Digital Storage Oscilloscopes Get Physical An Automotive Electronics Test Application

Breed Technologies is a full-range supplier of automotive safety systems with more than 11,000 employees worldwide. The company’s products, used on over 400 models from 51 car companies, include safety restraints, air-bag systems, automotive electronics, sensors, steering wheels, and interiors.

Engineering research, development, and validation are performed in several laboratories located at the company’s corporate headquarters in Lakeland, FL. One of these facilities, the Prototype Inflator Test Laboratory, gathers data from functional and chemical testing performed on air-bag inflator modules. Analysis of the data is used to verify and validate the performance characteristics of the modules.

Computers and data acquisition equipment are connected to several sensors in and around the inflators to measure, filter, analyze, profile, and store the data acquired during the tests. The reliability and accuracy of the test systems are critical because of obvious safety issues and because each inflator can be tested only once.

The laboratory experienced difficulties with previous PC-based data acquisition techniques. A multiplexed digitizer card introduced up to 100-µs time skew across channels, and the sample rate was too low to characterize the newer and faster R&D designs. Further concerns about accuracy, aliasing, calibration, and reliability led to a decision to replace the PC-based system.

Improved sample rates and longer record lengths were paramount considerations as the increasingly sophisticated designs placed additional demands on test. The fast rise time of each inflator’s pressure had to precisely meet the desired profile.

In an actual collision, an inflation either too fast or too slow means the air bag may not deliver the maximum possible protection. Variables such as propellant burn rate, bag inflation rate, peak pressure, and the gas discharge rate must be carefully controlled and matched to the vehicle in which the design will be used.

Even more difficult, the inflation profile must be held constant over manufacturing tolerances, vehicle age, and an extreme range of environmental conditions from Siberia to the Sahara. In addition to exceptional accuracy, the evaluation criteria for new instrumentation included ease of hardware maintenance and calibration, configuration flexibility, and expandability.

Most oscilloscopes and transient recorders are designed for the highest possible speed rather than the highest precision. For example, most digital oscilloscopes use an 8-bit analog-to-digital converter (ADC) which can resolve one part in 2⁸ (1/256 or 0.4% of the full scale range.) This quantization error means that any signal variation of less than 0.4% may not be detectable, and a peak value cannot be read more closely than 0.4%.

Other uncertainties include amplifier gain, offset, linearity, temperature drift, and noise. A typical digital scope only specifies DC gain that is 1% at best and sometimes as much as 3%, leaving several other significant error sources unspecified.

While acceptable for most electronic signals, such a wide margin of error is inadequate for physical measurements that demand high precision, repeatability, and statistical confidence. To ensure that the new instrument would not be a limiting factor, the lab specified a DSO containing 12-bit digitizers for each channel. By resolving one part in 2¹², or 4,096, the quantization error could be reduced 16 times to 0.024%.

A rate of up to 1 MS/s on each channel captures great detail during the milliseconds of actuation, with accurate time correlation across channels. Table 1 shows a comparison of digitizer resolutions. The greater detail provided by the 12-bit digitizer fills the requirement for high accuracy and repeatability.

Another consideration in physical measurements is the input amplifier. While electronic measurements demand the highest bandwidth possible, virtually all transducer signals are confined to less than 1 MHz.

Excessive bandwidth actually compromises transducer accuracy by passing additional noise. Since most transducer signals are in the millivolt region, they are very sensitive to EMI and RFI pickup. High-frequency interference radiated into the signal cables passes to the digitizer where aliasing reflects its energy into apparent lower frequencies. Unfortunately, it is common to find that the noise and aliasing components are much larger than the desired signal.

Breed’s new DSO is equipped with controlled-bandwidth, low-noise input amplifiers that preserve the resolution and dynamic range of the digitizer. It features less than 0.2% maximum static error (including all gain, offset, linearity and temperature compensation errors as specified by IEEE 1057) which is comparable to the best transducers available.

