Hot-Wire Anemometry

•Purpose:
to measure mean and fluctuating velocities in fluid flows
Principles of operation

 Consider a thin wire mounted to supports and exposed to a velocity U.

When a current is passed through wire, heat is generated (I2Rw). In
equilibrium, this must be balanced by heat loss (primarily convective) to
the surroundings.

• If velocity changes, C urrent I Sensor dim ensions:

convective heat length ~1 m m
diam eter ~5 m icrom eter
transfer coefficient
will change, wire
temperature will
change and
W ire supports
eventually reach a Velocity U
(S t.S t. needles)
new equilibrium. S ensor (thin w ire)
Governing equation I


dE
Governing Equation: W  H
dt

E = thermal energy stored in wire

E = CwTs
Cw = heat capacity of wire
W = power generated by Joule heating
W = I2 Rw
recall Rw = Rw(Tw)
H = heat transferred to surroundings
 Governing equation II

 Heat transferred to surroundings

( convection to fluid
H + conduction to supports
+ radiation to surroundings)

 Convection  Qc = Nu · A · (Tw -Ta)

Nu = h ·d/kf = f (Re, Pr, M, Gr,a ),
Re = r U/m

 Conduction f(Tw , lw , kw, Tsupports)

 Radiation f(Tw4 - Tf4)

Simplified static analysis I

 For equilibrium conditions the heat storage is zero:

dE
 O W  H
dt
and the Joule heating W equals the convective heat transfer H

 Assumptions
 Radiation losses small
 Conduction to wire supports small
 Tw uniform over length of sensor
 Velocity impinges normally on wire, and is uniform over its entire length, and
also small compared to sonic speed.
 Fluid temperature and density constant
Simplified static analysis II

Static heat transfer:

W = H I2Rw = hA(Tw -Ta)  I2Rw = Nukf/dA(Tw -Ta)
h = film coefficient of heat transfer
A = heat transfer area
d = wire diameter
kf = heat conductivity of fluid
Nu = dimensionless heat transfer coefficient

Forced convection regime, i.e. Re >Gr1/3 (0.02 in air) and Re<140 

Nu = A1 + B1 · Ren = A2+ B2 · Un
I2Rw2 = E2 = (Tw -Ta)(A + B · Un) “King’s law”

The voltage drop is used as a measure of velocity.

Hot-wire static transfer function

 Velocity sensitivity (King’s law coeff. A = 1.51, B = 0.811, n = 0.43)

2,4 5

dU/dE/U volts^-1
2,2
E volts 4

3
1,8

1,6 2
5 10 15 20 25 30 35 40 5 10 15 20 25 30 35 40
U m /s U m /s

Output voltage as fct. of velocity Voltage derivative as fct. of velocity

 Directional response I

Probe coordinate system

U
x
Uy

Uz

 Ux z

Velocity vector U is decomposed into normal Ux, tangential Uy and binormal Uz

components.
Directional response II

 Finite wire (l/d~200) response includes yaw and pitch sensitivity:

U2eff(a) = U2(cos2a + k2sin2a) q =0
U2eff(q ) = U2(cos2q +h2sin2q ) a =0
where:
k , h = yaw and pitch factors
a , q = angle between wire normal/wire-prong plane, respectively, and
velocity vector

 General response in 3D flows:

U2eff = Ux2 + k2Uy2 + h2Uz2
Ueff is the effective cooling velocity sensed by the wire and deducted from the
calibration expression, while U is the velocity component normal to the wire
Directional response III

 Typical directional response for hot-wire probe

(From DISA 1971)

Directional response IV

 Yaw and pitch factors k1 and k2 (or k and h) depend on velocity and flow
angle

(From Joergensen 1971)

Probe types I

 Miniature Wire Probes

Platinum-plated tungsten,
5 mm diameter, 1.2 mm length

 Gold-Plated Probes
3 mm total wire length,
1.25 mm active sensor
copper ends, gold-plated
Advantages:
- accurately defined sensing length
- reduced heat dissipation by the prongs
- more uniform temperature distribution
along wire
- less probe interference to the flow field
Probe types II

For optimal frequency response, the probe should have as small a thermal
inertia as possible.

