Adaptive Canceller Noise Measurements

Rich Bradley, NRAO, Charlottesville and Rick Fisher, NRAO, Green Bank, WV



To verify the RFI attenuation as a function of interference-to-noise ratio (INR) predicted by Equation 17 and shown in Figure 4 (top) of Barnbaum and Bradley (1998, Astron. J., vol. 116, p2598) we measured this quantity on the prototype adaptive canceler illustrated in their Figure 5. We also measured the noise added by the reference channel for a range of INRs in the main and reference channel as predicted by Equation 23 and shown in the bottom of their Figure 4. Finally, we measured the noise spectrum distortion caused by the frequency-dependent attenuation of noise in the reference channel by the tapped delay filter.

The test setup injected independent noise and a common simulated RFI CW signal into the main and reference channels. The strength of the CW signal could be changed independently in the two channels. The output of the adaptive canceler was analyzed by a 1024-channel integrating FFT spectrometer.

RFI Attenuation vs INR

Figure 1 shows the measured values (circles) of RFI attenuation by the adaptive canceler as a function of interference-to-noise ratio in the reference channel. The curve is the predicted relationship from Equation 17 of Barnbaum and Bradley. It wasn't clear to us what noise bandwidth to use in computing the INR so the zero point of the horizontal axis was taken as a free parameter. The visual fit of the curve in Figure 1 corresponds to a bandwidth of about 35% of the bandwidth of the full canceler passband. The may have a connection to the resolution bandwidth of the 9-tap adaptive filter. As predicted, the INR in the main channel had little or no effect on the RFI attenuation. These measurements were taken with a main channel INR of +5 dB.

Figure 1. RFI signal attenuation as a function of interference to noise ratio (INR) in the reference channel.

Noise Added by Reference Channel

Figure 2 shows the noise added to the canceler output by the reference channel as a function of reference-to-main channel ratio of INRs. A value of 1.0 corresponds to no added noise. The three symbols correspond to noise measurements averaged over different regions of the baseband spectrum. Crosses are for 6% of the spectrum surrounding but excluding the CW RFI signal; stars are for 80% of the spectrum excluding the CW signal and the end channels; and circles are for the lowest noise 45% of the spectrum well away from the CW signal. See Figure 3 below for a sense of the shape of the noise spectrum.

Using the 35% bandwidth fraction that gave the best curve fit in Figure 1 we estimate that the main channel INR was +5 dB. Given that, the prediction of Equation 23 of Barnbaum and Bradley falls roughly just above the broadband noise values (stars). As expected, the excess noise begins to fall off as the RFI becomes very weak in the reference channel, but the decrease is not substantial for the chosen strength of the RFI in the main channel.

Figure 2. Noise added to the system as a function of the ratio of interference signal-to- noise ratios (INR) in the reference and main channels of and adaptively canceling system. The three symbols are measured noise enhancements at frequencies near the interference signal (crosses), well away from the RFI (circles), and averaged over most of the spectrum (stars).

Noise Spectrum Distortion

Figure 3 shows the distortion of the noise spectrum due to frequency dependent noise from the reference channel injected into output of the canceler. Any excess noise above a value of 1.0 is from the reference channel. This spectrum is for the case of main channel INR = +5 dB and reference channel INR = +11 dB.

Figure 3. Distortion of the system noise spectrum by adaptive cancellation of a relatively weak, narrowband interference signal.

Last modified December 30, 1999

NRAO Green Bank Home Page

Rick Fisher's Home Page