An Introduction To Temperature Control
©2003 Tellurex Corporation
1462 International Drive
Traverse City, Michigan 49686 (231) 947-0110
One of the attractive features of thermoelectric (TE) technology,
is that it offers an incredible degree of controllability.
With a properly ‘tuned’ controller, it is possible
to maintain systems well within 0.1° C of set point. Unfortunately,
many off-the-shelf solutions for temperature control are not
well suited to the thermoelectric world because they were designed
for heating or cooling hardware that is very different—and
often far less responsive—than TE devices. This leaves
designers groping for alternatives. This guide is intended
to offer designers some practical guidance in exploring the
vast range of possibilities.
Types of Control
Basically there are two types of temperature control: thermostatic
and steady-state.
Thermostatic Control
With thermostatic control, a thermal load is maintained between
two temperature limits. For example, in a cooling application,
the controller may energize TE cooling power when the thermal
load rises to 30° C, then turn off cooling power when the
temperature cools to 27° C; the system would, therefore,
continually vary between 27 and 30° C (see Figure 1). The
difference between the high and low temperature limits, would
be the system's hysteresis—in this case, 3° C. Thermostatic
control is typically the least costly alternative and should
be considered whenever a user can tolerate some appreciable
variation in operating temperatures.
In the world of thermoelectric technology, thermostatic control
is usually accomplished by switching TE power electronically
rather than with electro-mechanical devices such as snap disks
or relays. This is due to three reasons:
1. The devices are switching direct rather than alternating
current and, therefore, the mechanical contacts are more vulnerable
to the pitting and premature wear which result from arcing.
2. Using electronic means, you don't have to worry about ratings
for the typical number of switching operations before failure
(which indicate the likely lifespan for a mechanical device)—you
can switch electronically as often as you want. Because of
this, it is easier to provide thermostatic control with minimal
hysteresis. In some systems, the hysteresis can be kept well
under 1° C. With mechanical snap discs on the other hand,
hysteresis is often around 5-9° C or more.
3. Using electronic circuitry, it is easier to design a thermostatic
controller that has an adjustable set-point.
Because thermoelectric technology uses DC voltage sources,
the component of choice for switching in electro-thermostatic
controllers, is the power MOSFET (i.e., metal oxide semiconductor
field effect transistor). Relatively inexpensive power MOSFET's
can be found with ‘on’ resistances as low as 0.028
ohms. This means that very little power is lost in the switching
device. For example, if you were driving two typical TE devices
connected electrically in parallel and collectively drawing
10 amps, there would only be 2.8 Watts (P=I2 R) dissipated
in this power MOSFET with a voltage drop of just 0.28 VDC (V=I
R). Of course, some designers naturally gravitate toward solid
state relays in this sort of situation, but these devices are
seldom used with this technology because they are significantly
more resistive than MOSFET's, drop more voltage, dissipate
a lot more power, and are bulkier
.
There are times when thermostatic control of TE systems is
inadvisable. For example, if the required hysteresis is too
small, the system may have to cycle on and off in a matter
of seconds to maintain the system within the desired range.
While the controller can certainly meet these requirements,
this sort of cycling is, unfortunately, mechanically stressful
to the TE modules and will lead to their premature aging and
failure (due to differential rates of expansion and contraction
within the module). Thermostatic control of TE systems should
generally be limited to applications where the controller will
cycle over a period of at least several minutes or more.
Steady-State Control
Whenever a system must be maintained within fairly tight limits,
some form of steady state controller should be considered.
A steady-state controller is designed to continually hold a
thermal load at the set-point temperature with very little
variation around the set-point. If the steady-state condition
is disrupted by some sudden change in ambient conditions, the
controller will quickly bring the thermal load back to a steady-state
condition (provided that the system has sufficient heating
or cooling capacity). To achieve steady-state control with
thermoelectric technology, the controller must vary the amount
of current to the TE modules so that there is just enough power
to maintain the system at the desired temperature. To accomplish
this, the controller must be able to make instantaneous adjustments
in TE current in response to changes in the ambient environment
or thermal load.
