Skip to main content

Full text of "NASA Technical Reports Server (NTRS) 20080004810: Augmented thermal bus wih multiple thermoelectric devices individually controlled"

See other formats


United States Patent [ 19 ] [li] Patent Number: 5,349,821 


[ 45 ] Date of Patent: Sep. 27, 1994 



[75] Inventor: Dean S. Schrage, Strongsville, Ohio 

[73] Assignee: The United States of America as 

represented by the Administrator of 

the National Aeronautics and Space 
Administration, Washington, D.C. 

[21] Appl. No.: 81,891 

[22] Filed: Jun. 25, 1993 

[51] Int. Cl.* F25B 21/02 

[52] U.S. a 62/3.7; 62/3.2 

[58] Field of Search 62/3.2, 3.7 

[56] References Cited 


2,844,638 7/1958 Lindenblad 136/4 

2,903,857 9/1959 Lindenblad 62/3 

3,304,726 2/1967 Beck 62/3.7 

3,438,214 4/1969 Schmittle 62/3 

3,481,393 12/1969 Chu 165/80 

4,310,047 1/1982 Branson 164/64 

4,610,142 9/1986 Davis 136/4 

4,848,090 7/1989 Peters 62/33 

5,022,928 6/1991 Buist 136/212 

5,269,146 12/1993 Kerner 62/3.7 

Primary Examiner — John M. Sollecito 

Attorney >, Agent, or Firm — Gene E. Shook; Guy M. 



The present invention is directed to an augmented ther- 
mal bus. In the present design a plurality of thermo- 
electric heat pumps are used to couple a source plate to 
a sink plate. Each heat pump is individually controlled 
by a model based controller. The controller coordinates 
the heat pumps to maintain isothermality in the source. 

5 Claims, 6 Drawing Sheets 

U.S. Patent 

Sep. 27, 1994 

Sheet 1 of 6 


FIG. 1 

U.S. Patent 

Sep. 27, 1994 

Sheet 3 of 6 


FIG. 3 

U.S. Patent 

Sep. 27, 1994 

Sheet 4 of 6 


FIG. 4 

Sheet 5 of 6 


G. 5 

U.S. Patent 

Sep. 27, 1994 

Sheet 6 of 6 



5 , 349,821 




The invention described herein was made in the per- 
formance of work under a NASA contract and is sub- 
ject to the provisions of Section 305 of the National 1Q 
Aeronautics & Space Act 1958, as amended, (42 U.S.C 


The present invention is directed to a thermal bus for | 5 
dissipating heat by using a plurality of individually con- 
trolled thermo-electric heat pumps(TEHP). Each heat 
pump individually controls a region on a source. The 
orchestrated control of all the TEHP unit, is performed 
by a model based controller. 20 

In thermal bus arrangements found in the prior art, a 
baseplate housing electronics is coupled to a coldplate 
through a thermo-electric heat pump. Integral with the 
coldplate is a fluid loop attached to radiator panels 
which discharges the energy convected at the cold- 25 
plate. The working fluid in the coldplate acts as a shunt 
to couple the electronics to remotely located radiator 
panels. Each component has an associated thermal resis- 
tance which, as a function of design, is a measure of the 
temperature drop across that component for a given 30 
heat load. Waste heat is dissipated in these systems 
through conduction. The waste heat in prior art systems 
is not upgraded to a higher temperature, therefore these 
systems require large radiator panels to reject the heat. 

While the prior art design of a thermal bus is simple 35 
and reliable, the overall effectiveness of the device is 
diminished by the radiator and liquid inventory weight 
restrictions, in conjunction with limited coldplate iso- 

When feedback control has been applied in the prior 
art thermal bus units, it has taken the form of PID con- 
trollers. Although the PID controller takes temperature 
variation (set point temperature minus the actual) into 
account, in an attempt to maintain isothermality, it is a ^ 
limited control mechanism. 

