you will notice that this methodology allows you to further enhance productivity by reducing the test process from five cycles to three.

“You’ve effectively reduced your test time by 66 per cent, and you’re making the most of your instruments by cutting instrument downtime to zero. The bottom line is that by taking advantage of the advanced parallel testing capabilities of NI’s software-defined automated test platform, you can use one set of instruments (that is, one test station) to test multiple devices in parallel. This, in turn, means you can test the same number of devices with fewer test stations, thereby drastically reducing the capital cost of test equipment without sacrificing test throughput,” explains Satish Mohanram, business development manager, National Instruments India.

READ
FPGA Development Boards

Use composite measurements, reduce measurement time
In order to reduce measurement time, “The idea is to perform composite measurements instead of configuring each measurement separately. Because composite measurements calculate multiple measurement results on a single burst, they are more efficient than when each measurement is performed sequentially. Multicore processors enable this to be done in parallel, resulting in a faster overall test time when using a software-defined approach,” suggests Tarun Gupta, business development manager at National Instruments.

Consider the case of measuring both power and error-vector magnitude (EVM) of a WLAN transmitter. Typical measurement times for these individual measurements are listed in Table III. In this case, each 802.11a/g EVM measurement is calculated using 16 OFDM symbols. The gated power measurement is performed on the portion of the burst from 20 to 120 µs.

An example: Testing smartphones
Assume that you are testing smartphones and running three different tests on each: a power consumption test that uses a programmable source measure unit (SMU), a GSM test that uses an RF vector signal generator (VSG) and an audio quality test that uses a dynamic signal analyser (DSA). Assume that each of these tests takes one unit of time to execute. On a typical test system that tests one phone at a time sequentially, you would test these phones at a rate of one device under test (DUT) every three units of time. Testing three phones would take nine (3×3 = 9) units of time.

The sequential test example in Fig. 2 shows that each test instrument is left unused for six of nine time units. That’s a 66 per cent downtime per instrument! Imagine how much more efficient your test system could be if you use this downtime to start testing the next phone. This is precisely what you can do with parallel testing.

READ
Seven Ways to Effectively Use Today’s Feature-Heavy Scopes

Fig. 2: Performance gains from implementing parallel test with autoscheduling
Fig. 2: Performance gains from implementing parallel test with autoscheduling

9CA_table-3Table III shows that when you perform key 802.11a/g measurements such as EVM and power as a composite measurement on a single burst, the total measurement time is significantly lesser than the sum of each individual measurement time.

Ensure upgradability, optimise time and investments
Being smart during the initial phase of chipset design can go a long way to reduce the pain of any change to be made on the hardware at later stages. By using readymade libraries in the design or simulation software, a designer can optimise the time at the initial stages itself.

Sadaf Arif Siddiqui, technical marketing specialist, Agilent Technologies, explains how the W1917EP WLAN baseband verification library helps in this respect. “The W1917EP WLAN baseband verification library is a new Layer 1 simulation reference library for Agilent SystemVue. The blockset, reference designs and test-benches of the W1917 assist the design and verification of multi-format radios, by providing configurable physical-layer waveforms for networking standards 802.11a and 802.11ac,” he says.

A new breed of spectrum analysers is now available that provides a different approach to signal analysis. For instance, X-Series spectrum analysers are able to run MATLAB and 89600 VSA software inside the analyser itself. Moreover, it provides a library of measurement applications that can be used with the spectrum analyser. For instance, one such measurement application enables the X-Series signal analysers to emulate the FSP/FSU/FSE spectrum analysers or ESU EMI receiver in remotely controlled, automated test equipment systems. Another application in the same library enables users to check the electromagnetic interference by running pre-compliance radiated and conducted emissions tests for both commercial and military standards.

As all these applications are available from a single point, they help test engineers to achieve measurement integrity and drive productivity.

Get started
Now that you have got an idea about the numerous factors that can optimise your test system, you can put these to use. You could also devise your own tweaks to get more performance out of your system by balancing trade-offs.


The author is tech correspondent at EFY Bengaluru

LEAVE A REPLY