It should be noted that many systems designers make the mistake
of thinking that they can achieve steady-state temperature
control by simply adjusting the TE power supply to a level
which yields the desired temperature. This approach can only
work if ambient conditions are absolutely constant. Without
real temperature control, any variations in the operating environment
(or load for that matter) will cause changes in the temperature
of the thermal load. The key to effective steady-state control,
is that the amount of TE power be made dependent upon the ambient
and load conditions which exist at any particular moment.
Steady-state control is usually achieved with some variant
of a proportional controller. Proportional controller design
is based upon the reality that set-point potentiometers and
temperature sensors ultimately use electrical voltages as indicators
of temperature. In its most simple form, a proportional controller
merely amplifies the difference between the set-point and sensor
voltages and provides an output (in this case, TE power) which
is proportional to that difference. Thus the amount of TE power
is a function of the difference in temperature between the
desired set-point and the measured temperature of the thermal
load. For example, the set-point voltage in a particular system
might be 2.73 V; if the temperature sensor develops a voltage
of 2.83 V and the proportional amplifier has a gain of 100,
the proportional amplifier would have an output of 10 VDC:
The controller would then translate this amplified error voltage
(i.e., 10 V) into a proportional amount of TE power. In a simple
proportional controller, therefore, the precise amount of TE
power at any given point in time, is a function of the amplified
error between the set-point and sensor voltages. If a source
of heat in the ambient environment causes greater error, the
amount of TE cooling power immediately increases, then gradually
decreases as the system gets back closer to the set-point.
It should be noted that with a basic proportional controller,
some error must exist between the set-point and sensor voltages
in order to have any power provided to the TE modules—if
there is no error, there is no voltage difference to amplify,
and as a result, no TE power generated by the control system.
In practice, a simple proportional controller will bring the
system to a temperature where there is just enough error—just
enough amplified error voltage—to yield the level of
TE power required to stay at that temperature. For example,
in a given ambient environment, if:
1. 2 V of amplified error voltage is required for the controller
to provide enough cooling power to maintain a 15° C temperature;
2. it takes a 0.1° C temperature difference to create
an amplified error voltage of 2 V, and
3. the set-point is adjusted to a voltage corresponding to
15° C,
The system will settle to a steady-state temperature
of 15.1° C.
If the ambient temperature rises, it will take more cooling
power—and thus more error—to maintain the temperature
near the set-point, and the system will settle to a somewhat
higher temperature. If the ambient temperature decreases (but
remains above the set-point), it will take less cooling power
(and less error) to maintain the temperature near the set-point—the
system will then settle to a temperature which is closer to
the set point. The only way that there will be no error between
the set-point and sensor temperatures, is if no TE power is
required to reach and maintain the set-point.
The amount of error which exists when the system reaches a
steady state, is called ‘steady-state error’ (see
Figure 2). It should be noted that the greater the gain of
the proportional amplifier, the less steady-state error will
be required to generate a given level of TE power; we can thus
decrease steady-state error by increasing proportional gain.
It might appear, therefore, that a designer would want the
maximum amount of proportional gain to decrease steady-state
error to the lowest achievable level. Excessive proportional
gain, however, makes a temperature control system too sensitive
and causes it to over-respond to changes in conditions (see
Figure 3). This results in system oscillation around the set-point,
with the temperature swinging back and forth continually. Generally,
the proportional gain will be set so that steady-state error
is minimized while assuring stable operation of the system.
It must be noted that proportional control alone, can only
approach steady-state operation. In reality, assuming that
sufficient power is available to reach the set-point, the temperature
of the system will vary somewhat as ambient conditions change
(see Figure 4). Assuming that the system always has enough
capacity to reach the steady-state condition, the system will
vary anywhere between the set-point (where there is no TE power)
and the temperature at which the
steady-state error yields full TE-power. In many TE systems
using only proportional control, maximum steady-state error
will only be a few tenths of a degree and this extent of inaccuracy
can be easily tolerated. This is often the case, for example,
when controlling the air temperature of an enclosed space.
In other systems, however, the maximum steady-state error may
be tens of degrees—a level of potential error which is
seldom acceptable. This might be the case, for example, when
the system is directly controlling the temperature of liquids
or metallic solids. In these cases, because temperature responds
so quickly to changes in TE current, the proportional gain
may have to be decreased to a point where steady-state error
becomes substantial. In situations where the level of steady-state
error is clearly unacceptable, additional hardware (or software
in the case of a digital controller) must be added to compensate.
