Home Search Collections Journals About Contact us My IOPscience

A sensitive nuclear quadrupole resonance spectrometer for 2-60 MHz

This article has been downloaded from IOPscience. Please scroll down to see the full text article.

1982 J. Phys. E: Sci. Instrum. 15 814

(http://iopscience.iop.org/0022-3735/15/8/005)

View the table of contents for this issue, or go to the journal homepage for more

Download details:
IP Address: 132.198.151.69
The article was downloaded on 10/09/2010 at 18:34

Please note that terms and conditions apply.

J . Phys. E: Sci. Instrum., Vol. 15, 1982. Printed in Great Britain

RESEARCH PAPERS

A sensitive nuclear quadrupole resonance characteristics of particular devices. On the other hand, unlike
the super-regenerative and marginal oscillators it does not
spectrometer for 2-60 MHz
display an intrinsic amplification of both the signal and the noise
arising from the way the amplitude of oscillation is determined
F N H Robinson by the nature of the feedback loop. As a result it also requires a
Clarendon Laboratory, University of Oxford, Parks Road. low-noise detector. This is. however. a relatively straightforward
Oxford OX1 3PU design problem and so we have based our instrument on a
limited oscillator circuit.
Received 11 January 1982. in final form 15 March 1982 Frequency modulation presents insuperable problems of
base-line drift, and spurious signals. in wide tuning-range circuits
Abstract. The paper describes the design, construction and which are needed to search for new resonances. Provided that
performance of a continuous-wave spectrometer for use at ferromagnetic contamination is rigorously excluded from the
temperatures between 77 K and 80°C. The limiting sensitivity tank circuit the bisymmetric Zeeman modulation scheme with
for chlorine resonances with narrow lines is 1 pmol and second-harmonic detection of Verwieck and Cornwell (1 96 1) is
compounds of molar volume up to lo4 cm3 can be studied. The free from this objection (Robinson 1980a). With a cw oscillator
corresponding figures for nitrogen resonances are 300 prnol it has the further advantage that it gives very clear spectra with
and IO3 cm3. The continuous tuning range with a single coil is the chart-recorder trace representing the true absorption line
2.5 : 1 and the resolution is a few PPM. It uses bisymmetric rather than its derivative, even though the lines may
Zeeman modulation, there are no critical adjustments and no occasionally display inverted wings whose origin has been
elaborate extra equipment is required. discussed by Watkins and Pound (1952). The modulator we use
has been described elsewhere (Robinson 1980b).
The oscillator circuit uses a long-tailed pair of bipolar
1. Introduction transistors (Faulkner and Holman 1967) as the limiter, and
Although Fitzky (1974) has reported a signal-to-noise ratio of these transistors also serve as a moderately efficient detector
5 :1 obtained from 200 pmol of chlorine using a 10 s phase- (Robinson 1974); however, in order to avoid loading the tuned
sensitive detector (PSD) time constant, nuclear quadrupole circuit and degrading its Q the limiter is preceded by a broad-
resonance (NQR) spectrometry is reputed to be a relatively band low-noise amplifier with a FET input stage. The gain in this
insensitive technique requiring large samples. Yet Pound (1 950) amplifier minimises the influence of the moderate detector
in his pioneer paper gave a theoretical estimate of the s/N ratio efficiency on the overall noise temperature and, at 30 MHz, the
which would suggest that sharp chlorine resonances ought to be electronic contribution to this temperature is only about 200 K.
observable with only one or two pmol. W e shall describe an It increases at higher frequencies but falls to negligible values at
instrument which goes some way to closing the gap between the lower frequencies where, of course. sensitivity is most important.
best current practice and theoretical expectation. The presence of this amplifier with a small, mainly
For a given resonance the sensitivity of a spectrometer capacitative, input admittance allows the tank circuit to be very
depends on the Q of the coil and its filling factor and on the loosely coupled. through a few picofarads, to the active
noise temperature of the oscillator and its capacity to drive electronics. As a result, very little RF current flows in the line
enough RF voltage across the coil to approach saturation. These from the tank circuit to the electronics and if the main tuning
last two requirements conflict since low-noise circuits will not capacitor is placed near the coil in the cryostat, loss in this line
usually handle high RF voltages. has little effect on the Q nor does its stored magnetic energy
The super-regenerative oscillator gives a large voltage but, reduce the effective sample filling factor. Both these factors
although it has some advantages in studying broad lines, it is improve the sensitivity.
difficult to see how a circuit operating in such a highly nonlinear The amplifier restricts the RF voltage that can be handled
mode can yield a low noise temperature. A marginal oscillator without overloading, especially at high frequencies, but
gives a low noise temperature if the loop gain is near unity and nevertheless one can manage, without increasing the noise
the feedback circuit is very nearly linear with a gain that falls figure. amplitudes from 0.05 up to 2 V across the coil, even at
slowly and monotonically with increasing amplitude. This is 50 MHz, and this is sufficient to saturate most lines at 77 K, if
difficult to achieve over a wide range of levels and frequencies not at room temperature. It is easy to increase the voltage at low
without continual critical adjustment. The limited oscillator frequencies but to do so at high frequencies would be more
(Robinson 1959) has the same low noise temperature as a difficult.
properly adjusted marginal oscillator but does not need The tuning capacitor is a modified 1 5 0 p F air-spaced
continual adjustment nor does it depend critically on the exact trimmer and it is driven by a shaft passing through a simple