Important considerations:
• Wire length should be as short as possible (spatial resolution;
want probe length << eddy size)
• Aspect ratio (l/d) should be high (to minimise effects of end losses)
• Wire should resist oxidation until high temperatures (want to operate
wire at high T to get good sensitivity, high signal to noise ratio)
• Temperature coefficient of resistance should be high (for high
sensitivity, signal to noise ratio and frequency response)
• Wires of less than 5 µm diameter cannot be drawn with reliable
diameters
Probe types III

 Film Probes
Thin metal film (nickel) deposited on quartz
body. Thin quartz layer protects metal film
against corrosion, wear, physical damage,
electrical action

 Fiber-Film Probes
“Hybrid” - film deposited on a thin
wire-like quartz rod (fiber) “split fiber-film
probes.”
Probe types IV

 X-probes for 2D flows

2 sensors perpendicular to each other. Measures within
±45o.

 Split-fiber probes for 2D flows

2 film sensors opposite each other on a quartz cylinder.
Measures within ±90o.

 Tri-axial probes for 3D flows

3 sensors in an orthogonal system. Measures within
70o cone.
Hints to select the right probe

 Use wire probes whenever possible

ü relatively inexpensive
ü better frequency response
ü can be repaired

 Use film probes for rough environments

ü more rugged
ü worse frequency response
ü cannot be repaired
ü electrically insulated
ü protected against mechanical and
chemical action
Modes of anemometer operation

Constant Current (CCA)

Constant Temperature (CTA)

Constant current anemometer CCA

• Principle:
Current through
sensor is kept
constant

• Advantages:
- High frequency
response

• Disadvantages:
- Difficult to use
- Output decreases with velocity
- Risk of probe burnout
Constant Temperature Anemometer CTA I

 Principle:
Sensor resistance
is kept constant by
servo amplifier

 Advantages:
- Easy to use
- High frequency
response
- Low noise
- Accepted standard

 Disadvantages:
- More complex circuit
Constant temperature anemometer CTA II

 3-channel StreamLine with Tri-axial

wire probe 55P91
Modes of operation, CTA I
 Wire resistance can be
written as:
Rw = Ro(1+a o(Tw-To))
Rw = wire hot resistance
Ro = wire resistance at To
ao = temp.coeff. of resistance
Tw = wire temperature
To = reference temperature

 Define: “OVERHEAT RATIO” as:

a = (Rw-Ro)/Ro = a o(Tw-T0)

 Set “DECADE” overheat resistor as: RD = (1+a)Rw

Modes of operation, CTA II

 The voltage across wire is given by:

E2 = I2Rw2 = Rw(Rw - Ra)(A1 + B1Un)
or as Rw is kept constant by the servoloop:
E2 = A + BUn
2,4

 Note following comments 2,2

to CTA and to CCA:

E volts
2
- Response is non-linear:
- CCA output decreases 1,8

- CTA output increases 1,6

5 10 15 20 25 30 35 40
- Sensitivity decreases U m/s
with increasing U
CTA output as fct. of U
Dynamic response, CCA I

Hot-wire Probes:
 For analysis of wire dynamic response, governing equation includes the term due
to thermal energy storage within the wire:
W = H + dE/dt
The equation then becomes a differential equation:
I2Rw = (Rw-Ra)(A+BUn) + Cw(dTw/dt)
or expressing Tw in terms of Rw:
I2Rw = (Rw-Ra)(A+BUn) + Cw/a oRo(dRw/dt)
Cw = heat capacity of the wire
ao = temperature coeff. of resistance of the wire
Dynamic response, CCA II

Hot-wire Probes:
The first-order differential equation is characterised by a single time constant t :
t = Cw/(aoRo(A+BU n)

The normalised transfer function can be expressed as:

Hwire(f) = 1/(1+jf/fcp)

Where fcp is the frequency at which the amplitude damping is 3dB (50% amplitude
reduction) and the phase lag is 45o.
Frequency limit can be calculated from the time constant:
fcp = 1/2pt
Dynamic response, CCA III

