AC-02-6-1

Degrading Chilled Water Plant Delta-T:

Causes and Mitigation

Steven T. Taylor, P.E.

Member ASHRAE

ABSTRACT ·
Q = m c p ∆T
Q ( Btu ⁄ h ) = 500 GPM ∆T (IP units) (1)
Variable-flow chilled water plants are designed to main-
tain a relatively constant delta-T, the difference between return Q ( kW ) = LPS ∆T (SI units)
and supply chilled water temperature. But in almost every real In almost every real chiller plant, delta-T falls well short
chiller plant, delta-T falls well short of design levels. The result of design levels, particularly at low loads. The result is higher
is that flow and load do not track, usually requiring that addi- pump and chiller energy usage. Many papers have been writ-
tional chillers be brought on line to maintain flow requirements ten on the subject of “low delta-T syndrome” (Kirsner 1996,
even though none of the chillers is fully loaded. Both pump 1995; Lizardos 1994; Sauer 1989; Fiorino 1996; Avery 1997;
energy and chiller energy increase accordingly. Many design Mannion 1988). Most are oriented toward how to keep delta-
and retrofit measures have been tried to resolve the problem, T high. This paper also will address causes of degrading delta-
but they are sometimes expensive and not always successful. T along with mitigation measures, but it goes on to show why
In this paper, the author argues that while many causes of delta-T degradation will almost always occur in chilled water
degrading delta-T may be eliminated, in most plants it is not systems and how to design around that eventuality to maintain
chiller plant efficiency despite degrading delta-T.
possible to avoid degrading delta-T under all operating condi-
tions. Several design and operational techniques are presented THE ENERGY IMPACT OF DEGRADING DELTA-T
both to minimize degrading delta-T and to design plants to be
Figure 1 shows a hypothetical chiller plant serving
efficient despite degrading delta-T.
several buildings in a large facility, such as a university
campus, office complex, district cooling system, or industrial
INTRODUCTION facility. The system is piped in a traditional primary-second-
ary manner with some tertiary pumps at remote buildings.
In most variable-flow chilled water plants, it is assumed
If the delta-T in this system is low, at least two problems
that delta-T, the difference between return and supply chilled
result: increased pump energy usage and either an increase in
water temperature, will remain relatively constant. Because chiller energy usage or a failure to meet cooling loads.
the load is directly proportional to flow rate and delta-T (Equa- The increase in pump energy is obvious. According to
tion 1), if the delta-T is constant, it follows that flow rate must Equation 1, any reduction in delta-T must cause a proportional
vary proportionally with the load. Most variable-flow systems increase in chilled water flow rate. Pump energy, theoretically,
are designed based on this assumption and usually fail to is proportional to the cube of the flow rate, so any increase in
perform well if the delta-T does not stay relatively constant. flow will have a much higher increase in pump energy. In real

Steven T. Taylor is a principal at Taylor Engineering, Alameda, California.

THIS PREPRINT IS FOR DISCUSSION PURPOSES ONLY, FOR INCLUSION IN ASHRAE TRANSACTIONS 2002, V. 108, Pt. 1. Not to be reprinted in whole or in
part without written permission of the American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc., 1791 Tullie Circle, NE, Atlanta, GA 30329.
Opinions, findings, conclusions, or recommendations expressed in this paper are those of the author(s) and do not necessarily reflect the views of ASHRAE. Written
questions and comments regarding this paper should be received at ASHRAE no later than January 25, 2002.
Figure 1 Typical chilled water plant and distribution system.

systems, actual pump energy impact will be less than this theo- the common leg or in the primary and secondary supplies from
retical relationship suggests,1 but the impact is significant which flow in the common leg can be deduced. When flow in
nonetheless. the secondary exceeds the primary, as indicated by flow in the
common leg moving from the secondary return toward the
The impact on chiller energy usage is more complex to
secondary pumps, another primary chilled water pump and
determine and will be a function of how the chillers are
chiller are started. A pump and chiller are shut off when flow
controlled. There are two basic chiller start/stop control strat-
in the common leg exceeds that of one pump, with some addi-
egies, one based on system flow rate and the other based on
tional margin to prevent short cycling.
load. Ideally, the two strategies would be effectively the same
since flow and load should track in a variable-flow system. The load-based strategy measures system load or indirect
However, when flow and load do not track, when delta-T falls, indications of load such as return water temperature. Chillers
neither strategy works ideally. are started when the operating chillers are operating at their
maximum capacity. Chillers are stopped when the measured
The flow-based strategies stage chillers and primary load is less than the operating capacity by the capacity of one
chilled water pumps in an attempt to keep the primary system chiller.
flow larger than the secondary system flow. In this way, the
So what happens when delta-T falls below design levels
secondary supply water temperature is equal to the primary
and flow and load are no longer in synch?
water temperature leaving the chillers. Flow is often sensed in
The flow-based control system will always make sure
1.
loads are met by starting additional chillers and pumps to keep
The cubic relationship between pump energy and flow assumes
the primary system flow larger than the secondary flow. But
that pressure drop varies as the square of fluid velocity, an
assumption valid only for fully developed turbulent flow in this means that chillers are not fully loaded when delta-T is
systems with fixed geometry. In real systems, pressure drop in below design. For example, assume the system was sized for
most system elements varies less that this since flow is not fully a 14°F delta-T on both the primary and secondary sides. If the
turbulent; velocities are more typically in the transitional region system were at 50% load but the actual delta-T was only 7°F,
between turbulent and laminar flow at design load conditions and
in the laminar flow regime at low loads. Control valves open and
all the chillers and primary pumps in the plant would have to
close, changing system geometry and, hence, its flow character- operate to keep the primary flow up. This wastes pump energy
istics. Most variable-speed pumping systems maintain a mini- and chiller energy since the chillers would all be operating at
mum differential pressure setpoint, further changing the pressure 50% of capacity, less than the 65% to 85% range where effi-
drop/flow relationship. Finally, motor efficiency and (to a lesser ciency is typically maximized for fixed-speed chillers.
extent) variable-speed drive efficiency drop off at lower loads. All
of these factors combine to make part-load pump energy savings The load-based control system would not start a new
less than what may be expected from the ideal cube-law relation- chiller until the operating chillers were loaded. As delta-T
ship. degrades, secondary flow increases, causing water in the