0022-3735/82/080814 + 10 $02.00 0 1982 The Institute of Physics

A sensitice nuclear quadrupole resonance spectrometer

greased 0 ring seal at the head of the cryostat. Figure I shows associated with the RF head. a small variable capacitor on the
the complete spectrometer with the worm-gear capacitor drive in PCR which sets the RF level. Although no attempt has been made
place and a modulating coil attached to the tail. The top plate is to stabilise the RF level. which therefore rises rather faster than
maintained at 30 f 5 OC by a small thermostat. This prevents linearly with increasing frequency. the level control can. at a
condensation on the RF lead where it passes through the top slight loss in sensitivity near the ends of the scan. be left
plate using a small press-fit PTFE leadthrough. There is also a unadjusted over an octave scan (see figure 4) but it is preferable
lead to a diode thermometer. attached to the tank-circuit to reset it four or five times during a complete scan which
assembly. and a knob to operate a switch that inserts extra usually means only once or twice a day. It is. in any case.
capacitance and extends the tuning range to 4 : I . The small desirable to have a level which rises linearly with frequency. as
diecast box ( 1 15 x 60 x 30 mm) houses the printed circuit board this maintains a constant saturation parameter.
(PCR) on which is mounted the oscillator circuit together with R F Some of the results which can be obtained with this
and LF buffer amplifiers and power lead filters. A coaxial cable instrument are now presented and where numerical values for
leads directly from this box to a frequency counter and a 4-core the s / N ratio are mentioned this will be in terms of the ratio of
screened cable leads to the power supply and eventually to the the peak signal height to a visual estimate of the RMS noise. since
PSD. This cable also contains a lead which controls a varicap this is the ratio predicted by theory. The noise estimate is to
diode that can be used to give very slow band-spread tuning some extent subjective and so the accuracy of these figures is
over a limited range. The DC voltage on the signal lead responds probably worse than +25%.
to the RF level and is used to operate a meter that can be
calibrated by connecting a signal generator across the tank 2. Experimental spectra
circuit. The overall length of the cryostat. from the tail to the top Figure 2 shows the complex spectrum due to "Cl in carbon
plate. is 280 mm, the tail diameter is 25 mm and the main tube tetrachloride at 77 K. It illustrates the clarity and resolution of
diameter 50 mm. the spectra that can be obtained from samples approaching
The ancillary equipment required comprises the modulator. conventional size. in this case 0.6 cm3. or 1 g. of material. The
a PSD with a second-harmonic reference. a chart recorder. a PSD time constant is 1 0 s and the frequency markers are at
frequency counter. a stabilised PSU giving 2 2 7 V at 30 mA and I O kHz intervals with wider markers every 100 kHz. Only the
a motor to drive the capacitor. with a range of speeds from l O - 3 resolution of the chart recorder prevented the use of I kHz or
to I O RPM. Apart from tuning there is only one other control finer markers.
Figure 3 shows lines obtained from a composite sample at
77 K containing 30pmol each of sodium chlorate (3 mg) and
potassium chlorate (3.5 mg) in a small coil 8 mm long wound on
a 2 mm diameter thin-walled PTFE tube. The upper traces show

Figure 2. The spectrum of CC14 at 77 K near 40 MHz.

Figure 3. Lines due to "Cl and "CI in 30pmol of sodium and

Figure 1. A general view of the spectrometer. potassium chlorate.

815
F N H Robinson

the 35Cllines with a 1 s PSD time constant and the lower traces
show the 37Cllines with a 5 0 s time constant. Since the 37Cllines
are six times less intense than the 35Cllines and the ratio of the
square roots of the bandwidths is 7 : 1: we might have expected
similar signal-to-noise ratios. In fact. the 35Cl lines give 1 0 : 1
while the 37Clgive only 5 :l. The extra noise with the long time
constant probably shows that we are approaching a limit set by
the overall stability of the system. With a 10 s time constant the
s/N ratio for 35Cl is, as expected, 30:l and corresponds to a
limiting sensitivity of 1 pmol. The accessible sample volume
within the coil former is 14 mm3 and so we see that we could
expect useful signals from as little as 10 ,umol of a compound of A B c
molar volume up to 1000 cm3 (for purposes of comparison the
molar volume of NaClO3 is 40 cm3).
Figure 4 shows some of the lines obtained at 77 K with a 1 s
time constant during a continuous scan from 55 to 27 M H z
I !
made without adjusting the RF level. The self-supporting coil has
an internal diameter of 10 mm and a length of 24 mm which
gives a usable sample volume of over 1.5 cm3 and it easily
accommodates four separate specimen tubes containing:
(i), 71 mg (530pmol) of N-chlorosuccinimide (NCS);(ii), 62 mg D E F 5
(440 pmol of chlorine per resolved line) of hexachlorobenzene
Figure 4. Lines obtained in a continuous scan from 55 to
(HCB);(iii), 27 mg (370 ,umol of chlorine) of p-dichlorobenzene
27 MHz with a composite sample.
(PDCB);and (iv) 39 mg (370pmol) of sodium chlorate.
Line A, at 54.1 MHz is due to 35Cl and line B at 42.6 MHz
to 37Clin NCS. The three lines C near 38.4 M H z are due to “C1
in HCB. Lines D and F at 34.8 and 27.4 MHz are due to 35Cland
37C1in PDCB and line E is the 30.6 MHz 35CI line in sodium
chlorate. Line G shows the improvement in line F that can be
effected by re-adjusting the level at the end of the scan. From
line A. at 54.1 M H z to line F a t 27.4 MHz: the baseline drifts by
about twice the height of the strong signals. which is easily
within the range of the chart recorder.
The signal-to-noise ratio for line D is 15 :1 from 370pmol
of chlorine so that even with this short time constant the limiting
sensitivity is 25 ,umol and useful signals would be available from
compounds of molar volume up to lo4 em3, say 500 carbon
atoms per chlorine. We have not been able to check this directly
but we have been able to study the phase change in cholesteryl
chloride (C27H45C1) between room temperature and 77 K. This
is a very light fluffy wax with an effective density of less than
0.25 g cm-3 giving a molar volume over 1600 cm3.
Furthermore the single broad line at 40 OC splits into four lines
4-
each 8 kHz wide at 77 K so that we can certainly detect a rather in
0 0
0

