INSTITUTE OF PHYSICS PUBLISHING MEASUREMENT SCIENCE AND TECHNOLOGY

Meas. Sci. Technol. 15 (2004) 1835–1847 PII: S0957-0233(04)77658-4

Dynamic response of constant

temperature hot-wire system in
turbulence velocity measurements
J D Li
School of Architectural, Civil, and Mechanical Engineering, Victoria University of
Technology, PO Box 14428, MCMC, Melbourne 8001, Australia

Received 4 March 2004, in final form 20 May 2004

Published 6 August 2004
Online at stacks.iop.org/MST/15/1835
doi:10.1088/0957-0233/15/9/022

Abstract
Analytical solutions have been derived for the first time for temperature
fluctuations along a hot wire when it is exposed to turbulent velocity
fluctuations of different frequencies and wave numbers. The solutions are
used to analyse the dynamic response of the hot wire alone first. The results
show that attenuation to the velocity fluctuations exists at high frequencies
and depends on the relative end-conduction heat loss from the hot wire to its
supports. A simplified circuit for a hot-wire anemometer is then used to
analyse the frequency response of the combined hot wire and its anemometer.
It is found that even though the frequency response of the anemometer (as
that determined from a square wave test) can be very high, the cut-off
frequency of the combined system is low, depending on the hot-wire length
to diameter ratio, Reynolds numbers, overheat ratio and the hot-wire
material used. A preliminary investigation of the spatial resolution of the hot
wire to the velocity fluctuations at different wave numbers is also presented.
Keywords: hot wire, dynamic response, turbulence measurements

Nomenclature kw thermal conductivity of the hot wire at mean

temperature
θw instantaneous hot-wire temperature, a function of t R0x electrical resistance of the hot wire per unit length at
and x ambient temperature
θ normalized hot-wire temperature Rx electrical resistance of the hot wire per unit length
θa temperature of the ambient air R0 total electrical resistance of the hot wire at ambient
θm mean temperature of the hot wire temperature
α thermal coefficient of electrical resistivity of the hot Rw total electrical resistance of the hot wire at the mean
wire temperature Tm
ρw density of the hot-wire material a overheat ratio of the hot wire
Cw specific heat of the hot-wire material Re Reynolds number
τ time constant of the hot wire Nu Nusselt number
x coordinate along the hot wire Pr Prandtl number
l half the hot-wire length U mean air velocity
d diameter of the hot wire u velocity fluctuation
t time
I mean electrical current flowing through the hot wire 1. Introduction
i electrical current fluctuation
kf thermal conductivity of the fluid at the mean hot-wire Hot wires have been used to measure velocities of turbulent gas
temperature flows for many years. Many people have analysed the mean

0957-0233/04/091835+13$30.00 © 2004 IOP Publishing Ltd Printed in the UK 1835

J D Li

temperature distribution along the hot-wire filament, Betchov five orders of magnitude less for the hot-film wall shear stress
(1949), Corrsin (1963), Champagne et al (1967), Hinze (1975), probes. They also found that fD increases with the mean
Freymuth (1979), Perry (1982) and Li et al (2004), to mention velocity flowing across the hot wires and the overheat ratio
a few. In these analyses, the solution of the mean temperature that the hot wires were operated at.
along the hot wire was generally given. The solution was then Morris and Foss (2003) analysed the frequency response
used to investigate the effect of the hot-wire length to diameter of a constant temperature hot-wire anemometer system nu-
ratio on the uniformity of the temperature distribution. In merically. In their analysis, the temperature distribution along
order to have a reasonably uniform temperature profile along the hot wire was solved for a particular hot wire at a speci-
the wire (away from its two ends), it is generally accepted that fied Reynolds number. In order to minimize the influence of
2l/d needs to be larger than 200. the anemometer on the hot wire itself, the cut-off frequency of
The dynamic response of the hot wire has been the anemometer was assumed to be very high. Their results
analysed by Freymuth (1979), Hinze (1975) and Perry (1982) of hot-wire attenuation on turbulent velocity at high frequency
analytically. In Freymuth (1979), the hot-wire filament confirmed the results of Freymuth (1979), which were based
was assumed to be infinitely long. Perry (1982) analysed on intuitive arguments. They also drew a similar conclusion
the dynamic response of the hot wire by considering the to Freymuth (1979) that changing the overheat ratio of the hot
perturbation in velocity, temperature and electrical current wire will not affect this attenuation. This is in contrast to the
separately. Hinze (1975) analysed the dynamic response of the findings from Khoo’s group.
hot wire by using the perturbations in the electrical resistance Borgos (1980), in his review on the electrical testing
of a hot wire, and the perturbation in hot-wire temperature was of the cut-off frequency from the anemometers, listed the
connected to its resistance. In Hinze (1975), the profile of the following limitations from such a test: (1) flow effects
temperature fluctuation was assumed to be similar to that of caused by unsteady boundary layers; (2) spatial resolution;
the mean temperature in order to determine the contribution (3) nonlinearity of the system with large velocity fluctuations;
of the heat conduction to the electrical resistance fluctuation. (4) the effect of thick coatings that are sometimes used over
thermal sensors and (5) only indirect information about the
Freymuth (1977a, 1977b) was the first to study the
effect of heat conduction. The limitations (2) and (5) can
dynamic response of the hot-wire system, which includes
be investigated by solving the differential equation of small
the hot wire itself and the anemometer for operating the
temperature fluctuations (less than 10%) for a simplified hot-
hot wire at constant temperature. In this pioneering
wire system.
work, Freymuth (1977a, 1977b) solved the combined energy
In this paper, theoretical solutions for the temperature
balanced equations of the hot wire and coupled to the
fluctuations along a hot wire are derived for given turbulent
electrical circuit. In considering the energy balance for
velocity fluctuations. To the best of our knowledge, this is the
thin wires, Freymuth (1977a) assumed that the temperature
first time that such solutions are given. The solutions are used
profile along the wire is uniform. From the solutions to these to study the hot-wire attenuation of velocity fluctuations at high
coupled equations, Freymuth (1977a, 1977b, 1978) suggested frequencies. The results confirm the conclusions of Freymuth
a method for determining the cut-off frequency of the hot-wire (1979) on this attenuation. By using the theoretical solution,
anemometer by using a small square or sine-wave voltage the effect on this attenuation from hot wires of different length
perturbation. to diameter ratios, Reynolds numbers, hot-wire materials and
In turbulence measurements, the hot-wire anemometer overheat ratios is investigated systematically. In contrast, the
is normally tuned to achieve a uniform frequency response results of Morris and Foss (2003) were for a particular hot wire
over the range of interest. The cut-off frequency is in general (a 1 mm active region of 5 µm tungsten filament with 50 µm
determined, using the method suggested by Freymuth (1977a), copper plated support stubs) at a single length to diameter ratio
by a square or sine wave test, which introduces a small (200) and a single mean flow velocity (7.5 m s−1).
perturbation to the electrical resistance in one of the two The hot wire is then connected to a constant temperature
arms of a Wheatstone bridge. However, during turbulence anemometer with simplified circuit, and the frequency
measurements, the fluctuations in the output voltage can only response of the combined system is investigated. It is found
be caused by the velocity fluctuations over the hot wire and that, similar to Khoo et al (1999), the cut-off frequency of the
the resistance fluctuations can only be a consequence of this system is mainly controlled by the hot wires rather than
velocity variation. Thus, the question should always be asked the anemometer itself, provided that the cut-off frequency of
as to how much could one be confident about the cut-off the latter is reasonably high. It is also found that the overheat
frequency determined by such a square/sine wave test. Khoo ratio, the Reynolds number and the length to diameter ratio
et al (1998, 1999), Chew et al (1998) and Teo et al (2001) of the hot wire can all influence the cut-off frequency of the
have studied the dynamic response of a hot-wire anemometer system. The spatial resolution of the hot wire for turbulent
for near-wall hot-wire/film probes experimentally. By using fluctuation along the wire is also studied preliminarily. It
a specially designed rotating disc to produce a known velocity is found that spatial resolution is not uniform across the
fluctuation at various frequencies, they were able to determine whole frequency range for a given wave number and length to
the frequency response of the marginally elevated and flush- diameter ratio.
mounted hot wires. They found that the cut-off frequency fD
of the hot-wire system, determined using velocity perturbation, 2. Equation for temperature along the hot wire
is much less than the cut-off frequency fS determined using
square or sine wave methods. Their results show that fD is By considering the heat generation, heat convection, heat
one order of magnitude less than fS for hot wires and up to conduction and heat accumulation in a hot-wire filament, it

