Aerospace and Electronic Systems Magazine December 2017 - 50

Sensitivity and Linearity Requirements of an OTH Radar Receiving System

Figure 7.

Noise spectral density (Panel 1: left) and signal power (Panel 2: right)
for a single measurement run.

the discrepancies have modest impact in practical systems [2], and
relative to other uncertainties in the model.
The external signal and noise can be collected simultaneously
with the system reconnected to its antenna. To sort external noise
from external signal requires transformation to the spectral domain
where the phenomena can be seen to partially separate. An example
of such a collection is shown in Figure 7. The resultant spectrum
can be thought of as being constructed from two separate spectral
phenomena, a frequency dependent base noise envelope (red) and
a multitude of narrowband signals (HF spectrum users) that are additive to this noise. Automated algorithms can be developed to estimate the noise and large signals, relying on the fact that signals do
not occupy all of the spectrum, and the noise floor at the frequencies
where user signals exist can be estimated by interpolation between
unused channels. Due to the large dynamic range involved careful
choice of windowing function in spectral estimation is important.
The practice of separating noise and users this way, has been established in these types of system for many decades, e.g. [6].
Care must be taken to ensure the spectrum measurement system
is not corrupted by the same issues that we are seeking to quantify
in the receiver system under test. As a precursor to this analysis,
we developed a custom made instantaneously wideband spectrum
collection system to do this robustly [2]. The system relies on both
RF filtering and signal processing methods to reduce the possibility
of corruption by internal noise limiting and distortion products. The
collection system avoids some of the biggest broadcast signals by
introducing a severe high pass filter at the start of its reception chain.
In characterizing the environment for skywave OTHR there is usually no need to collect below 5 MHz, which allows heavy suppression (40 dB+ insertion loss) of strong signals found below 5 MHz.
A D-TA 3290H direct digital receiver and D-TA 1000 data recorder form the basis of the system but utilizes two important customizations. In the front end of the receiver, an adaptive equalization filter
stage provides a selectable amount of frequency dependent insertion
loss, and is dynamically controlled simultaneously with automated
flat gain levels within the RF front end. Second, a mechanism for accurately characterizing and subtracting low variance, system internal
noise power spectral density estimates from low variance external
power spectral density estimates, allows external spectral estimation
at much lower levels than would normally be possible.
48

Figure 8.

Twenty-four hours of external noise spectral density (top) and signal
power (bottom) made at a measurement point in a receiver chain.

Separate signal processing for the noise spectral density estimate (red line in Figure 7, panel 1) and signal level (Figure 6, panel
2) is done to account for the different behavior of signal processing
losses on noise and signal. The measurement system is calibrated
to an accuracy of 1 dB against COTS noise and signal generators
system claiming similar accuracy [2].

EXAMPLE
An example of 24 hours of spectral data collected at a system measurement point is shown in Figure 8. Both estimates are shown as
functions of frequency and time of day.

IEEE A&E SYSTEMS MAGAZINE

DECEMBER 2017

Table of Contents for the Digital Edition of Aerospace and Electronic Systems Magazine December 2017