broad line in a compound whose molar volume per resolved line

is 6400 cm?.
Although the signals shown in figure 4 obtained with a Figure 5. The spectrum of hexachlorobenzene at high
relatively fast sweep are enough to indicate the presence and resolution.
frequencies of the lines, they are not clear enough to show their
width and possible structure. Figure 5 therefore shows the
improvement that can be made by using a longer time constant.
50 s, and a slower scan. The three lines from HCB can now be volume of 1.5 cm3 by no means filled by the H M r . The signal-to-
located and their widths measured to better than 1 kHz. noise ratio is 18 : 1 and we have also observed the lines at 2.438
Although the spectrometer noise temperature and sensitivity and 2.9125 MHz in 20mmol of urea, which has an
deteriorate above 45 MHz. it is still usable up to 130 MHz and, asymmetric field gradient, with a signal-to-noise ratio of 12: 1.
with the aid of switched capacitance. we have (Weaver and The HMT result corresponds to a limiting sensitivity of about
Robinson 1981) been able to observe all four lines, from 30.6 to 1 mmol of nitrogen with a 1 s time constant so that this
119 MHz in the spectrum of *09Biat 77 K in triphenylbismuth would give a useful line with a 10 s time constant and one might
without removing the sample or disturbing its temperature. hope to detect nitrogen lines in compounds of molar volume up
By changing a few passive components in the circuit, the to lo3 cm3. Since nitrogen lines are easily saturated, it might be
basic oscillator which covers 15-60 MHz can be arranged to possible to improve the latter estimate by using a larger sample
operate down to below 2 MHz and this allows the observation of and coil. The direct resonance technique. though simpler and
I4N lines which are of course much weaker than 35CI lines. more flexible does not. of course, compare in sensitivity for low-
Figure 6 obtained with a 1 s time constant shows the frequency lines with the double resonance techniques developed
3.407 MHz line obtained at 77 K from 715 mg (20 mmol of N) by Edmonds and his colleagues (Edmonds et a1 1974).
of hexamethylene tetrammine (HMT) in a coil wound with fine The high sensitivities that can be achieved mean that in
wire on a 10 mm former 25 mm long which gives a sample many cases only very small samples will be required and this

816
A sensitiue nuclear quadrupole resonance spectrometer

less and cannot exceed

s
_-- ,U0 + m ) ( ~ m- + 1)
g2p2hv312(I
N
-
3z
a1Vo
( kTAv(2I+ 1)