1836
 Dynamic response of constant temperature hot-wire system

can be shown that the equation for instantaneous temperature Here overbar means steady conditions. For a hot wire 2l long
along a hot wire of constant diameter is given by (Perry 1982) with large supports, such as the hot-wire prongs, the boundary
conditions of the mean temperature are
∂θ ∂ 2θ
τ = 2 + (Q − ξ )θ + Z. (1)
∂t ∂η θ =0 at η = ±1
Here ∂θ (4)
=0 at η = 0.
θw − θa x ∂η
θ= η=
θa l These boundary conditions are symmetrical at η = 0, and the



4I 2 R0x α l 2 4I 2 R0x l 2 solution to equation (3) with boundary conditions (4) is
Q= Z=
πkw d πkw θa d  

 Z cosh(Y η)
kf l 2 l2 θ= 1 − (5)
ξ = 4Nu τ = ρw Cw . Y2 cosh(Y )
kw d kw
Here we have assumed that the centre of the hot wire is at where
x = 0, the ends of the hot wire are at x = ±l where the hot 
wire connects to the prongs and the electrical resistance of the Y = ξ − Q.
hot wire depends on its temperature according to
Solution (5) is the same as that given by Corrsin (1963)
Rx = R0x (1 + α(θw − θa )) by using the same boundary conditions as in (4). The
 l (2) dimensionless group Y is the Biot number of the hot wire.
Rw = 2 Rx dx. The mean temperature of the hot wire can be determined from
−l
(5) as
In equation (1), we use the Nusselt number recommended by
 
Fand (1965) Zθa tanh(Y )
  θm = θa + 1 − . (6)
Nu = 0.35 + 0.56Rew 0.52 Pr0.3 . Y2 Y

Here Rew is the Reynolds number based on the hot-wire The overheat ratio of the hot wire is defined as
diameter and Pr is the Prandtl number. According to Holman Rw − R0
(1992), the above relationship is valid for 0.1 < Rew < a=
R0
10 000, which would cover the Reynolds number range in
most of the turbulent measurements so far. It should be noted where R0 = 2lR0x . The mean temperature in (6) is related to
that the above Nusselt number of convective heat transfer for the overheat ratio a as
flow across a cylinder is for steady flows at different mean a
velocities. In analysing the dynamic response of the hot wire, θm = θa + .
α
the flow considered is unsteady. So far a correlation between
Reynolds numbers and the Nusselt number for unsteady This shows that, for a given hot wire, the mean temperature
inflows across a cylinder is not available. As a referee of the hot wire is determined by the overheat ratio. Using
pointed out, strictly, one has to solve the coupled flow and (6), the mean electrical current through the hot wire can be
heat conduction equations in order to determine properly the determined. Once the electrical current is known, Z and
thermal response of the hot wire with contributions from both Y can be determined from their definitions, and the mean
the conduction and convection components. The recent work temperature profiles along the hot wire can be established.
of Lange et al (1999) solved such a problem for hot wires in free The profiles are the same as those plotted in Freymuth (1979).
stream and close to the wall numerically with prescribed steady In general, the profiles are uniform near the centre of the
inflow velocity at given Reynolds number. The numerical hot wire for reasonably large Biot numbers. However, for
code may also be used for solving the heat convection for flow small Biot numbers, the mean temperature profiles are close
across a cylinder with prescribed unsteady inflow velocity. An to parabolic shapes. When measuring turbulent velocity using
investigation in this direction would be most welcome. In this a hot wire, a uniform temperature distribution is desirable.
paper, we assume that the above given relationship between This is because the temperature distribution along the hot wire
the Reynolds numbers and the Nusselt number can be applied is a balance between the heat generation and the heat losses
to the unsteady incoming velocity. from convection and end conduction. A calibration of the
hot wire is to relate the heat generation (through the electrical
3. The mean temperature distribution at steady current flow along the hot wire) to the heat losses due to
conditions convection. Because the heat losses of a hot wire are from
both convection and end conduction, one way to increase the
Under steady conditions, equation (1) becomes hot-wire sensitivity is by using longer wires so that the losses
due to end conduction are only a small fraction of the total loss.
∂ 2θ An empirical value in practice is 2l/d
200 (Champagne
+ (Q − ξ )θ + Z = 0. (3)
∂η2 et al 1967).

1837
J D Li
  1
4. The solution for the temperature fluctuation 1
C2 = e−Ys sinh[Ys (1 − η)]M(η) dη
2Ys sinh(2Ys ) 0
For a hot wire, the temperature along the wire can fluctuate  1 
when the velocity passing it and/or the current through it −eYs
sinh[Ys (1 − η)]M(−η) dη .
changes. For small perturbations, the equation for the hot- 0
wire temperature fluctuations can be written as When M(η) is symmetrical, C1 = C2 .
∂θ
∂ 2θ
The electrical resistance fluctuation rw of the hot wire is
τ = + (Q
− ξ
)θ + (Q − ξ )θ
+ Z
(7) connected to the integration of the temperature fluctuations
∂t ∂η2
along the wire, and can be expressed as
where the prime means fluctuations, θ
= θw
/θa and τ is the  l  1
time constant of the hot wire. In (7), we have assumed that rw = R0x α θw
dx = R0x lαθa θ
dη. (11)
the ambient temperature θa is constant since we are concerned −l −1
here with only the turbulent velocity measurements using hot Using (10) and the constants C1 and C2 , it can be shown that
wires. Using the definition for Q, Z and ξ , equation (7) can be  1
r̃w 1

reorganized as = αθa θ̃ dη
R0 2 −1
∂θ
∂ 2θ
  
τ = + (Q − ξ )θ
+ 2(Z + Q θ)i + χ θu. (8) αθa 1 cosh(Ys η)
∂t ∂η2 = − 1 M(η) dη. (12)
2Ys2 −1 cosh(Ys )
Here i is the current fluctuation, u is the velocity fluctuation,
and using the expression for Nu, it can shown that Because the function cosh(Ys η) is symmetrical at η = 0,
this result shows that only the symmetrical part of M(η) can
U dNu 0.29Rew 0.52 make a contribution to the electrical resistance fluctuations of
χ =− ξ ≈− ξ.
Nu du 0.35 + 0.56Rew 0.52 the hot wire. The current fluctuation i is a function of t while the
In equation (8), the fluctuation of ξ
due to the temperature velocity fluctuation u can be a function of both t and x. Because
fluctuations (through the variation in kinematic viscosity ν and of this, ĩ will be a function of s only and symmetrical with
thermal conductivity kf of air) has been neglected. Estimation respect to η = 0 while ũ will be a function of both s and η.
has been made of its magnitude. It is found that in general the
error from neglecting the term is about 1% of the second r.h.s. 5. Dynamic response of the hot wire
term in equation (8).
By taking the Laplace transform of (8), we obtained To determine the frequency response and the spatial resolution
2
of a single hot wire, let us assume that, for a particular non-
d θ̃

