WO2003044560A1

WO2003044560A1 - Method for processing a frequency modulated interrupted continuous wave (fmicw) radar signal

Info

Publication number: WO2003044560A1
Application number: PCT/FI2002/000900
Authority: WO
Inventors: Pentti Karhunen; Jarkko Korte
Original assignee: Vaisala Oyj; Lilja, Aki
Priority date: 2001-11-21
Filing date: 2002-11-13
Publication date: 2003-05-30
Also published as: FI20012274A0; FI110966B; AU2002338986A1

Abstract

The present invention relates to a method for processing a Frequency Modulated Interrupted Continuous Wave (FMICW) radar signal: The method comprises following steps: generating a signal with a frequency changing over a set bandwidth, gating the signal on and off with a predetermined frequency (fg) thereby forming on-timeslots and off-timeslots, sending the gated signal to a desired target with help of a transmitter and an antenna, receiving a reflected signal during the off-timeslots of the gated signal, and detecting and gathering desired properties from the received signal during the off-timeslots of the gated signal. In accordance with the invention a desired portion of the off-timeslot is detected and/or processed during each off-timeslot of the gated signal in order to obtain information from a desired distance and in order to minimize the noise of the signal related to desired distance.

Description

Method for processing a Frequency Modulated Interrupted Continuous Wave (FMICW) radar signal

The invention relates to a method according to the preamble of claim 1 for processing a Frequency Modulated Interrupted Continuous Wave (FMICW) radar signal.

The invention is related to radar systems utilizing frequency-modulated interrupted continuous wave modulation (FMICW). This modulation method is sometimes referred also as interrupted FMCW, or IFMCW.

The invention is suited for situations in which FMICW radar signal is used to search for targets, either discrete or continuous, in a continuous range of distances, or to track several targets at various distances.

FMICW radar is a radar employing a pulsed frequency sweep (see, for example,

Principles and Applications of Millimeter- Wave Radar, Currie and Brown, editors, page 683). h such a system an interrupted linear frequency sweep is transmitted, typically with a 50 % transmit duty cycle.

Typically this kind of radar method comprises the following steps. A signal with a frequency changing over a set bandwidth is generated. This signal is gated on and off with a predetermined frequency (f_g) thereby forming on-timeslots and off-timeslots. This gated signal is sent to a desired target with help of a transmitter and an antenna. The reflected or back-scattered signal is received during the off-timeslots of the gated signal, and desired properties from the received signal is detected and gathered during the off-timeslots of the gated signal.

When applying FMICW to a wind profiler radar, typical operating parameters might be such that the range resolution is 150 meters and 3 km is the targeted maximum observation range. In such a case, a 1-MHz linear frequency sweep would be transmitted, and the sweep would be interrupted with a 20 μs pulsing sequence. As a result, the 3 -km range would be at the optimal point with respect to the pulsing frequency; the 20-μs pulses transmitted would arrive to the receiver 20 μs later, the time lag corresponding to round-trip time to 3 km and back, later. As a result, all of the 20-μs pulse scattered power would enter the receiver. For other heights, however, the situation is not as favourable. The time-frequency plot for a 20-μs pulse, 20-μs reception timing with 1 MHz linear frequency sweep is shown in Figure 1.

Because the returned signal is mixed with the transmitted signal's continuous frequency sweep in the FMICW receiver, different beat frequencies will correspond to different ranges. Illustrated in Figure 2 is the time-frequency plot after mixing the received signal with transmitted one. The mixing may happen in one step (homodyne) or several steps via intermediate frequencies (heterodyne).

In Figures 1 and 2, three exemplary radar returns are shown: one matching the pulse modulation frequency, 3000 meters, one return that is significantly closer than the matching distance, and one return that is further away from the radar than the pulse modulation matching distance. These three targets can be regarded as part of a geophysical scattering density targets, or as multiple point targets in a seeker or tracking radar.

Illustrated in Figure 1 are the transmitted signal and the returns from 3 heights, 1200, 3000, and 4200 meters. The return from 1200 meters is delayed 8 μs w.r.t. the transmission, and the return from 3000 meters is delayed 20 μs w.r.t. the transmission. The return from 4200 meters is delayed 28 μs.

fn the FMICW system, the signal is allowed to go into the receiver only during periods of no transmission. The baseband signal is formed by mixing the received signal with the no-transmission parts of the frequency sweep. As a result, the baseband return of the system from the three heights, 1200 meters, 3000 meters, and 4200 meters are obtained. The returns from these heights are illustrated in Figure 2.