• Hot-wire Probes:
Frequency response of film-probes is mainly determined by the thermal
properties of the backing material (substrate).
The time constant for film-probes becomes:

t = (R/R0)2F2rsCsks/(A+BUn)2
rs = substrate density
Cs = substrate heat capacity
ks = substrate heat conductivity

and the normalised transfer function becomes:

Hfilm(f) = 1/(1+(jf/fcp)0.5)
Dynamic response, CCA IV

Velocity
A m plitude re sponse and ph ase lag
U 2

6 dB/octave

3 dB
30°
U 1

Tim e

R esistan ce
R 1

0.63 (R -R )
1 2

Frequency
R U ppe r frequen cy lim it f = ½ 
2

Tim e


 Dynamic characteristic may be described by the response to

- Step change in velocity or
- Sinusoidal velocity variation
Dynamic response, CCA V

 The hot-wire response characteristic is specified by:

(From P.E. Nielsen

and C.G. Rasmussen,
1966)

For a 5 µm wire probe in CCA mode t ~ 0.005s, typically.

(Frequency response can be improved by compensation circuit)
Dynamic response, CTA I

 CTA keeps the wire at constant

temperature, hence the effect of thermal
inertia is greatly reduced:
Time constant is reduced to
t CTA = t CCA/(2aSRw)
where
a = overheat ratio
S = amplifier gain
Rw = wire hot resistance

 Frequency limit:
fc defined as -3dB amplitude
(From Blackwelder 1981)
damping
Dynamic response, CTA II

• Typical frequency response of 5 mm wire probe (Amplitude

damping and Phase lag):

(From Dantec MT)

Phase lag is reduced by frequency dependent gain (-1.2 dB/octave)

Velocity calibration (Static cal.)

 Despite extensive work, no universal

expression to describe heat transfer from
hot wires and films exist.
 For all actual measurements, direct
calibration of the anemometer is
necessary.
Velocity calibration (Static cal.) II

 Calibration in gases (example low turbulent free jet):

Velocity is determined from

isentropic expansion:

Po/P = (1+(g -1)/2M 2)g /(g- -1)

a0 = (g RT0 )0.5
a = ao/(1+(g -1)/2M 2)0.5
U = Ma
Velocity calibration (Static cal.) III

• Film probes in water

- Using a free jet of liquid
issuing from the bottom of
a container
- Towing the probe at a
known velocity in still
liquid
- Using a submerged jet
 Typical calibration curve

 Wire probe calibration with curve fit errors

E 1 v.U E rro r (% )
2.340 0.500

2.218 0.300

2.096 0.100
E 1 (v) E rror (% )
1.975 -0.10 0

1.853 -0.30 0

1.731 -0.50 0
4.0 76 11.1 2 1 8.17 25 .2 2 32.2 7 39.32 4.076 11.12 18 .17 25 .2 2 3 2 .2 7 39.32
U velo city U v elocity

(Obtained with Dantec 90H01/02)Calibrator)

Curve fit (velocity U as function of output voltage E):

U = C0 + C1E + C2E2 + C3E3 + C4E4
Dynamic calibration/tuning I

 Direct method

Need a flow in which sinusoidal velocity variations of known amplitude

are superimposed on a constant mean velocity

- Microwave simulation of turbulence (<500 Hz)

- Sound field simulation of turbulence (>500 Hz)
- Vibrating the probe in a laminar flow (<1000Hz)

All methods are difficult and are restricted to low frequencies.

Dynamic calibration/tuning II

 Indirect method, “SINUS TEST”

Subject the sensor to an electric sine wave which simulates an instantaneous
change in velocity and analyse the amplitude response.

10
3

10
A m p li t u d e ( m V r m s )

A m p li t u d e ( m V r m s )
10
2

10
2
-3 dB
-3 dB

10 10

1 1
2 3 4 5 6 2 3 4 5 6

10 10 10 10 10 10 1 10 10 10 10 10 10
Frequency (Hz) Frequency (Hz)

Typical Wire probe response Typical Fiber probe response

Dynamic calibration/tuning III

 Indirect method “SQUARE

WAVE TEST”
Subject the sensor to an electric f =
1
c
1.3 
sine wave which simulates an 0.97 h
w

h
instantaneous change in velocity
and analyse the shape of the
anemometer output
t

w
0.15 h

(From Bruun 1995)

