Exhaustive testing with random test waveforms ensures that your electronics will not only pass standard EMC tests, but operate flawlessly in the field.
The automobile is a rough environment for electronics. Not only must systems operate over a wide range of temperatures, supply voltages can also vary. In addition, supplies can also be quite noisy when devices such as the starter motor are operated.
These noise signals have been characterized and documented, and are often used to test how well the electronics cope with them. However, using standard test waveforms often will not identify problems. The reason for this is that using such waveforms offers only one instance of many different possible noise signals that an automotive electronic control unit (ECU) may encounter. In the real world, ECUs will experience many different variations of noise waveforms, and only specific instances may result in an ECU failure.
The solution to this dilemma is to break down standard test waveforms into their component parts and then randomly vary the voltage and time parameters of those components. By running a functional test while applying each of these variations, you have a better chance of finding the exact combination that will cause problems in the field and ensure that your electrical systems are robust and will comply with standards’ requirements. Exhaustive testing with random test waveforms can help you find those unexpected circumstances that testing only with standard test waveforms will not detect.
Breaking Down a Test Waveform
A typical test waveform is shown in Figure 1. This waveform simulates how the battery supply voltage might vary when a driver starts an engine or how the voltage varies in a start/stop environment. As you can see, we’ve assigned variables to the significant voltage values—U0 through U7—and the time values—T0 through T9. These parameters can then be assigned minimum and maximum values as well as a step size.
Generating Random Test Waveforms
Once we have assigned these variables to points on the test waveform, we can randomly assign actual values to them. We can use a simulation model that generates a pseudo-random sequence to assign those values using whatever randomization scheme seems most appropriate (e.g., uniform distribution, 1/x2, etc.). A pseudo-random sequence is sufficiently random for these purposes but is completely repeatable. Repeatability is important since, when a single test waveform in a very long sequence of waveforms causes a failure, you want to be able to reproduce that waveform in order to troubleshoot the problem.
Fail Faster by Testing Earlier
Finding errors in the field or late in the development cycle can be costly. There have been numerous cases recently of such defects found in the field that could have been identified earlier through extensive testing early in development. The point is that errors caused by conditions that are not tested for cannot be found.
This is even more critical now as automotive electronics feature more complex functionality, with multiple system buses and unpredictable noise. This is particularly true where these functions are provided by microcontroller-based designs, giving a significant risk of system failure with low and transient supply voltage conditions.
There are many standards used by OEMs around the world that require vehicle electronics be rigorously tested for power supply-related issues. These include a wide range of ISO standards (e.g., ISO 16750) as well as many OEM-defined standards (e.g., Ford CI210, 220, 230, 260, 270; General Motors GMW3172; Chrysler DC-10615; BMW/Daimler/VW LV124, Jaguar Land Rover CI-265, and many others).
Each OEM’s EMC testing standards have similar aims, but each sets out to build on previously written standards, some of which are based on test methods dating back decades. As a result, each standard varies to a reasonable degree. These traditional standards generally have a limited number of test patterns. What these standards have in common is that they specify conformance tests that focus on aspects of electrical supply testing at d.c. or relatively low frequencies, aiming to address particular use cases such as engine cranking, intermittent connection or relay chatter.
As more components develop possible failure conditions, bigger, more important system issues start to develop, due to the intermittent resetting of controllers, or non-communicating controllers on system-wide communications buses. Consequently, these standards generally set out ways to test both components and systems under controlled conditions in a repeatable manner to verify the likelihood that a system (or sub-system) will perform correctly during everyday use.
By running hundreds or even tens of thousands of test cycles while randomly varying each waveform parameter from minimum to maximum, you perform an exhaustive test of the ECU. Using this technique, the ECU will experience in the lab nearly all of the supply voltage conditions that it will experience in the field. You can run this series of tests overnight or over a weekend or longer while recording both the test waveform parameters and the test results. With this data, you’ll know exactly what combinations of parameters and waveforms contribute to failures.
This test technique is a big advance over traditional techniques for low voltage testing. Randomization allows you to test for a much wider variety of supply voltage conditions than traditional tests, and will enable you to identify problems that traditional techniques cannot. The result of this exhaustive testing will be more robust electronics and fewer field failures.
Hardware and Software Requirements for Low Voltage Testing
To run exhaustive tests using waveforms such as those described above, you need specialized software-controlled hardware that can generate a wide range of electrical waveforms. The input to this system would be a constant voltage from either a power supply or a battery. The system output would comprise voltage waveforms that duplicate a wide range of scenarios experienced by ECUs during cranking, start/stop or similar events. To properly simulate these waveforms, the system should be able to supply 120 A continuously, with surges up to 140 A.
In addition to supplying test waveforms, another beneficial feature would be the ability to record and replay waveforms. With this feature, test engineers could capture waveforms from prototype vehicles, analyze them, and then modify them in order to perform more effective tests. The modified waveforms could then be uploaded to the test system, perhaps as .csv files, to be replayed during testing.
Why Can’t I Just Use a Power Supply?
The question is often asked “Do I really need such a sophisticated test system? Why can’t I use a power supply to do this test?” While you can use computer-controlled power supplies to perform some of these tests, specialized test equipment has the features you need to perform these tests properly.