The frequency sweep described in Figure 1 is continuous in phase throughout the full 5 ms cycle. Thus, the fragments of 1.6 kHz (corresponding to 1200 meters), 4.0 kHz (corresponding to 3000 meters) and 5.6 kHz (corresponding to 4200 meters) signals shown in Figure 2 will also exhibit continuous phase throughout the frequency sweep, given the fact that the 5 ms sweep is much smaller in time than the decorrelation time of the scatterers in the volume studied.

Given the phase-continuous character of the returns, each of the fragments in each of the frequencies in Figure 2 creates a sinusoidal signature that is broken in time. The intermittent 1.6 kHz signal will have time to generate 8 cycles of intermittently broken sinusoid during the 5 ms sweep, the 4 kHz signal 20 cycles and the 5.6 kHz signal 28 cycles. These sinusoids get summed in the receiver, producing a voltage that contains various intermittent frequency components relating to various heights.

By taking digital samples during the receiver-on times a time series of 125 points (if one sample is taken during one RX interval) is generated. The various frequencies in this time series are separated by means of a Fourier transform. For example, the 8^th, 20^th, and 28^th sample in this Fourier transform correspond to the heights 1200, 3000 and 4200 meters. If the receiver is, as is preferably the case, a receiver capable of separating the in-phase and quadrature components in the base-band, the corresponding Fourier transform will be complex and the frequencies extracted in the given example are negative.

It is an object of the present invention to overcome the drawbacks of the above- described techniques and to provide an entirely novel type of method for processing a Frequency Modulated Interrupted Continuous Wave (FMICW) radar signal.

The goal of the invention is accomplished by detecting and/or processing only a desired portion of the off-timeslot during each off-timeslot of the gated signal in order to obtain information from a desired distance only and in order to minimize the noise of the signal.

More specifically, the method according to the invention is characterized by what is stated in the characterizing part of claim 1. The invention offers significant benefits over conventional techniques.

The invention improves the detection capability of an FMICW radar outside the distance that matches the pulsing period. Thus, it will improve the usability of the FMICW technique for detecting continuous profiles of volume scatterers, which is the case in weather radars, wind profilers, and other geophysical radars. It will also improve the usability of FMICW in seeker radars and reduce the need for pulse frequency scanning, because a larger distance range is detectable with one realization of pulsing frequency.

In the following the invention is described in greater detail with the help of exemplifying embodiments illustrated in the appended drawings in which

Figure 1 shows a frequency/time graph of the principle of the FMICW-modulation.

Figure 2 shows another frequency/time graph of the principle of the FMICW- modulation.

Figure 3 is a frequency/time graph of the principle of the invention;

Figure 4 is a multiple graph presentation of the signals obtained by the invention;

Figure 5 is another multiple graph presentation of the signals obtained by the invention.

This document describes a novel method associated with frequency-modulated interrupted continuous wave modulation (FMICW). The method enhances the radar return from distances not matching the pulsing frequency of the system, thus improving the capability of FMICW systems to make observations in a continuous range of distances. Traditionally, FMICW systems are used as seekers in a way in which the pulsing frequency is scanned.

As shown in Figure 2, signal from different heights in the FMICW system consists of intermittent frequency components. The intermittence is dictated by the pulsing parameters and the height in question. If one sample per pulse is taken, integrated over the reception period, the sample will contain noise and the signals from the various heights. Let us look at one, the first, of the reception periods in more detail.

Now, when one sample per pulse is taken, after integrating over the reception period, each frequency component in the signal after integration will contain a signal contribution and a noise contribution. The signal contribution results from the integrated voltage over the duration of each line shown in Figure 3. However, the noise content in that frequency bin will correspond to the integration over the whole 20 μs period. While the signal contribution to the signal will remain coherent over several pulses, the noise contribution will not, and the noise will thus have an adverse effect to the signal via disturbing the phase and the voltage of the actual, wanted signal.

Let us look at one continuous 5-ms sweep and the corresponding output from the FMICW receiver. Let us have an object, either discrete or a volume scatterer, at a distance of 1200 meters, corresponding to the 1,6 kHz signal in Figures 2 and 3. If we have a receiver with 1-MHz bandwidth, which is a matched bandwidth value with respect to the transmitted waveform, we have a full representation of the signal with 1-

MHz quadrature sampling in the baseband. We will not consider the effects of the receiver chains filters' finite transition bands or aliasing phenomena here. In a practical device, when taking these phenomena into account, a sampling bandwidth wider than 1

MHz might be chosen. However, it will not change the situation essentially - even in that case the same fraction of time would be digitally integrated as in these examples.

