32 A. Doan et al. progression of the signal’s spectral content at each step along the data acquisition chain. As noted, the nominal signal has distinct peaks at low frequencies corresponding to the target modal frequencies. The signal maintains this quality until the digitization (quantization) step where it experiences digital clipping. Once clipping is experienced, the low frequency peaks of interest are washed out and the data is no longer useful for the intended FTO. Figure 4.7 illustrates the flow of a signal through the two clipping gates common to the Artemis-1 DAQ system. The first gate is the sensor range. A sensor will clip if the total level within the sensor bandwidth (ie. nominal signal plus out-of-band operational signals) exceeds its internal voltage capacity. Sensor clipping has a large effect on signal quality as, once clipped, the electrical signal typically requires a long recovery time for the signal to decay back to its correct state. Assuming the original excitation did not clip the sensor it will travel to the digitizer which introduces the next gate. Prior to digitization, the signal can be amplified using an operational amplifier. The amount of gain is set based on the desired resolution of the output signal. As the voltage limits remain constant, this signal amplification induces a new clipping limit (digital range). The digital range of a signal is typically set at or below the sensor range. As Fig. 4.7 shows, if this is the case and no other mechanism exists in between the two gates to decrease the signal level, digital clipping may occur. This was found to be the common source of clipping on Artemis-1. Wherever such clipping problems were uncovered, the investigation sought to answer the question: what alterations should be made to the DAQ system in order to remove signal clipping and thus help ensure modal extraction success? To answer this question requires a complete understanding of the signal at risk and its intended purpose. For example, the first solution that may come to mind in this case would be to increase the digital range (decrease the gain amplification). At first glance that would be a workable solution. Unfortunately, since gain is inversely proportional to resolution, simply increasing the gain may lead to an intolerable loss in digital resolution of the nominal signal. Recall however, for modal extraction the bandwidth of interest is only low frequency modal response. Therefore, a better solution would be to apply an appropriate low-pass filter to remove the higher frequency vibro-acoustic noise causing clipping prior to digitization. This solution is illustrated in Fig. 4.8 by way of altering the anti-aliasing filter to cut-off at a much lower frequency then currently implemented. Figure 4.9 illustrates the effect of the change to the simulated signal DAQ process. As noted the updated filter alleviates the clipping risk and helps maintain nominal signal quality in the low frequency range of interest. Fig. 4.7 Signal Range Gates Clipped Signal Example Fig. 4.8 Effect of Modification to Anti-Aliasing Filter on Clipping
RkJQdWJsaXNoZXIy MTMzNzEzMQ==