Coherent Leakage (TN-003)


Coherent leakage in a radar receiver system occurs when a spurious signal at one of the radar's RF or IF frequencies leaks back in to the receive path. Leakage may occur due to several reasons, including interactions between the transmit and receive subsystems of the radar. This article will briefly cover the known causes of coherent leakage, how to detect it, and what can be done to fix a coherent leakage problem.

Reasons for coherent leakage

Interaction between the Tx and Rx signal paths

In a coherent radar, one of the main causes of coherent leakage is interaction between the transmit and receive signal paths. Such interactions occur due to the imperfections of RF components. As an example, we will consider the CHILL transmit and receive chains.

Simplified view of transmit and receive chains, showing the interactions that lead to coherent leakage

The figure shows a simplified version of the transmit and receive chains. In the transmit chain, a 50 MHz IF signal is generated by the "COHO", or Coherent Oscillator. This is band-pass filtered and mixed with the Stable Local Oscillator ("STALO") signal at 2775 MHz. Note that a power splitter is used to couple the STALO output to both the transmit and receive mixers. The mixer output contains the sum frequency at 2825 MHz and the difference frequency at 2725 MHz. We are interested in the difference frequency, so a band-pass filter is used to select it. Examining the band-pass filter reflection coefficient, one finds that the filter reflects energy outside its pass-band. This reflected signal enters the mixer RF port, and appears at the LO and IF ports, with some attenuation. The MDC-169 mixer used on CHILL has an RF to LO coupling of about -30 dB, so the 2825 MHz signal appears at a level of about -30 dBm at the LO port.

The power splitter is a Mini-circuits ZA4PD-4, which has a port-to-port isolation between -25 and -50 dB, depending on the ports used. We will consider the best case, and assume -50dB isolation. Now, the unwanted 2825 MHz signal is attenuated to about -80 dBm. This enters the receiver's first mixer LO port, along with the 2775 MHz signal, resulting in a 50 MHz product. Assuming that the mixer conversion loss is about 10 dB, the IF port will have a 50 MHz signal at about -90 dBm. This is significantly above the receiver's -113 dBm noise floor (at 1 MHz bandwidth), and results in quite a severe coherent leakage. The digital downconverter will bring the 50 MHz signal down to DC. Examining a spectrum of the digitized signal reveals the DC component standing out from the noise floor.

Tx to Rx radiative coupling

The transmit and receive chains are often placed close together, to better utilize the common signals between the two. Improper shielding, poor quality or loose connectors and faulty coax cables can result in coupling between the two parts of the RF/IF hardware, resulting in coherent leakage.

Digitizer clock signal interference

If the digitizer clock signal, or some reference from which it is derived, is harmonically related to the radar IF, severe coherent leakage results. This type of coherent leakage is difficult to control, since the digitizer clock and reference are usually high level signals (+10 dBm or more), in an effort to reduce clock jitter. An example of this is CHILL's previous receiver design, where a GPS-derived 10 MHz clock is used to generate a 40 MHz sampling clock using frequency multipliers, and the final IF is 10 MHz.The frequency multipliers do not sufficiently attenuate the 10 MHz product from the 40 MHz clock, resulting in a spurious component that got coupled in to the IF signal.

Detection of leakage

The first signs of coherent leakage in a radar receiver is an apparent increase in noise floor. The computed mean velocity does not appear random in areas of low SNR (it tends to get biased to zero), and the Normalized Correlation Power (NCP) tends to be relatively high (above 0.5) even in areas of low SNR, or when the transmitter is shut off.

Examining the power spectrum of the increased "noise" reveals a component at DC. This coherent component results in the high NCP and velocity bias.

In order to detect the cause of the coherent leakage, a microwave spectrum analyzer may be connected to various points in the receiver chain, such as the digitizer input, to determine the source of the problem. Use of narrow resolution bandwidths is recommended, since coherent leakage generally tends to produce very weak signals and can easily be buried under the spectrum analyzer noise floor.

What to do about it


Radiatively coupled coherent leakage is generally the hardest to track down. Any worn out connectors or cables should be replaced, and the transmit and receive portions of the RF/IF chain should be shielded and isolated from each other.

Proper Reference Frequency source design

The frequency references used on the radar must be carefully designed, to prevent any unintended interference between the references and the measured radar signal. Keep in mind that filters cannot infinitely attenuate unwanted frequencies, and most filters are reflective, and can cause energy in their stop-bands to end up in unwanted parts of the design. Wenzel Associates, in their Blue Tops line of filters (typically used on clock/frequency synthesis chains) tend to use attenuator pads on the filters to absorb some of the reflected energy.

Another approach that is commonly used is to judiciously apply isolators to the various points on the frequency synthesis chain, as well as the Tx/Rx chains. Typically, the isolators are not used on the main Tx/Rx paths, in order to avoid losses. Instead, they can be applied to the mixer LO ports. In the example shown above, if isolators are used on each output of the power splitter, the reflected 2825 MHz signal from the mixers gets absorbed by the isolator.

A rather expensive approach is to use diplexers in place of bandpass filters. The unwanted frequency arm of the diplexer is terminated, which absorbs the unwanted frequency.

Image rejection mixers (IRMs) also help to reduce interference from other channels. IRMs will attenuate the unwanted image frequency by their image rejection ratio, which is typically 25-30 dB. They do have a higher insertion loss, due to the 90° hybrids at the LO and IF ports.

Pulsed IF on Transmit

The approach used on the current CSU-CHILL radar transmit/receive chain is to not operate the "COHO" in CW mode. In this case, the COHO (which in reality is a digital waveform synthesizer) generates a shaped IF burst when the radar is transmitting. When the radar is receiving echoes from targets, there is no IF signal, and therefore, no energy at 2825 MHz. In addition, isolators are fitted to the LO distribution splitter, to reduce cross-talk between polarization channels.