|
Basics about data acquisition in crash testing
General
The data acquisition system is an extremely important device in crash testing. Basically, there are two type of information that tells an
engineer what happened during the crash. One is the high speed film that allows a visual inspection of the crash test itself from different angles. The second is the data acquisition system. Without
the data that is captured by such a system the whole test is useless. Therefore, a data acquisition system has to be extremely reliable. However, there are more things to such a system. If you are
interested in things like that or if you want to know how such a system works you should read this page.
History
When the automobile engineers started with crash testing, the data acquisition systems were big and analog. Basically these systems
worked like the good old tape recorders. The signal from a transducer was converted and amplified and then stored in analog format. Analog means that the signal level that was stored on the tape was
analog to the real signal that was recognized by the transducer. After the test the analog signals could be replayed and the analog signals were viewed with e.g. an oscilloscope or a strip chart
recorder. In the seventies the first digital systems were introduced. Digital means that the signal from the transducers was converted in a digital representation of the actuel signal level. This
made the analysis of a signal much easier since most of the computers worked in a digital way (if you believe it or not but there have been analog computers in the past!).
Digitizing an anolog signal was in the beginning pretty difficult. First of all it was necessary to have good amplifier that applies a
certain gain to the real signal because the real signals are very small (sometimes only a few millivolts). Then it was necessary to digitize the amplified signal with an A/D-converter. After the
A/D-converter the digital signal had to be stored in memory – or on e.g. a tape recorder (somehow it is strange to store digital data on an anlog tape but this is exactly how it was done). The
whole system used to be very big. A unit that had the size of a VCR was able to acquire data from one channel. It is obvious that these systems could not be installed in a crash vehicle. All the
transducer data was conducted in analog way from the sensor to the off-board data acquisition system. Sometimes over a distance of more than 200 meters!
This method had a lot of disadvantages:
- The analog signal is very small and feeding this fragile signal through a long cable does not make the signal better.
- The cables were very thick and heavy since each transducer needed its own set of wires (usually two for the signal, two for the bridge supply, and
one four ground)
- It was not possible to run a test without a cable.
In the seventies a new technology was introduced. For the first time it was possible to separate the amplifier and digitizing sections
from the recording section. The digitized data was transferred from the A/D converter via a so called PCM data stream (pulse code modulation). It would be too much at this point to explain PCM in
detail. However, it enabled the engineers to put the whole digitizing section of the data acquisition system on-board a vehicle. The data was transferred during the crash via PCM (cable or radio
link) to the off-board storage. The on-board systems got the name Encoders (since they encoded the measurement signal to PCM) and the off-board systems had the name Decoders (since they decoded the
PCM into something useful). One of the first on-board data acquisition systems on PCM technology was the K1222 from Kayser-Threde, Munich Germany (first installed at the BAST in Germany in 1978).
As time went by the technology changed dramatically. Soon it was possible to even have the storage of test data on-board and by the end
of the 80’s the PCM systems were replaced by digital recorders with everything on-board. With systems like that it was possible to run complete crash tests without a link to off-board devices.
Today, the on-board systems are small, compact, ruggedized and offer an extremely large amount of memory. There are different vendors of
such systems out there, but there are real differences!
Is ruggedized and ruggedized the same?
Well, an interesting question. As discussed before, the on-board data acquisition system have to work reliable in a very harsh
environment. During the impact the systems are exposed to accelerations that can reach up to 100 g. What does that mean? One g is equal to the gravity (acceleration) of our planet earth. Each body
with a mass of e.g. 1 kg is exposed to a force of 1 g times 1 kg. One g is equal to 9.81 m/s2. The force in Newton (N) is equal to 9.81 N. Lets say your body mass is 75 kg. The means you
are attracted by the planet earth with a force of 735,75 N. If you are accelerated with 100 g rather than 1 g the force will be 73575 N. This does not tell you anything. Well this is the
physical explanation. To say it more popular you will feel like weighing 7500 kg or 7.5 tons (I know this is physically spoken not correct but it helps to understand what is going on).
A typical crash test system today has a weight (again, this is not correct – a mass would be the right term) of 3.5 kg.
Accelerating this with 100 g means that the system appears to weigh 350 kg. In order to withstand this high loads it is necessary to make such a system ruggedized.
However, ruggedized is not ruggedized! A lot of manufacturers in the past thought: We have a very good data acquisition system! If
somebody wants to use it in crash tests – no big deal! We will put it into a ruggedized housing and it will work! Well, it is not as simple as that. The today’s market show that all these
companies who had this idea failed (no matter how famous their name was). Only smaller companies like Kayser-Threde are still in this market because they designed their systems from the beginning as
crash test systems – and it makes a differnce!
Complying with Regulations
The data acquisition systems that are used in crash tests have to comply with certain regulations such as SAE J211 or ISO 6487.
Basics about digital data acquisition
There are some steps necessary to convert an analog signal into a digital value. All these steps are usually integrated into a data acquisition system.
