This is the second post in a multi-part series on how to make successful measurements, i.e., how to get “good data”. In the first post we made the case that good data is about time, money, performance, results, and reputation. We also talked about why we care about good data at IC2. In this post we will discuss how to make good measurements of a fluctuating phenomenon such as shear stress, pressure, vibration, or sound.
Fluctuating Data Is Dynamic Data
Note that we are using the term “fluctuating data” which is essentially equivalent to “dynamic data”. These two terms mean that the quantity being measured is changing over time. One of the reasons we use “fluctuating data” is to avoid confusion that can happen when we use the term “dynamic pressure data” because “dynamic pressure” is also defined as the kinetic energy per unit volume of a fluid. If we refer to time varying pressure data as “fluctuating pressure data,” we avoid the ambiguity associated with “dynamic pressure data”.
However, be aware that “dynamic data”, “dynamic data systems”, and “dynamic pressure sensors” are commonly used terms in the scientific data acquisition community, and all of these terms are referring to something that is varying with time.
Scoping The Discussion
For the purposes of this discussion, we will not go into all the details related to selecting the right sensor, installing the sensor properly, and choosing the right data acquisition system components. So, our starting point is that you have the right sensor, it is installed properly, your data acquisition system is capable of digitizing at a rate that is appropriate for the measured frequency range of interest, has appropriate anti-aliasing filters, and has ample dynamic range. We also will not discuss how long the data acquisition event needs to be or how synchronized the acquisition needs to be across channels. These are requirements of your system that are dependent upon how the data will be processed after it is acquired, or while it is being acquired.
Of course, you should verify that the system is in fact giving you the level of synchronicity specified, but other than that, as long as it is able to acquire the amount of data required for an acquisition event, these factors do not play into the criteria for making sure that the acquired data is good data.
What we will focus on, once you have all of the above capabilities in place, is how you make sure that you are using them properly to collect good data and that you don’t waste precious testing resources (time and money) because you didn’t do it right the first time.
Specifically, what we will focus on is:
- Making sure that the sensors and data acquisition system are working as expected
- Optimizing the dynamic range of your measurement
- Selecting the appropriate input coupling
- Avoiding clipping (overrange)
- Avoiding underrange
Dynamic range is the ratio between the largest and smallest values that a certain quantity can assume, and is often expressed as the decibel difference between those values.
These are the attributes over which the data acquisition system operator has control once everything is set up and the test is underway.
Sanity Check: Are My Sensors Working?
The first point of focus on how to acquire good data, is to know that our sensors and supporting instrumentation are working properly. We assume, at this stage, that all of the sensors have been calibrated and we have the calibration information properly associated with the data acquisition channels to which the sensors are connected so that the voltages that we measure with the data acquisition system can be properly converted into engineering units. We also assume that steps have been taken to ensure that the correct sensors have been connected to the correct data acquisition system inputs.
This means that we just need a way to monitor the health of the instrumentation on an ongoing basis, ideally acquisition event to acquisition event, so we know that the measurement system is continuing to function as it did when it was initially set up.
A great way to do this is to have real-time displays that show what the sensors are “seeing”. For dynamic data, we are interested in acquiring time-series data and typically we are interested in the frequency content of that data. So, an oscilloscope-like time-series display along with a spectrum-analyzer-like frequency-domain display typically provide ample information for viewing sensor performance. Historically, these displays would have been provided by using rack-and-stack hardware, that is, actual oscilloscopes and spectrum analyzers, but with today’s data acquisition hardware and software, these displays are more typically provided on your computer display.
A seasoned data acquisition system operator has the experience to look at time and frequency domain signals and spot anomalies.
For example, a time series display during an ambient state can show whether the sensor is noisy. For a very low noise sensor, it can sometimes be difficult during an ambient condition to assess whether or not the sensor is “dead”. However, during a non-ambient condition, it should be quite obvious whether the sensor is dead or alive. Time-series displays can also be very useful for sensors that can be calibrated in place. For example, when a microphone is placed in a sound calibrator that produces a known frequency and sound pressure level, it is very useful to be able to view a stable and undistorted sine wave on a time-series display so you know that your sensor is seeing a legitimate signal before you perform the calibration (i.e., acquire the calibration data).