It is with the addition of an integral amplifier that we can
virtually eliminate steady-state error in a proportional control
system. The integral amplifier essentially integrates the amplified
error voltage in the controller; this means that its output
keeps increasing (due to the charging action of the integrating
capacitor) as long as an error voltage is present. Once the
amplified error is eliminated, the integral amplifier simply
holds its output steady until a new error voltage becomes present.
The voltage output of the integral amplifier is added to the
output of the amplified error voltage from the proportional
amplifier to determine the amount of TE power required. (Note
that if the system overshoots the set point, the integral amp
can integrate in the opposite polarity to provide corrective
action).
In theory, upon system power up, the amplified error voltage
from the proportional amplifier will contribute the bulk of
the TE power while the integral amplifier output is charging
up. Ultimately, the output of the integral amplifier takes
over and becomes the driving force in sustaining the necessary
level of TE power—and in eliminating all steady-state
error. (In reality, some amount of error will still be evident
because of offset errors in amplifiers and other subtle error-producing
phenomena in the circuitry, but these discrepancies will be
inconsequential in most systems). The higher the gain of the
integral amplifier, the faster it can respond to steady-state
error potential. If the integral gain is inadequate, it will
be very slow to compensate; if the integral gain is too high,
it will tend to over-correct and the system will oscillate.
Typically the gain of the integral amplifier is set so that
upon power-up, the system will ‘ring’ one time
(i.e., it will overshoot the set-point, then undershoot it,
then settle to the set-point).
A related problem in many TE systems, is that it takes a prolonged
period of time to reach the set-point. In these cases, an integral
amplifier can ‘saturate’ and ‘lock-up’ at
full output. This can cause the TE system to overshoot the
set-point by a considerable amount and then stay in this condition
for several minutes until the amplifier gets out of its locked-up
state; only then can it integrate again to bring the system
back toward the set-point. For this reason, it is usually desirable
for the integral amplifier to have "anti-windup" circuitry;
this inhibits integral action until the system gets close to
the set-point temperature. Typically, this will prevent the
integral amp from saturating.
A controller which uses only proportional and integral amplifiers,
is known as a "PI" controller.
Unfortunately, while integral amplifiers can effectively correct
for steady-state errors, they tend to make the control system
more unstable. For this reason, a derivative amplifier is often
employed to brings things into more optimal control. Where
the proportional amplifier keys only on the amount of change,
a derivative amplifier responds to the rate of change; the
faster that the proportional voltage changes, the greater is
the output of the derivative amplifier. It is thus used to
provide an anticipatory response to sudden variations in load
or ambient conditions.
Let's look at an example. We are cooling a cold plate which
has reached a steady-state temperature of 5° C. You suddenly
place your hand on the plate which causes the temperature to
rise quickly. If you only have proportional control, the TE
output would simply be a function of the amount of error which
results. With the addition of the derivative component, however,
the rapid change of temperature brings additional cooling power
on line to respond to the disruption in a more forceful and
effective way. Note, too, that as the system is being brought
back to the set-point, if it starts cooling too quickly, the
derivative amplifier will apply ‘the brakes’ and
decrease cooling power, thus minimizing or preventing overshoot.
The higher the gain of the derivative amplifier, the greater
will be its instantaneous response to change. The gain must
be set to a level which effectively minimizes the effect of
system disruptions, but not so high that it tends to over-compensate
or the system will oscillate. High derivative gain will also
tend to slow the system down in initially ramping to the set-point;
in some cases this is desirable or even necessary, in others,
it is not. A controller which uses proportional, integral,
and derivative amplifiers, is known as a "PID" controller
and this is the most common form of steady-state control.
With steady-state control, there are two approaches which
are taken to vary the amount of power provided to the TE device:
1) pulse-width modulation (a.k.a., ‘PWM'), and 2) linear
control. Pulse-width modulation provides the most efficient
means, because the driving transistor (usually a power MOSFET)
is either fully on or fully off. In either state of conduction,
very little
power is dissipated within the transistor. With PWM, the TE
modules are pulsed on and off at a specific frequency with
the controller merely changing the percentage of ‘on’ time
vs. ‘off’ time (i.e., the duty cycle) during the
pulse train (see Figure 5).