Traditionally, these PID controllers have been used 
in the prior art, to offer a single point of control. The 
concept can be extended to several PID loops, thereby 
controlling several set points in the plane of a source. 
However, the extension of PID controllers in this fash- 
ion, does not compensate for conduction heating, in the 
plane of the source. The individual heat control actions 
between PID units would not be coordinated. There- 
fore, in performing heat control, thermal bus designs in 55 
the prior art do not take the heating or cooling that is 
provided by the other TEHP units, into account. As a 
result, these systems could only offer coupled heat con- 

It is, therefore, an object of the present invention to 60 
create a thermal bus with a higher level of isothermality 
in a heat source or baseplate. 

It is a further objective of the present invention to 
upgrade the waste heat produced by the system thereby 
reducing the surface area required for a radiator. 65 

It is still a further object of the present invention to 
individually control areas of a heat source relative to a 
sink by using a plurality of thermo-electric heat pumps. 


It is still yet a further object of the present invention 
to achieve decoupled temperature control of a thermal 
bus by using a model based feedback control system. 


U.S. Pat. No. 4,848,090 by Peters discloses an appara- 
tus and a method for controlling the temperature of a 
semiconductor device. 

U.S. Pat. No. 3,481,393 by Chu discloses a cooling 
system for modular packaged electronic components. 

U.S. Pat. No. 3,438,214 by Schmittle discloses a ther- 
moelectric temperature control system for cooling and 
heating of a substance flowing therethrough. 

U.S. Pat. No. 4,610,142 by Davis discloses an appara- 
tus and method for controlling the temperature of a 
reagent refrigerator. 

U.S. Pat. Nos. 2,844,638 and 2,203,857 by Lunden- 
blad, both disclose a thermoelectric heat pump which is 
made of a compact thin panel construction. 

U.S. Pat. No. 4,310,047 by Branson discloses a device 
for maintaining an objective at a given temperature by 
means of a thermoelectric heat pump held between a 
thermally conductive member and a heat sink. 

U.S. Pat. No. 5,022,928 to Burst discloses a TEHP 
comprising p-type and n-type semiconductor film con- 
ductive elements selectively patterned on substrates. 


The present invention is an augmentation of a con- 
ventional single-phase thermal bus with an interstitial 
thermo-electric heat pump(TEHP). The thermo-elec- 
tric heat pump is a solid-state direct energy conversion 
device. Since the TEHP does not have any moving 
parts it is structurally and thermally robust making it 
uniquely suited for temperature cooling in a hostile 

In the present invention, a modular thermal bus in- 
cludes target electronics which are mounted on a base- 
plate forming a source. A plurality of TEHP devices are 
compression mounted between the source and a sink. 
The sink has a fluid loop which is coupled to radiator 
panels. Heat is transported from the source, through the 
heat pumps to the sink where the fluid loop coupled 
between the sink and the radiator panels uses the radia- 
tor panels to dissipate the heat. 


FIG. 1 displays a schematic of a thermo-electric heat 

FIG. 2 displays a schematic of a TEHP-assisted ther- 
mal bus. 

FIG. 3 displays the connection between the cold 
plate of the TEHP-assisted thermal bus and a feedback 

FIG. 4 displays a conceptual graph of the TEHP- 
assisted thermal bus without TEHP assistance. 

FIG. 5 displays a conceptual graph of the TEHP- 
assisted thermal bus with TEHP assistance. 

FIG. 6 is a block diagram of the feedback control 
method performed in accordance with the present in- 


FIG. 1 displays the thermo-electric heat pump mod- 
ule. A hot junction 10 and a cold junction 22 sandwich 
the terminal leads 12, 14, 16, a p-type junction 18 and a 
n-type junction 20. The hot junction 10 is connected to 

5 , 349,821 


a baseplate or source, and the cold junction 22 is at- 
tached to a coldplate or sink. When power is applied 
between the terminal leads 12, 14 and 16, heat will flow 
across an adverse temperature gradient from the hot 
junction 10 to the cold junction 22 thereby dissipating 5 
the heat developed by the electronic components. 