These equations segregate, in separate brackets, terms which

refer to the sample, the coil and the electronic circuit. The coil is
described by Q and U and since Q is proportional to the coil
diameter d while U varies as d 3 it is clear that when saturation is
possible the signal for a fixed number NO of nuclei varies as l j d .
When saturation is not possible it varies as d-’/* so that in either
case the coil should be as small as possible. When, however, the
nuclear density No/U is fixed, equation (1) gives a signal
proportional to d 2 and now as much material as possible should
be used in a coil just large enough to accommodate it. If this
leads to a size where, according to equation (2). saturation is no
longer possible the signal will finally increase only as d’j2 so that
to double the signal would then require 64 times as much
material. It is also clear that in every case the sample volume
should be as nearly equal to U as possible, i.e. a high filling
factor is desirable and designs which sacrifice filling factor to
w w w other considerations will always lead to lower sensitivity. The
f
0
.P
0
W
\o circuit described here, by placing the coil near the tuning
0 0
capacitor, leads to a high filling factor since there is little
magnetic energy stored in the wire from the tank circuit to the
Figure 6. The nitrogen line in hexamethylene tetrammine. active electronics.
In equation ( 3 ) the circuit is characterised by V:/T,B and
ideally we desire a value of V I which will saturate any
resonance. In our circuit V I is restricted to less than 2 V which
extends the range of substances that can be studied and also saturates almost any line in a coil of a few cm3 volume at 77 K if
means that only small single crystals will be needed for Zeeman not at room temperature. One could increase V1 without
studies. It also allows one, when searching for an unknown line, degrading T,, a t lower frequencies and this might possibly be
to incorporate a small sample of a standard material to act as a worthwhile. However, sensitivity is most important when
check on the spectrometer performance and as an aid to searching for a new line and this is best done ,at 77 K to take
adjustment. If the standard sample frequency also has a well advantage of the increased Boltzmann factor so that henceforth
known dependence on temperature it can at the same time be we shall base our discussion on equation (1) rather than
used as a thermometer in direct contact with the main sample. equation (3).
The bandwidth B is related (for a 6 dB roll-off3 to the PSD
time constant z by B= 1/4r and so we can reduce B, at the
3. Theoretical sensitivity expense of taking longer to make a frequency scan, by
If. in a coil of quality factor Q and effective volume U, we place increasing z. The limit is set by the overall stability of the circuit
a sample at a temperature T containing A$ atoms giving a which. in our case, corresponds to 5 - 100 s. This leaves only T,
resonance of width AV at a frequency v with a relaxation time to characterise the circuit and so, once the coil has been
T I , due to nuclei of isotopic abundance CL. spin I and optimised. the over-riding consideration is a low noise
gyromagnetic ratio g , making transitions between the sublevels temperature. T o see how much room is left for improvement we
+m and +(m- 1) and if we use a square-wave modulation
-
consider three examples.
scheme and a detector of bandwidth B and noise temperature Line D in figure 4 is due to 370pnol of chlorine at 77 K
T,. then the signal-to-noise ratio predicted by Pound (1 950) is, giving a line width of 1.5 kHz at 35 M H z with a relaxation
in SI units. time T I = 1 s. It was observed with z= 1 s and B=0.25 H z
using a coil of measured Q=350 and an effective volume
g/3hv’12(I + m)ll2(Z- m + l)’I2 U=3 x m3 ( 3 cm3). Equation (1) with x=i,g=0.537
and I = m = 9 gives
N 6n k T ( T i A ~ ) ” ~ ( 2 11)+
1.1 x 10-18iVo/T~’2
and since No = 223 x lo’* and experimentally s/N= 18, this gives
T , -270 K. Since 77 K arises from the tank circuit the
electronic circuit contributes about 200 K, not far from the
where p=5.05 x J T-’ is the nuclear magneton, h is
estimate obtained in 0 4 from the circuit parameters. Certainly
Planck‘s constant and k is Boltzmann’s constant. To achieve
the electronic contribution is somewhere between 100 and
this limiting sensitivity the RF voltage V I across the tank circuit
300 K and, even were we to eliminate it entirely. ~ 7 ewould not
must satisfy
increase s/N by more than about 2 : 1.
Uh2Av The small coil used to obtain the results shown in figure 3
cv: has about f the linear dimensions of the coil discussed above.
W e therefore expect the limiting sensitivity to be 5 pmol rather
where C is the tuning capacitance, otherwise the sensitivity is than 25 ,umol. The measured value is better. 3 pmol. There are

817
F N H Robinson

two reasons; first the sample is different, sodium chlorate rather To ensure this the transistors are chosen so that their base-
than p dichlorobenzene, and second the level was optimised to emitter voltages at the same current are matched to within a few
obtain figure 4 whereas figure 3 was obtained with the level set millivolts and. to ensure monotonic detection at low levels, the
when the frequency was 55 MHz. This result is again consistent transistor requiring the higher forward bias is used as T3.
with an electronic noise temperature contribution between 100 At low levels the capacitative input impedance of T3 has
