Solution: The Optical Vector Analyzer provides multiple options for filtering: the time domain filter and the smoothing filter. In this application note, we will discuss the Time Domain Filter.

Time Domain Filter

The time domain filter is applied to the raw Jones matrix data. It allows the user to select the portion of the time domain information to be included in the frequency domain representation of the Jones matrix. To adjust and apply this filter, click the Time Domain button and a new window will appear (see Figure 1).

OVA Time Domain Window

Figure 1: OVA Time Domain Window.

To set up an effective filter, it is necessary to frame the impulse response, making sure all relevant data is contained in the filter. The filter width determines the spectral resolution (or resolution bandwidth). In turn, the spectral content of your signal determines the width of the impulse response. So by setting the filter to just surround the impulse, and not include any surrounding noise, you are necessarily choosing the correct resolution bandwidth with which to measure your device.

If you initially set your filter, without looking at the lowest levels of the response, you may miss some important information (see Figure 2).

OVA Time Domain Filter set without zoom

Figure 2: An example of setting the OVA Time Domain Impulse Response Filter without zooming into the lowest levels of the impulse response. Notice that the filter (in purple) cuts out a portion of the signal as shown in red.

As a result, when setting the filter, you should zoom in to the lowest levels of the "grass" to find the outside peaks. It typically takes multiple times of using the Y-axis zoom tool to bring the data into view.

After you have an accurate representation of your low level data and determine where the outside peaks are, it is necessary to investigate to see if those peaks are part of the signal or just noise. To determine if they are actually part of the impulse response, return to the main screen and take multiple scans of the same data. Scan once with approximately 10 averages. After the scan, copy the data into Matrix B. Then do a second scan with 25-30 averages. That data will be placed in Matrix A. Display both matrices on the graphs, and change the view to Time Domain (Amplitude). Zoom in on this data, as you did in the original filter.

In the example shown below (Figure 3), the same device was scanned two different times: once with 16 scans (red) and a second time with 64 scans (blue). Notice how the peak appearing at 1.61 ns dropped significantly as more scans were averaged.

OVA impulse response with 16 vs 64 scans

Figure 3: The same DUT scanned 16 times (red) versus 64 times (blue). Notice how the peak at 1.61 ns disappears with more scans. This is noise and should be left out of our time domain filter.

Now, when we return to the Time Domain window to set the actual filter that we want to use, we know that the peak appearing at 1.61 is noise, and we can leave it out of our filter settings. Once in the window, we have several ways of controlling the filter shape. The filter is a flat-top Gaussian function. The flat center region has a value of unity between the two point defined by the vertical yellow cursors. We use these yellow cursors as a guide to the eye, but the purple is the actual filter. Notice that on Figure 4 below, which is the same DUT as above in Figure 3, we placed the yellow cursors inside of the peak appearing at 1.61 ns, since we found it was noise. However, because of the shape of the filter (shown in purple), the peak is still included in the filter.

OVA time domain filter - big sigma

Figure 4: Despite the yellow cursors being set to keep out the noise peak at 1.61 ns, the filter shape (in purple) still contains the peak.

In an effort to keep the noise peak out, we can control the shape of the filter using the sigma value, which is the width of the Gaussian edges on the filter. To tighten the filter function, set sigma to a smaller value. In this case, we changed sigma by an order of magnitude and were successful in eliminating the noisy peak (see Figure 5).

OVA Time Domain Filter- small sigma, tight filter

Figure 5: By decreasing the value of sigma, we are able to tighten the filter and eliminate the unwanted noise.

Getting the correct time domain filter is crucial in obtain valid, useful measurements.

Return to top of page.

Related Links:
Product Support Home

Document ID: 0615006003
Last Update: 06/15/2006

Other questions, contact LUNA SUPPORT
Phone: 1.866 LUNAOVA (1.866.586.2682)
E-mail: support@lunatechnologies.com