One of the problems in providing PWM for TE devices, is that
most off-the-shelf PID controllers will provide pulses at only
1 Hz (i.e., cycle per second) or less, and this low frequency
drive is more thermally stressful to the TE modules than pulsing
at higher frequencies. At Tellurex, we recommend that TE modules
be pulsed at 2000 Hz or greater when using PWM to minimize
the effects of thermal expansion and contraction. Unfortunately,
few manufacturers provide off-the-shelf options that can meet
this specification. Fortunately, controllers are available
which have an analog output; with a fairly simple interface
circuit (see Figure 6), the analog level can be converted to
a proportional pulse-width modulated drive signal at the desired
frequency. Note that when using PWM, the ‘on’ voltage
should never be allowed to go beyond the VMAX specification
for the TE module or performance will suffer greatly.
Despite its efficiency advantages, PWM cannot be used for
steady-state control in many cases because of the potential
for electromagnetic interference (EMI) from the high current
TE pulses and/or because poor transient response in the available
power supply can cause audible high frequency noise. In these
instances, some sort of linear drive circuit is created to
provide a variable DC level for powering the TE module. Most
designers employ bipolar junction transistors (BJT's) in these
instances to provide a linear drive current for the TE modules.
Unfortunately, these BJT drivers can dissipate a lot more power
than pulse-width-modulated MOSFET's. For example, if you use
a TE module which is rated for 5 amps, 12 VDC at full power,
the worst case for the BJT drive transistor would be when 6
V is provided to the TE module at 2.5 amps. Here the TE device
would dissipate 15 W—as would the drive transistor. The
BJT driver, therefore, would have to be mounted to a heat sink
which can effectively manage 15 W of heat. Compare that to
a power MOSFET with a 0.028 ohm ‘on’ resistance
running at 50% duty cycle (the equivalent in PWM); here the
driver would only dissipate an average of 0.175 Watts. It's
also important to note in this example, that if multiple TE
modules were used, there would be a worst case power dissipation
in the BJT driver of 15 W for every TE module employed. Unfortunately,
in some applications, linear drive is the only reasonable alternative,
and the electrical inefficiencies must be tolerated in order
to serve more important objectives.
In recent years, Tellurex electrical engineers have pioneered
a means of creating linear drive using power MOSFETs. While
this still is not as efficient as PWM, the much better saturation
conductivity of the MOSFET, makes it far more suitable than
the traditional BJT driver. This is seen when the TE system
requires full power from the cooling module. In this circumstance,
less power is dissipated in the drive transistor, and the voltage
drop across the driver is significantly lower than with a BJT.
When using a properly-selected MOSFET, all but a small fraction
of a volt from the available supply, can be applied to the
TE device for full-power operation. In a BJT-controlled circuit
at saturation, a full volt may be dropped across the transistor
resulting in a many-fold increase in power dissipation within
the driver. Thankfully, too, only minimal additional circuitry
is required for this improved solution to linear drive.
Digital Controllers
While most of the preceding discussion has focused on controllers
built from analog electronic components, much of the available
control hardware is now digitally based using microprocessors
or micro-controllers. Digital PID controllers use software
algorithms to interpret the differences between the measured
temperature and the set point; it is software which amplifies,
integrates, and derives the proper amount of TE power to hold
the thermal load in steady-state operation. Digital PID controllers
basically mimic the functioning of their analog equivalents.
There are both advantages and disadvantages associated with
digital controllers. Perhaps their greatest strength comes
into play when a number of identical TE systems are to be fabricated;
once the correct tuning has been determined for the prototype,
the gain settings can be easily programmed into each controller
to assure consistent performance among all units. In contrast,
analog controllers typically employ trim pots for the gain
settings (unless the gains are locked in with fixed resistor
values), and every unit must be tuned individually. Another
attractive feature of most digital controllers, is an LED or
LCD display that can show the set-point and measured temperatures
precisely. A lot of these controllers can also accommodate
a variety of sensor options and many units can even tune themselves
to provide satisfactory performance. It is not uncommon, too,
for these circuits to offer a choice between steady-state (i.e.,
PID) and thermostatic control (although steady state will usually
win out in TE systems).