2 AC-02-6-1
 TABLE 1 Setpoints are often adjusted downward by building engineers
Coil Performance at Low SAT Setpoints trying to resolve a comfort problem quickly, although this is
seldom the real source of the problem. In the author’s experi-
CHW % of ence, coils are seldom undersized. Cooling shortages are more
Leaving Air Temperature Flow Delta-T, Design typically due to undersized zone terminal boxes or ductwork
Setpoint, °F Rate, gpm °F GPM to those boxes and to the inability of the supply fan to over-
54 80 13 100% come the resulting high pressure drop to these zones.
53 104 11 130% A similar effect can be caused by improper sensor or
controller calibration. The controller may be set to a proper
52 143 8.5 179%
setpoint on the dial, but improper calibration may cause the
51 208 6.5 260% system to try to achieve a much lower actual setpoint.
50 327 4.3 409% Mitigation: Check setpoints regularly and adjust to
49 cannot be attained design levels or higher. If the control system is digital, “lock”
setpoint ranges in software to prevent operators from setting
Based on a six-row 100 fpf coil, 78°F entering dry-bulb overly low setpoints. Calibrate sensors and controllers semi-
63°F entering wet-bulb.
annually or annually. This is particularly true when pneumatic
common leg to flow from the secondary return back into the controls are used, but even the RTDs and thermistors used with
secondary pumps. This causes the secondary supply water modern digital control systems drift over time and must be
temperature to rise, which in turn causes coil performance to recalibrated regularly.
degrade, which in turn causes control valves to open more to The problem of exceeding design flow rates when valves
demand more flow, which in turn causes ever increasing flow are wide open can be resolved by installing automatic flow
in the secondary and ever warmer supply water temperatures. control valves, which are self-powered valves that maintain a
Eventually, coils will starve, their control valves will be wide preset maximum flow regardless of the differential pressure
open, and temperature control is lost. The system controlling across them (within certain limits). These valves, however,
chiller staging would be oblivious to these problems; it would add considerable cost and pressure drop to the system, both
not start more pumps and chillers since the operating chillers due to the valves themselves and to the strainers that are often
were not fully loaded. required in front of them to prevent valve clogging. Also, these
The solution to these problems lies in first maximizing valves can only prevent flow from exceeding design rates; an
delta-T as much as possible but then designing the plant to improper setpoint can still result in considerable bypass of
accommodate the low delta-Ts that will inevitably occur. water above what is actually “used” in the coil at low-load
conditions even though flow is limited.
DEGRADING DELTA-T: CAUSES AND MITIGATION
Use of Three-Way Valves
The causes of degrading delta-T can be broken into three
categories: Three-way valves always have a negative impact on
delta-T and almost always can and should be avoided.
• Causes that can be avoided by proper design or opera- Three-way valves theoretically maintain constant flow
tion of the chilled water system and, hence, delta-T will vary proportionally with the load. In
• Causes that can be mitigated or resolved but through fact, three-way valves will cause an increase in flow at part
measures that may not result in overall energy savings load and thus have an even worse impact on delta-T than one
• Causes that are inevitable and simply cannot be avoided might think.
One reason for this is the often missing or unbalanced
CAUSES THAT CAN BE AVOIDED valve in the coil bypass leg of the three-way valve connection.
This valve is absolutely required except where coil pressure
Improper Setpoint or Controls Calibration
drops are very low (e.g., reheat coils). When properly
Probably the most common cause of low delta-T is balanced, the valve is throttled down to match the coil pressure
improper setpoints on controllers controlling supply air drop so that when the three-way valve is in either the flow-
temperature off of cooling coils such as those in VAV systems through-coil or flow-through-bypass positions, the pumping
and other central fan systems. When the setpoint of a cooling system will see the same pressure drop. It seems that most
coil is set too low, the controller causes the chilled water valve three-way valve installations the author has observed either
to open fully since it is unable to attain the setpoint no matter have no valve in the bypass or the valve is wide open, indicat-
how much chilled water flows through the coil. Table 1 shows ing that it was never balanced. The impact of either condition
how even a modest drop in supply air temperature setpoint is that at part load, when flow is bypassed around the coil, a
from 54°F to 51°F can cause coil flow rate to more than double hydronic short-circuit occurs, i.e., supply water is directly
and delta-T to drop in half (see also example in Kirsner 1995). injected into the return. Flow increases above design rates and

