| This paper presents the results of a study into the use of distributed digital signal processing (DSP) at the instrument level in a VXI and PXI based test system and the effects on test time. One of the limiting factors in testing mixed signal or analog devices using standard bus based instruments is the transfer speed from the instrument to the controlling computer of large amounts of waveform data. This is important as these types of tests use non-deterministic, quantized signals that must be mathematically processed to extract test information. This processing can either be done at the instrument or at the central controller. If the processing is done at the instrument then only the results are transferred to the controller. If the controller does the processing then the raw data must be transferred to and from the instrument. Using two instruments, one in VXI and one in PXI, this paper measures the effects of typical tests contrasting the measurements as done in the central processor as opposed to a distributed DSP processor in each instrument. For each acquisition instrument, tests were implemented by capturing the data and moving it to the controlling computer where it was processed to extract test results, or by using the instruments on board DSP so only the final test results were se nt to the controlling computer. The study results show that a significant improvement in test time can be made by selecting “smart” instruments for the test system when using PXI or VXI based instruments. |
There is a drive in the test industry to move to a standard open automated test equipment (ATE) architecture to reduce the cost of test[1]. The consumers of these test systems are challenging the manufacturers to move to an open standard for test systems[1][2] and using their economic strength as driving force to a common architecture[3]. Although there are currently several open architectures being proposed, the two most common currently in use today as a general purpose ATE system are based on VXI[4] or PXI test busses.
The first generation of VXI and PXI instruments kept the cost of hardware down and the instruments simple by putting a minimum amount of intelligence within the instruments. However, data throughput became an issue in these types of systems[5]. The ATE system must stimulate the device under test (DUT), capture and process the results[6]. If the instrument is not intelligent enough to process that data in place then the entire data set must be transferred to a host computer for processing.
The purpose of this paper is to present the results of instrument experiments using PXI and VXI instruments to study the limitations and solutions of using a standard test bus to implement a mixed signal test and measurement system. There are several challenges to using the standard VXI or PXI test busses. This paper explores and quantifies the use of distributed DSP to enhance test time and improve test performance by capturing a waveform and performing various measurements on these waveforms.
Standard Test Bus Challenges
Distributed DSP
A standard test system architecture consists of device stimulus and capture instruments and a system controller tied together with the test bus. Normally the stimulus and capture instruments are divided into analog and digital because of the very different requirements of each. The analog instruments may further be divided to separate instruments based on speed and resolution. For this simple discussion we shall omit the power distribution. A basic block diagram of this architecture is shown in figure 1.
Figure 1. Test System Architecture
One method of implementing this architecture is to use very simple instruments with a powerful system controller providing the intelligence for the system and providing tight system control. The main advantage to this type of system is that the software provides most of the instrument functionality while the local instrument hardware provides only rudimentary stimulus and capture capabilities. In effect, in this model, the system controller software is most of the instrument functionality.
In this central processing model, all of the data is transferred to or from the system controller for processing. Digital stimulus, like a memory test pattern, must be created in the system controller and downloaded to an instrument. Digital capture data must be sent to the system controller for comparison to deterministic data. Waveforms must be generated in the system controller and downloaded to the analog stimulus instrument (normally an arbitrary waveform generator). Analog captured data must be analyzed in the system controller for non-deterministic testing. With large amounts of data being transferred between the controller and instruments in this central processing model, the test time is often limited by the bus data throughput.
Another method of implementing this test system architecture, the distributed DSP model provides intelligence within the instruments. This approach puts a much lower demand on the bus speed and central system controller and only requires simple system setup to perform complex tests. In this model the instruments become similar to stand alone, traditional instruments.
In this distributed DSP model, all of the data is processed in place and only the setup and results are transferred to or from the system controller for display and logging. Digital stimulus, like a memory test pattern, is created in the instrument with only simple setup and handshaking commands. Digital capture data is compared directly in the instrument and only a pass/fail result is transferred to the system controller. Waveforms would be generated in place in the analog stimulus instrument using on board DSP. Analog non-deterministic captured data could be analyzed in the analog capture instrument DSP.
An Analog Capture Instrument
It was theorized that the distributed DSP method of implementing this ATE architecture would result in significantly faster test times and be much simpler to implement and use. The simplicity would result from fewer commands being needed to control the instrument and less data moving back and fourth between the instrument and the controller. To test this theory, two standard test bus architectures, VXI and PXI, were chosen.
For the comparison between these two architectures a similar analog capture instrument (Digitizer) was chosen from each. For the VXI capture instrument, the ZT1428VXI was chosen to provide 8-bit analog capture resolution at 1 GS/s. For the PXI capture instrument the ZT430PXI was chosen to provide 12-bits of analog capture resolution at 200 MS/s.
The tests for each of these instruments were implemented in one of two ways:
- The instrument was used to capture the data and move it to the controlling computer where it was processed to extract test results.
- The instrument’s on board DSP was used to make the measurement and only the final test results were sent to the controlling computer.
Experimental Results
VXI Capture Instrument
The VXI test bus provides a common platform for instrumentation. Although VXI is one of the oldest of the ATE test busses, with a broad range of available high performance test instruments, it still is a bus of choice as a basis for ATE systems. A ZT1428VXI digitizing oscilloscope was chosen for the VXI capture instrument. This instrument samples at 8 bits of vertical resolution at up to 1 GS/s. For all of these experiments the maximum 1 GS/s sample rate was used. This instrument is a drop in replacement for the popular HP E1428.
