Hello, I have created a Matlab simulation of a series of band-limited sinusoidal pulses indexed 1:N, spaced evenly in "slow time" each with a starting phase that increments linearly from pulse to pulse corresponding with frequency, f. Each pulse is created in "fast time" with a sample index from 1:M and undergoes complex downconversion. I and Q is filtered through a FIR after the complex multiply shifting to baseband. I then take the FFT of the resulting M x N matrix along the n dimension, so a mesh plot can be made of n bins vs. m bins. What I observe for a full-scale signal is an artifact at the image frequency I think I have seen in the past, but not given much though into. I would expect some residue at the image if the FIR filter had insufficient stop band rejection relative to the noise floor, but the amplitude of this artifact does not appear to change linearly with the pulse amplitude, or the stop band rejection of the filter, I actually can't get rid of it. I am careful, I think, there is nothing aliased along the pipeline. Plotting the image frequency bin (down the fast time m-axis) the artifact is two humps fairly low in amplitude, one time coincident with the leading edge of the pulse, the other with the falling edge. I can't make it go away even with a 100dB FIR stop band. This seems like the "rabbit ears" phenomenon observed in channelized receivers in the bandpass channels that are adjacent to (offset from) the center frequency of the pulse. In my case, the large signal that is in the filter stopband is the image. Maybe there is a special filter that is used in complex downconversion that addresses this issue. Thanks!

Hello JohnnyM, The rabbit ears are the transient response of the channelizer bins (or offset frequency filters) to an input signal not in its bandwidth. Every filter sees the incoming signal and starts responding till it recognizes that it is not the signal the filter was designed to pass.. The transient occurs as the offset filters ring up (fill) and as they ring down (empty). The amplitude of the transients decrease with spectral distance from the input tone burst signal's center frequency.

fred h   

Fred, 

Your explanation seems reasonable.So every complex baseband converter has this issue? They all will have an image that must be filtered, and it is offset from the FIR center by Fsample/2. Is there a way to control the amplitude? It seems if I play around with slightly different skirt shapes, there is some effect on on the shape of the distortion in the time domain, but I have failed to totally eliminate it. 

Thanks, John M. 

John,

The rabbit ears can by modified by controlled transitions in the pulses... controlled rise time and controlled fall time... makes amplifies less efficient. This a classic problem in radar threat receivers and interference between hoppers and nearby center frequency receivers. see the attached pdf

Rabbit_ears_16_Channels_new.pdf  

fred h

Fred, 

Thanks for the explanation.

John M.

Quick Guess - When the Pulse starts or stops you have a DC offset in I or Q signal.

Joakim