AC-02-6-1 3
differential pressure in the system drops, possibly starving and middle row), rather a standard “full row” coil where the
other coils. water enters only into the last row of the coil (the row at the air
Even when the bypass valve is properly balanced, flow discharge side of the coil). The dual-row coil has lower tube
will only remain at design rates when the control valve is fully velocity and, hence, lower water-side pressure drop than the
open or fully closed to the coil. When the control valve is in standard coil. But (as pointed out in the laminar flow discus-
between these two extremes, flow will always increase sion below), high water velocity will improve coil part-load
because pressure drop through each circuit varies approxi- performance. Figure 7 shows how the dual-row coil has a
mately as the square of the flow rate. Therefore, when the lower delta-T at part load than full row coils, and flow falls
valve is in the 50% mixing position, the flow rate will approx- into the transition and laminar flow regimes sooner because of
imately double, not remain constant as desired. This will the lower initial velocity.
further degrade delta-T. Mitigation: Document the plant design delta-T and
Mitigation: The bottom line is: never use three-way chilled water supply temperature well and ensure that design-
valves in variable-flow systems, except perhaps for one or two ers select coils for equal or higher delta-Ts. Selecting coils
valves to ensure that pumps are never dead-headed. This based on a somewhat higher than design chilled water supply
includes one of the most common misapplications: the so- temperature (e.g., 1°F to 2°F higher) is prudent to account for
called end-of-line three-way valve (see Buildings A and C in coil fouling over time. Some exceptions should be allowed,
Figure 1) intended to keep water constantly moving through however. While high delta-Ts are favorable in reducing water-
the circuit. The argument for end-of-line three-way valves is side costs and energy usage, they are detrimental to air-side
that they will ensure chilled water will be available immedi- (fan) energy usage. Achieving larger water-side delta-Ts
ately on demand by any coil in the system. The concept prob- requires more heat transfer area (more rows and fins), which
ably comes from domestic water recirculating systems, which results in higher fan pressure drops and higher fan energy. The
are designed to make sure people do not have to wait to get hot ideal balance between water-side and air-side considerations
water at a lavatory. But the concept is not applicable to most will vary with each fan system, and it may be that in some
chilled water systems simply because it seldom matters if coils cases, delta-T should be sacrificed for reduced air-side pres-
have to wait a few seconds or a few minutes to get chilled sure drop. A detailed discussion of this trade-off is beyond the
water when they call for it. For instance, assume a system scope of this paper (see Taylor et al. 2000).
located one mile from the central plant has been operating on Coils should almost always have “full row” headers to
economizer, then the outdoor air temperature rises 1°F so that maintain high part-load delta-Ts, although again, some excep-
some chilled water is needed to maintain desired supply air tions should be considered. The question is whether the reduc-
temperatures. At typical piping velocities, it may take only tion in pump head achieved with the dual-row coils will offset
about 10 minutes for water to travel the one mile from the plant the higher flow rates required at part load. The answer depends
to the coil. Surely, the building will not go out of control in 10 on the specifics of the system design, how pumps are
minutes. Had the system had its own chiller, it might take that controlled, whether the coil is part of the circuit determining
long to start the system and cool down the mass of water within pump head, how many hours the system is at each part-load
it. For the more typical close-coupled chilled water distribu- point, etc.
tion system, the waiting time will be much less than 10 Coil selection parameters need not be the same for each
minutes and even less likely to be of concern. Thus, unless coil. The best design may be to use dual-row coils on the
some process load requires instantaneous chilled water, or if hydraulically longest runs to reduce pump head, then compen-
the chilled water plant serves buildings very far away, end-of- sate by using higher delta-T coils on circuits closer to the
line three-way valves are not necessary and should be avoided. pump where excess head is available.

Improper Coil Selection Improperly Selected Control Valves

Probably the most common coil selection error is simply Two aspects of two-way control valve selection are
selecting the coil for a delta-T that is lower than the plant particularly critical for variable-flow systems: valve size (Cv)
design delta-T. Buildings in a multi-building campus are often and actuator size (shutoff capability).
designed by different engineers. For some reason, some engi- Proper sizing of control valves is critical in a two-way
neers “are not with the program” and do not use the correct valve system, particularly when pumps are uncontrolled
delta-T in their coil selections. For example, a plant may be (riding their curves). Many designers feel that oversizing
designed for a 14°F delta-T, but the engineer of one building control valves is no longer a concern in modern systems with
served by the plant selects coils based on a “standard” 10°F direct digital controls using PID control loops and variable-
delta-T simply out of ignorance or laziness. speed drives to control system pressure. This is certainly
Low delta-T also can also be caused by coils selected to partly true, but no amount of control magic can compensate for
minimize coil water-side pressure drop. One way to reduce a grossly oversized valve.
coil pressure drop is to select a “dual row” coil, one that is Oversized control valves cause the controller to “hunt,”
piped with the water entering at two coil locations (last row alternately opening and closing the valve, over- and under-

