Improvement in SNR and/or penetration depth can be achieved in medical ultrasoundby using long coded waveforms, in a similar manner as in radars or sonars.However, the time-bandwidth product (TB) improvement, and thereby SNRimprovement is considerably lower in medical ultrasound, due to the loweravailable bandwidth. There is still space for about 20 dB improvement in theSNR that will yield a penetration depth up to 20 cm at 5 MHz (O'Donnell,1992). The limited TB additionally yields unacceptably high range sidelobes.However, the frequency weighting from the ultrasonic transducer's bandwidth,although suboptimal can be beneficial in sidelobe reduction.The purpose of this study is an experimental evaluation of the above considerations,using artificial tissue phantoms. Specifically as a first step, the effect ofthe transducer's impulse response has been identified. Attenuation has alsobeen taken into consideration and its effect on the degradation of thecompressed signal characteristics after both matched and mismatched filteringhas been measured. Various codes that have been used successfully in radarsystems, such as chirps, m-sequences and polyphase codes have been examined.Since most of these binary codes have a larger bandwidth than the transducerin a typical medical ultrasound system can drive, a more careful code designhas been proven essential. Simulation results are also presented forcomparison.This paper presents an improved non-linear FM signal appropriatefor ultrasonic applications. The new coded waveform exhibits distinctfeatures, that make it very attractive in the implementation of codedultrasound systems. The range resolution that can be achieved is comparableto that of a conventional system, depending on the transducer's bandwidth andcan even be better for broad-band transducers. The range sidelobes, at thesame time, are well beyond the typical dynamic range of an ultrasound image.The energy of the sidelobe region is also reduced by lowering the distantsidelobes caused by the ripples of the spectrum's amplitude. The compressedsignal-to-noise ratio loss is only -...dB. The effect of frequency shift onthe matched filter due to attenuation has shown good stability of the newcode. The new code is compared with the conventional approach of a linearchirp that is frequency weighted with a Chebyshev window for sidelobereduction. Additionaly, pulse compression on the receiver does not require anyadditional weighting or filtering and can be performed by a simple correlator.