and 300 K. little effect on the pre-amplifier nor does it much affect the
The nitrogen line in figure 6, with a signal-to-noise ratio of behaviour on the negative half-cycle. At higher levels, however,
18 :1 at 77 K, was obtained. using a 1 s time constant and a coil on the positive half-cycle, although the base-to-emitter
with Q = 200 and U= 2 x m3, from 1.23 x lo2’ nuclei with admittance is reduced by the negative feedback due to R3 (when
g=0.403, I = m = 1, which give a line of width 900 Hz at T 4 is off) it is still necessary to charge the collector-base
3.4 MHz and relaxation time TI = 11 s. Equation (1) gives capacitance of T3. Now the long-tailed pair transistors are
sjN= 1.28 x 1 0 - 2 0 N ~ j T ~and -
’ 2 so we obtain T, 7 7 K . It now
looks as though most of the noise comes from the tank circuit
subject to two stringent requirements, they must have good high-
frequency properties and also very low l/f noise since this
and indeed at this frequency we expect the electronic appears in the detected output of T J . The only satisfactory
contribution to be small, determined mainly by the FET input device that we have found is the PNP transistor type BCY71
stage of the amplifier and certainly much less than 100 K. which combines a highf, > 300 MHz with very low Ijfnoise. It
It therefore appears that at 34 MHz there is room for a has, unfortunately, about 6 p F collector-base capacitance and,
modest improvement but that at 3.4 MHz we are very nearly up at 50 MHz, this has a reactance of only 500 R . Thus to achieve
against the theoretical limit. The most profitable avenue to a drive of 2 V or so to the limiter, T2 must be able to swing over
explore would be t o increase Vl so that larger samples could be 4 mA and so must run with a standing current of about 8 mA.
used at room temperature. This problem, which limits the high-frequency performance, is,
of course, quite insignificant below 20 MHz.
4. Circuit design The typical shunt impedance Z of a tank circuit exceeds
Figure 7 shows the basic circuit less a few of the bias lo4 L2 which is much higher than the source impedance R , that
components. Transistors 1 and 2 form a broadband mismatched optimises the noise temperature of the amplifier and so the two
pair amplifier (Cherry and Hooper 1963) and, if gl is the mutual input capacitors Ci and Ci (mainly the input capacitance of T I )
conductance of T I , the gain is A =glR2 with a high-frequency can be used to transform this impedance and the RF voltage to a
roll-off mainly determined by the collector-base capacitance of lower value. With a step down ratio of 4 : 1 the input to TI is
T2, which is in parallel with R2. This amplifier has a low output only 0.5 V when the RF voltage V1 across the tank circuit is 2 V
impedance and drives the long-tailed pair T3. T4. If the RF input and it is this feature that allows us to achieve a reasonable tank
to the base of T3 has an amplitude exceeding about 100 mV the circuit voltage without degrading the noise performance. In
current through T4 is switched on and off. A fraction practice Ci is about 2 p F and if the variable capacitor C2 which
(determined by C1 and C2) of this clipped current is fed back to sets the RF level has a maximum value of 60 p F the capacitor C1
the tank circuit and sustains oscillations whose amplitude is need only be 1 pF so that the entire circuit is very loosely
linearly proportional to the Q of the circuit and so responds coupled to the tank circuit with the advantages that we have
directly to changes in Q due to NQR absorption. On the negative mentioned in 0 1.
half-cycle T3 is switched off, but on the positive half-cycle the Apart from Johnson noise in the tuned circuit the principal
current through T3 continues to rise at a rate governed by R 3 , sources of noise are RF noise from the amplifier, low-frequency
even after T4 is off. Thus the collector current of T3 contains a noise in the detector and RF noise fed back to the tank circuit
component which is a detected signal. and a RF input of from the collector of T4. Provided that glR1 exceeds about 3:
amplitude G I produces the same LF output as a low-frequency the two significant sources of amplifier noise are those
input U I / Z . T o obtain stable oscillations it is essential that the associated with T I and with the base current of Tz, Noise in T I
transconductance from the base of T3 to the drain of T4 falls can be represented in terms of an equivalent noise resistance
monotonically with increasing input and this requires that the somewhat less than 2/3gl and an equivalent noise conductance
ratio of the standing currents in T! and Td is between 3 and 0.3. somewhat less than W2c2/3gl where C is now the input
capacitance of TI. Base current noise from T2 increases the
noise resistance by g2/2/3g:? where gz is the mutual conductance
and /3 the DC current gain of T2. The detector behaves as though
it had an equivalent noise resistance R &= n2R&where RL is the
low-frequency noise resistance of the long-tailed pair. This is
approximately equal to R3 + R,, where R n is the noise resistance
of T3 at the detection frequency. The effects of llf noise are
minimised because the base of Td is grounded and the base of T!
returned to ground through a RF choke. For the transistor type
BCY71 at 100 Hz R, = 250 R over quite a wide range of emitter
currents. Detector noise appears as an addition to the equivalent
noise resistance which is therefore