4 AC-02-6-1
shooting the setpoint. The overall average flow is higher than ing calibrated balancing valves, will not work because they
desired, and thus delta-T is reduced. can only balance the system under one operating condition;
Undersizing the actuator is another common problem. the system will not maintain balance under variable-flow
Valve/actuator combinations usually have two ratings: the conditions as control valves open and close.
close-off rating is the maximum differential pressure across
the valve against which the valve and actuator can completely No Control Valve Interlock
close. The dynamic close-off pressure rating is the maximum Controls are usually designed for low cost and simplicity.
differential pressure for modulating (as opposed to two-posi- As a result, it is not uncommon for valves to remain under
tion) applications; above this pressure, control through the control even when the associated air handler is off. The
entire stroke will no longer be smooth and the design turn- controller will futilely try to achieve the desired space or
down ratio will not be achieved. Often valves are selected only supply air temperature and ultimately cause the control valve
to achieve the required close-off rating, but they cannot to fully open, bypassing cold supply water into the return with
provide modulating duty under high differential pressures. little or no temperature rise. This is a much more common
The result is hunting and excess water being forced through problem with pneumatic control systems, where one or more
the coil. EP valves must be added and interlocked to the supply fan at
Two-position (as opposed to modulating) control valves considerable expense, than for digital controls where valve
are often blamed for low delta-T problems. These valves, interlocks can be made at little cost through software.
commonly used on small fan-coil units, cause flow through Mitigation: Control valves must be interlocked to shut
the coil to be either “on” (at design rate or higher) or “off”. off flow when the associated air handler shuts off.
When on, it is argued that the flow will exceed that required by
the load under all conditions except design conditions, and Improperly Piped Coils
thus water will leave the coil at less than design leaving water
temperature. But in fact the impact on delta-T is not great, and It is not uncommon to find chilled water coils piped back-
it may even be better than for modulating flow under low-load wards. Instead of being piped in a counterflow arrangement,
conditions. The leaving water temperature at low load is not they are piped in a parallel-flow arrangement with water enter-
low because the valve will only stay open long enough to ing the coil on the same side as the entering air. A coil piped
satisfy the load. Assuming flow is balanced (i.e., limited to counterflow can achieve “overlapping” temperature ranges
design flow by flow control valves or a self-balancing system with the supply air, e.g., the leaving water temperature can
layout), delta-T is negatively affected only by the lower enter- enter at 44°F and leave at 60°F while the supply air enters at
ing air temperature to the coil. But coil entering air tempera- 80°F and leaves at 55°F. With parallel-flow piping, the two
ture is fairly constant for most fan-coil installations because ranges cannot overlap: leaving water temperature will always
they usually supply only air returned from the space with little be a few degrees cooler than the leaving supply air tempera-
or no outdoor air. Thus, the delta-T will not degrade signifi- ture. Thus, if 55°F is maintained, flow must be much higher
cantly at part load. On the other hand, if flow is controlled by and the return water temperature will only be in the low 50s.
a modulating valve, delta-T will increase at first (see Figure 7), The desired delta-T would be impossible to attain.
but the transitional/laminar flow effect at low load will ulti- Mitigation: Coils must be piped counterflow.
mately cause a drop in delta-T. Thus, the selection of two-posi-
tion or modulating valves should be based on other Improper Tertiary Connection and Control
considerations (e.g., cost, thermal performance), not their Figure 1 shows a tertiary pump at Building C. Tertiary
impact on delta-T. pumps are most commonly used:
Mitigation: Valve and actuator selection is beyond the
scope of this paper. See Taylor (1996) for a detailed discussion • when the building requires greater pump head than the
of valve selection issues. secondary pumps can deliver, e.g., at remote buildings
If valves do not have adequate close-off capability, one (in this case, the pump is often referred to as a booster
way to mitigate the problem is to control system differential pump) and
pressure using a variable-speed drive on the pump. The differ- • to maintain a warmer supply water temperature to the
ential pressure sensor should be located as far out into the building. In this case, the pump is often referred to as a
system as possible, usually at the hydraulically most remote blending pump.
coil, and the setpoint should be only as high as necessary to
deliver design flow rates at design conditions. Usually, the two-way control valve in the secondary
If two-position valves are used, flow must be balanced return line is modulated to maintain the supply water to the
using either automatic flow-limiting valves or by designing building at a given setpoint. This setpoint must be above the
the system to be reasonably self-balancing, e.g., using reverse- chilled water temperature supplied to the building or the valve
return or oversized headers. Manual balancing valves, includ- will simply be opened fully by its controller in the futile

AC-02-6-1 5
 TABLE 2
Coil Performance with Increasing CHWS Temperature

Leaving
Chilled Water
Entering Chilled Water Flow Delta-T, Temperature,
Temperature,°F Rate, gpm °F °F
42 30 16.7 58.7
44 34.5 14.7 58.7
46 41 12.3 58.3
48 53 9.5 57.5
Based on an eight-row 96 fpf coil, 77°F entering dry-bulb/ 62°F
entering wet-bulb, 55°F leaving air temperature.

attempt to maintain an unachievable setpoint. This bypasses

chilled water from the secondary supply to the return through
the common leg, severely reducing delta-T when the load on
the building is low. If there is no flow-limiting device in the
connection to the secondary mains (e.g., an automatic flow
control valve), this bypass flow can far exceed the design flow
to the building because of the low pressure drop in the Figure 2 Check valve in common leg. Check valve only
common leg. This hydronic “short-circuit” also causes a drop allows tertiary (building) flow to exceed
in differential pressure across the secondary system, possibly secondary flow. Chilled water supply cannot
starving other buildings that do not have tertiary pumps. short-circuit to return. Tertiary pump is in series
Mitigation: For retrofit applications, a simple solution is with secondary pumps at low building flows, so
to ensure that the setpoint of the controller maintaining supply tertiary pump must be variable speed driven to
water temperature to the building is several degrees above the keep from over-pressurizing building.
chilled water temperature being delivered to building. To do
this, the latter temperature should be constantly monitored and
ing loop, it also slightly increases flow in the secondary loop.
the setpoint constantly adjusted to maintain the desired differ-
Furthermore, this control is indirect and may inadvertently
ential, in case the supply water temperature from the second-
starve coils if they require design water temperature to meet
ary system changes (e.g., reset).
loads. Clearly, this is not a good control strategy.
This control will not help improve delta-T if the coils
downstream are not maintaining a high delta-T. To resolve this Another possible retrofit solution is to place a check valve
problem, some designers move the temperature sensor in the common leg as shown in Figure 2. This ensures that
controlling the two-way valve from the supply line to the chilled water supply will never be bypassed through the
return water line. The control valve is then modulated to main- common leg to the return. Building water flow will always
tain the return water temperature at design levels, e.g., 58°F. equal or exceed that supplied from the secondary system.
If return water is too cold, it is essentially recirculated back Placing anything in the common leg, in particular a check
into the building to absorb more heat. This control can be valve, is considered heresy by some designers because the two
effective if low delta-T is caused solely by reduced coil effec- pumping systems are then no longer hydraulically indepen-
tiveness at low loads and all coils are experiencing similar dent. This author’s response: So what? Designers have been
loads. Feeding the water back to the coils in this case delivers taught that hydraulic independence has some intrinsic value,
the same effect as the pumped coil in Figure 8 (discussed in but its value is generally that it prevents series coupled pumps
next section). The fact that the load is low allows the water from overpressurizing two-way valves. In this case, the check
temperature to the coil to be warmer than design, so the coil valve will cause the secondary and tertiary pumps to operate
should not be starved. However, while instinctively it may in series when the tertiary flow is low and chilled water is still
make sense that recirculating water will increase the return in demand (the chilled water valve on secondary return is
water temperature, the return water temperature is driven open). This could overpressurize the tertiary loop and over-
more by the entering air temperature and the coil effective- power some control valves. But the problem is resolved if the
ness, not the entering water temperature. Table 2 shows that tertiary pump is variable speed driven; it will simply slow
for a given load, increasing entering chilled water temperature down if the secondary system differential pressure is transmit-
results in a lower, not higher, leaving return water temperature. ted to the building due to the check valve. The use of variable-
So recirculating water not only increases the flow in the build- speed drives makes hydraulic independence of little value.

