Sometimes a COLA (Constant Output Level Amplitude) signal is not available, or was not recorded, during a Sine Data Reduction test. This article describes how to create the COLA signal from a measurement or control channel that was acquired during the sine sweep.
Using an accelerometer measurement from a control channel for a COLA signal can be a challenge. Unlike a COLA signal, which is a clean sinusoidal wave (blue in Figure 1), an accelerometer signal will have noise (green in Figure 1):
Figure 1: Accelerometer control signal (green) overlaid with ideal sine wave COLA signal (blue)
This noise makes it difficult to determine the exact frequency of the sine wave. This article explains how to overcome the noise and generate a COLA signal from an accelerometer measurement in a sine reduction test.
Agenda 1. Overview of Challenge 2. Getting Started 3. Time Signal Calculator 4. Conclusion
1. Overview of Challenge
Offline Sine Reduction requires knowing the instantaneous sweep frequency in order to process recorded data. This is typically done by recording a COLA (Constant Output Level Adapter) which outputs the Sine frequency with a noise-free constant amplitude value. This is important as when amplitudes are low or noisy on a control channel, it may be difficult to resolve the exact frequency from time domain data. However, if no COLA signal is available, it is still possible to use offline sine data reduction by using one of the existing channels as a COLA signal.
In most cases, however, it is necessary to filter the control channel to remove high frequency noise and harmonic distortion to create a useable COLA signal. The effect of this is clear in Figure 2, which shows the unfiltered control channel, even with upsampling applied.
Figure 2: Processed COLA signal (red) based on the unfiltered control channel (green).
Note: In this example, the control channel is used to demonstrate the ability in Simcenter Testlab to use non-COLA data to process into a frequency vs. time signal, but any other trace that is reasonably unaffected by resonances, anti-resonances or distortions can be used.
2. Getting Started
To use the control channel for COLA and also process it into a spectrum, a duplicate of the control channel must be made. This is accomplished in "Time Data Selection" workbook by checking the control channel and clicking the “duplicate channel” function. (Figure 3).
Figure 3: Creating a COLA channel from the control channel data.
The data set with the copied channel must then be saved to use for processing. Click the “save as” button and save the dataset into a new run name.
2. Time Signal Calculator
To apply a filter to the control channel, the Time Signal Calculator Add-in must be activated. To do this navigate to “Tools -> Add-ins…” in the main menu of Simcenter Testlab. This will open the Add-ins pop-up (Figure 4).
Figure 4: Add-ins pop-up.
Find the “Time Signal Calculator” option, check the box, and click “OK” to activate the add-in.
The Time Data Selection worksheet should now have the Time Signal Calculator pane available, see Figure 5.
Figure 5: Time Data Selection worksheet with the Time Signal Calculator add-in loaded.
To filter the control channel highlight the “Formula” cell in Row 1 of the calculator and press the button with “f(x)” showing on it (Figure 6).
Figure 6: Location of the “insert a function” button in the Time Signal Calculator pane.
A list of all available functions appears (Figure 32). Locate and select the “FILTER_BP” function and click “OK.” The formula arguments entry menu should then appear (Figure 7).
Figure 7: Function selection menu.
For a filter function, the following can be filled in:
Function1: Enter channel identification or point identification name (this would be the channel being copied for COLA processing)
freqlo and freqhi: Enter band frequency for filter (these numbers are dictated by the start and end frequencies of the sine sweep)
Filtermode: Can be 0 or 1, depending if you direct or zero phase filter is desired
Press “OK” when finished. The formula will now be visible in the “Formula” cell. The software will automatically enter the next available channel number in the “Id” cell, but this can be changed if desired. It is also recommended to enter a channel name (in this case “Filtered COLA”) to identify the new channel more easily as shown in Figure 9:
Figure 9: Time signal calculator with formula entered.
At this point, as recommended above, the new dataset should be saved as a new run by clicking “Save as…” (see Figure 10).
Figure 10: Using the “Save As…” button to save the dataset as a new run.
Once saved, the view column will change from orange to green, as seen in Figure 11.
Figure 11: Once saved the calculated channel turns green.
At this point, navigate to the Offline Sine Data Reduction of the Sine Data Reduction knowledge article and follow the steps to process the filtered control channel to frequency vs. time.
As seen in the top figure of Figure 12, the filtering has improved the COLA signal, but there is still some tweaking that needs to be done.
Figure 12: Top: After applying a band-pass filter to the COLA signal (red). Bottom: Improved frequency vs. time trace (red) using filtered control channel and 75% holdoff.
The bottom of Figure 12 shows a much-improved frequency vs. time trace by using 75% holdoff in the “COLA Settings…” menu.
Once a clean frequency vs. time trace has been established, follow the steps above and process the time data into spectra. It may take a few iterations to get the best possible result.
While oversampling and having a COLA signal are very important to getting the best processed data, it is still possible to get a reasonable result using a processed data channel when the COLA is not recorded. Figure 13 shows the deviation of the online sweep data versus the offline reduction data.
Figure 13: Comparison of sine sweep data from original online swept sine test (blue trace) with oversampled control channel offline processed data (magenta trace) and undersampled control channel offline processed data (cyan trace). Top Graph - Full frequency range, Middle and Bottom Graphs - Two different zoomed in frequency ranges.
In the figure above the following are compared:
Blue trace: Original online sine sweep data from control channel
Magenta trace: Oversampled control channel from offline processed data
Cyan trace: Undersampled control channel from offline processed data
Note that while the filtered control channel processed data is not a perfect match to the original spectrum generated during the sine sweep, it is close enough to be useful when an oversampled COLA recording is unavailable.