Professional Documents
Culture Documents
This content has been downloaded from IOPscience. Please scroll down to see the full text.
(http://iopscience.iop.org/0957-0233/1/4/002)
View the table of contents for this issue, or go to the journal homepage for more
Download details:
IP Address: 128.192.114.19
This content was downloaded on 07/09/2015 at 22:24
C D H Williams
Department of Physics, University of Exeter, Stocker Road, Exeter EX4 4QL, UK
Abstract. The noise and bandwidth behaviour of the constant temperature bolometer
detector is analysed with a view to optimising practical systems. It is shown that
careful adjustment of the time constants in this type of detector system is essential
to avoid serious and unnecessary degradation of the attainable performance. It is
concluded that a system with a critically damped response to a heat pulse input is
optimal from the point of view of sensitivity and useful signal bandwidth. The use of
feedback to enhance artificially the bandwidth of a sensor is considered but it is
found that any attempt to compensate for a slow sensor element using this
technique degrades the sensitivity seriously. Finally, some general guidelines for
the design of sensitive detector systems are suggested.
An exact model of the CTD dynamics is precluded by the where the bolometer temperature coefficient of resistance
non-linear nature of the system. An analytic treatment is X = drD/d$D.
must therefore be confined to a description of the effect
of small changes w h in the input power on the output
voltage zjE which will experience a perturbation v, about 2.3. Noise sources and bias vB
its quiescent value VE. The linearised system model of
Each component in the circuit contributes noise to the
the circuit given in figure l(b) will be described in this
system; but if stray capacitance is ignored, a step vindi-
section starting with the CTD in static equilibrium.
cated in section 3, then the individual sources can be
lumped together and their combined effect reproduced
2.1. Static equilibrium by the two noise equivalent electromotive forces no and
n , shown in figure l(b). Here n, represents the amplifier
At equilibrium a bias current I D flows in the detector input noise and the thermal noise of R , and R 2 , no is the
elevating its temperature gDand causing a net heat loss noise due to the bolometer and Ro. It is sometimes useful
to the surroundings ZARD. The amplifier, assumed to to be able to apply an external electrical signal, vB in
have negligible offset errors and a large input impedance, figure l(b), to the system. This can be a DC bias V, to
has an open loop DC gain GA so that the output is given adjust the bolometer operating point or a small pulse ub
by to check the dynamic response.
323
C D H Williams
h'
"b
I I I I I I
Id
3. System responsivity and sensitivity
324
Noise in bolometer systems
So far the treatment has been quite general but now equation has coincident roots, that is when b2 = 4c which
it is necessary to simplify matters slightly to let us draw means that a. = o z / 2 G A a j . The effects of different
some useful conclusions without needing to present an damping levels are illustrated by using a parameter
unreasonable number of graphs and tables. To start with, 6 = 2 G A o o o j / o z so a system having 6 = 1 is critically
stray capacitance shunting the resistive elements of the damped. Figure 3(u) comprises the frequency response
bridge, as mentioned in section 2.3, is to be ignored. This curves of representative cases, the corresponding noise
is possible not only because the resistors comprising the spectra are given in figure 3(b). The crucial importance
bridge are usually of low value, but also because careful of the correct adjustment of w o is illustrated by figure 4.
layout can ensure that stray capacitances affect each arm If 6 < 1 (overdamped) changing 6 affects the signal band-
of the bridge equally and thus cancel each other. Lead width and total noise power equally and the sensitivity
capacitance arising fom a remotely positioned detector of the system remains almost constant. However, when
can be dealt with by a capacitor in parallel with Ro. A 6 > 1 (underdamped) the rise in the signal bandwidth
typical low-noise operational amplifier used in a CTD caused by increasing 6 is insufficient to compensate for
circuit might have an input noise of about 3 nV Hz-'l2 the increased noise and, as can be seen from figure 4(c),
-corresponding to a noise equivalent resistance RG of the sensitivity falls dramatically. This result explains one
about 500 SZ at room temperature. The relevant transfer of the rather surprising aspects of CTD system behaviour
functions H e , and H , are of similar size and behaviour encountered in practice-when the operational amplifier
so if RD 11 Ro (the value of RD and Ro in parallel) is signi- is replaced by a higher quality device (higher gain-
+
ficantly smaller than RG ( R , 11 R2), as can usually be bandwidth product) the sensitivity can actually decrease.
arranged, then the noise in the system is dominated by If the widest possible bandwidth that can be obtained
n,. It will also be assumed that n, is a white, Gaussian without a significant reduction in sensitivity is desired,
noise source; above the l/f-noise corner, about 100 Hz then the critically damped 6 = 1 system must be selected.