6 AC-02-6-1
Simply deleting the common leg is similar to adding the check chilled water when it is “on” as it does when it is “off.” This
except that now the secondary and tertiary pumps are always is not always clear to the designer; he or she is directed to
in series. Again, with variable-speed drives, this should not deliver so much chilled water to a process device, but it is not
cause any hydraulic problems (Rishel 1988). clear what happens inside the device and whether it includes
For new systems, by far the best solution is not to have any modulating or shutoff controls. When controls are lack-
tertiary pumps or, rather, not to have conventional secondary ing, chilled water delta-T falls whenever these process
pumps. Figure 3 shows the same building complex as Figure systems are not at full load.
1 but with secondary pumps moved out to what was formerly Mitigation: The designer should work with the process
the tertiary pump position. The common leg for the system is equipment supplier to determine if controls are present, and if
still back at the central plant. Building pumps are sized to not, whether external, field mounted shutoff valves may be
handle the head from the plant to and through the building then installed.
back to the plant. This design not only eliminates the problems Some process equipment does not require the low chilled
associated with bypass connections, it reduces pump energy water temperatures typical of HVAC loads. If the equipment
by allowing secondary pumps to be sized just for the head can be served with water in the 50s, supply return chilled water
required for the building they serve, rather than the highest- instead of supply water using a pump piped in a primary-
head circuit for which the secondary pumps in Figure 1 must secondary (or secondary/tertiary) manner as in Figure 4. This
be sized. For large coils (shown in the Figure 3 as Building B), will increase system delta-T even further and will require no
two-way control valves may be eliminated and the variable increase in plant chilled water pumping capacity.
speed driven pump can be used to directly control flow
through the coil (pump speed is controlled by the same CAUSES THAT CAN BE RESOLVED
controller that would have controlled the two-way valve). This BUT MAY NOT RESULT IN
reduces pump head by eliminating the control valve pressure OVERALL ENERGY SAVINGS
drop (typically on the order of 10 to 15 feet) and by allowing
cube-law energy savings through the whole range of flow Laminar Flow
through the coil (typical VFD control maintains a fixed mini- Heat transfer coefficient on the water side of the coil is
mum differential pressure that allows only linear reductions in primarily a function of flow turbulence, which is described by
pump power at low flows). Blending, if desired to maintain the Reynolds number, a dimensionless number defined as
higher supply water temperatures (e.g., for process loads), can
still be accommodated by adding a blending bypass valve (as VDρ
Re = ------------ (2)
µ
shown in Building C in Figure 3).
where V is the tube velocity, D is the tube diameter, ρ is
Uncontrolled Process Loads density of the water, and µ is the viscosity.
A chiller plant in an industrial environment may serve The water-side heat transfer characteristics of coils are
process loads in addition to cooling coils. Some process equip- not well understood. Many papers on low delta-T problems
ment has no flow control devices and, hence, uses as much have attributed the problem in part to a sudden drop in heat

Figure 3 Chilled water plant with remote secondary pumps.

AC-02-6-1 7
Figure 4 Process load piped to CHW return. Many process
loads do not require chilled water in the 40s. If
warmer water is acceptable, piping the load in a Figure 5 Heat transfer factor as flow varies. This coil was
primary-secondary manner can improve plant selected for 3 fps design velocity with 5/8 in.
delta-T. tubes. In this case, the coil never experiences fully
developed turbulent flow; the design condition is
transfer coefficient as flow goes from the turbulent regime to already in the transition region. Laminar flow
the laminar flow regime when the Reynolds number drops occurs at 0.5 to 0.8 fps, roughly 20% to 25% of
below about 2000. However, this is based on a simplistic coil design flow. Data obtained from coil
model that neglects the fact that coil tubes are not infinite in manufacturer selection program correlated to
length. Coil bends at the end of each row cause flow to be measured coil data under low flow conditions.
turbulent even when the velocity is below that which would
normally correspond to laminar flow in long tubes. According
to ARI Standard 410 (ARI 1991), for coils with smooth tubes,
there are three flow regimes:

• Fully developed turbulent flow (Re ≥ 10,000). This high

a Reynolds number seldom occurs in coils (or in typical
HVAC piping systems in general).
• Transitional flow (2100 ≤ Re < 10,000). This is the most
common flow regime for typical coils and HVAC pip-
ing.
• Laminar flow (Re < 2100). This occurs at low flow rates
(low loads).

Figure 5 shows this effect on heat transfer factor J

(defined as St Pr2/3 (µs/µ) 0.18 where St is the Stanton number,
Pr is the Prandtl number, and the subscript s refers to the condi- Figure 6 Heat transfer resistance as flow varies. This coil
tions at the inside surface of the tube) for two typical coils, one is the same as the one in the previous figure with
12 feet long and one 2 feet long. At high, turbulent flow rates, heat transfer factor converted to percent of
J is the same for both coils. As velocity decreases into the tran- design the film heat transfer resistance at the
sition region, the heat transfer factor begins to fall, but less so inside surface of the tube and Reynolds number
for the shorter coil because the tube bends tend to keep flow converted to percent of design flow rate.
more turbulent. At the onset laminar flow, the heat transfer
factor begins to rise. Figure 6 shows the same data with the
heat transfer factor converted to percent of design of the film flow conditions, it accounts for almost 90% of the overall
heat transfer resistance at the inside surface of the tube and resistance.
Reynolds number converted to percent of design flow rate. There is no sudden rise in heat transfer resistance as tradi-
Film heat transfer resistance is only a small portion of the over- tional theory predicts. However, the significant increase in
all air-to-water heat transfer resistance at the design flow rate, film resistance at low flows would still support the notion that
but as water velocity falls, this resistance rises until, at laminar delta-T in the laminar flow region should fall. But there is

