Deconvolution noise in DS-radio channel sounding and a polynomial compensation method

We have shown previously that matched filter deconvolution (MFD) is a powerful inverse filtering technique for increasing time resolution in direct-sequence (DS) type radio channel sounding. A drawback of the technique anyhow is an SNR floor that is caused by deconvolution noise (DN). The DN is correlative with the channel impulse response (IR) and can appear also in noiseless channels. We suggest that the reason for the DN is in spectral zeros of the sounding code. The concept is studied by deriving an approximation for the MFD IR. We observe that the accuracy of this expression is proportional to the deconvolution constant. Based on this scenario we then suggest a spectral interpolation technique for compensation of the DN. We simulate its performance as the function of uniformly distributed delay and Gaussian amplitude components (UDDGAC) channel tap number and post channel SNR and record the respective instantaneous radio channel impulse response estimate (IRCIRE) SNR. Our simulations indicate that the higher the channel tap number is, the higher degree curve fit should be applied to replace the missing frequency domain data. For example we note that the fifth degree polynomial fit can yield IRCIRE SNR improvement varying between 4 dB to 11 dB in a line-of-sight (LOS) channel having 9 to 1 in average 20 dB smaller UDDGAC taps respectively. If the degree of the spectral fit is smaller than what is required by the actual channel taps, we note that the spectral interpolation can potentially deteriorate the MFD performance