There definitely are some drawbacks associated with digital
control—not the least of which is cost. It would be very
difficult to create a custom, stripped-down, digital controller
inexpensively while it is very possible to do so with analog
components. In TE systems, digital controllers also don't tend
to perform quite as well as their analog counterparts, particularly
with respect to the derivative component. Analog controllers
react instantaneously to any changes in the environment; their
ability to respond is limited only by the overall cooling or
heating capacity available and the soundness of the PID tuning.
Digital controllers, on the other hand, take periodic measurements
then change the TE power as appropriate—all while managing
other functions like dumping display data and polling input
buttons. Microprocessors can only do one thing at a time, albeit
with a great deal of speed. In most applications, the trade-off
in responsiveness can be easily accommodated for the advantages
offered by digital technology, however, in situations which
require more ideal performance—particularly in responding
to sudden ambient changes—analog controllers will almost
always have the advantage.
Every design situation is different and the choice of an analog
or digital solution will depend on the unique demands of the
application. There is one general ‘rule of thumb', however:
whenever a digital display is required, digital control options
will almost always deliver the best cost advantages.
The Cost of Control
In virtually every TE application where temperature control
is required, cost of control is a major consideration. Unfortunately,
many TE system designers unconsciously drive up costs by convincing
themselves that what is merely desired, is required. The reality
is that, for every ‘inch’ of quality, precision,
or flexibility in controller performance, a disproportionate
price must be paid. Wise designers will—very early in
the process—examine their control objectives and separate
those items which reflect necessity, from those which would
simply be "nice to have". If minimal cost is important,
designers must make every possible sacrifice in system specifications
and avoid the temptation to add unnecessary performance requirements
or superficial ‘bells and whistles’. Among the
most important considerations are these:
Is an adjustable set-point required? The simplest and least
costly controllers will be factory-calibrated for a single
set-point with no method of re-adjustment (i.e., other than
perhaps re-calibration) for the end user. Providing a set-point
range for end-user adjustment will add cost to the control
system with the extent of that increase dependent upon other
control or user-interface objectives.
Is there a need for visual indicators for set-point or temperature?
The simplest and least costly controllers will have a ‘blind’ set-point
range (i.e., there is no visual indication of the precise set-point
temperature) and no display of the resulting temperature. For
example, the set-point knob might be labeled with an arrow
and the word "Cooler" to show the user which direction
to turn the knob for adjustment; the user would then monitor
the temperature with a thermometer and adjust, as necessary,
until the desired performance is achieved. If some visual indication
of set-point is required, it can be provided in the least costly
manner by using a printed adjustment scale and indicator knob;
this decision, however, will impact upon the choice of temperature
sensor and the controller will require additional calibration
to assure modest accuracy in the correspondence between the
knob setting and the actual set-point which results. Because
of the sensors employed in such controllers (usually IC's),
these units also tend to require additional circuitry to minimize
the impact of noise on system accuracy.
A digital display will typically add great cost to the controller.
This is due not only to the cost of display-related hardware,
but also because some modest sources of error can no longer
be effectively ‘hidden’ from the user. As a result,
the controller manufacturer will need to go to extra lengths
to ‘trim out’ error
.
How much accuracy and repeatability is required? One of the
biggest challenges in designing electronic controllers, is
in coping with the variations of components within tolerance.
For example, while a resistor with a tolerance rating of 1%
and a temperature coefficient of 50 ppm may be non-problematic
in most electronic circuits, in an instrumentation or control
application, the variations can result in set-point shifts
of 5° C or more. Similarly, offset errors in operational
amplifier IC's are inconsequential in many applications, but
can contribute to serious inaccuracies in instrumentation and
control circuits. This reality means that control designers
sometimes must rely upon extreme measures to achieve accuracy
and repeatability, and the cost of a controller will correlate
highly with the degree of conformance demanded. The difference
in cost between a controller which is accurate to within 2-3° C
and one accurate to less than 0.1° C, is immense. As requirements
for accuracy and repeatability tighten, the controller will
need more and more provisions for calibration and error reduction.
In extreme cases, some of the components on the control board
may require temperature regulation themselves to minimize variations.
In coming to terms with these issues, designers should avoid
falling into the trap of specifying absurd performance requirements.