for typical devices, this-is a good approximation to the
behaviour of most operational amplifiers. Given a suit-
ably constructed bolometer, the form of CTD described
30
in this paper is expected to have an upper frequency
limit in the range 10-100 MHz imposed by the amplifier
performance. These approximations greatly simplify the 20
expression for S which becomes -
53
a 10
L"
z
-1 0
-m
The first system to consider has R 1 = R 2 with VB = 0 -10
and GA b 1, so R, = Ro and the values for the system c-
%
parameters of interest are:
5 -20
- 30
6
325
C D H Williams
5. Asymmetric bridge; zero offset better than this. A virtual earth current detector elimi-
nates the effects of shunt capacitance by keeping the poten-
Having chosen a critically damped system on the basis tial difference across it constant and zero (Hamilton
that it offers the best overall performance it is tempting 1977). In an analogous fashion electrothermal feedback
to wonder whether the CTD can be further improved by can be used to nullify the heat capacity of a bolometer
modifying the circuit in some way. Perhaps the simplest and enhance the frequency response. An amplifier of gain
potentially beneficial change is altering the bridge ratio bandwidth product GAwocan be used to attain an over-
so that R I # R 2 . It is convenient to define all useful signal bandwidth of - ( G A ~ o ~ j ) 1 /Sherlock
2.
(1984) gives the details but the principles can be under-
stood more easily from Bode diagrams like those given
in figure 5 which demonstrate how the first-order pole in
and since GA $ 1 and V, = 0 the detector response can be cancelled by an amplifier
with its frequency response appropriately tailored by
wth=wi+(1 - 2 p ) o j ul = GAwOWjp/lD feedback.
b= C = G ~ O o W j 2 p (-
l p) (16~-d) The expressions given earlier (equations (9d, e)) for
the coefficients b and c of the characteristic equation
remembering that the system is critically damped so b2 = suggest that a considerable enhancement of the system
4c and bandwidth is possible when p > q . Inspection of
b equation ( 2 ) reveals that for static equilibrium to be at-
w, = 0.403 - = 0.201 0 t h (17) tained in the absence of offsets q must be slightly larger
2
than p but application of a DC bias VB to the bolometer
a little manipulation of the equations gives an expression network can reduce the value of and achieve the condi-
for the sensitivity tion p > y. The usable values of y are constrained be-
0.41Op2wf cause a stable system must have 0 t h > 0, this requirement
s=-0.201 32 p2w2
- ___ - ’ (18) is always met if 0 < q < 0.5. Bearing all this in mind the
Ikn? (5n) - I;z[wi + (1 - 2p)wj]2 expressions for the parameters of the bandwidth en-
The value of p that maximises the sensitivity depends on hanced CTD system are
the relative values of mi and wj. If cui > w j then p must
be as large as possible-that is unity. In the case when
wth = mi + (1 - 2y)wj = GAWOwjq/lD
wi < w j the maximum value that p can take is con- = + Oth) + GAoO(p - q )
+
strained by the need to keep wi (1 - 2p)oj positive.
c=oo@th(l + ( p - y ) ) + 2 G ~ W o w j ~- V( )l (19~-d)
However, a large value of w j is also desirable and inspec-
tion of equations (16) and (17) reveals that any value of the condition for critical damping becomes
p greater than 0.5 not only compromises the signal band-
[(Oth - O O ) - GAwO(p - q)12 = 8GAwoojq(l - q ) (20)
width but, in more serious cases when oj > mi, can also
result in instability as 0 t h becomes negative. An impor-
tant feature of the CTD is its ability to give an absolute
value for changes in the net heat loss from the sensor
element. Systems with p = 0.5 (Sherlock 1984) offer the 1
326
Noise in bolometer systems
so the signal bandwidth is given by the same but w, is increased to 20,. With the circuits
adjusted to keep q = 0.5 in each case, the sensitivity drops
U, = 0.201 [20,, f 2J2GAwOwju(1 - y)] (21) by 13 dB. Taking into account the change in bandwidth
and the sensitivity this means that increasing the bandwidth by this amount
reduces the signal to noise ratio by 23 dB.
7. Conclusions
I I , 1 I
The author is grateful to M G Brown, G M Wyborn
and Professor A F G Wyatt for commenting on the draft
manuscript.
References
327
C D H Williams
Sherlock R A 1984 Frequency response optimisation of the Wyatt A F G, Sherlock R A and Allum D R 1982 The
constant temperature detector system-a detailed root- temperature of phonon pulses injected into liquid 4He
locus analysis. J . Phys. E: Sci. Instrum. 17 386-93 J . Phys. C: Solid State Phys. 15 1897-915
Sherlock R A and Wyatt A F G 1983a The effect of self- Ziegler M 1934 The construction of a hot-wire anemometer
heating on the dynamical response of bolometric with linear scale and negligible lag Proc. Koninlijke
detectors J . Phys. E: Sci. Instrum. 16 669-72 Akademie van Wetenschappen te Amsterdam 15 ( 1 )
Sherlock R A and Wyatt A F G 1983b The dynamics of the
constant temperature detector system J . Phys. E: Sci.
Instrum. 16 673-82
328