For a wire probe (1-order probe response):

Frequency limit (- 3dB damping): fc = 1/1.3 t
Dynamic calibration

Conclusion:
 Indirect methods are the only ones applicable in practice.
 Sinus test necessary for determination of frequency limit for fiber and film probes.
 Square wave test determines frequency limits for wire probes. Time taken by the
anemometer to rebalance itself is used as a measure of its frequency response.
 Square wave test is primarily used for checking dynamic stability of CTA at high
velocities.
 Indirect methods cannot simulate effect of thermal boundary layers around sensor
(which reduces the frequency response).
Disturbing effects (problem sources)

 Anemometer system makes use of heat transfer from the probe

Qc = Nu · A · (Tw -Ta)
Nu = h · d/kf = f (Re, Pr, M, Gr,a ),

 Anything which changes this heat transfer (other than the flow variable
being measured) is a “PROBLEM SOURCE!”

 Unsystematic effects (contamination, air bubbles in water, probe

vibrations, etc.)

 Systematic effects (ambient temperature changes, solid wall proximity,

eddy shedding from cylindrical sensors etc.)
Problem sources
Probe contamination I
 Most common sources:
- dust particles
- dirt
- oil vapours
- chemicals
 Effects:
- Change flow sensitivity of sensor (DC
drift of calibration curve)
- Reduce frequency response

 Cure:
- Clean the sensor
- Recalibrate
Problem Sources
Probe contamination II
 Drift due to particle 20
contamination in air

(Um-Uact)/Uact*100%
10
5 mm Wire, 70 mm Fiber and 1.2
mm SteelClad Probes 0
w ire

-10 fiber

steel-clad
-20
0 10 20 30 40 50
U (m /s) Poly. (steel-
clad)
Poly. (fiber)

(From Jorgensen, 1977)

Wire and fiber exposed to unfiltered air at 40 m/s in 40 hours

Steel Clad probe exposed to outdoor conditions 3 months during winter
conditions
Problem Sources
Probe contamination IV
 Low Velocity
- slight effect of dirt on heat transfer
- heat transfer may even increase!
- effect of increased surface vs. insulating effect
 High Velocity
- more contact with particles
- bigger problem in laminar flow
- turbulent flow has “cleaning effect”
 Influence of dirt INCREASES as wire diameter DECREASES
 Deposition of chemicals INCREASES as wire temperature
INCREASES

* FILTER THE FLOW, CLEAN SENSOR AND RECALIBRATE!

Problem Sources
Probe contamination III
 Drift due to particle contamination in water
Output voltage decreases with increasing dirt deposit

10

% voltage reduction
theory
1 fiber
w edge

0,1
0,001 0,01 0,1 1
Dirt thicknes versus sensor
diam eter, e/D

(From Morrow and Kline 1971)

Problem Sources
Bubbles in Liquids I
 Drift due to bubbles in water

(From C.G.Rasmussen 1967)

In liquids, dissolved gases form bubbles on sensor, resulting in:

- reduced heat transfer
- downward calibration drift
Problem Sources
Bubbles in Liquids II
e
 Effect of bubbling on
portion of typical
calibration curve

 Bubble size depends on

- surface tension
- overheat ratio
- velocity

 Precautions
155 175 195 cm /sec
- Use low overheat!
- Let liquid stand before use! (From C.G.Rasmussen 1967)

- Don’t allow liquid to cascade in air!

- Clean sensor!
Problem Sources (solved)
Stability in Liquid Measurements
 Fiber probe operated stable in water

(From Bruun 1996)

- De-ionised water (reduces algae growth)

- Filtration (better than 2 mm)
- Keeping water temperature constant (within 0.1oC)
 Problem sources
Eddy shedding I
• Eddy shedding from cylindrical sensors
Occurs at Re ~50

(From Eckelmann 1975)