For example, the test system needs to both source and sink current. General-purpose power supplies normally only source current from the positive pin and do not actively drive the negative pin downwards pulling current into the power supply unit (PSU). Power supplies that can both source and sink output current are very rare and tend to be very expensive. Additionally, PSUs may offer one or the other option, i.e., either sourcing or sinking and sourcing, but not both in the same supply.
Another consideration is the slew rate. Power supplies with controlled slew rates are very hard to find, and it is pretty much impossible to find a PSU with both controlled rise and fall times. Power supplies traditionally offer very slow slew rates in excess of 50 ms on lower cost units and, on more expensive units, this may drop to low numbers of a few ms.
To run exhaustive tests using the waveforms described, the slew rate must be constant and configurable (depending on the test waveform being output). Also, for some tests, you may want the option of allowing the impedance of the equipment under test to define the fall time of the output of the test waveform. Some specialized hardware is capable of slew rates as fast as 3 μs, but you rarely need rise or fall times that fast. Typically, the fastest slew rate is 10 μs, and for many waveforms, a 500 μs slew rate is fast enough.
Transient Response and Output Noise
In addition to poor output slew rate, power supplies may also have poor transient response. If a power supply has poor transient response, there will then be a small dip in the output voltage when the supply tries to output a very fast, high-current pulse. To prevent this from causing erroneous test results, the test system needs a transient load response of at least 500 μs. For some tests, a transient load response of 10 μs is desirable.
Most DC supplies on the market today, especially those capable of delivering the currents required by these low-voltage tests, are switched-mode, or simply switching, power supplies. They are smaller, more efficient, and cost less than linear supplies, but they can be noisier. This noise is undesirable and may affect test results.
For applications that require very low noise output, you’ll need to power the test system with a battery. Batteries are inherently low noise and, if the test system has a linear output stage, the test system will not add any switching noise.
Sophisticated Waveform Generation
Waveform generation is one of the weakest areas for off-the-shelf power supplies. While some offer basic, pre-programmed waveform shapes such as cranking waveforms, they do not offer all of the waveforms now specified by the automotive OEM EMC test standards. In addition to outputting standard test waveforms, users also want to add test waveforms so that they can more accurately simulate the electrical environment that their ECUs will encounter.
Furthermore, using a power supply’s standard waveforms does not allow users to randomize the waveforms test parameters in order to perform exhaustive testing. Nor does it allow users to “window in” on areas of susceptibility as indicated by the standard tests. A test system that allows users to do these things will find susceptible units faster and help them troubleshoot problems faster.
To allow users to add custom waveforms to the test system’s library, the system should allow users to capture data with data acquisition systems or oscilloscopes and then upload the waveforms to the system as .csv files. This will allow users to test ECUs using waveforms actually created by a vehicle and not limit them to the waveforms defined by the standards.
Finally, a desirable feature is the ability to modulate or control the test system output with an external signal. The purpose of this feature is to allow users to add noise or high frequency waveforms to the output in a controlled way. This function is rare for power supplies.
Don’t Forget System Test
The low-voltage test techniques described here are extremely useful at the module level. If you perform low-voltage robustness testing early in the design cycle of an electronic controller, and find that your design is susceptible to noise, you can fix the problem and avoid expensive redesign later. However, they can and should be used throughout the development process, such as during HIL testing.
A key opportunity exists for use at the breadboard, LabCar, or yellow board phase. These test systems are frequently under used or used simply to duplicate HIL functionality. However they represent a significant opportunity to introduce rigorous robustness testing not only of the functional or software based performance but also to identify electrical issues related to ground offsets, sneak circuits, quiescent current issues etc.
Simulating cranking signals will enable you to find system problems that would affect the system’s electrical robustness and stability. For example, noise on a power line might get coupled to a signal line in a wiring harness. By finding this type of problem at the breadboard phase, you can take steps to correct the problem before building prototype vehicles. Ideally the low voltage testing should be combined with testing for the electrical issues mentioned above.
Exhaustive robustness testing really works. This technique allows you to find and fix more problems than traditional test techniques, and in the end, this is a win for you and a win for your customers.
The techniques mentioned here reference automotive electronics testing and compliance. They are also well suited for aerospace or military electronics testing and compliance.
Maurice Snyder is the founder and President of Spes Development Co in Ann Arbor, MI specializing in equipment for EMC testing and real-time hardware-in-the-loop testing. He has a PhD in Electrical Engineering and is the former Chair of the IEEE SE Michigan Section, former membership Director and an IEEE Life Senior member. He can be reached at firstname.lastname@example.org .
Brett Dowen is Managing Director of add2 Co Ltd, based in the UK, a designer and manufacturer of EMC and HIL test equipment. He is a chartered engineer, has an Honours Degree from Nottingham University in electrical and electronic engineering and has worked in automotive embedded design and test since 1977. He can be reached at email@example.com .
Simon Clarke is Technical Director of add2 Co Ltd, based in the UK. He studied Electrical & Electronic engineering at Aston University and has been designing and testing automotive embedded systems for over 35 years. He can be reached at firstname.lastname@example.org.