However, in that case, there would more samples to integrate for each case.

Figure 4, upmost row of plots, shows the time domain signal of a (simulated) target at

1200 meters, with noise. The right-hand plot shows the result of using a 5000-point fast

Fourier transform to extract the different frequency- and, at the same time, range - components from the signal. The signal-to-noise ratio for this target and method is 15.9 dB, measured over the resolved bandwidth of 200 Hz. It occurs at the 1.6 kHz baseband frequency, at the 8^th frequency bin, which corresponds to the 8^th height bin, 1200 meters.

The signal consists of 20 zeroes, corresponding to the transmission time during which the receiver is blanked, followed by 20 samples of signal and noise, followed by 20 zeroes, etc (see also Figure 5). The time unit in left column of plots in Figure 4 is 1 μs, and the frequency unit in the right column of plots is 1 MHz. In this discussion, it is assumed that the blanking is achieved in the digital domain, making complete nulling of the transmission-time signal possible. Note that the 1-MHz bandwidth is larger than the pulse repetition frequency, 25 kHz, resulting in replicas of the signal ±25 kHz from the centermost frequencies.

Figure 4, middle row of plots shows the effect of integrating over the period of 20 μs, the full reception time, and using a 125-point fast Fourier transform to extract the different range components from the signal. The length of the Fourier transform was decreased without loss in signal-to-noise ratio, which is now 15.7 dB. The pulsing effect is not anymore shown due to decreased Nyquist frequency in the baseband. The integration has been realized as a sum of values over each of the reception periods, resulting in increased voltage amplitudes. It is noteworthy that the signal studied here, 1200 meters, corresponds to a rotating phasor that turns less than 5 degrees during one reception period. It is thus possible to sum all the samples from the reception period without losing significantly signal voltage due to summing rotating phasors.

The two upmost rows in Figures 4 and 5 can be regarded as belonging to the prior art. Figure 4, last row, shows the effect of integrating according to this invention. In this case, a discrete Fourier transform, which can also be implemented as using only one of the output components of the fast Fourier transform, is used to extract the frequency component corresponding to 1200-meter range. The difference is that instead of using the full reception-time integrated signal, an integrated signal in which the integration time has been matched with the two-way delay between the transmission and the said target's echo reception is used. In this case, only the 8 first samples are integrated during each reception period to produce the time series shown in the left plot in the last row. Eight-sample integration was chosen for this DFT, because it matches the amount of time at the beginning of the reception period with target-scattered signal. This can be seen in Figure 3, only the 8 first microseconds in the reception period contain scattered signal from the target at 1200 meters. The amount of noise in this time series is less than the noise in the previous time series, because the last 12 samples has been omitted in the integration, and the noise in those samples does not contribute to the output of the 8- point integrator. The signal itself remains intact, because all the information from that range has been used, as shown in Figure 3. The signal-to-noise ratio resulting from partial reception time integration for this exemplary case is 19.5 dB, 3.8 dB higher than the SNR for the fully integrated reception period.

In Figure 5 the data on the second row results from adding all the 20 points of the signal in the upmost plot per reception period. The data on the last row results from adding 8 points per reception period, resulting in less noise. Similar scheme is used for all the height separately.

The essence of the invention described herein is to construct several digital integrators, preferably by accumulating a sum continuously, and use a variable integration time for each range. Thus, for example, 150-meter range would be detected by using 1 -point integration, i.e., the first sample, of each reception period with a DFT that resolves the first non-zero frequency from the 125-point time series consisting of said first samples. 300-meter range would be detected by using 2-point integration, i.e., the first two samples summed, of each reception period, and subjecting the resulting 125-point time series to a DFT that resolves the second non-zero frequency component from the 125- point time series. This scheme is continued up to the pulse matching height, 3000 meters in this case, for which the full 20-point integration result is used with a DFT resolving the

non-zero frequency component.