Select small sensor diameters/ Low pass filter the signal
 Problem Sources
Eddy shedding II
• Vibrations from prongs and probe supports:
- Probe prongs may vibrate due to eddy shedding from them or
due induced vibrations from the surroundings via the probe
support.
- Prongs have natural frequencies from 8 to 20 kHz
Always use stiff and rigid probe mounts.
 Problem Sources
Temperature Variations I
• Fluctuating fluid temperature
Heat transfer from the probe is proportional to the temperature
difference between fluid and sensor.
E2 = (Tw-Ta)(A + B·Un)
As Ta varies:
- heat transfer changes
- fluid properties change
Air measurements:
- limited effect at high overheat ratio
- changes in fluid properties are small
Liquid measurements effected more, because of:
- lower overheats
- stronger effects of T change on fluid properties
 Problem Sources
Temperature Variations II
• Anemometer output depends on both velocity and
temperature
Hot-wire calibrations at diff. temperatures Relative velocity error for 1C temp. increase

(From Joergensen and Morot1998)

2,4 -1,5
2,3
T=20 -1,7
2,2
2,1 T=25 -1,9
2,0
T=30 -2,1
1,9 Tdiff=10 C
1,8 T=35 -2,3
1,7 T=40 -2,5
1,6
1,5 -2,7
5 10 15 20 25 30 35 40 0 10 20 30 40

When ambient temperature increases the velocity is measured too low,

if not corrected for.
 Problem Sources
Temperature Variations III
Film probe calibrated at different temperatures
 Problem Sources
Temperature Variations IV
• To deal with temperature variations:
- Keep the wire temperature fixed (no overheat adjustment), measure
the temperature along and correct anemometer voltage prior to
conversion
- Keep the overheat constant either manually, or automatically using a
second compensating sensor.
- Calibrate over the range of expected temperature and monitor
simultaneously velocity and temperature fluctuations.
 Measurements in 2D Flows I

X-ARRAY PROBES (measures within ±45o with respect to probe axis):

• Velocity decomposition into the (U,V) probe coordinate system

U = U1·cos 1 + U2·cos2

V = U1·sin1 - U2·sin 2

U = U1·cos1 + U2·cos2

V = U1·sin1 - U2·sin2

where U1 and U2 in wire coordinate system are found by solving:

Ucal1 ·(1+k1 )·(cos(90 - 1)) = k1 U1 + U2
2 2 2 2 2 2

Ucal2 ·(1+k2 )·(cos(90 - 2)) = U1 + k2 U2

2 2 2 2 2 2

Ucal1 ·(1+k1 )·(cos(90 - 1)) = k1 U1 + U2

2 2 2 2 2 2

Ucal2 ·(1+k2 )·(cos(90 - 2)) = U1 + k2 U2

2 2 2 2 2 2
 Measurements in 2D Flows II
• Directional calibration provides yaw coefficients k1 and k2

U c1,U c2 vs. A ngle K1,K 2 vs. An gle

34.68 3.0 00

29.14 0.6 00

23.59 0.2 00
U c1,Uc2 K1 ,K2

18.04 -0.20 0

12.49 -0.60 0

6.9 45 -1.00 0
-40.00 -24.0 0 -8.000 8.0 00 24.00 40.00 -40.0 0 -24.00 -8.00 0 8.0 00 24.00 40.00
An gle (d eg) An gle (deg)

(Obtained with Dantec 55P51 X-probe and 55H01/H02 Calibrator)

 Measurements in 3D Flows I

TRIAXIAL PROBES (measures within 70o cone around probe axis):

3
35 ° x

P ro be ste m
55 °

1
35 °
45 °

2
 Measurements in 3D Flows II

• Velocity decomposition into the (U,V,W) probe coordinate system

U = U1·cos54.74 + U2·cos54.74 + U3·cos54.74
V = -U1·cos45 - U2·cos135 + U3·cos90
W = -U1·cos114.09 - U2·cos114.09 - U3·cos35.26

where U1 , U2 and U3 in2 wire coordinate system are 2found by solving:

U1cal ·(1+k1 +h1 ) ·cos 35.264= k1 ·U1 + U2 + h1 ·U3
2 2 2 2 2 2 2

U2cal ·(1+k2 +h2 )·cos 35.264 = h2 ·U1 + k2 ·U2 + U3

2 2 2 2 2 2 2 2 2

U3cal ·(1+k3 +h3 )·cos 35.264 = U1 + h3 ·U2 + k3 ·U3

2 2 2 2 2 2 2 2 2

left hand sides are effective cooling velocities. Yaw and pitch
coefficients are determined by directional calibration.
 Measurements in 3D Flows III

• U, V and W measured by Triaxial probe, when rotated around its

axis. Inclination between flow and probe axis is 20o.