For example, some people will demand stability and accuracy
to within 0.01° of set-point when they: 1) can’t
dependably and accurately measure to that standard, and/or
2) will have gradients across the thermal load that exceed
this specification. Just ask for what you really need and you
will keep costs down substantially. Don't be tempted to pad
your requirements with an additional margin of error—and
if you deal with controller designers, ask them to do the same.
If everyone substantially pads their performance standards,
the resulting controller can cost a great deal more than necessary
in the end.
Is there a need for precise limits at either end of the set-point
range? Virtually any analog temperature controller which has
an adjustable set-point range, will have to be calibrated to
provide anything approaching a consistent range from unit to
unit. In the simplest case, the range will be calibrated at
the center-point of the range with a variation at the endpoints
from unit to unit of 2-3° C; or perhaps one of the endpoints
is calibrated with the other varying 4-6° C from unit to
unit. If this degree of variation is excessive for a given
application, further calibration will be required to achieve
the desired performance; the cost will be directly related
to the degree of accuracy and stability which is demanded.
Here again, in extreme cases, some of the components associated
with the set-point voltages, may require temperature regulation
themselves to minimize variations.
If you're going to use a digital controller and you want precise
limits, you will need to select a model which provides for
programming the set-point range. Naturally, the repeatability
of the range from unit to unit will be very good.
Will the system be single-mode (i.e., just heating or just
cooling) or dual mode (i.e., switchable between heating and
cooling modes)? Of course, one of the attractive features of
TE technology, is that it can be used for either heating or
cooling and it is often the technology of choice in situations
where either mode of operation may be required depending upon
ambient conditions. Dual mode control, as one might expect,
comes at a higher price, particularly in steady-state systems
or those requiring automatic switching between modes. Mode
switching in a temperature-controlled system is more than a
simple matter of toggling the polarity to the TE devices; typically,
other signals will have to be re-routed, as well. In proportional
controllers, gain settings must also be changed whenever the
mode is toggled. Generally speaking, the more complex the controller,
the more signals or settings will require switching to change
modes between heating and cooling. It should be pointed out
that this is one area where digital controllers have a distinct
advantage; they can just alter the values of a few mathematical
variables to get much of the job done.
If frequent automatic mode switching will be done, mechanical
(e.g., relay) polarity switching can become problematic; the
high direct currents involved can ultimately lead to pitting
of switch contacts and premature failure of the switching device.
One way to deal with this is to employ circuitry which holds
TE power off whenever the mode will be switched; thus there
will never be a voltage across the switch when it is toggled.
Another approach is to use a power MOSFET bridge instead of
a relay; usually this is more costly, but you never have problems
with contact wear. When employing a transistor bridge, however,
you must design your circuitry to prevent the creation of transient
short-circuit conditions through the MOSFET's; transistors
are switched in pairs and you must make sure that one pair
is fully off before the other pair is energized.
Does the potential for sales volume justify a custom controller
for the application? In many high volume applications—particularly
those which do not require a display or those in which a modest
amount of error is tolerable—a custom controller may
be an excellent choice. The controller can be designed to provide
the most cost-effective solution while providing only those
features which are essential. In most cases, a prototype controller
can be tuned to the TE system and those settings can be translated
into hard-wired gain settings for the production version of
the control board. From that point onward, trimming of controller
settings will be confined to those areas of the controller
where it is absolutely necessary for delivering a specified
level of performance (e.g., calibration of the temperature
sensor, fixing the set-point range, etc.). This approach can
result in substantial savings. Even when there is a need for
displays and high-accuracy circuitry, custom solutions can
sometimes yield significant cost reductions in high-volume
applications.
It should be noted in this discussion, that low-cost, stripped-down
designs do not tend to be flexible; they are spec'ed for a
particular situation and redirecting them to another purpose
can involve significant costs and effort. For example, if you
want to use a ‘bare bones’ controller from one
application with a different thermal load that requires another
set point or range, many (perhaps most) of the fixed-value
resistors on the controller's circuit board will have to be
changed. This can involve a great deal of challenging and tedious
mathematical analysis, as well. When the controller is thermistor-based—and
many of the lower-cost circuits will be—if you decide
to specify a different thermistor, values for most resistive
components will have to be re-calculated. Certainly, it is
possible to create flexible architectures for custom controllerers,
but the greater the adaptability demanded, the more likely
it is that costs will soar geometrically. At some point of
complexity, the hardware and labor costs will commend going
with an off-the-shelf solution that has ‘all the bells
and whistles’.
