The Need for Embedded Intelligence in ATE

Introduction

Since the discovery and invention of electronic semiconductor materials in the late 1950's, improvements in semiconductor manufacturing yield has roughly followed Moore’s Law, which predicts the doubling of electronic device density every one to two years [1]. For example, figure 1 shows the growth in the number of transistors in Intel microprocessors since 1971 [2].

Figure 1: Transistor count in Intel microprocessors

According to Gordon Moore, cofounder of Intel, the cost to manufacture integrated circuits is nearly independent of the number of components on the device. He states that costs and complexities of new devices are only limited by manufacturing yields. As improvements in manufacturing have led to better yields, the optimum complexity has grown accordingly. As a consequence of the exponential growth in electronic device complexity, a second limiting factor not foreseen by Moore has emerged. The cost and complexity of comprehensive testing has grown to become a significant portion of the total cost of electronic component manufacturing [3]. Consequently, more sophisticated electronic testing techniques are required to keep pace with the advances in electronics technology.

Cost Of Test

There are two components to the cost of test for an electronic device: (1) the manufacturing cost of testing each device, and (2) the design, development and maintenance engineering cost of the actual test. The Automatic Test Equipment (ATE) industry continuously strives to minimize these two elements of the cost of test. Manufacturing test cost per device has been restrained by advances in test automation and speed, offsetting the exponential increase in testing requirements. Unfortunately, test development costs have increased dramatically due to the increased scope and complexity of the more sophisticated testing.

Modular Instruments

Modular instruments are well suited to ATE applications that require small sizes, high data transfer throughput, and highly integrated system software. The instrument-on-a-card concept can be applied to address both requirements of decreased test time per device and lower test development costs. An ideal highly integrated modular ATE system will perform sophisticated parallel processing to rapidly test, characterize and diagnose failures.

A limiting factor in test time using standard bus-based modular instruments is the transfer speed from the instrument to the controlling computer. Electronic component testing requires the processing of waveform signals to extract test information. When the processing is performed at the instrument using embedded intelligence, only the results are transferred to the controller. When the controller performs the processing, all raw data must be transferred to and from the instrument, significantly increasing data throughput requirements. Benchmarks have demonstrated better than 10X improvements in test time by performing standard test and measurement functions using instruments with embedded intelligence [4]. Without embedded intelligence, the greatly increased data transfer volume often negates the backplane bus speed advantage of the modular architecture.

Distributed parallel test and measurement processing is a key aspect necessary to increase test throughput. Parallel processing enables multiple intelligent test instruments to perform signal and test processing simultaneously, with the host computer providing only test sequencing and data logging. Unfortunately, in practice many modular instruments fall short of their benchtop counterparts by lacking advanced signal processing and measurement functions. This omission slows the testing throughput, and places an added burden on the ATE test developer to add the standard test and measurement capabilities necessary for automated test.

Embedded Intelligence

When selecting test equipment, ATE developers must scrutinize analog and mixed-signal specifications. Other, equally important specifications that are sometimes overlooked include the standard test and measurement processing functions. Instrument test and measurement functions require embedded intelligence to perform added functionalities necessary for test, including sophisticated signal processing and transformation, pass/fail characterization, and the application of standardized test methodologies. These processing techniques can be used to rapidly identify and diagnose component failures.

In addition to decreasing data-transfer requirements, signal processing performed at the instrument enables sophisticated test and troubleshooting techniques. For example, standard intelligent oscilloscope processing algorithms include waveform measurements, math operations, digital filtering, Fourier transforms, waveform averaging, envelope detection, and equivalent-time sampling. Taking embedded intelligence another step further applies the results of the signal processing to limits or masks for pass/fail testing. Statistics of intermittent or infrequent failures can be rapidly characterized with an instrument that can identify, count and record failures. The final step in applying embedded intelligence is to use this pass/fail methodology with industry-standard compliance limits or masks. An intelligent instrument will perform sophisticated tests and test sequences in parallel with other instruments, independent of the host processor.

ATE Examples

The following two examples describe common tests that are rapidly performed using intelligent instruments, but are very difficult and time consuming to implement and execute without the embedded intelligence.

Telecom Mask Testing

Telecommunication standards require specific compliance testing for all telecom network equipment. Transmission signal quality specifications are particularly important. Standardized tests include the verification of signal integrity for parameters of the digital data transmission stream, including pulse width, risetime, falltime, overshoot and undershoot. The captured signal must fit within a predefined template called a pulse mask. For example, figure 2 shows a standard T1 network pulse mask. The captured raw data is aligned and tested repetitively for thousands of pulses to detect any non-compliant signal pulses (signals outside the compliance mask) and to measure pass/fail statistics. Mask testing is also widely used for testing non-standard digital data transmission systems using customized masks.

Test equipment that employ embedded intelligence can rapidly perform mask testing at rates exceeding thousands of sweeps per second. When selecting test equipment for digital network testing, the test instruments must include the following capabilities: pulse width triggering, reference waveform storage, mask testing, and limit test statistics. A library of standard masks or the ability to create custom masks may also be necessary for certain applications. During test, captured signals are tested against a specified mask, the instrument keeps track of pass/fail statistics, stores non-compliant waveforms, and either stops or continues the test upon failure. Unattended or long-term testing is accomplished through the use of these automated actions upon failure.

Figure 2: T1 Network Pulse Mask

Video Signal Capture and Analysis

Capturing and analyzing video signals using test equipment without the proper embedded intelligence can be time consuming and error-prone. Being able to isolate and analyze a specific line of video is a requisite feature of any video test equipment. Common video signal tests include measuring specific line parameters, sync timing, and amplitude and frequency characteristics of the video signal.

Test equipment that employ embedded intelligence can synchronize to specific frames, fields and lines of the video signal, and perform measurements upon gated portions of that captured waveform. Video triggering capabilities isolate single lines of video, and embedded signal processing is used to perform standard tests. For example, figure 3 shows data capture of NTSC Line 50 of a color bar test pattern signal.

Figure 3: NTSC Color Bar video line capture

The amplitude of a video waveform is defined in IRE units. An IRE unit is a relative unit of measure equaling 1/140th of the peak-to-peak (pp) video amplitude. For a typical 1 Vpp video signal, one IRE unit is nominally 7.14 mV. For video signals with non-typical signal levels, the test equipment will normalize the sync triggering and measurement levels to the actual amplitude. Gated measurements are used to check sync width, sync and color burst levels, frequency response, chrominance-to-luminance gain, luminance non-linearity, and differential gain and phase. When selecting test equipment for video testing, the test instruments must include the following capabilities: video triggering, gated measurements, selectable standards (NTSC, PAL, SECAM), and non-standard video capture setup for military applications, medical displays, security cameras, etc.

Conclusion

The techniques described in this paper demonstrate the significant advantages and cost savings of embedded intelligence in ATE. Embedded intelligence within an instrument provides a comprehensive test solution beyond simple data acquisition, enabling faster test development and shorter testing times. The embedded signal processing features must be considered when selecting test equipment. By selecting and integrating instruments that provide embedded signal processing, an ATE test developer can significantly reduce the total cost of test.

References

[1] G.E. Moore, Cramming more components onto integrated circuits, 1965. Electronics, volume 38, number 8.

[2] I. Tuomi, The Lives and Death of Moore's Law, 2002, First Monday, volume 7, number 11.

[3] A. Armutat, New Test Sequencing Instruments Lower Cost of Test for Device Manufacturers, 2005, EDN.

[4] J.A. Mielke, Improving Performance in a VXI or PXI Test System Using Distributed DSP, 2004, IEEE AUTOTESTCON Proceedings.