Solvent suppression using a rise-time-compensated high-pass filter

Abstract

The ideal water-suppression technique m ight be expected to provide (i) a high degree of water suppression without eliminating resonances near or beneath the solvent, (ii) a constant amp litude and linear phase response over the entire spectral width, and (iii) an uncompromised signal-to-noise (S/N) ratio. Ease of implementation is also an important consideration. Numerous solvent-suppression techniques have been proposed, and there are excellent reviews in the literature (I). The most widely used approaches are pulse techniques capable of achieving water-suppression factors of 1000 or more (2). However, these techniques sacrifice performance with regard to amp litude response, phase response, or signal strength. Hardware approaches to solvent suppression, which m ight offer improved performance in these areas, have been unable to achieve the dynamic range enhancement provided by the pulse techniques. For example, one early hardware implementation (3) used a tunable notch filter to selectively remove frequencies at or near the solvent signal prior to digitization, achieving a water-suppression factor of 25. In this work we describe preliminary results of an analogue method that suppresses the solvent signal on resonance by three orders of magn itude in a single hard-pulse acquisition. The method is easy to implement, produces no amp litude or phase distortions in the spectrum except for frequencies immediately adjacent to the solvent, and suffers no signal-to-noise penalty. It takes advantage of the fact that the large decaying dc signal produced by the solvent on resonance can be effectively attenuated by a high-pass filter. However, since the starting point of the FID in at least one of the two quadrature channels must be nonzero, the finite rise time and subsequent overshoot of the filter in response to such an impulse produce an additional signal due to ringing at the cutoff frequency. We compensate for this effect and achieve suppression factors comparable to those attainable using pulse techniques by biasing the filter on prior to acquisition, as described in more detail below. F igure 1 shows four spectra, each obtained in a single acquisition on a sample of 5 A4 H20 and 5 m M alanine in D20. Measurements were performed on a 400 MHz GE Omega using a conventional one-pulse sequence. Spectra were obtained with the water on resonance using a 5 ps pulse ( 30”), a 4 kHz sweep width, an SK block size, and a gain of 15. F igure la was acquired without the filter and is shown magn ified by