− [τ s + (ξ − Q)]θ̃ = −2(Z + Q θ)ĩ − χ θ ũ. (9) dimensional wave number β = kl,
dη2
Here ∼ means the Laplace transform and s is the Laplace ũ = ũ1 (s) cos(βη). (13)
variable. For a given s, equation (9) is a second-order ordinary Then it can be shown, after much algebra, that
differential equation (ODE) with variable coefficients. The
r̃w
= P ĩ
+ Aũ
1 (14)
general solution of (9) (Kreyszig 1993) can be written as
 where

eYs η η −Ys η
θ̃ = C1 eYs η + C2 e−Ys η + e M(η) dη r̃w ĩ ũ1
2Ys 0 r̃w
= , ĩ
= , ũ
1 =
 R0 I U
e−Ys η η Ys η
− e M(η) dη. (10)
2Ys 0  
2Q tanh(Ys )
Here C1 and C2 are constants to be determined by using the P =− − 1
Ys2 Ys
boundary conditions for θ
, and
2  
 2Q tanh(Ys ) tanh(Y ) τs
Ys = τ s + (ξ − Q) + − + (15)
τs Ys3 Y3 Ys2 Y 2
M(η) = −2(Z + Q θ)ĩ − χ θ ũ. 
χQ β


The boundary conditions for θ are such that (because we A= sin(β) − 2 (Y tanh(Y ) cos(β)
βYs2 Y 2 Y + β2
have assumed that the ambient temperature is constant and the
β
temperature for the supports of the hot wire is the same as that + β sin(β)) − 2 (Ys tanh(Ys ) cos(β)
of the ambient) Ys + β 2


β
θ̃ (η = −1, s) = θ̃ (η = 1, s) = 0. + β sin(β)) + ((Ys − Y )(tanh(Ys )
2[(Ys − Y )2 + β 2 ]
Using these conditions, the constants C1 and C2 in (10) − tanh(Y )) cos(β) + β(1 − tanh(Ys ) tanh(Y )) sin(β))
can be determined as β
  1 + ((Ys + Y )(tanh(Ys )
1 2[(Ys + Y )2 + β 2 ]
C1 = e−Ys sinh[Ys (1 − η)]M(−η) dη 
2Ys sinh(2Ys ) 0
 1  + tanh(Y )) cos(β) + β(1 + tanh(Ys ) tanh(Y )) sin(β)) .
− eYs sinh[Ys (1 − η)]M(η) dη
0 (16)

1838
 Dynamic response of constant temperature hot-wire system

1.1 as suggested by Beljaars (1976). It seems that this is too

low since the results in figure 1 show that the attenuation
1
continues beyond ω = 100ωl to 1000 ωl. The plateau at high
frequencies is 92.4% of that at low frequencies. This means
that the hot wire can resolve only 92.4% of the magnitude of
Dynamic Response

0.9
the velocity fluctuations at high frequencies for this particular
hot wire and Reynolds number. In terms of energy, this is
0.8 only 85%. This attenuation starts at fairly low frequency
fl = ωl /2π ≈ 160 Hz and finishes at f ≈ 160 kHz.
0.7
Unfortunately, this covers the frequency range in most of the
turbulence measurements today.
In calibrating the hot wires, both static and dynamic
0.6
calibration methods exist. The static calibration method
ω1 100ω1 1000ω1
calibrates the hot wire at different mean velocities (f = 0).
0.5 Figure 1 shows that this will not be able to capture the plateau
100 101 102 103 104 105 106 107 108
ω at the high frequencies. Also, this attenuation cannot be
corrected by using the existing ‘dynamic calibration’ method
Figure 1. The dynamics response (or attenuation) of the hot wire
only.
(Perry 1982) either. In the existing dynamic calibration,
the shaking frequencies of the calibrator at different mean
velocities would in general be at about 1 Hz, far too low a
When β → 0,
  frequency to capture the plateau at the high frequencies shown
χ Q tanh(Ys ) tanh(Y ) τs in figure 1. Given the very large frequency range where the
A → A1 = − + . (17)
τs Ys3 Y3 Ys2 Y 2 transition in attenuation occurs, it would be very difficult to
If we consider the hot wire at constant temperature all the time, come up with a scheme to calibrate the hot wire so that this
the electrical resistance fluctuation rw would be zero as that transition and the plateau at the high frequency can be captured.
assumed by Freymuth (1979). Later, we will show that this It seems that the only feasible solution would be to calibrate
cannot be true for the commonly used hot-wire anemometers. the hot wire using the existing methods and correct the hot-
For the time being, let us accept this assumption. Then the wire results using its frequency response such as that shown
dynamic response of the hot wire to the velocity fluctuations in figure 1.
is
ĩ
A 6. Attenuation of the hot wire to velocity
=− . (18)
ũ
1 P fluctuations (β = 0)
Freymuth (1979) considered the dynamic response of the hot
wire for velocity fluctuations with time only, this corresponds To investigate the effect of the hot-wire length and other
to β = 0 in (18) and the dynamic response of the hot wire parameters on the attenuation at zero wave number β = 0,
becomes (note that P is a function of s only) the ratio of the plateau at high frequencies to that at low
frequencies needs to be determined. Using (19), this is defined
ĩ
A1
=− . (19) as
ũ
1 P
A1 /P |ω→∞
Figure 1 shows the dynamic response of the hot wire according ε1 = (20)
A1 /P |ω→0
to equation (19) for a 5 µm platinum wire with 2l/d = 200,
a = 0.5 and Rew = 3.2 (This corresponds to a mean velocity with
of U ≈ 14 m s−1). Here the dynamic response is defined as A1 χ (1 − tanh(Y )/Y )
the magnitude ratio between the electrical current fluctuations =
P 2Y 2 + 2Q(1 − tanh(Y )/Y )
i
and the velocity fluctuation u
. The results in the figure ω→∞

have been normalized by the response at ω = 2πf = 0 to A1

take into account the hot-wire calibration (f is the frequency P ω→0
in Hertz). The figure shows that the dynamic response of the χ [−3 tanh(Y ) + Y (3 − tanh2 (Y ))]
hot wire is flat at low frequencies and reaches another lower = .
−4Y 2 (tanh(Y ) − Y ) + 2Q[−3 tanh(Y ) + Y (3 − tanh2 (Y ))]
plateau at high frequencies. This is due to the attenuation of
the heat waves along the hot wire as pointed out by Freymuth Here the physical meaning of ε 1 is the same as the ε1
used in
(1979). This attenuation was also investigated by Bremhorst Freymuth (1979) for the attenuation at high frequency. Note
and Gilmore (1978) and Teo et al (2001), and was attributed to that the actual amount of attenuation is 1 − ε1.
the so-called ‘end/side loss’ from the hot-wire filament to its Figure 2 shows the attenuation according to equation (20)
supports. This attenuation starts at ωl ≈ 10/τ ≈ 40 D/(2l)2 at different length to diameter ratios for platinum wires at
as suggested by Beljaars (1976) and Freymuth (1979). Here Rew = 3.2 with a = 0.5 and 0.01, respectively. It can be seen
D = kw /ρw Cw is the thermal diffusivity of the hot-wire from the figure that the attenuation becomes less as the length
material. Also shown in figure 1 is ω = 100 ωl , the radial to diameter ratio increases. This is because the attenuation is
frequency where the attenuation is supposed to be complete related to the heat conduction from the hot wire to its prongs.

