A
Anonymous
Guest
I have been doing some tests with digital filters. These tests assume that I will be integrating A-D samples in batches to produce 30 samples per second for the filters to work on. So each point in the input data for these tests has already been integrated for 1/30 of a second. There is approximately one seconds worth of input data. The target signal is 200 msec wide and is centered in the graph. Noise was added to the signal before the integration.
This is a test of low pass filters that perform the same function as the integrator in an analog PI detector. There would be other filters used in addition to these.
All these filters have the same cutoff frequency of about 4.5 Hz. So signals with a period less than 6 samples will be attenuated. Signals wider than 6 samples (200 msec) should not be attenuated.
The first graph is the 30 Hz input data.
The other graphs show the outputs of 1, 2, 4, and 6 pole filters.
The signal gets delayed (shifted to the right) by the filters. The delay for the 6 pole filter is about 6 samples. This delay limits how may poles it is practical to use in search mode. However additional poles could be used in pinpoint mode where the coil is not expected to be moving as fast.
Robert
This is a test of low pass filters that perform the same function as the integrator in an analog PI detector. There would be other filters used in addition to these.
All these filters have the same cutoff frequency of about 4.5 Hz. So signals with a period less than 6 samples will be attenuated. Signals wider than 6 samples (200 msec) should not be attenuated.
The first graph is the 30 Hz input data.
The other graphs show the outputs of 1, 2, 4, and 6 pole filters.
The signal gets delayed (shifted to the right) by the filters. The delay for the 6 pole filter is about 6 samples. This delay limits how may poles it is practical to use in search mode. However additional poles could be used in pinpoint mode where the coil is not expected to be moving as fast.
Robert