FIG. 2 displays a schematic view of the TEHP- 
assisted thermal bus. A coldplate or sink 26 is coupled to 
a fluid loop formed by 60 and 62 which enables fluid to 
circulate through the coldplate, thereby dissipating 10 
heat. A plurality of quadrants 30, 32, 34, 36, 38 and 39 
house a plurality of thermo-electric heat pumps denoted 
by 40, 42, 44, 46, 48, and 49, respectively. A negative 
polarity busbar 64 is also attached to the coldplate 26, 
thereby enabling a negative voltage across the coldplate 15 

The thermo-electric heat pumps 40, 42, 44, 46, 48, and 
49 are sandwiched between the coldplate 26 and the 
baseplate or source 28, thereby offering a conductive 
heat pathway. Electronics are located in the matching 20 
quadrants 50, 52, 54, 56, 58 and 59 of the baseplate 28, 
thereby enabling each quadrant to be controlled by a 
TEHP unit denoted by 40, 42, 44, 46, 48 and 49, respec- 

A busbar for carrying a positive current 66 is also 25 
attached to the baseplate 28. The busbar for positive 
current 66 and the busbar for negative current 64 are 
connected through a power supply 68. By placing a 
positive polarity across the source or baseplate 28 and a 
negative polarity across the sink or coldplate 26, volt- 30 
age is applied across the TEHP devices 40, 42, 44, 46, 48 
and 49. The applied voltage would cause heat to flow 
from the source or baseplate 28 to the sink or coldplate 
26 across an adverse temperature gradient. 

FIG. 3 displays a schematic view of the coldplate 26 35 
with quadrants 30, 32, 34, 36, 38, and 39 each controlled 
by a controller 109, through connections 80, 82, 84, 86, 

88, and 89 respectively. 

FIG. 4 displays a schematic flow diagram of the 
methodology used in the controller 109 presented in 40 
FIG. 3. The controller 109 receives an output of tem- 
perature readings generated by thermocouple or therm- 
istors located in the quadrants 50, 52, 54, 56, 58 and 59 
of the baseplate, 28 shown in FIG. 3. The temperature 
readings are transported through the hard-wired con- 45 
nections 90, 92, 94, 96, 98 and 99, depicted in FIG. 3. 
The temperature readings are inputed into the control- 
ler 109 shown in FIG. 4, at location 160, as an array of 
process temperatures. 

According to the present invention, temperature, 50 
voltage and current measurements are taken of each 
quadrant of the source. These parameters along with 
several other parameters define the process of the aug- 
mented thermal electric heat pump, which is denoted by 
150. 55 

A mathematically based model of the augmented 
thermal bus is also maintained by the controller 109. 
The modelled process is denoted by 155. Both the ac- 
tual process 150 and the modelled process 155 produce 
arrays of temperature outputs at 160 and 165, respec- 60 
tively. The process temperature array at 160 and the 
modelled temperature array at 165, are summed at 170. 
When there is a difference between the process temper- 
ature array 160 and the modelled temperature array at 
165, an error array occurs at 175. This error array repre- 65 
sents a deviation from the expected operation of the 
system as described by 155. The error in temperature 
175 is then summed with a set point temperature array 


at 115. The set point temperature array is an array of 
temperature values set for quadrants 50, 52, 54, 56, 58 
and 59. The expected inverse model array 120 repre- 
sents a deviation from the set point temperature of the 
baseplate. The expected inverse model array 120 serves 
as the input for the inverse of the modelled process 125. 
The inverse of the modelled process augmented with a 
conditioning filter, 125 produces an output 130 that tries 
to correct for the expected inverse model array 120. 
The output from the inverse of the modelled process 
130, is an array of voltages or currents used to control 
the TEHP devices, thereby compensating for the ex- 
pected inverse model array, denoted by 120. The volt- 
age change at 130 will serve as an input to both the 
actual process at 140 and the modelled process at 145. 
Changing the voltage or current inputs 140 and 145 will 
result in an increase of the pumping capacity of the 
TEHP units, to accommodate for the change in isother- 

In the case where there is no disturbance in the opera- 
tion of the augmented thermal bus, the process tempera- 
ture array at 160 would equal the model temperature 
array at 165. When these two values are summed there 
would not be any error at 170 therefore, there would be 
no error temperature at 175. Also, the value fed into the 
inverse model 125 would be equal to the value of the set 
point temperature array 110. Since the set point temper- 
ature array at 110 are the values that the inverse model 
125 expects as input, there is no variation in the output 
voltage at 130. Consequently, there would not be any 
change in the inputs to the process at 140 or the mod- 
elled process at 145. In the scenario given above, the the 
pumping capacity of the individual TEHP units would 
not be altered. 