1839
J D Li

1 1

0.95 0.95
Platinum wire

a = 0.5

ε1
0.9
ε1

0.9

Tungsten wire

a = 0.01
0.85 0.85

0.8 0.8
101 10 2 10 3 101 102 103
I/d I/d

Figure 2. The dependence of attenuation on hot-wire length to Figure 4. Hot-wire attenuation of platinum and tungsten wires.
diameter ratio.

1 1

0.98

0.95
Dynamic Response

0.96

Platinum wire
ε1

0.9
0.94

0.85 0.92

Tungsten wire

0.9
0.8 100 101 102 103 104 105 106 107 108
100 10 1
10 2
10 3 ω
Re
Figure 5. Dynamic response of platinum and tungsten hot wires.
Figure 3. The dependence of hot-wire attenuation on Reynolds
number.
The figure shows that the attenuation decreases as the Reynolds
As the length to diameter ratio increases, the heat losses from number increases. This is because, as the Reynolds number
the end conduction become relatively less than those from increases, the convective heat losses increase. This again
convection, and so does the attenuation. results in relatively smaller conductive heat losses.
Figure 2 also compares the attenuation at the overheat
Figure 4 shows the attenuation of the platinum and
ratios a = 0.5 and 0.01 for the given hot wire. It can be seen
tungsten wires at different l/d with Rew = 3.2 and a = 0.5.
that the overheat ratio does have some effect on the attenuation,
The figure shows that the tungsten wire gives more attenuation
especially at small l/d. The figure shows that a higher overheat
than the platinum wire. This is because tungsten has a thermal
ratio will result in less attenuation.
The results for a = 0.01 have been compared with the conductivity kw = 174 W K−1 m−1, which is higher than
results from Freymuth (1979) at a ≈ 0. When plotted as ε 1 kw = 71.6 W K−1 m−1 for platinum. Because of this, the
versus Y (the Biot number), no distinction can be made between relative heat losses from end conduction for the tungsten
the two results. Thus the present theoretical results at a ≈ 0 wire will be higher than those for the platinum wire at a
agree with the results of Freymuth (1979), which were based given Reynolds number. However, the thermal diffusivity
on intuitive arguments for very small overheat ratios. Morris of tungsten is higher than that of platinum. This means that
and Foss (2003) found that their results at a = 1.7 agree with the frequency at which the tungsten wire starts to attenuate the
the results of Freymuth (1979) at a ≈ 0. The present results velocity fluctuations will be higher than that for the platinum
in figure 2 show that the attenuations at a ≈ 0 and a > 0 are wire for the same 2l/d. Figure 5 shows the dynamic response
different, especially at smaller 2l/d. of the platinum and tungsten wires. The figure shows that the
Figure 3 shows the effect of Reynolds number on the frequency at which the tungsten hot wire starts to attenuate is
attenuation for platinum wire with 2l/d = 200 and a = 0.5. indeed higher than that of the platinum wire.

1840
 Dynamic response of constant temperature hot-wire system

it can be shown that

r̃w /R0
B1 =
ẽout /E out
r̃
(M

s 2 + M
s + 1) Z − KZ0 (Ra + Rw )
=
w = −
ẽout KRa R0 (Rc + Zb )
(22)
where
Z0 = Ra Zb − Rc Zw
Z = (Ra + Zw )(Rc + Zb )
Zb = Rb + Lb s
Figure 6. Schematic diagram of a constant temperature hot-wire Zw = Rw + Lw s.
anemometer (CTHWA). Also, it can be shown that the electrical current fluctuation is
related to that of the output voltage
7. The frequency response of the hot wire with its  
ĩ
Ra + Rw (M

s 2 + M
s + 1) Z − KZ0
anemometer C1 =
= 1+ .
ẽout Ra + Zw KRa (Rc + Zb )
In order to connect the temperature perturbations with the (23)
hot-wire output voltages, a simplified circuit for a constant
Combining (14), (22) and (23) for β = 0, we obtain
temperature hot wire anemometer (CTHWA) such as that

A1
shown in figure 6 is assumed. ẽout = ũ
. (24)
In figure 6, Ra , Rb , Rc and Rd are the four resistors B1 − C1 P 1
forming a Wheatstone bridge, Lw is the inductance from The frequency response of the CTHWA, for a perturbation in
the hot-wire probe cable, Lb is a balance inductance to be the hot-wire resistance, is
adjusted during frequency response optimization, I and I1

ẽout 1 −KRa R0 (Rc + Zb )
are the currents in the two arms of the bridge, Eout is the

= = .
r̃w B1 (M s + M
s + 1) Z − KZ0 (Ra + Rw )


2
voltage output of the CTHWA, Eb is the offset voltage and G1
and G2 are the gains of the two-stage amplifiers, respectively. (25)
Following Freymuth (1977a), it is assumed that, for the two- Figure 7 shows the Bode diagram and the impulse response
stage amplifiers, the gain–bandwidth products are GBP1 and of equation (25) for a typical CTHWA with the following
GBP2, and the first-order time constants are M1 and M2, parameters: Rb = 10Rw , Ra = 100 , Rc = 1000 , Lw =
respectively. The time constants are related to the gains G1 15 µH, Lb = 1.0032Rb Lw /Rw , K = 1000, G1 = G2 ,
and G2 by GBP1 = GBP2 = 107. Here, we have assumed that the
Wheatstone bridge is in perfect balance at steady conditions.
G1 G2
GBP1 = GBP2 = . The hot-wire resistance Rw was calculated for a 5 µm platinum
2πM1 2πM2 wire 1 mm long (2l/d = 200) with a = 0.5. Figure 7
and the zero and pole diagram shown later were produced
The output voltage fluctuation eout from the two-stage amplifier
using MATLAB. The impulse response is similar to the
is related to the current and hot-wire resistance fluctuations in
square wave test normally used for tuning the frequency
the hot wire as
response of the CTHWA. The inductance Lw used here is
d2 eout deout for a hot-wire cable about 3 m long (Perry 1982). The
M

