FIR Filter Overflow

The FIR filter implementation in the OSP is designed for unity-gain. If the coefficients for the FIR are not scaled correctly, or there are transients in the input waveform to the FIR, then an overflow can occur. If such transients are expected, you should attenuate the data or adjust the pre-filter gain so that the overflow does not occur. When the FIR overflows, the data is clipped at the rail where the overflow occurred and the OSP block continues to process data. NI-FGEN returns an error when the status of the signal generator is checked. NI-FGEN ignores all overflows for the first 50 IQ samples of a generation and for 50 samples after the pre-filter offset or pre-filter gain is changed. If an overflow occurs during these 50 samples, the data is clipped but an error is not returned. Overflows are common during the first 50 IQ samples of a generation because the filter is initialized with zeros. If the first point of generation is a full scale value, then this is seen as a transient step response by the FIR filter and an overflow can occur. Overflows are also common when the and pre-filter offset or pre-filter gain is changed during generation because sudden changes to the gain or offset are seen as transient step responses by the FIR filter.

Tip  To disable error reporting caused by OSP overflows, set the OSP Overflow Error Reporting property or the NIFGEN_ATTR_OSP_OVERFLOW_ERROR_REPORTING attribute.