Spectral Dynamics - Puma Answers

CATS

• What does the limit parameter 'Sigma Clipping' refer to?

Answer: A random based time domain signal's magnitudes are generally characterized by probability distributions. These distributions typically define the probability fraction (i.e. amount of "time" or likelihood) that the signal amplitude occurs within a given amplitude range. In particular, random vibration contain peak magnitudes that occur according to a Gaussian distribution. This means that the signal amplitudes spend most of time near the low amplitude limits and, very infrequently, the time history may have much larger peaks. The frequency content of the random signal can be shaped to match the control system's needs, but the time history basically follows a Gaussian definitions. The generated time domain output can have a peak to RMS ratio of up to 6.5:1 . Since most output devices, as well as shaker amplifiers, are sensitive to peak values, the user may establish the maximum allowable amplitude excursions as a multiple of the signal's RMS level (sigma). This allows the user to tailor the peak to RMS ratio of the drive time history waveform.
Sigma clipping is used to prevent high magnitude voltage signals from being sent to the shaker system. It is used to protect either the shaker amplifier from an over voltage condition or possibly to limit the instantaneous acceleration seen on the test article. Sigma clipping limits the peak values of the drive signal by establishing a maximum level that is set by the user relative to the 1 sigma, or RMS level of the drive waveform.
Sigma clipping should be utilized with care since the effects of severe signal truncation may produce unintended consequences. Various internal digital processing windows and output smoothing filters generally help smooth the effects of sigma clipping, but extreme clipping limits can cause control problems in certain situations. Sigma clipping can produce non-linear effects when extreme limits are utilized (e.g. less than 3.0 or so). Extreme clipping effects are similar to increased noise or broadband non-correlated excitations (i.e. structure rattling). One will find that a sigma clip level in the 3 to 4 range usually produces good control and good peak to RMS ratios.

Back to FAQs

• What is the 'Data Scan' option and how is it utilized?

Answer: The Data Scan feature is found under the Setup dialog list of options. This feature is designed for cases where the time channel data stream contains spurious "noise" like spikes (i.e. local maxima) that appear unrelated to the physical data measurement. The spikes are usually a single data point that generally exceeds the expected waveform profile. These spikes are sometimes observed when utilizing very low amplitude signals (i.e. approximately 1 millivolt full scale) in the presence of noisy ambient environments. The spikes can cause unwanted effects in post processing displays such as the time domain peak envelope processor found under the composite plot type option.

Setup and Operation:

The data scan feature is configured on an individual channel basis. The user needs to supply a Trigger Reject Threshold (%) as well as a Multi-Pass Reject Count. A threshold of 0.0 will deactivate the scanner for the specified channel. The reject count indicates how many passes should be performed on each time domain data frame. Each pass can remove a single spurious "spike". Multiple passes are only necessary if multiple spikes are observed in the channel's displayed time domain waveform. The reject threshold specifies the rejection criterion based on an adjacent neighbor measurement. Once a spike has been identified, the neighbor data point values are examined to determine if the spike's value exceeds the threshold amount from the neighbor values. If the spike's value exceeds the specified threshold, then the spike value will be replaced with an arithmetic average of the neighbor data points.

Example:

Consider the (time) data value sequence:

... 10., 15., 60., 15., 10., ..

The local spike is determined to be the value 60.0 (i.e. local maxima).

If the threshold is specified as 100%, then the rejection criterion would be:

15. + 100% = 30.0

for both the left and right neighbor data values. Since the spike is 60.0, then that value would be replaced with the average value of:

(15.0 + 15.0)/2.0 = 15.0

Note that both neighbor points are utilized in the rejection criteria. Note also that only a single rejection is performed. If there is a possibility of multiple spikes in a data frame, then the user should specify a higher rejection count (greater than 1). One should generally keep the count low, however, (such as <= 5) in order to not inadvertantly alter the true character of the time domain waveform.

If the rejection count is greater than 1; then once a spike has been corrected, a complete scan is performed to determine the next local maxima. The rejection processing is performed exactly the same as done for the first pass. If there were no rejected spikes during a scan, than all additional scans (if any are pending) will be skipped.

Back to FAQs