+ M
+ eout = Kein inductance Lb has been tuned so that the frequency response

dt 2 dt 

di di1 is optimal according to Bruun (1995). Figure 8 shows the
ein = − Rw i + I rw + Lw + Rb i1 + Lb
dt dt (21) zero and pole diagram for this CTHWA. The figure shows
di1 that the system has four poles and one zero. The poles
eout = (Rb + Rc )i1 + Lb
dt close to the imaginary axis have a damping of 0.522 and
di
eout = (Ra + Rw )i + I rw + Lw an overshot of 14.6% (as suggested by Bruun (1995)) at
dt the natural frequency fn = ωn /2π = 250 kHz. Figure 8
where ein is the voltage fluctuation at the input to the first-stage shows that the system is close to a second order one near
amplifier, K = G1 G2 the total gain of the two amplifiers, its first two poles and the cut-off frequency fc ≈ 1.3 fn ≈
M
= M1 + M2 , M

= M1 M2 and i1 is the electrical current 320 kHz for damping factor of 0.522. (Here the natural
fluctuation for I1 . In (21), the fluctuation of the hot-wire frequency fn is the frequency shown in the inset of figure 8,
resistance rw cannot be assumed to be zero, since this will lead and the cut-off frequency is that determined from the −3 dB
to the conclusion that the frequency response of the CTHWA line in the frequency response). This cut-off frequency is close
does not depend on the characteristics of the amplifiers, the to that determined from the impulse response according to (see
resistors Rb and Rc and the inductance Lb as the last equation figure 7)
(21) shows. This is contradictory to experience. By taking 1
fs =
the Laplace transforms to (21) and solving the linear system, 1.3τs

1841
J D Li

Bode Diagram
50

Magnitude (dB)
0

–50

Phase (deg) –180

–360
100 101 102 103 104 105 106 107
Frequency (rad/sec)

x 105 Impulse Response

10

GBP=10000000
5 Lb=1.0032LbRb/Rw
Amplitude

–5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time (sec) x 10–5

Figure 7. The Bode diagram and impulse response of a CTHWA.

x 106 Pole-Zero Map 3

4 CTHWA only
2
Hot wire only
1
3
0
2 –1
Frequency Response (dB)

–2
1
Imaginary Axis

–3
Hot wire and CTHWA
0 –4 combined
System: B
Pole: –5.17e+005 – 1.34e+0061 –5
Damping: 0.522
–1 Overshoot(%): 14.6
Frequency (rad/sec): 1.57e+006 –6

–7
–2
–8

–3 –9

–10
–4 101 102 103 104 105 106
–9 –8 –7 –6 –5 –4 –3 –2 –1 0 f (Hz)
Real Axis 6
x 10
Figure 9. Frequency response of a hot wire and its anemometer.
Figure 8. Zero and pole diagram for a CTHWA.

shown. The overshoot of the CTHWA near its cut-off

as given by Freymuth (1977a). Here τs is the time when
frequency can clearly be seen. It can be seen from figure 9
the dynamics response has dropped to 97% of its peak value.
that even though the cut-off frequency of the CTHWA is
The cut-off frequency of fc = 320 kHz is close to that of the
320 kHz, it is only 13 kHz for the hot-wire system. This
best of the commercially available hot-wire anemometers. For
example, the TSI IFA300 anemometer claims fully automated cut-off frequency (determined from the −3 dB line) is more
operation with a frequency response of 300 kHz or greater. than one and half orders of magnitude less than that of the
Figure 9 shows the dynamic response in decibels for the anemometer itself. The numerical results of Morris and Foss
hot wire (equation (19)), the CTHWA (equation (25), with (2003) did not show this difference in cut-off frequencies
the parameters the same as those given for figure 7) itself and determined by a square wave test and those determined by
the hot-wire system (equation (24), a combination of the hot velocity perturbations. Their figure 15 shows no cut-off (cut-
wire and the anemometer). Note that the dynamic response off here means that their response ratio φ has dropped to less
here means the ratio in magnitude between the output voltage than 71%) even at f > 3 MHz, while their cut-off frequency


fluctuation eout and the velocity fluctuation u
for the hot- for the hot-wire anemometer is 3.8 MHz. By comparing the
wire system, it means the ratio in magnitude between eout and cut-off frequency fS determined from a square/sine wave test
rw for the CTHWA and the ratio in magnitude between i
and the fD determined from the velocity perturbations for the
and u
for the hot wire only. The parameters for the 5 µm marginally elevated and flush-mounted hot wires, Khoo et al
platinum hot wire used are U = 10 m s−1 and a = 0.5. (1999) found that fS is one order of magnitude higher than
In figure 9, only the responses greater than −10 dB are fD . The current results confirm Khoo’s conclusions for hot

1842
 Dynamic response of constant temperature hot-wire system

3 3
Hot wire only
2 CTHWA only 2
CTHWA only
1 1

0 0
Frequency Response (dB)

Frequency Response (dB)

–1 –1

–2 –2

–3 –3
Hot wire only
–4 –4
Hot wire and CTHWA
–5 –5 combined with a = 0.5
Hot wire with 0.6
–6 increasing a = 0.3 0.7
–6
0.5 0.8
0.7 0.9
–7 0.9 –7 1.0
–8 –8

–9 –9

–10 –10
101 102 103 104 105 106 101 102 103 104 105
f (Hz) f (Hz)

Figure 10. The effect of overheat ratio on the frequency response of Figure 11. The effect of overheat ratio on the frequency response of
the hot wire system with GBP = 107 and a perfectly balanced bridge. the hot-wire system with GBP = 106 and an imbalanced bridge.

3
wires in free stream. Both the present results and those of Khoo
2
et al (1999) show that the frequency response determined using
the square/sine wave test in turbulence measurements could 1

be seriously misleading. As Khoo et al (1999) pointed out, 0

from a user’s point of view, the dynamic frequency response –1

Frequency Response (dB)

fD should serve as a more accurate indicator of the overall –2

frequency response of the hot-wire systems since fD signifies –3
the onset of amplitude attenuation when a hot wire is subjected –4
increasing 2l/d = 100
to direct velocity perturbations. –5 200
Figure 10 shows the dynamic response of the hot-wire 400
–6 800
system at different overheat ratios while the other parameters
–7
are kept the same as those in figure 9. The effect of changing
–8
overheat ratio on the frequency response of the CTHWA is
–9
minimum. However, the figure shows that as the overheat ratio
increases from 0.3 to 0.9, the cut-off frequency of the hot-wire –10
101 102 103 104 105 106
system increases from 8 kHz to 28.0 kHz. This increase in cut- f (Hz)