The types of tests chosen for this benchmark were selected to measure the affects of bus on test times without getting into complex or proprietary test methodologies. As such the two tests chosen were a simple peak-to-peak measurement and a RMS measurement on a waveform. The number of samples for the waveform measurements varied from 1024 to 8192. Also, the tests were made by averaging the captured waveforms 1, 16 or 64 times to increase accuracy. A PC was used as the system controller and the communication link from the PC to the VXI chassis was a National Instruments (NI) MXI-2. LabVIEW was used as the controlling language.
The raw test results for the six tests are shown in Table 1. The central measurements were made by capturing the data and moving it to central system controller in a “dumb” mode and processing the waveforms using the central controller to analyze and provide the test results. The distributed measurements were made using the instruments on-board DSP processor to analyze the data in a distributed DSP or “smart” mode and only transferring the results to the system controller.
| Test No. | Central Measurement |
Distributed Measurement |
| 1 | 30.5 ms | 7.6 ms |
| 2 | 2.8 ms | 1.8 ms |
| 3 | 2.2 ms | 1.6 ms |
| 4 | 1.9 ms | 1.5 ms |
| 5 | 119.2 ms | 25.7 ms |
| 6 | 1.8 ms | 1.5 ms |
Table 1. VXI Test Results
The VXI test results show that as expected when transferring small amounts of data there is only a small amount of difference in the test times. However, when large waveforms are transferred or, in the case of averaging, when multiple waveforms are transferred there is a much more significant difference in test time. The percent improvement for each test is shown in Graph 1.
Graph 1. VXI Test Results
For a VXI based system, the test results showed at least a 20% improvement in overall test time and greater then a 75% improvement for complex test cases. The typical test result improvement among the 6 test cases averaged a 42% improvement in test time.
PXI Capture Instrument
The PXI test bus is one of the newer platforms for ATE test system implementation. Although PXI still does not see the breath or performance of instruments as the VXI test bus, with new instruments being added at a rapid pace, it is fast becoming a contender as an ATE test bus. A ZT430PXI digitizing oscilloscope was chosen for the PXI capture instrument. This instrument samples at 12 bits of vertical resolution at up to a 200 MS/s acquisition rate. For all tests implemented the maximum 200MS/s sample rate was used.
The same simple peak-to-peak and RMS measurements chosen for the VXI tests were also chosen for the PXI measurements. However, because of the greater capture depth of this instrument the number of samples for the measurement varied from 1024 to 1M. Averaging was not used on these measurements. A PC was used as the system controller and the communication link from the PC to the VXI chassis was a National Instruments (NI) MXI-3. NI CVI based “C” programming was used as the controlling language.
The raw test results for the six PXI tests are shown in Table 2. The central and distributed DSP methods were done in the same manner as the VXI test time measurements.
| Test No. | Central Measurement |
Distributed Measurement |
| 1 | 31 ms | 24 ms |
| 2 | 842 ms | 83 ms |
| 3 | 8544 ms | 617 ms |
| 4 | 37 ms | 24 ms |
| 5 | 839 ms | 83 ms |
| 6 | 8556 ms | 611 ms |
Table 2. PXI Test Results
Like the VXI results, the PXI test results also show that when transferring small amounts of data there is a small amount of difference in the test times. And like the VXI, when a large number of waveform samples are used there is a much more significant difference in test time. The raw test times show that the improvements with complex test implementations is even more significant with PXI then with VXI. The percent improvement for each test is shown in Graph 2.
Graph 2. PXI Test Results
For a PXI based system, the test results were surprisingly similar to the VXI test results. Like VXI, each test showed at least a 20% improvement in overall test time. In four of the six tests there was greater then a 90% test time improvement. The typical test result improvement among the 6 PXI test cases averaged a 71% improvement in test time.
Summary
In a PXI or VXI based test system a typical 2X improvement can be achieved if the system is implemented using instrumentation capable of distributed signal processing. Depending on the tests executed, up to a 10X improvement can be realized. Even for small amounts of data, there was at least a 20% improvement in test times. Over the 12 tests using the two tests busses, an average improvement of better then 56% in test time was achieved.
With the competition for a standard ATE test bus showing no clear winners at this time, it is important that these results are consistent over the two forerunning test buses and be extended to any other test system architecture selected.
Further Work
The author would like to complete further work to verify that these results are also applicable to force instruments in VXI and PXI implementations. The choice of instruments would include an arbitrary waveform generator with on board digital signal processing where waveforms are either created in place or downloaded from the system controller.
References
[1] B.G. West, Open ATE architecture: key challenges, 2002, IEEE International Test Conference Proceedings.
[2] P.C. Jackson, EAF challenge to ATE, 2000, IEEE AUTOTESTCON Proceedings.
[3] S.M. Perez, The consequences of an open ATE architecture, 2002, IEEE International Test Conference Proceedings.
[4] J. Hou and W. Feng, A VXI-based automatic test solution for avionics: issues and implementation, 2000, IEEE AUTOTESTCON Proceedings.
[5] B. Wood, A System-Component Approach To Functional Test Systems, April 2004, Evaluation Engineering.
[6] J. Mielke, Frequency Domain Testing of ADCs, Spring 1996, IEEE Design & Test Of Computers.
