Chimney Effect on Natural Convection

Cooling of Electronic Components

B. Vermeersch(1) , G. De Mey(1) , M. Wjcik(2) , J. Pilarski(2) , M. Lasota(2) ,
J. Banaszczyk(1) , A. Napieralski(2) , M. De Paepe(3)
E-mail: bjorn.vermeersch@elis.ugent.be
(1)

Electronics & Information Systems, Ghent University, Sint Pietersnieuwstraat , Gent, Belgium.
Microelectronics & Computer Science, Technical University of Lodz, al. Politechniki , Lodz, Poland.
(3)
Flow, Heat and Combustion, Ghent University, Sint Pietersnieuwstraat , Gent, Belgium.
(2)

Summary The thermal impedance Zth (j) has been measured for a transistor fixed on a cooling fin. We have
put the setup inside a cilindrical chimney with an inner diameter around cm and various heights H ranging from
to cm. The enhancement of the natural convection cooling by the chimney is clearly demonstrated by means
of Nyquist plots of the thermal impedance. Depending on the insertion distance of the setup into the chimney, the
case-to-ambient thermal resistance Rth could be reduced by more than . We also propose a simple hydraulic
model for the chimney. The theoretical analysis predicts that 1/Rth is proportional to H 1/4 . The experimental
results agree quite well with this tendency.

Introduction

The use of a chimney to improve the natural air convection cooling of electronic components has
been the topic of only a few papers [][]. Nevertheless it should be pointed out that even a small
chimney can reduce the thermal resistance by . Although this is not spectacular at first sight,
in case the nominal junction temperature rise of a power component is C, a chimney can then
give rise to a reduction of C which has a non negligible influence on the reliability and the lifetime [, ]. A chimney has also the advantage of being a passive device. The major disadvantage is
that it needs relatively more space. However, it often happens that inside an existing rack or subrack
enough space is still available to insert a small chimney.
In this paper laboratory experiments will be described. The thermal impedance Zth (j) has been
measured as a function of the angular frequency for various heights of the chimney. If the impedance is shown graphically in a Nyquist plot, one can clearly see which part of the total thermal
resistance Rth = Zth (0) is due to the package and the cooling fin and which part corresponds to
the natural convection influenced by the chimney.

Experimental results

A schematic view of the experimental setup is shown on Fig. .

Chimneys with height H varying between and cm and a diameter D = 5 cm have been used.
A transistor was used as the heat source and mounted on a cooling fin and put at the bottom of the
chimney. The cooling fin having a height of mm was shifted partially into the chimney so that a
distance x was still outside the chimney. In our experiments x was used as an additional parameter
varying between x = 20 mm to x = +40 mm.
The thermal impedance is measured using the Tster equipment from MICRED company. From the
time t = 0 a constant power is dissipated into the junction of the transistor. The junction temperature T (t) is recorded continuously till steady state is obtained [][]. By Fourier techniques, the

3mm
H

D = 48 mm

chimney
thermally insulating material

(cardboard used in experiment)

front view

top view

Figure : Schematic layout of the chimney and transistor mounted on a cooling fin.

obtained transient heating curve or equivalently, the cooling curve can be converted into a complex thermal impedance Zth (j) []. Further details concerning the measurement and evaluation
are described in the provided references and will be omitted here.
Figure shows the thermal impedance for a chimney with a height H = 10 cm in a so called Nyquist
plot. This is a curve representing the imaginary part Im[Zth (j)] of the impedance versus its real part
Re[Zth (j)] using the angular frequency as a parameter.
Rth = 10.77 K/W
Rth = 10.69 K/W
Rth = 10.50 K/W
Rth = 9.67 K/W
Rth = 9.43 K/W

-2.5

-5

2.5

HF

Im(Zth(j)) [K/W]

MF

7.5

10

12.5

15

Re(Zth(j)) [K/W]

H = 10 cm

without
chimney

x = 0mm
x = 20mm

x = -20mm
x = 40mm

LF

Figure : Nyquist plot of the thermal impedance for various chimney positions (H = 10 cm).

All the results clearly show three circular arcs, corresponding to the low frequency (LF), mid frequency (MF) and high frequency (HF) range. By changing the distance x (Fig. ) only the LF arc is
changing. Hence it can be concluded that this arc corresponds to the natural convection cooling.
The MF part is related to the cooling fin on which the transistor is mounted while the HF part is
generated by the semiconductor chip and the package. The latter is approaching the origin under
an angle of , which is typical for planar heat sources [][].

