In the context of pulse radars, I'm wondering exactly why and how so-called "fill pulses" are used.
I understand having $N$ fill pulses allow operators to retrieve $N$ echoes within the coherent processing interval (CPI), which excludes the fill pulses, for ambiguous targets with a delay within $[PRI, (N+1)*PRI]$. Unambiguous targets also generate $N$ echoes, possibly coinciding with the range bin of ambiguous targets. The following figure illustrates my current understanding:
Assuming what I just said is correct, I'm still confused regarding:
- the weights, zero (cf. first quote at the end of this question) and non-zero, used for pulses integration downstream. Why weights and why zero weights specifically for fill pulses ?
- the possibility to separate unambiguous targets located within the same range bin as the ambiguous ones. This range bin would contain echoes from both fill and regular pulses for the ambiguous targets, and echoes only from regular pulses for the unambiguous target. How do people address this usually if possible ?
- whatever I'm missing regarding the motivation behind fill pulses. I've mostly read about ambiguous clutter and blanking period, which haven't really convinced me yet.
A few quotes and references leading me to these questions :
From "Radar Design Principles - Signal Processing and the Environment" by Nathanson, Reilly, Cohen (chapter 9, p. 426):
The cancellation of ambiguous-range returns can be improved by using fill pulses, i.e., pulses that are transmitted but given zero weight in the receiver processor. The benefit of fill pulses can be illustrated in the three-pulse canceler, in which the weights are W1 = 0, W, = 1, W, = - 1. In this example, the first pulse is a fill pulse, and the other two are combined a s in a two-pulse canceler. (...) The three-pulse canceler in this example would have a maximum im- provement factor of 27.7 dB, but with a fill pulse, the limitation would be 43.0 dB. The limitation from ambiguous-range clutter is illustrated in the following for rain and for sea clutter under various conditions of evap- oration ducts.
From "Pulse Doppler Radar" by Alabaster, (13.1.3, p.309):
A space charging time needs to be included whenever the radar switches from one coherent processing interval (CPI) to the next. During this time, the radar must transmit pulses on the new PRF (also known as fill pulses) but must blank the processing of any returns. The blanking period must be sufficient to allow all returns from the previous PRF to clear the system before processing returns on the new PRF may begin. In essence, the space charging pulses must be generated, but this time cannot be used by the receiver and processor; it is dead time.
From this online patent "Coherent integration of fill pulses in pulse doppler type sensors":
The fill pulses are used to “fill” up the range bins that have clutter reflection scatterers in them, and these Fill Pulses are not processed for detection by the traditional PDW processing. These Fill Pulses therefore represent a waste of radar RF power from a pure economics standpoint, although they are absolutely necessary to achieve good Doppler target detection performance in the presence of clutter. (...) The invention enables full recovery of all Fill Pulse energy through coherent addition of the Fill Pulses into the PDW processing for enhanced SNR and SCR. It also increases the length of the CPI to provide better Doppler resolution and hence better velocity measurement accuracy. The invention further increases the number of coherently integrated pulses in the farther out range intervals in the Clear Region to provide higher SNR and SCR at farther ranges. The invention also decreases the number of unreceived pulses, and helps avoid dwell to dwell interference.
The latter puts forward a Doppler filter definition revealing fill pulses:
$DF_i = \left( \sum_{k=N_{fp}}^{N_{TOT}} w_{i,k} x_k \right) + \left( \sum_{j=1}^{N_{fp}} w_{i,j} x^{'}_{j+N_{TOT}} \right)$ with $x^{'}_{j+N_{TOT}} = x_j + x_{j+N_{TOT}}$
I assume there is a typo here and the first sum should start at $k=N_{fp}+1$.
EDIT:
forgot to mention I also found in the literature pulses called "fill pulses" which amount to waveform diversity to address the blind close-range zone associated with longer CPI pulses:
From "Multifunction Phased Array Radar Pulse Compression Limits" by J.Y.N Cho:
Lengthening the transmitted pulse, however, creates its own problems. First, the nearby range corresponding to the pulse length becomes a blind zone, because reception is not possible during transmission. This blind zone can be covered separately using a much shorter fill pulse as long as the reduced sensitivity still meets the surveillance requirements for those ranges. If the pulse compression ratio is made too large (blind zone too large), then the sensitivity with the fill pulse would not be able to fulfill the surveillance requirements at the farthest reaches of the blind zone. This is a trade-off analysis that sets a limit on how much pulse compression (lengthening) can be used.