but the noise conductance is unmodified

b - v e
(5)
and these are associated with the usual reference temperature
Figure 7. The basic circuit. To = 293 K.
The main source of noise in the collector current of Td is
Johnson noise from the ’tail’ resistance R3 and since this is only

818
A sensitive nuclear quadrupole resonance spectrometer

connected half the time, the noise voltage developed across the measured noise temperature approaches the temperature of the
tank circuit is tank circuit. As Z increases we can use B larger step-down ratio
at the input without seriously increasing the noise, especially
Lc
2 (+
2 k T 0 d c ZCl I2- ) i l R 3 , if the coil is at room temperature, and thus we can increase
the RF level if necessary. Consider a 25 mm diameter coil at
3.4 MHz and at room temperature. W e expect Q- 200 and.
If IO is the standing current in each transistor and VI the RF
if this is tuned with a 100 p F capacitor, the shunt impedance is
voltage across the tank circuit, we have
2- 9 x lo4 0 . The noise resistance of the amplifier-detector will
be about 200 ,Q and the noise conductance quite negligible. Thus
we could step 2 down by 300: 1 without much degrading the
noise temperature and this would step the RF voltage down by
and so the feedback noise contributes 17 : 1 allowing the use of levels up to about 8 V. Although there
would then only be 0.5 V RF at the input to T I , this would
overload the detector T3 and it would be necessary to reduce the
gain by reducing R2.
to the overall noise temperature of the system. We have so far assumed that the amplifier is not being
The contribution from the amplifier is overloaded but, although it is justifiable to regard it as fairly
linear because of the negative feedback round Tz,in fact both T I
Ta=T~[(rn/Rs)+Rsgn]
(a quadratic device) and T2 (an exponential device), are being
and thus takes a minimum value driven about half-way along their complete characteristics and
are therefore appreciably non-linear at high levels. As a result
1,”noise generated by TI and especially by Tz can modulate the
if R , is chosen to be RF carrier and seriously degrade the noise performance. To
avoid this these transistors are biased using. large resistors (Rs
R , = (rn,’gn)1!2
and R 7 in the complete circuit shown in figure 8) in their source
The minimum is. however, very flat and an error 2 : 1 in R , only and emitter leads and also a -27 V negative supply rail. These
increases T, by 25?& We see that the minimum value of T , due bias resistors are only bypassed by small capacitors. of
to the first stage alone is negligible reactance at RF but high reactance near 100 Hz and
so both T I and Tz have considerable negative feedback (26 dB
for T I and 60 dB for T2) at low frequencies. This reduces the
low-frequency noise voltage between the gate and source, or
base and emitter, and so minimises cross modulation.
and that this is modified. for the complete amplifier and detector. The Zeeman modulating field, especially if it is coaxial with
to the RF coil, can generate voltages in this coil at odd harmonics of
the modulation frequency of the order of several mV. This pick
up, though it is ignored by the PSD, can easily overload the
circuit. It is therefore eliminated, see figure 8$ by placing a RF
which increases with R3 whereas the feedback noise ( 6 ) choke from the gate of T I to ground. If Ci = 3 p F and the choke
decreases with R 3 . There is therefore an optimum value of R3. inductance is as large as 2 m H the pick up is reduced by
Rather than seek this rather flat optimum algebraically we approximately l o w 8 .This choke and also the choke used at the
consider a numerical example. input to the limiter can themselves pick up the modulation. They
For the first field effect transistor 2-1 we use the UHF type therefore take the form of two very small ferrite-cored chokes in
BF256LB with g l - 4 m A V - ’ and C-3 pF. The second series, mounted in the PCB as an astatic pair.
transistor is type MPSH 1 1 which has a high f,(650 MHz), a The signal and noise level at the output of the detector
very low collector-base capacitance (less than 1 pF) and a transistor T3 is relatively large, the noise is about 50 nV per root
reasonably high DC current gain p > 50 and it is run at 8 mA so H z , and so the noise figure of the LF amplifier is not critical;
-
that g2 320 mA V-I. The value of R2 in the high-frequency
circuit is l o 3 ,Q so that A = 4 and, with R , = 2 5 0 and
however: by including this amplifier, with a voltage gain of about
100, in the RF box the signal is raised to a level where pick up in
R3 = 500 62, we obtain r, =830 62 of which more than half the output lead can be ignored as well. A buffer RF amplifier is
comes from the detector. For g, at 3 0 M H z we have also used to isolate the circuit from pick up on the lead to the
2.7 x mA V-’ and so the optimum source resistance is frequency counter.
-
R , 5.6 x l o 3 a.A typical value for 2 at this frequency is
35 x IO3 0 so that the input step down ratio should be about
Finally the outputs of integrated circuit power supply
stabilisers usually exhibit several hundred microvolts of noise
2.5 : 1 , although 4 : 1, by allowing a higher RF voltage with little and also occasional spikes due to transients on the mains. These
change in T,. might be better. The optimum value of T, is 90 K. can be partially eliminated by simple RC filtering outside the RF
With V1 = 2 V, IO= I mA. R3 =500 Cl and Z = 3 5 x lo3 we box but in practice we have found it most satisfactory to place a
obtain Tr 40 K giving an overall electronic contribution T, = further transistorised filter in the box itself. With these
-
T, + T f 130 K. This, with the tank circuit at 77 K, gives a total
noise temperature of about 200 K which is sufficiently close to
precautions there is very little trouble from extraneous noise
though, to be on the safe side? we use a basic modulation
our estimate 270 K, from the p-dichlorobenzene signal in figure frequency of 87Hz so that the detected second harmonic at
4, to give us confidence that our analysis is basically correct. 174 Hz is as far away as possible from harmonics of the 50 Hz
A t lower frequencies we can increase R2 and the amplifier mains.
gain, and decrease the current drawn by transistor 2. This
virtually eliminates all the amplifier and detector noise except
that due to transistor 1 and this is, in any case, small. Since the 5. Practical circuit
tank circuit impedance 2 is also increased, which decreases the Figure 8 shows the complete circuit including the varicap
already small term Tr. it is not surprising that at 3 MHz the (BB105)bandspread circuit, the RF buffer amplifier Ts ( B F X 8 9 ) .

819
F N H Robinson

Figure 8. The complete circuit.

the supply line filters Tg (BC 182L) and Tp (BC212L), and the LF wire sleeved with PTFE tubing, C3 is 4.7 n F in parallel with
buffer amplifier Tg (BCY71) and T7 (BC182L). The values of 68 pF. C4 is 2 x 4.7 n F (4.7 nF + 47 nF) in parallel with 68 p F
the unmarked components (with the LF version values in and CZ is a Mullard type 800 miniature trimmer. All the
brackets) are: L1 = L z , an astatic pair of 3 3 p H (1 mH) sub- resistors are miniature metal film (Mullard type MR25), most of
miniature RF chokes, Sigma type SC 3 0 ; R I = 680 R (1 kR), the capacitors up to 4.7 n F are Mullard miniature plate ceramics
R I= 1 kR (4.7 kR), R J~ 4 7 R, 0 R4 = 8.2 kR. Rs = 15 kR. though the 1 n F capacitors coupling Tz to T3 and Ts to the RF
Rrj ~ 4 . kQ,
7 Ry = 3.3 kR (10 kR), Re = 3 3 kR (100 kR), Rs = output lead are polystyrene. The 47 n and 100 n capacitors are
3.3 k R (10 kQ), Rio = 15 kR (22 kR), R I ]= 2 2 kR ( 3 3 k 0 ) . CI polyester and the electrolytic capacitors are miniature PCB
and C ; are fabricated by wrapping 0.25 mm wire over a 2 mm mounting types, rated at 3 5 V.

Figure 9. Printed circuit board layout.