On the horizontal axis the thermal resistance Rth between junction and ambient can be found at the
intersection with the horizontal axis, i.e. the impedance value at steady state Zth (0). To analyse the
natural convection cooling, it is more convenient to deal with Rth , as indicated on Fig. . Because
only the LF arc is determined by natural convection (with or without a chimney), Rth is defined
as the distance between the two intersection points of the LF arc and the horizontal axis. It must
be pointed out here that the measurement of the thermal impedance as a function of the angular
frequency offers us the possibility to define Rth unambiguously. From Fig. it is clear that the
best cooling is obtained for x = 20 mm and x = 0 mm, which correspond to the cases where the
cooling fin is half or just completely inserted inside the chimney. For comparison, the impedance
plot measured without a chimney is also shown on Fig. . The use of a chimney reduces Rth more
than for the optimal case.
Similar results are shown on Fig. for a chimney with a height H = 30 cm.
Rth = 10.77 K/W
Rth = 10.10 K/W
Rth = 9.66 K/W
Rth = 9.13 K/W
Rth = 9.04 K/W

-2.5

-5

2.5

HF

Im(Zth(j)) [K/W]

MF

7.5

10

12.5

15

Re(Zth(j)) [K/W]

H = 30 cm

without chimney
x = 40mm
x = -20mm
x = 20mm
x = 0mm

LF

Figure : Nyquist plot of the thermal impedance for various chimney positions (H = 30 cm).

The results are quite analogous to the previous ones. Again x = 20 mm and x = 0 mm turn out to
be the best cases. This time Rth has been reduced more than compared to the case without
a chimney.
Measurements have also been carried out for chimney heights of cm, cm, cm and cm.
All results were found to be similar as those shown on Figs. and .

Theoretical analysis

A chimney can be modelled by the equivalent hydraulic circuit shown on Fig. .

The hydraulic resistance of the chimney is obviously proportional to the height H and denoted by
H. is the resistance due to the outlet of the chimney. p is the buoyancy pressure source which
is also proportional to the height H. Hence p = H because it is assumed that due to the small
dimensions and the insulating material the chimney is made from, the entire chimney will be filled
with warm air. The air flow Q inside the chimney is then given by:
s
s
p
H
=
()
Q=
H +
H +

p = H

Figure : Equivalent hydraulic network for a chimney.

This air flow is also equal to the product of the cross-section area and the air speed v inside the
chimney. The latter is then hence easily found to be:
s
4
4
H
v =Q
=
()
2
2
D
D
H +
in which D is the diameter of the chimney. For the dimensions used in our experiments, the outlet
pressure losses turn out to be much greater than the losses inside the chimney or in other words
 H. Hence () can then be simplified to:
s

4
H
v

H
()
D 2

In order to estimate the convective heat transfer coefficient h from the air velocity v , the following
correlation is used [, ]:
ka
h = 0.664 Re 1/2 Pr 1/3
()
D
where Re = vD/ is the Reynolds number, Pr = 0.72 the Prandtl number for air, ka the thermal
conductivity of air ( 0.025 W/m-K) and the kinematic viscosity of air. It must be remarked here
that () is known as the correlation for forced convection at moderate Reynold numbers, whereas
the experimental setup is just a natural convection cooling. In this simplified theoretical analysis we
consider the combination of natural convection assisted by a chimney as a mechanism giving rise
to an air velocity v , as it would have been realised by a fan. In that respect, () can still be used to
estimate the heat transfer coefficient. For our investigation we only have to remember from () and
() that:

h v , v H or h = H 1/4
()
with  a proportionality factor. The values Rth , defined on Figs. and , can be roughly written as:
Rth =

1
(h + h0 )Sc + hr Sr

()

where Sc is the total area of the cooling fin subjected to convective cooling and h0 is the heat transfer coefficient in the absence of a chimney (H = 0). Sr is the so called radiative area of the cooling
fin, i.e. the area which radiates heat to the ambient. hr is the corresponding heat transfer coefficient
due to radiation only.
From () one gets obviously:
1
= (h + h0 )Sc + hr Sr = H 1/4 Sc + h0 Sc + hr Sr
Rth

()

Figure shows a plot of measured 1/Rth values versus H 1/4 for different insertion distances x.

0.120
0.115

1 / Rth

0.110
0.105

x = -20mm
x = 0mm
x = 20mm
x = 40mm

0.100
0.095
0.090
0.085
0.080
0.0

1/4

H
0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Figure : Thermal conductance as a function of the chimney height H.