As an example of a case where a disturbance does 
occur, in FIG. 3, a first quadrant 50, a second quadrant 
52 and a third quadrant 54 are controlled by a first 
TEHP unit 40, a second TEHP unit 42 and a third 
TEHP unit 44. If the second TEHP unit 42 were to fail, 
it would cause quadrant 52 to increase in temperature. 
Due to conduction heating in the plane of the source, 
quadrants 50 and 54 would also increase in temperature. 
The thermocouple senses this activity and reports a T1 
representing the first temperature, a T2 representing a 
second temperature and a T3 representing a third tem- 
perature back to the controller through the hard wired 
connection 90, 92 and 94, respectively. These three 
values, T1 the temperature of quadrant 50, T2 the tem- 
perature of quadrant 52 and T3 the temperature of 
quadrant 54 would be fed back into our model at 160. 
When the process temperature array [Tl, T2, T3] at 160 
are combined with the model 160 temperature array 
[Tim, T2m, T3m] at location 170, the first temperature 
of the process (Tl) will cancel the first temperature of 
the model (Tim). The third temperature of the process 
(T3) will cancel the third temperature of the modelled 
process (T3m). However, the second temperature of the 
model (T2), will not cancel the second temperature of 
the model (T2m). Therefore there will be a non-zero 
value T2e at 175, which represents an error in the aug- 
mented thermal bus. The error temperature array [0, 
T2e, 0] will be combined at 115 with the set point tem- 
perature array [Tlsp, T2sp, T3sp], As a result of the 
combination at 115, the inverse model 125 will note a 
deviation in the set point temperature for the second 
quadrant, since T2e will be subtracted from T2sp. Since 
the inverse model 125 will receive a set point tempera- 
ture that is lower than what it expects (T2sp-T2e) for 


the second quadrant, the inverse model 125 would vary 
the voltage outputs to the TEHP units surrounding that 
quadrant would require a contingency instruction; after 
knowing that TEHP/q2 failed the controller would 
define new objectives such as to control the average 
temperature of all quadrants in the vicinity of the failed 
quadrant to the average set point. 

In our example, these two quadrants are physically 
represented by unit 40 and 44 of FIG. 4. By increasing 
the pumping capacity of units 40 and 44, voltage varia- 
tions would be fed back into the process 150, through 
input 140 and to the modelled process 155, through 
input 145. Assuming that the actual process 150 was just 
combined with the set point temperature back into the 
inverse process 125, we would experience coupled tem- 
perature control. However, by combining the modelled 
temperature 165 with the actual temperature 160 and 
then feeding the temperature error 175 back into the 
inverse of the model 125, decoupled cooling is achieved 
in the augmented thermal bus system. 

A mathematical description would include the fol- 

T1 — temperature in the first quadrant 
T2 — temperature in the second quadrant 
T3 — temperature in the third quadrant 
ql — the inverse model of the first region 
q2 — the inverse model of the second region 
q3 — the inverse model of the third region 
x — a scaler multiple 
y — a scaler multiple 

In the coupled case, the process temperature array is 
multiplied by the matrix of the process 150 to produce 
the following output: 














<?3 _ 



In this coupled representation, the inverse process 
will respond to Tl, T2 and T3 individually through ql, 
q2, and q3 without ever accounting for the temperature 
changes or disturbances occurring in other regions. 
However, by adding the inverse model 125 and devel- 
oping q as the inverse of the modelled process 155, the 
following conceptual formulation would result: 

\ ... 



X 912 

}913 ~ 

2 ... 




















_ 71^31 





Where ql2 and ql3 are some multiple of q22 and q33, 

The result is a decoupled cooling process. The tem- 
perature in quadrant 1 (Tl), the temperature in quadrant 
(T2) and the temperature in quadrant three (T3) will be 
adjusted taking the other temperature regions into ac- 
count, as displayed by the resulting decoupled vector, 
inequation 1 (Tlqn-bxT2qi2-hyT3qi3). 

