in gain does not change the S/N ratio of the spectrum, since Paci and co-workers (1) have recently reported that selecboth the noise and signal are equally enhanced and digitized. tive-excitation pulse sequences (2-4), which are exten-Since digitizers are preceded by audio filters, which usually sively used in the case of diamagnetic systems, can improve reduce the receiver bandwidth to the spectral region of interest, the detection of cross peaks in the case of far-shifted, fastthe desirable minimum receiver gain changes only slightly relaxing signals in paramagnetic metalloproteins. However, with the spectral width. If, for example, the gain was set at this view is inconsistent with current ''state-of-the-art'' 256 for a spectral width of 20 ppm, it could be reduced to NMR methodology and instrumentation for the study of 128 for a spectral width of 80 ppm, since the noise level paramagnetic proteins.
follows the square root of the bandwidth. Figure 1 shows the To briefly demonstrate this, I will consider the same exam-S/N ratio obtained for a broad far-shifted signal in the spectrum ple used by the authors of the above-mentioned article. Diof Cu 2 Co 2 SOD as a function of the receiver gain used. The copper, dicobalt, superoxide dismutase (Cu 2 Co 2 SOD) is a data were collected at 600 MHz on a Bruker AMX spectromeparamagnetic, dimeric protein, about 32 kDa, which is exter, using a 16 bit AD converter on an approximately 0.5 mM tremely suitable for NMR investigation (5). The NMR spec-Cu 2 Co 2 SOD sample in 90% H 2 O, 10% D 2 O. The carrier was trum shows broad, fast-relaxing signals that are shifted far positioned at the water frequency, and a spectral window of downfield (up to 70 ppm) with respect to the diamagnetic 125 kHz (dwell time of 4 ms) was used. Spectra were acquired envelope. In general, the value of T 2 dramatically influences with 128 scans, with presaturation of the water signal and a the detection of these signals, which may be barely observfast repetition rate (approximately 10 acquisitions per second). able with respect to the noise, and so optimization of S/N There is almost no improvement on increasing the receiver ratio is important when observing broad and relaxed resogain from 64 to 128, and definitely no improvement from 128 nances. The rationale behind the use of selective-excitation to 256. No dependence of the S/N ratio on the excitation pulse sequences is that the S/N ratio of the hyperfine-shifted profile was observed with offset frequencies in the range 0signals is limited by the presence of the very intense signals 40 kHz. Analogous results are obtained if the experiment is of the diamagnetic part of the protein.
performed with alternative pulse sequences. The S/N ratio in a single-transient acquisition can be im-
The data reported in Fig. 1 have been reproduced with proved by increasing the receiver gain of the spectrometer the superWEFT pulse sequence (6), which alter the relative so long as the preamplifier noise stays below the threshold intensities of signals on the basis of their relaxation rates. noise of the analog-to-digital converter (A/D). If the dy-When 20 transients per second are recorded, the receiver namic range of the digitizer is not fully used, the noise gain can be increased to 1024 without any observable signal originated by the digitizer itself will be added to the FID at distortion. An identical S/N ratio for hyperfine-shifted resoa relatively high level, resulting in a poor dynamic range. If nances is observed for receiver gain values ranging from two largely unequal noise levels are added, the resulting 128 to 1024. The experiment was also repeated using an offnoise will be only marginally higher than the larger of the resonance DANTE sequence for solvent suppression (7, 8). two, while the summation of two equivalent noise levels With this sequence, carrier position can be placed far from results in a 3 dB increase. By increase of the receiver gain, the solvent signal and a smaller spectral window can be the S/N ratio of the spectrum will improve until the receiver used. When the carrier was placed at about 31 ppm and a gain is such that the preamplifier noise matches or surpasses spectral window of 54 kHz (dwell time 9.2S ms) was used, similar data to those reported in Fig. 1 were obtained. the A/D's noise level.