off frequency is due to less attenuation at higher overheating Figure 12. The effect of 2l/d on the frequency response of the
ratio as mentioned before (figure 2). This finding is different hot-wire system.
from those of Freymuth (1979) and Morris and Foss (2003),
and agrees with the experimental results of Khoo et al (1999). Figure 11 also shows the frequency response of the hot-
Figure 11 shows the frequency response of a CTHWA wire system at different overheat ratios. The parameters used
with GBP1 = GBP2 = 106 for the two amplifiers. This for the 5 µm platinum wires are: 2l/d = 200 and U =
CTHWA is close to the anemometers using amplifiers of the 10.0 m s−1. It can be seen from the figure that the frequency
old generation (Perry 1982). The resistors Ra and Rc are the response of the hot-wire system increases with increasing
same as those used in figure 7, and are the values recommended overheat ratio. At a = 0.5, the cut-off frequency is about
in Perry (1982). It is found that the best frequency response 13 kHz and at a = 0.9 this is increased to 37 kHz. However,
of the hot-wire system can be achieved by using a slightly figure 11 shows that at a = 0.9, the system is close to marginal
unbalanced bridge with Rb = 9.9Rw . When using perfect stability. We think that this close to loss of stability at high
bridge balance, it is found that the frequency response of the overheat ratios is because the cut-off frequency of the CTHWA
hot-wire system becomes unstable at high overheat ratios (say is too low, and the time constants for the hot wire and that of
a > 0.6). This imbalance of the bridge is achieved in practice the CTHWA are too close to each other. It is recommended
by adjusting the offset voltage Eb and Perry (1982) suggested that amplifiers of the new generation (with GBP > 106) should
that R o = Rw Rc − Ra Rb is in general of the order of 100 2 always be used.
(p 62). The present results confirm his suggestion. To Figure 12 shows the frequency response of the hot wire
achieve an optimal frequency response for the CTHWA, the and its anemometer for platinum wires of different 2l/d. The
inductance Lb was re-tuned to 1.033Lw Rb /Rw . The cut-off parameters for the CTHWA are the same as those used for
frequency for the CTHWA at a = 0.5 is about 48 kHz for the figure 9. The overheat ratio a = 0.9 is fixed. The figure shows
parameters given. This is about double the 25 kHz normally that the cut-off frequency increases from 18 kHz to 101 kHz
achieved in practice, considering that only idealized electronic when 2l/d increases from 100 to 800. This increase in
components are used here. cut-off frequency as 2l/d increases is due to the decrease

1843
J D Li

3 3
CTHWA only 2
2

1 Hot wire only 1

0 0
Frequency Response (dB)

Frequency Response (dB)

–1 –1

–2 –2

–3 Platinum wire
–3 Hot wire system with 2l/d = 200
increasing mean velocity Ti/Ni wire
–4 U = 1.0 m/s –4
increasing
3.0 2l/d = 50
–5 10.0 –5 100
30.0 200
–6 100.0 –6

–7 –7

–8 –8

–9 –9

–10 –10
101 102 103 104 105 106 101 102 103 104 105 106
f (Hz) f (Hz)

Figure 13. The effect of Reynolds number on the frequency Figure 14. The effect of hot-wire materials on the frequency
response of the hot-wire system. response of the hot-wire system. The results for the Ti/Ni wires at
different 2l/d are also shown.

in attenuation of the hot wire at high frequencies for long

wires. This increased cut-off frequency is only for the time at V = 2.17 m s−1 (Rew ≈ 0.49). These results are still
variation of the turbulence at the hot wire. It should be noted higher than those of Khoo et al (1998, 1999) from experiments.
that this improved frequency response does not mean that However, it should be pointed out that Khoo et al (1998, 1999)
the spatial resolution of the hot wire is increased. On the defined fD as the upper frequency limit where amea = 0.9 aimp .
Here amea is the measured velocity amplitude and aimp is the
contrary, in a given turbulent flow, the spatial resolution of
imposed velocity fluctuation. In the present paper, the cut-
the hot wire will deteriorate as the length of the hot wire
off frequency is defined as the frequency where the energy
increases. For example, in turbulence measurements very
is half of that at the low frequency (−3 dB in figures 9–
close to the wall, Ligrani and Bradshaw (1987) and Khoo
14). If the criterion used in Khoo et al (1998, 1999)
et al (1997) have found that, in order to fully resolve the
is adopted for defining fD , the present results give fD ≈
turbulent fluctuations, the hot-wire length needs to be of
2.4 kHz at V = 2.17 m s−1 (Rew ≈ 0.49) and fD ≈ 2.53 kHz
the order of the viscous length scale (with 2l/d > 200).
at V = 2.7 m s−1 (Rew ≈ 0.61) for a 5 µm tungsten wire of
This suggests that in turbulence measurements, preliminary
1.2 mm long. These values are in remarkable agreement with
investigation should be conducted first to determine the highest
those given by Khoo et al (1998, 1999).
possible frequencies (after applying Taylor’s hypothesis) and
Li (2001) and Li et al (2004) suggested that titanium
the smallest turbulent length scales. If the length scale and nickel alloy (55% titanium and 45% nickel) could be
is the limiting factor, short wires should be used. The a good hot-wire material. It has mechanical and electrical
short wire may have a lower cut-off frequency according to properties comparable with platinum but a very low thermal
figure 12 but will have a better spatial resolution. On the other conductivity (kw ≈ 10 W K−1 m−1). Figure 14 shows the
hand, if the cut-off frequency is the limiting factor, longer wires frequency response of the hot-wire system using platinum and
should be used. Thus it can be concluded that, in turbulence Ti/Ni as the hot-wire materials. The overheat ratio a = 0.9
measurements, an optimized length to diameter ratio exists for is used and the other parameters for the CTHWA are the same
a hot wire at a given Reynolds number. Further study is needed as those used in figure 9 and U = 10 m s−1. It can be seen
to quantify this optimization. that, at 2l/d = 200, the cut-off frequency of the Ti/Ni wire is
Figure 13 shows the frequency response of a platinum 180 kHz. This is much higher than the 27 kHz for the platinum
hot wire with 2l/d = 200 at different mean velocity for the wire at the same 2l/d.
same CTHWA as in figure 9. The figure shows that the cut- In today’s turbulence measurements, the highest
off frequency of the hot wire increases as the mean velocity frequency of the turbulent flows would in general be much
(Reynolds number) increases. At U = 3.0 m s−1, the cut-off less than 180 kHz. In figure 14, the frequency response of the
frequency from figure 13 is about 9.2 kHz. This is increased to Ti/Ni wires at smaller 2l/d is also shown. It can be seen from
22 kHz at U = 30 m s−1. Khoo et al (1998, 1999) also found the figure that, even at 2l/d = 50, the cut-off frequency of
that the cut-off frequency fD increases with mean velocity the Ti/Ni wire is still higher than that of the platinum wire at
for a given hot wire. For example, Khoo (1998) found that 2l/d = 200. With shorter hot wires, it is generally accepted
fD = 2690 Hz at V = 2.7 m s−1 (Rew ≈ 0.61) and fD = that the spatial resolution will be improved. Thus, it can be
2345 Hz at V = 2.17 m s−1 (Rew ≈ 0.49). Thus the present concluded that, by using a Ti/Ni hot wire, wires with 2l/d less
results are consistent with the experimental results of Khoo than 200 can be used to improve the spatial resolution with still
et al (1998, 1999). The hot wires used in Khoo et al (1998, sufficiently high enough cut-off frequency.
1999) were 5 µm tungsten wires of 1.2 mm long. Using these Figures 11–14 show that the cut-off frequency of the
parameters and a = 0.5, the cut-off frequencies will be fD = hot-wire system can be improved by increasing the overheat
6.5 kHz at V = 2.7 m s−1 (Rew ≈ 0.61) and fD = 6.3 kHz ratio, length to diameter ratio or the Reynolds numbers. By

1844
 Dynamic response of constant temperature hot-wire system

combining this with the results shown in figures 2–5, it can be 1

concluded that the less the attenuation of the turbulent velocity 0.9
fluctuations at high frequencies by the hot wire, the higher the
cut-off frequency of the hot-wire system will be. 0.8