FIG, 5. displays a conceptual view of the temperature 
flow in a conventional thermal bus. The electronics 


mounted on the baseplate, denoted by 210 functions as 
a source, generating heat at a temperature 310. The 
baseplate denoted by 220 receives this heat at a slightly 
lower temperature 320. In the conventional thermal bus 
the baseplate denoted by 220 would be directly coupled 
to the coldplate denoted by 230. The coldplate would 
experience another loss in temperature 330, as a result of 
the conduction process. The loss in temperature would 
continue through the fluid loop 240 coupled to the 
coldplate, the radiator denoted by 250 and culminating 
in the effective sink denoted by 260. In each of these 
steps, the heat flow due to conduction has an associated 
temperature drop 340, 350, and 351, respectively. 

FIG. 6 displays the thermal bus with the thermo-elec- 
tric heat pumps, located between the baseplate and the 
coldplate. The electronics 400 and the baseplate 410 
receive the convectional temperature drop, as denoted 
by 500 and 510. However, in the TEHP assisted thermal 
bus, a TEHP denoted by 420 is placed between the 
baseplate 410 and the coldplate 430. Therefore instead 
of experiencing a temperature drop between the base- 
plate 410 and the coldplate 430, heat is pumped across 
an adverse temperature gradient from the baseplate 410 
to the coldplate 430. The result is the increase in tem- 
perature from the temperature in the baseplate 510, 
across an adverse temperature gradient 520, to the tem- 
perature of the coldplate 530. The fluid loop coupled to 
the baseplate denoted by 450 and the radiator 460, con- 
tinue to decrease the temperature from 550 to 560. The 
radiator 460 maintains the same temperature of the 
effective heat sink 470. The radiator temperature 560 is 
normally the same as the effective sink temperature 561. 

The effective heat transfer coefficient is equal to the 
third power of the power of the radiator temperature 
(h~T 3 ). As a result, radiating at a higher radiator tem- 
perature requires a radiator with a smaller surface area. 
Therefore, the overall result of adding the TEHP to 
upgrade the waste heat at 520, is a smaller radiator at 

Each TEHP unit will individually control an area on 
the source. When a small direct current is applied to the 
TEHP devices, thermal energy is pumped from the 
source or baseplate to the thermal sink or coldplate, 
across an adverse temperature gradient. As a result, a 
TEHP unit will be responsible for transferring heat 
from an individual quadrant on the source, to the sink. 
Each of the TEHP units within a quadrant will be indi- 
vidually controlled thereby enabling a controller to 
vary pumping operation of TEHP units either within 
the quadrant or surrounding the quadrant. As a result 
isothermality will be maintained as the fluid loop inte- 
gral with the coldplate, shifts to an increased sink tem- 

The operation of each of the TEHP units located 
within the predefined quadrants will be controlled by a 
model-based feedback controller. The control system 
maintains isothermality by adjusting the TEHP heat 
pumping capacity to compensate for scheduled and 
unscheduled disturbances in the thermal bus. The heat 
pumping capacity is adjusted while maintaining the 
local quadrant set-point temperature. 

In general, any off-design or contingency operating 
conditions are considered unscheduled disturbance. 
Usually these unscheduled disturbances are caused by 
parasitic heating in the target electronics or variations 
in the fluid loop inlet temperature. In general, any dis- 
turbances or variations in heat or power, may be consid- 














5 , 349,821 


ered a scheduled disturbance. For example, heating 
caused by the predictable degradations in the electronic 

The model-based control system used to control the 
TEHP units operate by first sensing a process tempera- 5 
ture in each quadrant on the source or baseplate. This 
process temperature array is combined with a modelled 
temperature array to create a temperature error array. 
The system then combines the temperature error array 
with the set point temperature array to produce the 10 
expected inverse modelled array, which is an array of 
set point temperature values, fed into an inverse model 
of the process. Any variation in the input of the inverse 
model, causes the model to produce a change in the 
voltage or current output produced by the inverse 15 
model. The result of a change of voltage or current is an 
increase the pumping capacity of the TEHP units, lo- 
cated in the region of the affected quadrant. 