As further protection from aliasing, Breed uses the high digitizing rate and long memory for heavy oversampling not possible in the previous system. The DSO’s differential inputs are used to avoid signal contamination from ground loops and to make strain gage and current shunt measurements.

With transducers measuring physical quantities such as pressure, it is convenient to view results directly in engineering units such as psi. This eliminates converting data points manually and remembering conversion factors.

The instrument chosen by Breed provides individual engineering units for each channel, allowing independent channel calibration for the transducer being used. Other advantages include:

Ease of calibration—the instrument is NIST traceable and calibrated by Breed’s in-house facility.

Nonvolatile storage—data from a destructive test is not lost in the event of a power failure.

PC independence—the system is not “married” to one particular processor or operating system.

Better confidence and reliability—Breed’s control software invokes local and system self-test during each test.

With the more powerful sampling rates, memory capabilities, and input configurations, the new data acquisition system accommodates a greater variation in signal transducers and test configurations than before. Tests once limited to 10 kS/s for 100 ms now can be set to faster sampling rates over longer periods of time. On the other hand, tests requiring low sampling rates over hours can be implemented quickly using self-contained internal timing.

About the Authors

Mike Hoyer, a product applications engineer, has been employed by Nicolet Instrument Technologies for more than six years. He has held various positions such as applications writer and applications specialist. Mr. Hoyer received a B.S.E.E. degree from New York Institute of Technology. Nicolet Instrument Technologies, P.O. Box 44451, Madison, WI 53744, (608) 276-5600.

James W. Griggs III is the manager of test-systems development at Breed Technologies. He received a Ph.D. degree in information science from the Center for Computer and Information Sciences at Nova University. Breed Technologies, P.O. Box 33050, Lakeland, FL 33807-3050, (941) 668-6684.

Sidebar

Enhanced Resolution

Several digital oscilloscopes offer a mode called hi-res or enhanced resolution that claims to increase 8-bit ADC resolution to as much as 12 to 13 bits. Vendors have implemented different algorithms; but in each case, the technique is based on digital filtering of high-speed ADC samples.

While appearing to be somewhat magical, the mathematics is relatively simple. Numerous consecutive readings are averaged or otherwise filtered to produce a higher-resolution result. For example, the average of four samples reading 5.0 V, 5.2 V, 5.2 V and 5.1 V is 5.125 V. The average value has greater numeric resolution than any of the samples. While this technique is mathematically valid, several limitations may degrade your actual results.

First, to achieve greater resolution, the samples included in the average must vary in value. As a result, one of these conditions must be true:

A dynamic signal across the number of points being averaged.

A noisy signal.

A noisy amplifier.

Ironically, this means that if you have acquired a clean signal with a quiet digitizer, resolution enhancement will be ineffective.

Second, the improvement is proportional to the square root of the number of samples. To enhance an 8-bit signal to the equivalent of 12 bits (a 16 times improvement), you must sample at 256 times the normal speed. For example, a 10-kHz bandwidth strain-gauge signal might be conservatively sampled at 100 kS/s to retain excellent signal detail. An 8-bit digitizer must oversample at 25.6 MS/s, using 256 times the memory, to achieve a similar result.

Third, the averaging of many sample points means that the bandwidth reduces drastically: Fast peaks and rise times will be distorted. A sharp peak will be averaged with many other samples and its amplitude reduced, resulting in lowered accuracy. The amount of error is difficult to estimate unless the scope allows you to view both the raw and processed data. Not all do.

Finally, although the apparent resolution is enhanced, the absolute accuracy cannot be improved. If the scope specifies a typical accuracy of 2%, a reading of 5.125 V actually indicates an input anywhere between 5.02 V and 5.23 V.

The pitfalls of enhanced-resolution techniques indicate that, for critical measurements requiring a high level of accuracy, repeatability, and confidence, there still is no substitute for a high-resolution digitizer. Modern systems designed for transducer measurements can deliver better than 0.2% accuracy and a 70-dB dynamic range. Some newer oscilloscopes also add internal transducer signal conditioning.

Table 1.