In general, the present results show that minimizing the 0.7

Spatial Resolution β = 0
conductive heat loss relative to the total heat loss can increase
0.6
the cut-off frequency of the hot-wire system. This conclusion
can also be used to explain the findings of Khoo et al (1998, 0.5
1999) that for the air flow generated from a rotating disc, at sin2(β)/β2 Decreasing 2l/d = 800
0.4 400
the same mean velocity, the cut-off frequency of the hot wire 200
100
increases as it is moved close to the wall. As the hot wire is 0.3

moved close to the wall, the heat loss is increased for a hot
0.2
wire experiencing the same mean velocity as that found by
Khoo et al (1997) and Lange et al (1998) for air flowing over a 0.1

highly conducting wall. This increased heat loss is due to the 0

0 1 2 3 4 5 6 7
higher thermal conductivity of the wall material than that of β
the air. The results of Lange et al (1998) (figure 9) show that
the temperature distribution around the hot wire close to the Figure 15. The spatial resolution of the hot wire at different 2l/d.
wall is skewed in the normal to the wall direction in the case
5
of a highly conducting wall. Because of this, the temperature
4
gradient between the hot wire and the wall is greater than that CTHWA only
3
between the hot wire and the free stream. This larger gradient
2
results in more heat being conducted away from the hot-wire Hot wire only
β=0
1
filament. The consequence of this is that more heat will be
Frequency Response (dB)

0
lost from the hot-wire filament to the surrounding air than that –1
to the hot-wire prongs through end conduction. Because of –2
Hot wire and
this, the relative end conduction heat loss becomes less as the –3 CTHWA β = 0
hot wire is moved close to the wall, a false ‘high’ velocity –4
is measured and the hot wire shows an ‘improved’ frequency –5
Hot wire only
response. –6 β = π/2
On the other hand, for air is flowing over a well-insulated –7 Hot wire and
wall, the heat loss to the surrounding air will be reduced –8 CTHWA β = π/2

between the hot wire and the wall as the hot wire is placed close –9

to it. This is because the temperature gradient between the hot –10
102 103 104 105 106 107
wire and the wall is less than that between it and the free stream ω (rad/sec)
(see figure 13 of Lange et al (1999)). This will result in less
heat loss directly from the hot-wire filament to the surrounding Figure 16. The frequency response of the hot wire at two different
air in comparison with a hot wire placed in free stream at the wave numbers.
same mean velocity. Because of this, a hot wire placed close
to a well-insulated wall will measure a false ‘low’ air velocity. along the hot wire, β = 0, and the hot wire cannot in general
Of course, in practice, it is very difficult to have a wall made resolve the spatial variation of the turbulence structures when
of materials having smaller thermal conductivity than that of the hot wire is longer than the structure’s length scale. This is
air, as pointed out by Lange et al (1999). This explains their the spatial resolution problem of the hot wire.
‘unexpected and contradictory’ numerical results that a false Figure 15 shows the spatial resolution of the hot wire at
‘low’ velocity will be measured near a well-insulated wall in different wave numbers β for different 2l/d at ω = 0. The
turbulent airflows. The present results show that, for a hot wire figure shows that as β increases, the spatial resolution of the hot
in airflow near a well-insulated wall, not only will the measured wire decreases dramatically, as would be expected. Figure 15
velocity be falsely low, but also the cut-off frequency will be shows that the spatial resolution is improved as the 2l/d
reduced. However, for turbulence measurements in the near decreases. The spatial resolution approaches
wall regions of water or oil channels (Alfredsson et al 1988), sin2 (β) sin2 (kl)
=
because of the high thermal conductivities of water and oil, the β 2 (kl)2
measured turbulence quantities will be closer to the true values as 2l/d → ∞. This is the function used by Wyngaard (1968)
(Khoo et al 1997) than those measured in airflows. in his analysis of the spatial resolution of three-dimensional
turbulence. Spatial resolution of this form is obtained by
8. Frequency response of the hot wire at wave assuming uniform mean temperature along the hot wire, which
numbers other than zero (β /= 0) can only be achieved at 2l/d → ∞.
Figure 16 shows the frequency response of the hot wire
So far we have been considering the frequency response of a with its anemometer at β = 0 and β = π/2. The parameters
single wire and it has been assumed that the velocity fluctuation used for the hot wire and its anemometer are the same as
is uniform along the wire (β = 0). When the fluctuation varies those used for figure 7. It can be seen from the figure that, for

1845
J D Li

those of Morris and Foss (2003) based on numerical solution

for a ≈ 0. Unlike those of Freymuth (1979) and Morris and
1.1 Frequency response of Foss (2003), the present results show that the overheat ratio
hot wire only β = 0
does have some effect on attenuation.
1
This attenuation of the velocity fluctuations starts at
0.9
fairly low frequency and finishes at very high frequency.
Unfortunately, this frequency range covers most of today’s
0.8
turbulent measurements. It seems that the current hot-wire
calibration methods, both static and dynamic, would not be
0.7 able to capture this attenuation. The only possible solution
would be to correct the hot-wire results based on the theoretical
0.6 Frequency response ratio for attenuation results. This could be accomplished by measuring
the hot wire system (see text)
the spectrum of the velocity fluctuations first. The spectrum
0.5 will then be corrected by using the attenuation for a given
hot wire at the known Reynolds number. This would give an
0.4
102 103 104 105 106 107
estimation of the magnitude of the errors in turbulent intensity
ω (rad/sec)
results based on static or dynamic calibration, and if desired the
errors in the power law estimation. It is expected, for flows
Figure 17. The ratio of the frequency response at two different like turbulent boundary layers, that this error will increase
wave numbers. as the wall is approached since the frequency of the energy
containing eddies (after applying Taylor’s hypothesis) will be
β = π/2, the hot wire will underestimate the turbulence energy at reasonably high frequencies.
across the whole frequency range. At low frequency, this The solution for the fluctuating temperature is integrated
underestimation is from the insufficient spatial resolution. At to obtain the electrical resistance fluctuation of the hot
high frequencies, both the insufficient spatial resolution and wire. This resistance fluctuation is then connected to a
the frequency response from the CTHWA underestimate the simplified constant temperature hot-wire anemometer. The
turbulent energy. response of the anemometer is determined for a given velocity
A close look at figure 16 reveals that, at β = 0, the fluctuation. It is found that the cut-off frequency of
attenuation of velocity fluctuation at high frequencies is greater the combined hot-wire system is much less than the cut-
than that at β = 0. To see this more clearly, figure 17 shows off frequency of the anemometer itself as determined by
the frequency response ratio of the hot-wire system (including a square/sine wave test. This finding agrees with the
the hot wire and its anemometer) between β = π/2 and experimental results of Khoo et al (1998, 1999), Chew et al
β = 0. Again, the parameters used for the hot-wire system (1998) and Teo et al (2001), but disagrees with the numerical
are the same as in figure 7. Also shown in the figure is the results of Morris and Foss (2003). Because of this, questions
frequency response of the hot wire only at β = 0. It can be should be raised about the cut-off frequency determined by
seen from figure 16 that the capability for spatial resolution of using the square/sine wave. In contrast to the numerical
the hot wire at high frequencies is reduced in comparison with results of Morris and Foss (2003) for a particular hot wire at a
that at low frequencies. Because of this, it can be concluded single length to diameter ratio and a single Reynolds number,
that the resolution of the hot wire to turbulence of a given small the present theoretical results allow us to study systematically
length scale depends on its frequency with reduced resolution the effect on the cut-off frequencies of the hot-wire system
at high frequencies. from various parameters, such as the length to diameter ratios,
Reynolds numbers, the overheat ratios and the materials used
for the hot wire. In general, it is found that minimizing
9. Conclusions and discussion the conductive heat loss relative to the total heat loss can
increase the cut-off frequency of the hot-wire system. The
The equations for the mean and fluctuation temperatures along present results show that the cut-off frequency of the hot-
a hot wire have been solved analytically. These solutions wire system increases with increasing Reynolds number and
can be used to investigate the frequency response and spatial overheat ratio. These are consistent with the experimental
resolution of the hot-wire system theoretically for the first results of Khoo’s group but are different from those of Morris
time. and Foss (2003). The present results also show that increasing
The solution for the temperature fluctuation is first used the hot-wire length to diameter ratio and using hot wires of
to investigate the attenuation of the velocity fluctuation at high low thermal conductivity will also result in higher cut-off
frequency by the hot wire. It is found that this attenuation is frequency. Experimental results are required to confirm these
directly related to the heat conduction loss of the wire to its conclusions.
prongs. The higher the relative heat conduction loss is, the A preliminary investigation on the spatial resolution of
greater the attenuation is. The relative conductive heat loss the hot wire for fluctuations along the wire is presented. The
of the hot wire can be reduced by increasing 2l/d, exposing results show that spatial resolution depends strongly on 2l/d.
the wire at a higher Reynolds number, operating the hot wire Also, for the same wave number, the spatial resolution is better
at a higher overheat ratio or using hot wires with low thermal at low frequency than at high frequency.
conductivity. The current theoretical results agree with the To further investigate the spatial resolution of the hot wire,
results of Freymuth (1979) based on intuitive argument and a procedure similar to that given by Wyngaard (1968) needs to