820
A sensitive nuclear quadrupole resonance spectrometer

The values of the bias and load resistors associated with TI

and TI are chosen so that the collector of T2 is at about + 7 V
and the drain of T I at + 15 V and since the mutual conductance
of T I (BF256LB) is 4 mA V - ' the gain of the HF amplifier is 4
and in the LF version 20. With R3 = 4 7 0 R the bias required by
T3 and T4 to give 0.8 mA each is about + 1.4 V and this is
obtained by driving 5 mA forward current through two small
diodes. 1S44. in series. The load R s is chosen to leave the
collector of Tq at about -12 V and the 8.2 kR load R4 causes
the collector of T3 to rise from -18 at zero RF level to about
-8 V at maximum level. The buffer amplifier T6T7has a gain of
100 above 100 Hz and unity gain for DC. Thus the DC level of
the signal output rises from about -16 V to -6 V as the RF level
increases. A 100pA FSD meter with a total series resistance of
100 kR, including the potentiometer used to back off the zero
level. is used to monitor this voltage and the resistance is split
into equal parts with a capacitor (IOpF) to ground at the centre
point. to prevent interference picked up by the meter from
entering the signal input to the PSD.
The whole circuit is built on a glass fibre PCB (89 mm x
58 mm) and since the layout of the oscillator components is
critical a diagram of this board is shown in figure 9. The view is
from the copper-clad side and the components (transistors.
resistors, capacitors etc) on the other side of the board are
shown as though it were transparent. Dots mark the holes where
the components are attached. crosses mark circuit pins to which
CI. C ; and the output leads are attached. The copper ground
plane is shown hatched. a link below the copper side joins bb
(the base lead of T3) and two links behind the board aa and cc
complete the positive and negative supply rail wiring.
Electrolytic capacitors and the transistors are ringed. Two A''
square aluminium alloy bars are bolted by three screws across
the board at AA' and BB'. These and a A'' post at C support the
PCB from the bottom of the diecast box. The bar BB' separates
the oscillator from the rest of the circuit since above the PCB it is
only traversed by resistors (shown asterisked in figure 8). Two
slots are machined in the bottom of the bar AA'. where it bolts
to the box. to allow return currents from the tank circuit a direct
access to the copper ground plane of the PCB. Figure IO shows
the PCB mounted in the box. The varicap circuit is omitted but it
bolts to the end of the box near where the RF lead from the tank
coil enters the box.
Figure 11. The tuning capacitor.

6. The tuning capacitor

Making a tuning capacitor which would run freely at 77 K
without contact noise and which had a large ratio of maximum
to minimum capacitance proved to be the most difficult task in
the whole design. The solution is shown in figure 11. The
capacitor is constructed using the rotor and stator from a
Jackson Bros. air-spaced 150 pF trimmer type C8. The stator is
grounded to the support frame and the rotor runs between two
ball bearings. with 4'' nylon balls in PTFE races, which are
lightly loaded by a phosphor-bronze leaf-spring. A 3 mm hole is
bored 25 mm deep into the end of the rotor shaft remote from
the drive and into this hole is inserted a thin (1.5 mm x 0.1 mm)
copper tape sheathed in a thin-walled PTFE tube. The tape is
soldered to the rotor through a small side hole at the bottom of
the bore. and the other end is fixed to an insulating support to
form the 'live' lead to the coil. With stops to restrict the rotor
movement to 180° the tape is never twisted by more than 90°
and. since it is only 1.5 mm wide and 30 mm long, its life is
Figure 10. The circuit in place in its case. indefinite.
The stator is fixed to the end plates by brass grub screws.
and this allows its position to be adjusted after the rotor

