limiting the dynamic range already at the RF level. At the IF level the signal is bandlimited 
and then the signal is down-converted to video in a quadrature I/Q demodulator. The I and Q 
signals are digitized to 8 bits per channel in dual AD converters running at a sampling rate of 
100 MHz. The digitized echo is stretched to a high duty cycle signal in the data buffer, which 
holds 8192 complex samples, before the data flows into the digital preprocessor. The digital 
preprocessor performs Doppler tracking, initial motion compensation, and prefiltering of the 
data to reduce the effective pulse repetition frequency. A digital range filter that accepts 
complex data at a 100 MHz sampling rate facilitating programmable low-pass filtering and 
downsampling is presently being developed and this filter will be inserted between the AD 
converters and the buffers. Finally the signal is transferred to a HDDT system for later 
processing at the off-line facilities at the Institute. 
II THE DIGITAL SIGNAL GENERATION SYSTEM 
The purpose of coding the transmitted pulses is in short that thereby a good signal-to-noise 
ratio and a fine resolution can be achieved simultaneously. There are many different coding 
schemes of which phase and frequency coding are most relevant to radars using power 
amplifiers in their saturated power range. 
Originally it was planned to use SAW (Surface Acoustic Wave) dispersive filters for 
generation of linear FM pulses in the KRAS radar as well as for pulse compression in the 
receiver. It was soon realized that to obtain resonable flexibility a number of pulse codes 
with different bandwidths and pulselengths were required. Likewise it was found that in 
situations where low sidelobes are required non-linear FM will give better performance than 
the usual linear FM chirp. However, since the SAW filters characteristics are fixed such 
flexibility can only be obtained by using a significant number of SAW devices. By using a 
digital data generation system it was found that the required flexibility could be supported by 
a single system. 
Likewise the application of digital pulse compression in the receiver gives a number of 
advantages though at the expense of quite heavy processing requirements. The most 
important property is probably that the dynamic range measured as the ratio between the 
largest point target and the noise floor is improved by the processing gain relative to the 
dynamic range of the AD converters themselves. 
There are a number of ways to generate the digital codes and similarly there are several ways 
to transform the digital numbers to analog signals. Assume that the desired analog signal is: 
S(t) = a(t) cos(2nfrr t+ (1) 
(where a(t) will be an on-off keying and ¢(t) is the phase modulation of the signal), then this 
signal can be generated by one of the systems outlined in figure 2. 
The system shown in figure 2(A) implements a direct phase modulation on the RF carrier. 
Such a system is fairly simple and very attractive when phase codes with only few different 
phasors and moderate sidelobes are required. This coding process, however, introduces 
phase errors whose effect must be included in the design, [Iglehart,1978]. Both figure 2(B) 
and 2(C) use DA converters and low-pass filters to generate the modulation function, the 
main difference being that the one channel implementation (B) requires a digital carrier 
frequency which means that to generate signals of bandwidth B the sampling rate must be 
larger than 2 B. This implementation has been proposed for a simple pulse generator using a 
one bit DA converter [Johnston,1984]. The quadrature modulator shown in fi gure 2(C) 
generates the complex modulation function m(t) = a(t)exp{jd(t)} at baseband, and hence the 
DA converters need only a sample rate larger than B. The latter implementation is well suited 
for high bandwidth signals with low sidelobes and was therefore chosen for the KRAS 
signal generator. 
93