These steps are:
- Amplification of the input signal to an optimum range of the analog to digital converter.
- Filtering of the input signal to avoid signal disturbances by noise or aliasing.
- Analog to digital conversion.
- Storage of the digital data.
Amplification of the input signal
Usually, electrical signals such as
voltages cover a wide range. Today it is possible with certain devices to measure voltages down to the Nanovolt range while other devices are able to handle kilo Volts or Mega Volts. It is obvious
that you need different devices to cover the whole range. Since analog to digital converters (ADCs) usually work in a typical range from -10 V to +10 V or -5 V to +5 V. Of course, there are also ADCs
out there that work in different ranges but we want to keep it simple. In addition, there are not only different ranges of ADCs available but also different principles (e.g. dual slope converters,
successive approximation converters, or sigma delta converters). We will talk about the A/D function later.
The job of the amplifier is to apply a certain gain on the input signal to have at the end of the amplification process a signal that
utilizes the input range of the ADC best. For example, you want to measure a signal of 1 mV (0.001 V). If you want to digitize this signal with a 5 V ADC you can apply a gain of 5000 to the signal.
The output of the amplifier will be 5 V in this case. For a dynamic signal (e.g. those acquired in crash tests) it is important that the gain is selected in a way that the largest expected signal
does not exceed the input range of the ADC!.
Filtering
As discussed before there are two reasons for applying a filter
to the signal. The first reason is simple to understand: A filter is able to remove high frequency noise from the input signal. This noise is sometimes picked up via the lines from the outside world
or just part of the measurement signal. A noise filter is fine for static measurements (e.g. a DC measurement with a digital multimeter). For dynamic measurements a noise filter is not so easy to
create since you basically do not know whether the high frequency part is noise or part of the actual signal. Crash test data acquisition is definitely a dynamic application. Dynamic means that the
signal levels are changing within a very short amount of time. Therefore, usually everything is measured from dynamic signals (including the noise) and the noise is then removed with data analysis
tools. However, there is one physical effect you have to take care of when you digitize dynamic signals. This effect is called the Aliasing Effect.
What does that mean? Well, you can find a
lot of explanation about this effect. The one I like most is the story about the old western movies. I am sure you remember watching this old western movies (and also new ones) where people where
travelling with horse carriages to the wild west. Sometimes you could see that the wheels started to move in the forwared direction when the carriage started to move. But once of a sudden you could
realize that the wheels came to a standstill or even turned backwards despite the fact that the carriage still went forward. This is because of the aliasing effect! What excatly happens? The film
camera does not take a contineous stream of pictures. It takes only 24 pictures per second. This is fast enough since our eyes will not notice individual pictures anymore once the rate goes above 15
pictures. What we see is a contineous movement. Think back to the wheels of the carriage. The camera takes one picture of the wheel then it waits a little bit and takes the next one. As long as the
next picture shows that the spoke has moved a little bit we can still see a forward motion.
But what happens if the camera and the wheel is synchronised in a way such as the next spoke is on the
next picture at exactly the same location as the last one was? Well, we get the impression that the wheel has come to a standstill because we will always see the same picture. When the carriage
accelerates we might even see the wheel turning backward and so on.
The question is: Has this something to do with data acquisition? The answer is yes because we will see a similar thing when we
digitize data. ADCs take sample from the analog signal at a certain sampling rate (remember the camera took a picture at a certain rate). As long as the sampling rate is much higher than the signal
frequency everthing is fine. However, when the signal frequency comes close or even exceeds the sampling rate we might see things that are not real (like a standstill of the wheel while the carriage
is still moving). In crash testing we are only interested in signals up to 1000 Hz. Therefore, the typical sampling rate is 10 kHz or above and we can be sure that we do not run into aliasing
effects. However, there might be noise and other signals that are of higher frequency. To make sure that these signals have no aliasing influence on the measurement signal we implement Anit-Aliasing
filters!
There are three parameters that typically define filter:
1. The cut-off frequency at -3 dB
2. The type of filter (Bessel, Butterworth, Chebyshev etc.)
3. The number of
poles
Cut-off Frequency
The cut-off frequency is the frequency where signals are already damped to an amplitude to -3 dB relatively to the measurement signal. This is an important
parameter since it is not possible to create filters that are open for signals of lets say 1000 Hz and block all signals above that. A filter introduces a damping characterstic over a certain
frequency range. Filters in crash test data acquisition are typical low-pass filters since they allow all signals below a certain frequency to pass.
Filter type
Filters are built
simply spoken with resistors and capacitors. Unfortunately, these components are not only blocking unwanted frequencies they also have an influence on the part of the signal you are interested in.
Some filters introduce a certain time delay to the signal and some others slightly change the amplitude of the signal. Therefore, it is important to select a filter that gives you the best compromise
for the application. For crash testing in most of the cases a so called Butterworth filter is used. If you are more interested in Butterworth filters we recommend this web site:
http://sepwww.stanford.edu/sep/prof/pvi/spec/paper_html/node14.html
to be continued!
|