82 1
F N H Robinson

assembly has been put together. The drive shaft of the rotor is leaky seal the interior of the cryostat is maintained at a slight
coupled through a short length of nylon rod and a Heli-Cal (R) overpressure of helium. Air is not suitable and even with helium
universal joint to a longer length of nylon rod which in turn is it is possible to cause a measurable change in the oscillator
coupled by a universal joint to the final drive shaft from the frequency by changing the pressure. despite helium's very low
worm gear. This gives a structure that maintains its alignment dielectric susceptibility.
and runs smoothly over a wide temperature range, and no jitter Although our modulator will provide up to 2 A at 30 V with
due to either contact noise or stiffness in the bearings is observed a variable frequency we rarely use more than 10% of this output
as the capacitor is tuned. The capacitor frame is supported from and since there appears to be little change in the signal-to-noise
the cryostat top plate by a 15 mm stainless steel tube concentric ratio with modulation frequency this is usually left set at 87 Hz.
with the drive shaft, and at the bottom of the frame a circular We have also found very little difference in the spectra obtained
plate (not shown) carries the thin-walled copper tube containing with transverse fields produced by a coil, such as that shown in
the sample coil. figure 1, and axial coils. The latter are more compact and have
A cover, with spring fingers, fits over the capacitor and gives less ohmic loss than the transverse coils and if a 1 mm radial
good thermal contact with the cryostat walls as well as a solid clearance is left between the tail of the cryostat and the coil
earth return contact. A single silver-plated stainless steel wire former, the coolant liquid can circulate freely over the tail. It is
leads from the live terminal of the capacitor to the RF not necessary to obtain a uniform modulating field, indeed. a
feedthrough bush in the top plate. non-uniform field reduces the spurious inverted wings on the
lines so that we frequently place the coil so that the specimen is
7. Adjustment only in the fringing field at the end of the coil.
The capacitor C[ is first chosen roughly to match the expected
shunt impedance of the coil to the amplifier, to give the
minimum noise figure, The input capacitance of the amplifier, 8. Conclusion
including strays, is about 5 p F so that CI = 2 p F gives a 2 : 7 We have shown that it is possible to construct a cw
step down ratio, It is better to make C[ smaller rather than spectrometer which. despite its rather limited RF level, gives high
larger than the first rough estimate since this allows the RF level sensitivity, both in terms of sample size and acceptable molar
to be higher without much degradation of the noise figure. In volume, by virtue of its low noise temperature. Temperatures
practice the value of Cj' is not very critical. A signal generator from 77 K to 8OoC are accessible with a simple cryostat and
can then be connected between the tank circuit terminal and there is no reason why. with a more elaborate cryostat. one
ground, to calibrate the level meter. The feedback capacitor C I should not reach lower temperatures. Above room temperature
is chosen next, so that with C2 at its maximum value the lowest there is little advantage in placing the capacitor near the coil
level desired at the highest frequency can be obtained. The since this will not have a very high Q and the line from the coil
circuit is then ready for use and will need no further adjustment, to the rest of the system can be kept short. With available
except setting C2, unless the tank coil is changed to one of devices there is little prospect of extending the upper frequency
radically different inductance and Q. It is now only necessary to limit of this circuit much above 100 M H z and. in a forthcoming
set the modulator and PSD controls and choose a sweep rate. paper. we shall describe a rather different system for use up to
The only troublesome source of interference that we have 700 MHz.
experienced comes from switching transients produced by the Although we have discussed the circuit solely in terms of
counter on certain ranges. These get into the RF circuit by pure NQR it can, of course, also be used as a sensitive NMR
radiation, not down the cable, since they are present when this spectrometer. For example. Bleaney et a1 (1 982) have recently
cable is disconnected and the counter is driven by a signal used it to study not only the zero-field resonance of the common
generator. They are easily eliminated by choosing a different isotope 51Vin SmV04 at temperatures between 1.6 and 4.2 K,
counter range. but also the quadrupole splitting of the NMR lines due to "V and
the rare (0.24% abundant) isotope 'OV. Their measurements give
8. The power supply, etc the ratio of the two quadrupole moments 50Q/''Q=4.019 (4)
The 5 2 7 V supplies are obtained using IOOmA, 2 4 V with an accuracy five times better than earlier measurements.
integrated stabilisers boosted to 27 V by diodes in the common
leads and this PSU is housed in a separate case. A second case
houses the level meter, the sweep unit for the varicap diode and References
two RC filters (47 a,1000 p F ) to remove spikes and most of the Bleaney B, Gregg J F and Wells M R 1982 The ratio of the
noise from the supply rails. nuclear electric quadrupole moments of "V and "V in Sm V 0 4
The phase-sensitive detector has one novel feature. Although J . Phys. C: Solid St. Phys. 15 L349
we have interpreted our results in terms of the usual 6 dB per Cherry E M and Hooper D E 1963 The design of wide-band
octave RC integrator we actually use a linear integrator transistor amplifiers
described by Grimbleby (1980) since for a given bandwidth this Proc. IEE 110 375
gives a faster response. Apart from a change in the resistance Edmonds D T, Hunt M J, iMackay A L and Summers C P 1974
and capacitance values to give time constants up to 500 s this The high sensitivity detection of pure NQR
follows his circuit exactly although it is realised using a single Adc. Nucl. Quadrupole Resonance 1 145
quadruple operational amplifier LM324.
Faulkner E A and Holman A 1967 An improved circuit for
With the final mechanical reduction gear (100 : 1) rigidly
nuclear magnetic resonance
attached to the cryostat the mechanical drive to this gear from
J. Sci. Instrum. 44 391
the motor presents no problems and vibration is easily
eliminated using three standard flexible couplings. It is however, Fitzky H G 1974 NQR resonance spectroscopy for analysis of
important to ensure that the shaft to the capacitor does not bind chlorinated organic compounds in industry
where is passes through the 0 ring seal; this is therefore rather Adu. Nucl. Quadrupole Resonance 1 79
carefully machined so that the radial clearance between the shaft Grimbleby J B 1980 Averaging filter circuits for signal recovery
and the bore is only slightly (2%) less than the chord of the 0 applications
ring. In order to prevent the ingress of moisture past this rather J. Phys. E: Sci. Instrum. 13 557

822
A sensitive nuclear quadrupole resonance spectrometer

Pound R V 1950 Nuclear electric quadrupole interactions in

crystals
Phys. Rev. 79 685
Robinson F N H 1959 Nuclear resonance absorption circuit
J . Sci. Instrum. 36 48 1
Robinson F N H 1974 Noise and Fluctuations (Oxford: Oxford
University Press)
Robinson F N H 1980a Base-line drift in NQR spectrometers
J. Magn. Resonance 39 155
Robinson F N H 1980b A bi-symmetric Zeeman modulator for
nuclear quadrupole resonance
J. Phys. E: Sci. Instrum. 13 96 1
Verwieck J F and Cornwell C D 1961 Radio-frequency
spectrometer with bidirectional wave frequency modulation
Rev;.Sci. Instrum. 32 1383
Watkins G D and Pound R V 1952 The pure nuclear electric
quadrupole resonance of NI4 in three molecular solids
Phys. Rev. 85 1062
Weaver J R M and Robinson F N H 1981 NQR of bismuth and
the quadrupole Hamiltonian
Phyr. Lett. 85A 389

823