8 AC-02-6-1
another factor occurring at the same time that more than effects are, therefore, unlikely to be a major source of degrad-
offsets this rise in heat transfer resistance: the low flow rate ing delta-T syndrome.
through the coil effectively “sees” an oversized coil, a large Many articles that referenced laminar flow as a problem
amount of heat transfer surface area relative to the amount of proposed as a solution the addition of a tertiary (or “quadri-
water running through the coil. The water stays in the coil ary”) pump at each coil, piped to maintain a constant water
longer and more heat transfer occurs, which causes tempera- velocity through the coil (Figure 8). This will ensure flow is at
ture rise to increase rather than decrease. Figure 7 shows delta- the design rate, nearly fully turbulent, eliminating any laminar
T as a function of space sensible load in a VAV system for three flow effects. However, first costs and maintenance costs will
coil types. The effect of increasing coil effectiveness due to the increase substantially and, ironically, pump energy costs will
reduced flow rates overcomes the increase in film resistance also increase substantially, counter to one of the primary goals
so that coil delta-T remains above design delta-T except at the for installing the coil pumps. The reason is that the coil pumps
transition point into the laminar flow regime. Contrary to are constant volume (constant energy) and, because of their
conventional thinking, delta-T below the onset of laminar small size, they are inherently less efficient than the larger
flow increases rather than decreases. secondary pumps. Also, total head on the system will increase
The point at which delta-T begins to fall is a function of due to the added valves, fittings, etc., required to install the
initial coil tube velocity, but it typically is at around 40% of coil pumps. These penalties are sufficient to more than offset
load for VAV systems and 50% of load for constant volume the reduced secondary pump energy provided by the higher
systems. The point is lower for variable volume systems delta-Ts at low load.
because the supply air temperature on these systems typically Tube diameter has only a minor impact on coil part-load
remains fairly constant at near design conditions (keeping performance. One would think that smaller tubes would
approach temperature difference constant even at reduced increase tube velocities for the same flow rate and, hence, keep
loads) and the airflow rate falls (reducing air-side heat transfer turbulence and associated heat transfer coefficients high.
coefficient). Thus, coil effectiveness is less than for constant Also, since velocity is inversely proportional to the square of
volume systems, requiring higher flow rates for the same the diameter for a given flow rate, Equation 2 indicates that a
space load. smaller tube diameter would reduce the flow rate at which
Mitigation: The laminar flow “problem” so often refer- laminar flow occurs. However, when tube size is reduced in a
enced in the literature does not appear to be a real problem. real coil, more tubes are used, so the velocity increase is not as
Yes, delta-T will fall near the transition to laminar flow, but it dramatic as one might expect. Figure 7 shows that typical coils
is still very near design delta-T. During most low-load flow with 1/2-inch tube diameter do outperform the coil with 5/8-
conditions, delta-T will be above design delta-T. Laminar flow inch tubes, but only slightly. The percentage of flow at which

Figure 8 Pumped coil. Coil is piped in a traditional

primary-secondary manner. The control valve is
Figure 7 Delta-T at part load. Coils shown were selected modulated in the same manner as it would
for a 13°F design delta-T. Load reduction is first without the pump (e.g., off of supply air
by reducing volume to 40%, then reducing temperature, as shown, or off of space
entering air temperature. Leaving air temperature.) The pump maintains constant flow
temperature is constant, as per standard VAV through the coil regardless of the operation of
systems. Performance determined by coil other pumps upstream or the position of the
manufacturer’s expanded simulation program, control valve. Constant flow eliminates laminar
allowing simulation below the laminar region. flow effects.