What sort of temperature sensors should be used? This is a
very important question and must be considered within the overall
objectives for TE system performance. For example, in a single
set-point application with no display, or in a situation where
a ‘blind’ set-point range can be used, a simple
thermistor offers a number of cost and utility advantages.
With a thermistor, the designer can maximize the volts per
degree ratio to a level which can offer much more noise immunity
than is possible with many other types of sensors—this
alone can help to eliminate a significant number of components
from the controller. If a digital readout or scaled set-point
graduation is required, integrated circuit sensors (e.g., LM324,
LM335, or AD590) may be the best choice because they offer
good linearity and relatively high accuracy after calibration
(although their relatively high mass rules them out in applications
which require greater sensitivity). Many designers prefer to
use thermocouples in their applications, but the need for cold-junction
compensation and low-noise circuitry will add significant expense—thermocouple-based
solutions never come cheaply. Of course, off-the-shelf digital
controllers will almost always come equipped with provisions
for thermocouples.
Is ramp rate an important consideration? Many designers seek
to maximize the ramp rate in reaching the set-point and they
need to understand that there are a number of important related
issues. Foremost among them, is the reality that excessive
ramp rates can actually destroy TE modules. Because of different
rates of thermal expansion and contraction in the materials
which make up TE modules, excessively high temperature changes
can cause solder joints to open—and a single open solder
joint renders the module useless. This issue is especially
important when the technology is used in heating applications,
because TE modules are more efficient as heaters than coolers.
The absolute maximum ‘safe’ ramp rate, is approximately
1° C per second, but only if the end temperature is held
for several minutes after it is reached. Thus, whenever TE
modules are used in a system in which excessive ramp rates
are possible, a suitable temperature controller must be employed
to limit the rate of change to a safe level. In some systems,
this can be accomplished through appropriate adjustment of
a derivative amplifier, in others a ‘ramp and soak’ controller
will be necessary (these latter controllers ramp the set-point
up or down to fit a programmed temperature profile).
What is the nature of available power for the controller?
If AC utility power is readily available, the user has many
more options and has a better chance to keep costs down. Most
off-the-shelf control solutions run on 120 or 240 VAC (although
you will still need a DC supply for the TE module). If the
only source of electrical power is DC (e.g., the 13.2 VDC in
most automotive systems), there will be far fewer options—in
many cases it will be necessary to go with a custom solution.
If steady-state operation is desired, can pulse-width modulation
be employed? In the vast majority of applications, pulse-width
modulation at 2000 Hz or greater can be employed with relative
impunity as long as the designer is careful to keep signal
wiring separate from power wiring and uses suitable shielding
with each. When done properly, control with pulse-width modulation
can be employed in a far more cost-effective manner than linear
control. If, on the other hand, the user is dealing with very
sensitive circuitry—particularly if it is to run from
the same power supply as the TE system—it may be necessary
to use linear drive; this is often the case in the laser diode
world, for instance. When making borderline calls, especially
in low-volume applications where the user cannot afford the
time to experiment with control options, it can make sense
to go with linear control from the beginning—there's
no sense flirting with EMI problems if you don't have time
to address them. In the vast majority of applications, however,
PWM will work just fine and it can be a costly mistake—especially
in high volume applications—to simply assume that linear
control will be required.
Tellurex Temperature Control Products
When it comes to temperature control, Tellurex Corporation
specializes in providing solutions for high-volume applications
which can benefit from custom controller design. In most cases,
we can adapt a previous design to new systems and then provide
circuit boards which are custom populated. In those cases which
require unique approaches, we can rely upon our extensive experience
to arrive at optimal solutions. To discuss your needs, contact
a sales representative at Tellurex (231-947-0110; sales@tellurex.com).
In Conclusion
While TE technology is very controllable, there are a great
many things to consider in arriving at a suitable solution.
Every desire comes at a cost, so each must be contemplated
thoroughly before proceeding. The more that cost is a constraining
force in choosing a solution, the more diligent the designer
must be in reigning in the ‘wish list’ to focus
on what is absolutelynecessary. necessary.
©2003 Tellurex Corporation
|