For the heights above the 3000-meter heights, only the end part of the reception period is used in the integration, to include all the signal information and to minimize the amount of noise in the corresponding frequency component. For example, for the 4200- meter signal shown in Figures 2-3, the 12 last samples during each of the 1-MHz sampled 20 μs reception period are used in conjunction with a DFT extracting the 28^th non-zero frequency component from the time series. One preferred embodiment of the present invention would be to use a DSP or FPGA system to compute the range-resolved signals, using cumulative sum accumulator at a high-enough frequency to resolve the transmitted waveform and multiplying the contents of said accumulator with complex weights that correspond to the DFT coefficient for that pulse and that range. The output of the multiplication would be input into another complex sum accumulator for each range, and the contents of those accumulators would be read at the end of each sweep. These values are the complex voltages corresponding to each height for that sweep. The DFT accumulators are nulled at the end of each sweep, and the integrator accumulators are nulled at the beginning of each reception period.

It is in the scope of this invention to use, instead of digitally summing all the samples during one pulse's reception period, a DFT to the range-matched subset of the original 1-MHz samples. This is useful in cases where targets at ranges commensurate with the sweep repetition period are detected. For example, in the exemplary case shown in this document, targets at several hundred kilometers would get damped if the summing integrator was used. This is due to summing complex voltages with a rotating phasor over a time in which the phasor rotates significantly. Thus, in the case of targets for away from the radar, the preferred method is to use a matched subset of the original samples as an input to a long DFT. In these cases summing the digital samples would result in loss of some of the signal amplitude due to summing a significantly rotating phasor.

The amount of detection improvement presented in the example corresponding to Figure 4, 3.8 dB, is due to the fact that the target being detected remains essentially coherent during the 5 ms sweep period, whereas the noise is fully incoherent, having no phase correlation between the samples. Thus, noise power is additive during the reception period, and in the preferred integration method 8/20 of the noise power is received, instead of the 20/20 in the full integration. The 8/20 factor corresponds to 4.0 dB, which essentially equivalent with the simulated result. To illustrate better the partial integration during the reception of pulses, Figure 5 shows zoom-ins of the plots in Figure 4.

A typical device in accordance with the invention comprises following elements:

i) a wide-bandwidth sampling system (connected to an analog-to-digital converter), ii) a digital logic that produces more than one sum of samples per reception period of one FMICW pulse, iii) a computing system that uses more than one time series, consisting of the said sums of samples, as input to DFTs or FFTs to extract one or more frequency components per time series, corresponding to same number of ranges.

Typically the length of each sum of samples and the frequency component extracted in the DFT or FFT is matched to correspond to the same radar range.

In this application, "wide-bandwidth" refers to a sampling bandwidth that is sufficient to resolve the transmitted bandwidth of the radar.

As one alternative integration system in accordance with the invention may comprise the following elements: i) a wide-bandwidth sampling system, ii) a digital logic that produces more than one subset of samples per reception period of one FMICW pulse, iii) a computing system that uses zero-padded time series, consisting of the said subsets of samples, as input to DFTs or FFTs to extract one or more frequency components per time series, corresponding to same number of ranges.

Typically, the length of each continuous subset of sample per reception period, and the frequency component extracted in the DFT and FFT, is matched to correspond to the same radar range. Although a typical duty cycle for the gating is 50%, in accordance with the invention the duty cycle can vary in a range of 10-90%.

Claims

Claims:

1. A method for processing a Frequency Modulated Interrupted Continuous Wave (FMICW) radar signal, which method comprises following steps: - generating a signal with a frequency changing over a set bandwidth,

- gating the signal on and off with a predetermined frequency (fg) thereby foiming on-timeslots and off-timeslots,

- sending the gated signal to a desired target with help of a transmitter and an antenna, - receiving a reflected or back-scattered signal during the off-timeslots of the gated signal, and

- detecting and gathering desired properties from the received signal during the off- timeslots of the gated signal, characterized in - detecting and/or processing a desired portion of the off-timeslot during each off- timeslot of the gated signal in order to obtain information from a desired distance and in order to minimize the noise of the signal related to the desired distance.

2. A method in accordance with claim 1 , characterized in that only the beginning of the off-timeslot is processed.

3. A method in accordance with claim 1 , characterized in that only the end of the off- timeslot is processed.

4. A method in accordance with claim 1 , characterized in that the transmit duty cycle res ting from the gating is about 50%.

5. A method in accordance with any of the previous claims, characterized in that the gating frequency is varied in the range between values 1 % and 50% of the bandwidth (B).

6. A method in accordance with any of the previous claims, characterized in that the method is used for geophysical remote sensing.