AC-02-6-1 9
laminar flow occurs for these particular coils drops from about Air-side fouling is usually minimized by good filtration.
29% of design for the 5/8-inch tubes to 27% for 1/2-inch tubes. Filters should be at least 30% dust-spot efficiency (or MERV
The higher pump heads caused by the higher water-side pres- 6 or 7 using the new ASHRAE 52.2 rating method).
sure drop associated with 1/2-inch tubes will almost surely
offset any pump energy savings due to this slight increase in Outdoor Air Economizers
performance. and 100% Outdoor Air Systems
The selection of 1/2-inch-diameter or 5/8-inch tubes will One issue very often overlooked as a cause of degrading
hinge on first costs versus pump head requirement at full load. delta-T on systems designed for high delta-Ts (e.g., above
The impact of this choice on delta-T is very small, and should 14°F) is the impact of integrated outdoor air-side economizers
not be a consideration. and 100% outdoor air systems. When the weather is cool but
not cold enough to provide 100% of the system cooling load,
Chilled Water Reset
these systems deliver 100% outdoor air but need a small
Chillers are more efficient at higher leaving water temper- amount of chilled water to meet cooling demands. Under these
atures, so resetting chilled water temperature upward at low conditions, the air temperature entering the coil is low, causing
loads can be an effective energy-saving strategy. However, correspondingly low return water temperatures. For instance,
high chilled water temperature will reduce coil performance a coil might be designed for 80°F entering air temperature with
and, hence, coils will demand more chilled water, lowering a chilled water return temperature of 60°F. When the outdoor
delta-T. Table 2 shows this effect: coil flow rate almost air temperature is 60°F, it is clearly impossible to maintain a
doubles when CHW supply temperature is increased from 60°F return water temperature. A coil on a VAV system
42°F to 48°F. designed for 44°F chilled water and an 18°F delta-T would
Mitigation: The best chilled water reset strategy will only be able to achieve an 11°F to 15°F delta-T at 55°F to 65°F
vary depending on the plant design, chiller performance char- outdoor air temperatures.
acteristics, and the nature of coil loads. Smaller plants, those Mitigation: The impact of low coil entering air temper-
with low pumping distribution losses, will usually benefit ature can be mitigated by using a lower design delta-T.
from chilled water reset. For large plants with high pumping However, this will increase pump energy under all operating
distribution losses, raising chilled water temperature will conditions, so it clearly is not a reasonable solution.
increase pumping energy more than it reduces chiller energy,
resulting in a net increase in plant energy usage. These plants ACCOMMODATING LOW DELTA-T
may benefit from lowering the chilled water setpoint even
The causes of low delta-T due to improper design, oper-
below design levels in mild weather. A complete analysis of
ation, or maintenance can and should be eliminated. But even
these various options is beyond the scope of this paper.
with a perfectly designed, operated, and maintained system,
CAUSES THAT CANNOT BE AVOIDED there are still factors, such as the impact of air economizers
and coil effectiveness degradation as systems age, that will
Reduced Coil Effectiveness result in degrading delta-T. Degrading delta-T is inevitable.
In the author’s experience, delta-Ts in real systems, due to
Coil heat transfer effectiveness is reduced by water-side
a combination of the factors listed above, can be expected to
fouling (e.g., slime, scale, or corrosion on the inside of coil
fall to about one-half to two-thirds of design at low loads.
tubes), air-side fouling (e.g., dirt buildup on coil fins), and air-
Figure 9 shows trend data for a large chiller plant in San Jose,
side deterioration (e.g., deteriorating fins). Any reduction in
California. The average delta-T was about half of the design
coil effectiveness increases the flow rate of water required to
delta-T.
deliver the desired leaving water temperature, thus reducing
delta-T. If delta-T degradation is inevitable, then chiller plants
Another related problem is a reduction in air flow caused must be designed to operate efficiently under high flow, low
by, for example, dirty filters. This reduces air-side heat trans- load conditions. There is nothing one can do about the increase
fer coefficients and reduces the overall space cooling capacity in distribution pump energy caused by low delta-T, but here
of the system. The space thermostat then causes the chilled are a few ideas on how the chiller plant can be designed so that
water valve to increase water flow rates to deliver colder chiller energy is not impacted by low delta-T. They fall into
supply air temperatures to compensate for low air flow, thus two categories: improve chiller low-load performance so that
reducing delta-T. premature staging does not affect energy use (item 1) and
providing more flow through operating chillers so that they
Mitigation: Water-side fouling is easily controlled with
may be more fully loaded before another chiller must be
water treatment at the time the system is filled. Since the
brought on line (items 2 to 4).
system is generally closed, water treatment need not be a large,
ongoing expense. If, however, the system has considerable 1. Variable-speed chillers. Variable-speed drives on chillers
leaks or otherwise requires considerable make-up water, water dramatically improve part-load performance during low
treatment should be automated to ensure its adequacy. ambient conditions when condensing temperatures can be

10 AC-02-6-1
 Figure 9 Degrading delta-T in large chiller plant. The design delta-T for this chiller
plant in San Jose, Calif., was 10°F. This range occurred only a few hours
per year.

Figure 11 Primary-only pumping. Bypass valve maintains

minimum chiller flow at low loads.
Figure 10 Variable speed vs. fixed speed chillers at part
load. These data are for a two-chiller plant with
equally-sized chillers and dedicated condenser 2. Primary-only pumping. An increasingly more common
water and primary chilled water pumps design in modern chilled water plants is the primary-only
interlocked to start with each chiller. Chiller variable-speed pumping scheme (Figure 11). This design is
models did not include ARI 550/590-98 less expensive to install and also reduces pump energy costs
tolerances. compared to primary-secondary systems, even for systems
without delta-T degradation (Taylor et al. 2000; Avery
2001). The pumps also can operate more efficiently when
reduced below design conditions. Even with parasitic delta-T does degrade because at low loads they have the
energy users such as interlocked condenser water and capacity to overpump flow through the chillers, i.e., the
primary chilled water pumps, it is more efficient to run two pump head not used in the system is available to force addi-
chillers than one at loads above 25% to 35% of plant capac- tional flow through the chillers, increasing the load they can
ity (see Figure 10). Thus, even if low delta-T at low loads handle.
causes more chillers to operate than are required to meet the 3. Check valve in the common leg. Figure 12 shows a check
load, chiller plant energy will be no higher and can be much valve installed in the common leg between the primary and
lower than if chillers were “maxed out” before the next secondary systems. The check valve always keeps the
chiller was started. primary flow greater than or equal to the secondary flow. If