In the case where the set point temperatures are all 
equal, individually varying the TEHP units will pro- 20 
duce a uniform isothermal baseplate. In the event that 
one of the TEHP units completely fails, the control 
system will perform load levelling to minimize varia- 
tions in isothermality. As a result of using model based 
feedback control, the TEHP unit transfers heat from a 25 
quadrant, taking into account the other individually 
controlled quadrants, thereby maintaining isothermality 
in the source. Therefore, convection heating in the 
plane of the source plate, is controlled with a decoupled 
control method. 30 

For the TEHP units to transfer heat across an adverse 
temperature gradient the TEHP requires a direct cur- 
rent with less than five percent ripple. A current gener- 
ator is used in the system to modulate the voltage and 
current. The nominal voltage and current requirements 35 
per device are on the order of 0-0.5 volts and 10 am- 
peres, respectively. The busbar requirements depend on 
the degree of control required. For applications where 
the individual TEHP units are controlled with a single 
power source, the baseplate and coldplate would serve 40 
as the busbar. While this design is very simple, redun- 
dancy and isothermality control are compromised. If 
groupings of devices or quadrants are powered sepa- 
rately, each quadrant would require a segregated busbar 
arrangement. Independent of the number of controlled 45 
devices, a single power source is multiplexed with peer 
transistors operating under the command of the model- 
based feedback control system. On command, the 
power is distributed to a specified quadrant of devices, 
at a specified level, for a specified time, after which the 50 
power is switched off and diverted to the next quadrant 
of devices. The frequency of modulation is dependent 
on the allowed switching losses and the thermal capaci- 
tance of the TEHP and adjoining structure, along with 
the nature of the thermal disturbance. Low weight 55 
structures have low thermal capacitance which would 
require an increased switching rate, which is desirable. 


However, an increased switching rate results in para- 
sitic switch heating, which is an undesirable feature. 

While the preferred embodiment of the invention is 
disclosed and described it will be apparent that various 
modifications may be made without departing from the 
spirit of the invention or the scope of the subjoined 

I claim: 

1. A thermal bus comprising: 
a heat source, 

a plurality of quadrants located in said heat source, 
said quadrants each including a heat sensing means 
for sensing heat in said heat source, 
a heat sink including a plurality of quadrants each, 
coupled to one of said plurality of quadrants in said 
heat source, 

a plurality of thermo-electric heat pumps each posi- 
tioned between one of said plurality of quadrants 
located in the heat source and said coupled quad- 
rant in said heat sink, thereby pumping heat from 
each of said plurality of quadrants in said heat 
source, to each of said coupled quadrants in said 
heat sink, and 

a model based controller individually connected be- 
tween each of said heat sensing means and each of 
said plurality of thermo-electric heat pumps 
thereby controlling said plurality of thermo-elec- 
tric heat pumps by sensing said heat. 

2. A thermal bus as claimed in claim 1 wherein said 
heat sensing means is a thermistor or a thermocouple. 

3. An augmented thermal bus comprising: 

a baseplate including a plurality of quadrants for 
housing electronic components therein, 
a coldplate with a single phase fluid loop coupled 
thereto, thereby providing a heat sink for said aug- 
mented thermal bus, 

a plurality of thermo-electric heat pumps coupling 
said baseplate to said coldplate, 
a power supply attached between said baseplate and 
said coldplate thereby creating a voltage potential 
across said plurality of thermo-electric heat pumps, 
a model based controller individually attached to 
each thermo-electric heat pump in said plurality of 
thermo-electric heat pumps, and 
a radiator coupled to said fluid loop for dissipating 

4 . An augmented thermal bus as claimed in claim 3 
wherein each of said thermo-electric heat pumps is 
coupled to each of said plurality of quadrants thereby 
individually controlling temperature in each of said 
plurality of quadrants. 

5 . An augmented thermal bus as claimed in claim 4 
wherein said power supply applies a direct current of 
less than about 5 percent ripple, a voltage of less than 
about 0.5 volts and a current of about 10 amperes.