5
Umeas
4
Velocity component, m/s

Vmeas
3
Wmeas
2 Res,meas
1 Uact
0,15
0 Vact

Meas. - Act. vel., m/s

0,10
Wact
-1 0,05
Up-Uact
Res,act
0,00
-2 Vp-Vact
0 30 60 90 120 150 180 210 240 270 300 330 360 -0,05 Wp-Wact
-0,10
Roll angle.
-0,15
0 60 120 180 240 300 360
Roll angle

(Obtained with Dantec Tri-axial probe 55P91 and 55H01/02 Calibrator)

 Measurement at Varying Temperature
Temperature Correction I
• Recommended temperature correction:
Keep sensor temperature constant, measure temperature and correct
voltages or calibration constants.
I) Output Voltage is corrected before conversion into velocity
0.5
Ecorr = ((Tw- Tref)/(Tw- Tacq)) Eacq.
- This gives under-compensation of approx. 0.4%/C in velocity.

Improved correction:
0.5(1±m)
Ecorr = ((Tw- Tref)/(Tw- Tacq)) Eacq.

Selecting proper m (m= 0.2 typically for wire probe at a = 0.8) improves
compensation to better than ±0.05%/C.
 Measurement at Varying Temperature
Temperature Correction II
• Temperature correction in liquids may require correction of
power law constants A and B:
(1±m) 0.2
Acorr = (((Tw-To)/(Tw-Tacq)) ·(kf0/kf1)·(Prf0/Prf1) ·A0
(1±m)
Bcorr = ((Tw-To)/(Tw-Tacq)) ·
(kf0/kf1)·(Prf0/Prf1) ·(f1/f0)n·(f0/f1) ·B0
0.33 n

In this case the voltage is not corrected

 Data acquisition I

• Data acquisition, conversion and reduction:

Requires digital processing based on

- Selection of proper A/D board
- Signal conditioning
- Proper sampling rate and number of samples
 Data acquisition II

A/D boards convert analogue signals into digital information (numbers)

They have following main characteristics:

• Resolution:
- Min. 12 bit (~1-2 mV depending on range)
• Sampling rate:
- Min. 100 kHz (allows 3D probes to be sampled with approx. 30 kHz
per sensor)
• Simultaneous sampling:
- Recommended (if not sampled simultaneously there will be phase lag
between sensors of 2- and 3D probes)
• External triggering:
Recommended (allows sampling to be started by external event)
 Data acquisition III

Signal Conditioning of anemometer output

EG
A ne m o m e ter O ffs e t A m plifie r

G (E (t)-E off)
E (t)-E off
E (1 )

t t

t
(From Bruun 1995)

• Increases the AC part of the anemometer output and improves

resolution:
EG(t) = G(E(t) - Eoff )

• Allows filtering of anemometer

- Low pass filtering is recommended
- High pass filtering may cause phase distortion of the signal
 Data acquisition IV

Sample rate and number of samples

• Time domain statistics (spectra) require sampling 2 times the highest
frequency in the flow
• Amplitude domain statistics (moments) require uncorrelated samples.
Sampling interval min. 2 times integral time scale.
• Number of samples shall be sufficient to provide stable statistics
(often several thousand samples are required)
Proper choice requires some knowledge about the flow aforehand
It is recommended to try to make autocorrelation and power spectra
at first as basis for the choice
 CTA Anemometry
Steps needed to get good measurements:

• Get an idea of the flow (velocity range, dimensions, frequency)

• Select right probe and anemometer configuration
• Select proper A/D board
• Perform set-up (hardware set-up, velocity calibration, directional calibration)
• Make a first rough verification of the assumptions about the flow
• Define experiment (traverse, sampling frequency and number of samples)
• Perform the experiment
• Reduce the data (moments, spectra, correlations)
• Evaluate results
• Recalibrate to make sure that the anemometer/probe has not drifted