1846
 Dynamic response of constant temperature hot-wire system

be adopted. In this case, for statistically stationary turbulent Freymuth P 1978 Theory of frequency optimisation for hot-film
flows, the frequency of the measured voltage can be connected anemometers J. Phys. E: Sci. Instrum. 11 177–9
Freymuth P 1979 Engineering estimation of heat conduction loss in
to the length scale in the streamwise direction by using Taylor’s
constant temperature thermal sensors TSI Q. 3 (3)
hypothesis. Wyngaard (1968) assumed that the turbulence is Hinze J O 1975 Turbulence 2nd edn (New York: McGraw Hill)
isotropic and using the three-dimensional spectra given by Pao Holman J P 1992 Heat Transfer 7th edn (New York: McGraw Hill)
(1965) as that of the fully resolved. In turbulent wall shear Khoo B C, Chew Y T and Li G L 1997 Effects of spatial resolution
flows, the turbulence will in general be non-isotropic, and thus on turbulence measurements in the very near-wall viscous
sublayer region Exp. Fluids 22 327–35
only qualitative conclusions can be drawn by using the spectra
Khoo B C, Chew Y T, Lim C P and Teo C J 1998 Dynamic response
for isotropic turbulence. Work along this line is currently of hot-wire anemometer: Part I. A marginally-elevated
under investigation. hot-wire probe for near wall velocity measurements Meas. Sci.
Technol. 9 751–63
Khoo B C, Chew Y T and Teo C J 2003 A model for the frequency
References response of a near-wall hot wire: velocity perturbation and
sine-wave voltage perturbation tests Exp. Therm. Fluid Sci. 27
Alfredsson P H, Johansson A V, Haritonidis J H and Eckelmann H 167–75
1988 The fluctuating wall shear stress and the velocity field in Khoo B C, Chew Y T, Teo C J and Lim C P 1999 Dynamic response
the viscous sublayer Phys. Fluids 31 1026–33 of hot-wire anemometer: Part III. Voltage-perturbation versus
Beljaars A C M 1976 Dynamic behavior of the constant temperature velocity testing for near wall hot-wire/film probes Meas. Sci.
anemometer due to thermal inertia of the wire Appl. Sci. Res. Technol. 19 152–69
32 509–18 Kreyszig E 1993 Advanced Engineering Mathematics 7th edn
Betchov R 1949 Theorie non-lineaire de l’anemometre a fil chaud (New York: Wiley)
Proc. K. Ned. Akad. Wet. 52 195–207 Lange C F, Dust F and Breuer M 1999 Wall effects on heat losses
Borgos J A 1980 A review of electrical testing of hot-wire and from hot-wires Int. J. Heat Fluid Flow 20 34–47
hot-film anemometers TSI Q. 3 (3) Li J D 2001 An investigation on the length to diameter ratio of
Bremhorst K and Gilmore G B 1978 Influence of end conduction on hot wire filament in turbulence measurements 14th
the sensitivity to stream temperature fluctuations of a hot-wire Australasian Fluid Mechanics Conf. (Adelaide University,
anemometer Int. J. Heat Mass Transfer 21 145–54 10–14 Dec. 2001)
Bruun H H 1995 Hot-Wire Anemometry: Principles and Signal Li J D, McKeon B J, Jiang W, Morrison J F and Smits A J 2004 The
Analysis (Oxford: Oxford University Press) response of hot wires in high Reynolds-number turbulent pipe
Champagne F H, Sleicher C A and Wehrmann O H 1967 Turbulence flow Meas. Sci. Technol. 15 789–98
measurements with inclined hot wires: Part 1. Heat transfer Ligrani P M and Bradshaw P 1987 Spatial resolution and
experiments with inclined hot wire J. Fluid Mech. 28 153–75 measurement of turbulence in the viscous sublayer Exp. Fluids
Chew Y T, Khoo B C, Lim C P and Teo C J 1998 Dynamic response 5 407–17
of hot-wire anemometer: Part II. A flush-mounted hot-wire and Morris S C and Foss J F 2003 Transient thermal response of a
hot-film probes for wall shear stress measurements Meas. Sci. hot-wire anemometer Meas. Sci. Technol. 14 251–9
Technol. 9 764–78 Pao Y H 1965 Structure of turbulent velocity and scalar field at large
Corrsin S 1963 Turbulence: experimental methods Handbuch der wavenumber Phys. Fluids 8 1063–108
Physik vol 9, part 2 (Berlin: Springer) pp 523–90 Perry A E 1982 Hot Wire Anemometry (New York: Oxford
Fand R M 1965 Heat transfer by forced convection from a cylinder University Press)
to water in cross flow Int. J. Heat Mass Transfer 8 995 Teo C J, Khoo B C and Chew Y T 2001 Dynamic response of
Freymuth P 1977a Frequency response and electronic testing for hot-wire anemometer: Part IV. Sine-wave voltage perturbation
constant-temperature hot-wire anemometers J. Phys. E: Sci. testing for near-wall hot-wire/film probes and the presence of
Instrum. 10 705–10 low-high frequency response characteristics Meas. Sci.
Freymuth P 1977b Further investigation of nonlinear theory for Technol. 12 37–51
constant-temperature hot-wire anemometers J. Phys. E: Sci. Wyngaard J C 1968 Measurement of small scale turbulence
Instrum. 10 710–3 structure with hot wires J. Phys. E: Sci. Instrum. 1 1105–8

1847