Frequency displays add another view of sensor health. For example, the power spectrum of a pressure sensor or wall shear stress sensor in the boundary layer of a wind tunnel typically has a certain shape to it. Once an operator becomes familiar with this characteristic, they can monitor the sensors once the wind tunnel starts running to see whether everything looks as expected. Once you are convinced that your sensor systems are operating properly, your focus now needs to be on acquiring quality data from those sensor systems.
Use AC Input Coupling: The Prerequisite for Optimizing The Dynamic Range of Your Measurement
For high-quality fluctuating measurements, it is most advantageous to use AC input coupling. AC coupling removes any DC offset that may exist at the input to the analog-to-digital converter (ADC). This optimizes the available dynamic range of your acquisition hardware. Consider the example of a fluctuating pressure sensor where the cavity behind the diaphragm is sealed. Any difference in static pressure, from whatever the condition was under which the cavity was sealed, will be seen as a DC offset by your measurement system. If the DC offset shows up as +2 V and your ADC operates over a ±10 V input range, then this offset will effectively reduce the upper range for making the fluctuating measurement to 8 V. By using AC coupling, the full 10 V range of your ADC will be available.
Setting Ranges: Optimizing The Dynamic Range of Your Measurement
The ADCs of your data acquisition system have a voltage range over which the digitization is performed. If you don’t use the full voltage range, you sacrifice the dynamic range of your measurement. We just discussed the importance of using AC coupling so that the full voltage range of the ADC is available. Setting ranges properly is how you take full advantage of the available range. If your input signal level exceeds the voltage range (called an overload or overrange condition), then your signal gets clipped. Most ADCs have the ability to set the range of the ADC. What really happens when you set the range is that a gain or attenuation is applied to the input signal prior to when it is digitized. For example, if your ADC operates over a ±10 V range and you specify a range of 1 V, a 10X amplification will be applied to the input signal prior to when it is digitized. Note that your system will bookkeep this range setting and account for it when it produces an output voltage so that the output voltage matches the pre-input-gain voltage that was input to the system.
Some systems have autorange capabilities where the ADC range can be set automatically. Since this capability has some preset methods of operation, it can have some drawbacks. For example, it may take an inordinate amount of time for the autorange function to perform its duty. Or, the autorange function may be overly conservative, providing data that is not using the full voltage range of the ADC.
In many applications, the best procedure is for the system operator to set ranges manually based upon what they observe that the sensors are seeing. Once again, real-time displays can be very useful to assist in this process.
To support manual ranging, three additional displays are recommended.
The first display is a real-time overload/overrange indicator. This is just a simple Boolean indicator that tells you whether or not the previous window of data (i.e., block of data extracted from the ADC) contained any overloads/overranges. If this indicator is on steady or flashing on and off, the range setting needs to be increased one step at a time until it stays off.
The second display is an RMS-voltmeter-like display that provides the overall level of the signal. The overall level of the signal combined with knowledge of what that level implies for a peak signal level can be a very useful and effective way to determine an optimal range setting for the measurement that is about to be made. For example, if the operator knows that the peak level of a sensor is typically 10X the RMS level of the sensor when the wind tunnel is operating at Mach 0.2, then they will set the range to be equal to 10 times the RMS level rounded up to the closest available range setting. The RMS multiplier is initially determined by selecting a range and seeing If it produces an overload. If it does, then the range should be increased one step at a time until no overloads are produced. If the selected range does not produce an overload, then the range should be decreased one step at a time until an overload is registered, or until the lowest range available has been selected. If a range is reached where overloads are registered then the range should be increased by one step.
After just a few acquisition events, an operator can get dialed in to setting optimal ranges effectively and efficiently.
A third display, a Probability Density Function (PDF), can provide conclusive evidence for overload detection. When overloads are occurring, there will be hard limits at the left and right edges of the PDF display where you would typically see tails tapering down to zero.
So, How Do You Know You Are Getting Good Data?
For dynamic data acquisition, we can now answer the question that is the subject of this series. You know you are getting good data because:
- you know the health of your sensors since you know what working sensors look like under various operational conditions and you can see that they look as expected in your real-time time-series and spectral displays; and
- you have set the input ranges of your ADCs optimally such that you are using as much of the ADC input voltage range as possible without registering any overloads.
For the next post in this series, we will discuss the basics of getting good data when acquiring static data.