Cancel
Showing results for 
Search instead for 
Did you mean: 

Shock vibration control: What is zero padding doing?

Valued Contributor
Valued Contributor

In Testlab Shock workbook, I was wondering if anyone could provide some insight into Pre Pulse Zero Padding (what exactly is it doing?) setting under “Advanced…”?

 

settings.png

 

We have found that some of our Shock Tests start off with a lot of high frequency noise riding on the accelerometer signals (Blue Curve - It doesn’t seem to be coming from the UUT, as we see the same signature in our data even if we place the accels on a flat, thick, aluminum plate). Excuse the hand drawn curve.

 

difference.png

 

 

But when we increase the Pre Pulse Zero Padding to be greater than 35%, most of the high frequency content seems to be removed from the signal…almost as if we applied a Low Pass Filter.  There still seems to be some residual low frequency signal in the data (Red Curve).

 

We’re trying to determine:  What exactly is Pre Pulse Zero Padding doing?...and Why does it remove the initial high frequency content?

 

 

1 REPLY

Re: Shock vibration control: What is zero padding doing?

Siemens Legend Siemens Legend
Siemens Legend

 

The cause of the initial ringing is as follows:

 

  1. Delay because of circular convolution:
  • First the time domain target (pulse) is converted to the frequency domain using a Fourier Transform (FFT).
  • Next the drive spectrum is calculated using the Inverse Transfer Function (ITF) between the drive voltage and target vibration (Volts/g function). This is actually applying a FRF filter (ITF) on the target spectrum. This can introduce delays (phase changes).
  • Finally the drive spectrum is converted to the time domain using inverse an FFT. FFT and inverse FFT assume a periodic signal. As a result of the assumed periodicity and the introduced delays, it is possible that signals which should appear after the drive time signal block is finished (or before the time block has started), appear at the start of the time block (or at the end of the time block).
  1. Ideal filter effect:
  • In the frequency domain it is possible to use an ideal filter (at the borders for the control frequency range, the spectra an ITF go from the actual value to zero from one line to the next).
  • In the time domain an ideal filter does not exist except if the filter effect / delay can be infinite.
  • As a consequence, once the drive spectrum is converted to the drive time block, slight distortions can appear in the drive time block (drive time block not starting and stopping exactly at zero).

 

Both effects are most noticeable in case the target pulse and ITF are high at low frequencies. A solution for these issues is to apply more zero padding.

 

With zero padding, time data values of zero are added at the beginning of the shock time trace. (magenta versus blue curve below). This helps reduce leakage and other effects that introduce ringing into the signal by making the shock signal more periodic.

 

zero.png

 

  • More pre pulse zero padding allows the part of the drive signal which would otherwise be transferred from the start of the block to the end of the block to stay in place.
  • Extra zero padding increases the block duration. This lower the frequency resolution (and allows to lower the minimum frequency). This way there will be no energy at the lower frequency of the target pulse and the negative effect of the ideal filter will be reduced.
  • More post pulse zero padding allows the part of the drive signal which would otherwise be transferred from the end of the block to the start of the block to stay in place.