According to () the theoretical model predicts a linear relationship. This tendency is observed relatively well in the measurements as seen on Fig. . Extrapolating for H = 0 (for x = 0) gives us
the value 1/Rth = 0.08 or Rth = 12.5 K/W. This value will be compared with () in case H = 0.
The area of the cooling fin has also been measured accurately. We obtained Sc = 0.00657 m2 . The
radiative area Sr was estimated by subtracting all surfaces radiating to each other and found to
be Sr = 0.00399 m2 . Using the following typical values h0 = 8 W/m2 K (natural convection) and
hr = 6 W/m2 K for the heat transfer coefficients, one gets then:
Rth =

1
= 13.05 K/W
h0 Sc + hr Sr

()

It must be emphasised here that the theory outlined so far is a largely simplified analysis so that a
precise agreement between theory and experiment cannot be expected. Still, the estimation obtained in () is quite close to the extrapolated value of 12.5 K/W derived from Fig. . The relation ()
gives also a good estimation how much can be gained by making a chimney higher.

Conclusions

We have investigated the influende of a chimney on the thermal impedance of electronic components. It was found that the low frequency part of the thermal impedance, corresponding to the
convective heat transfer, was changed by using a chimney. The higher frequency parts, related to
the chip and package, remain unaltered. As expected, the thermal resistance could be reduced
more than for the optimal case. A simple theoretical analysis revealed that the improvement
varies approximately proportional to H 1/4 , which was also confirmed experimentally. The practical
use depends of course on another very important parameter dealing with the availability of enough
space to insert a chimney inside an electronic rack or subrack.

Acknowledgements

B. Vermeersch is pursuing a PhD in Electronic Engineering at Ghent University, Belgium, as a Research Assistant for the Research Foundation Flanders (FWO - Vlaanderen). He wishes to thank FWO
for supporting the presented work. M. Wojcik, J. Pilarski and M. Lasota want to express their gratitude to the European Union for the financial support within the framework of the Erasmus Socrates
student exchange program.

References
[] Y. Kitamura, M. Ishizuka, S. Nakagawa. Numerical and experimental study of the effect of inclination
on the thermal behaviour of natural convection inside a thin electronic equipment casing. Proc. of the
Institution of Mechanical Engineers A: Journal of Power and Energy :, .
[] Y. Kitamura, M. Ishizuka. Chimney effect on natural air cooling of electronic equipment under inclination.
Journal of Electronic Packaging :, .
[] T.S. Fisher, K.E. Torrance. Free convection limits for pin fin cooling. Journal of Heat Transfer - Transactions
of the ASME :, .
[] A.V. Vairagar, S.G. Mhaisalkar, A. Krishnamoorthy. Electromigration behaviour of dual damascene Cu
interconnects - structure, width and length dependence. Microelectronics Reliability :, .
[] Q. Zhang, A. Dasgupta, P. Haswell. Isothermal mechanical durability of three selected Pb free solders:
Sn.Ag.Cu, Sn.Ag and Sn.Cu. Journal of Electronic Packaging :, .
[] V. Szkely, T. Van Bien. Fine stucture of heat flow path in semiconductor devices: a measurement and
identification method. Solid State Electronics :, .
[] V. Szkely. Thermal monitoring of microelectronic structures. Microelectronics Journal :, .
[] J. Banaszczyk, G. De Mey, M. Janicki, A. Napieralski, B. Vermeersch, P. Kawka. Dynamic thermal analysis
of a power amplifier. In Proc. THERMINIC, Nice, France, pp. , .
[] P. Kawka, G. De Mey, B. Vermeersch. Thermal characterisation of electronic packages using the Nyquist
plot of the thermal impedance. IEEE Trans. on Components and Packaging Technologies :,
.
[] B. Vermeersch, G. De Mey. Thermal impedance plots of micro scaled devices. Microelectronics Reliability
:, .
[] B. Vermeersch, G. De Mey. Influence of substrate thickness on thermal impedance of microelectronic
structures. Microelectronics Reliability :, .
[] B. Vermeersch, G. De Mey. Influence of thermal contact resistance on thermal impedance of microelectronic structures. Microelectronics Reliability :, .
[] F. Incropera, D. De Witt. Introduction to heat transfer. Wiley, New York, USA, p. , .
[] B. Staniszewski. Wymiana ciepla (in Polish). Panstwowe Wydawnictwo Naukowe, Warsaw, Poland, pp.
-, .