AC-02-6-1 11
 the secondary flow increases beyond the capacity of the speed-driven pump has sufficient flow and head capacity to
primary pumps, the check valve essentially places the overpump either of the two small chillers to compensate for
primary and secondary pumps in series with each other. low delta-T. At 50% plant load and delta-Ts as low as 50%
This causes flow through the chiller to increase along with of design, both pumps can operate through the two small
the secondary flow. Since the primary pumps were selected chillers or through the large chiller. As loads go above 50%,
for design flow and head, to increase flow beyond design the large chiller and one of the small chillers can still be
rates, the secondary pumps must speed up to “draw” water overpumped, although not to the same degree. This should
through the primary loop, assisting the primary pumps in be satisfactory since delta-T generally improves at higher
overcoming the increased pressure drop of the primary loads so the need for overpumping is reduced. Above 75%,
loop. Eventually, the secondary pumps will be at full speed all three chillers operate. Overpumping is still possible
and the system will be at the maximum flow rate it can since pump heads were oversized and pumps can ride out
achieve without bringing more chillers and primary pumps on their curves as they become in series with the secondary
on line. Pumps and chillers are staged on when the second- pumps due to the check valve. (The check valve is optional
ary pumps are at full speed and beginning to starve of flow. when primary pumps are oversized since the pumps can
Depending on the pump curves and other system details, provide increased flow without any help from the second-
adding the check valve can cause flow through a single ary pumps.)
chiller to increase to about 140% of design rates. This is 5. Low design delta-T in primary loop. If the primary loop is
unlikely to exceed any maximum flow rate established by designed for a smaller delta-T than the secondary loop,
the chiller manufacturer.2 For instance, if the chiller is chillers may be properly staged even with delta-T degrada-
sized for a 14°F delta-T, an increase in flow to 140% of tion in the secondary loop down to the point where the
design is equivalent to the rate associated with a 10°F secondary delta-T degrades to the reduced primary delta-T.
delta-T at full load, clearly within a manufacturer’s limits. While this design can mitigate the energy impact of prema-
Primary pumps also are unlikely to exceed the limits of ture chiller staging, it will result in additional primary pump
their curves with an increase of only 40% if selected as energy under all load conditions (even when system delta-
required for any non-overloading parallel-flow applica- T has not degraded) and in larger primary pumps, piping,
tion. etc. Accordingly, the other suggestions above may be more
Adding a check valve in the common leg does have one efficient and cost-effective at mitigating degrading delta-T.
drawback: if the primary chilled water pumps are shut off
or fail and the chiller isolation valves close (Figure 12),
any operating secondary pumps will be dead-headed and
eventually will overheat and damage seals or bearings.
This can be avoided by shutting off secondary pumps
should all primary pumps be off, but that is not always
practical on large campus projects.
While considered heresy by fans of traditional primary-
secondary systems, the check valve has proved successful
at reducing chiller plant energy by as much as 20% on
constant-speed chiller plants (Avery 1997). However, the
valve will most likely have little or no benefit for plants
with variable-speed chillers since the efficiency of these Figure 12 Check valve in the common leg.
chillers is so high at low load.
4. Unequally sized chillers/pumps. Figure 13 is an example of
how pumps and chillers can be unequally sized to provide
better low-load flexibility. At low loads, either variable-

2.
Manufacturers’ maximum flow rates are usually based on some-
what arbitrary maximum velocities designed to limit erosion. The
author has experience with a plant that has routinely overpumped
chillers at rates far above manufacturers’ recommendations for
hours on end for over 20 years with no apparent problems.
Limited laboratory tests also suggest that in clean systems, little
erosion will occur at velocities well above what might be expected
in real systems (Sturley 1975). Hence, occasional excursions
above manufacturers’ limits are probably not cause for concern,
and the high cost and pressure drop of flow-limiting valves can be
Figure 13 Unequally-sized chillers and pumps.
avoided.

12 AC-02-6-1
CONCLUSIONS Avery, G. 2001. Improving the efficiency of chilled water
plants. ASHRAE Journal, May.
Degrading delta-T syndrome is an affliction common to Fiorino, D.P. 1996. Twenty-five ways to raise your chilled
almost all chilled water plants. The reasons for the problem are water temperature differential. ASHRAE Transactions
becoming better understood, and many can be resolved by 102 (1).
proper design and component selection and proper operation Kirsner, W. 1995. Troubleshooting chilled water distribution
and maintenance. But some of the causes of low delta-T are problems at the NASA Johnson Space Center. Heating/
either impossible or not practical to eliminate. Therefore, the Piping/Air Conditioning, February.
system must be designed to accommodate low delta-Ts in an Kirsner, W. 1996. The demise of the primary-secondary
efficient manner while still meeting all coil loads. This can be pumping paradigm for chilled water plant design. Heat-
done by using variable-speed-driven chillers, which are so ing/Piping/Air Conditioning, November.
efficient at part load that under all but the lowest load condi- Lizardos, E.J. 1994. Optimizing delta-T. Engineered Sys-
tions, it is more efficient to run more chillers than are required tems, September.
to meet the load. Thus, additional flow resulting from degrad- Mannion, G.F. 1988. High temperature rise piping design for
ing delta-T will have no impact on chiller energy use. To miti- variable volume systems: Key to chiller energy manage-
gate degrading delta-T for fixed-speed chiller plants, the ment. ASHRAE Transactions 94 (2).
design must allow the chillers to be overpumped (supplied Redden, G.H. 1996. Effect of variable flow on centrifugal
with more than design flow) so that they can be more fully chiller performance. ASHRAE Transactions 102 (2).
loaded before staging on the next chiller. Installing a check Rishel, J.B. 1988. Twenty-five years’ experience with vari-
valve in the common leg of the primary-secondary connection able speed pumps on hot and chilled water systems.
is one way to force increased flow through chillers since it ASHRAE Transactions 94 (2).
places the primary and secondary pumps in series. Other Sauer, J.M. 1989. Diagnosing low temperature differential.
options include sizing primary pumps for increased flow ASHRAE Journal, June.
either using unequally sized pumps or with a lower design Sturley, R.A. 1975. Increasing the design velocity of water
primary loop delta-T. and its effect on copper tube heat exchangers. Corro-
sion/75, Paper No. 58. The National Association of Cor-
REFERENCES rosion Engineers.
Taylor, S.T. 1996. Fundamentals of HVAC control systems.
ARI. 1991. Standard 410, Forced circulation air-cooling ASHRAE Self-Directed Learning Course.
and air-heating coils. Air Conditioning and Refrigera- Taylor, S., P. DuPont, M. Hydeman, and C. Eley. 2000. Cool-
tion Institute. Tools™ Energy Efficient Chilled Water Plant Design
Avery, G. 1997. Controlling chillers in variable flow sys- and Performance Guide. Pacific Gas & Electric Com-
tems. ASHRAE Journal, December. pany.

AC-02-6-1 13