Delivering Tons to the
Register: Energy Efficient Design and Operation of
Residential Cooling Systems
Jeffrey Siegel, Lawrence Berkeley National Laboratory
lain Walker, Lawrence Berkeley National Laboratory
Max Sherman, Lawrence Berkeley National Laboratory
work presented in this paper shows how proper air conditioning equipment,
location, sizing, installation and operation can improve performance, save
on energy bills, and reduce peak demand. A residential heat and mass
transfer model, REGCAP, was used to determine the effect of several
parameters on energy consumption, peak electrical demand and air
conditioner performance. These parameters included placing the entire air
conditioning system within the insulated envelope of the house, reducing
air conditioner capacity, correct installation (refrigerant charge and
evaporator airflow), and alternative operating strategies (thermostat
setback versus constant thermostat set point). Our results indicate that a
properly sized and installed air conditioner has either equivalent or
improved performance compared to an oversized poorly installed air
conditioner that is typical of residential construction. This paper
examines a recent innovation in bringing the HVAC system inside the
thermal and air leakage envelope by locating the system in a cathedralized
attic that is insulated and sealed at the roofline and is well connected
to the house. Both field measurements and simulation results show that
houses with ducts located in cathedralized
attics have dramatically increased cooling performance and lower energy
consumption than houses with ducts in conventional attics. However,
the marginal benefit of improving an air conditioning system once it is in
a cathedralized attic is small: the largest part of energy savings come
from insulating and sealing the attic.
Residential central air conditioning systems use about 11 x 109 kWh/year
California and are also responsible for a significant part of the peak
load conditions in California and the Southwest (calculated from
information given by Wenzel et al. (1997) and California Dept. of Finance
(1997)). Many of the systems that contribute to this load have poor
performance and high energy use. The energy and comfort advantages of
having ducts inside the conditioned space are already widely accepted by
engineers. However, some architects, builders, and homeowners continue to
resist interior ducts because of concerns about aesthetic, space, and
construction issues. Thus, in much of the Southwestern United States, it
is common practice to put ducts in attics that can reach 60�C (140�F) or
even hotter (Carison et al. 1992; Parker et al. 1997a; Walker et al.
Recently, a movement has begun to change the way that attics are built in
hot and dry climates where moisture problems are not an issue (Rudd &
Lstiburek 1998). The idea is to seal and insulate attic at the roof line,
but not to seal and insulate between the attic and the house. This "cathedralizes"
the attic, bringing it inside of the conditioned space. Such astrategy
allows the ducts to be brought inside without compromising the concerns of
builders, architects, and homeowners. The main purpose of this paper is to
determine the effect of cathedralized attics on air conditioner
performance, energy consumption, and power demand. The effects of
refrigerant charge, evaporator air flow), oversizing (relative to ACCA
Manual J), duct leakage, and thermostat operation are considered. This
work represents a continuation of air conditioner performance work
reported in Walker et al. (1998) and Siegel (1999). In addition to the
study of cathedralized attics, this paper adds to the previous work by
including a model of air conditioner energy consumption and peak power, a
more sophisticated house loads model, as well as an examination of
different thermostat operating conditions.
In order to study the effects of cathedralized attics and other
parameters, several simulation runs were performed with the REGCAP
computer simulation program. REGCAP is designed to estimate dynamic
cooling performance in houses with ducts in the attic. REGCAP links an
coupled attic heat and mass (i.e. air flow) transfer model with a house
heat and mass transfer model. REGCAP also includes a duct model that
accounts for leakage and conduction losses as well as flow between the
attic and the house when the air handler is off. A recent addition is a
complete air conditioner model that models energy consumption and capacity
and includes the effects of deviations from manufacturers recommended air
flow and refrigerant charge (Proctor 1999). Details regarding the
structure of REGCAP and required inputs are described by Siegel (1999) and
Walker et al. (1998).
REGCAP has been verified with measured data from seven different houses at
a variety of weather conditions and locations (47 comparisons over all).
Some of this verification work has been described elsewhere (Siegel 1999;
Walker et al. 1999). More recent verification studies have focused on
using REGCAP for unvented cathedralized attics. All of the verification
shows a similar pattern. Specifically, the house and attic temperatures
are predicted within 1�C (<3 average absolute difference in
temperature for the house and <2 for the attic) over the whole day with
the following caveat: if nighttime cloud cover is substantial and this
data was not recorded (and consequently input into the model) that the
model underpredicts attic temperature slightly (<2�C) during the
The duct supply and return temperatures are both predicted very closely
(within 0.5�C or 4 average absolute difference from the measured
temperatures) when the air handler is on, with the exception of two sites
for which the predicted air conditioner capacity varies sharply from the
measured values and thus affects the supply temperature prediction. When
the air handler is off, REGCAP does not do as well at predicting duct
temperatures, as it does not account for flows between different zones in
the house or possible thermosiphon flows. This will be addressed in future
versions, but does not affect the analysis for this paper because the duct
system performance is dominated by the capacity of the air conditioner,
not the initial temperature of the duct.
The equipment model predicts energy consumption and capacity very closely
for all sites (<4 of measured capacity) with the exception of two sites
that potentially had incorrect data on the nameplate or an operating
problem that was not reflected in the input data. One of these simulations
overpredicts capacity by about 10 , the other underpredicts by a slightly
The conclusion from the verification work is that REGCAP performs very
well for both conventional (vented) and cathedralized (unvented) attics.
Its major limitation is that it represents the attic, house, supply and
return ducts each with a single zone and thus can not model intrazonal
flows. This is only a limitation in such situations where such flows are
important such as modeling pollutant transport or heat transfer when the
air handler is off.
The situation that we choose to examine in this paper is much simpler than
either of these cases.
The prototype house that was used as the basis for the simulations for
this study was a 186 m2 (2000 ft2) single story slab on grade house with
the ducts, air handler and cooling coils located in the attic space. This
house with a conventional attic was used for earlier work and is well
described in Siegel (1999) and Walker et al. (1998). One important
difference is that the house used in the earlier work had an asphalt
shingle roof; this house has a concrete barrel tile roof.
This research also used a modified version of the prototype house: the
conventional attic was replaced with a cathedralized attic. The ideal
cathedralized attic is fully inside the thermal and pressure envelope of
the house. It has a perfectly sealed (to outside) attic, transfer grills
to allow for pressure relief between the attic and the house, and
insulation at the roofline. Recent tests performed by LBNL on 4 homes with
cathedralized attics in Las Vegas indicate that although well insulated at
the rooflines, the attics were actually about half inside and half outside
the pressure boundary of the house. There were also no transfer grills
between the house and the attic. This is because the builders are in the
process of changing construction methods, the procedures for sealing
attics are not well developed, and the habit of sealing the attic from the
house is slow to change. For this research, we simulated an attic that is
in between what we saw in the field and the ideal case. The simulated
cathedralized attic has twice as much leakage between the house and the
attic as between the attic and the outside. This case represents an
improvement over very early cathedralized attic building practices, but
assumes some imperfection that is inherent in other parts of residential
construction. In the end, this distinction is not as important as other
effects that will be discussed in the paper. Specifically, the difference
between the energy consumption of an ideally cathedralized attic and
poorly cathedralized attic is only about 5, because most of the energy
savings result from the insulation location.
The following four test cases were used for both conventional attics and
cathedralized attics. The parameters are summarized in Table 1. These are
similar to cases used in earlier research (Walker et al. 1998) with the
exception of the fact that a more thorough reading of air conditioner
literature (Blasnik et al. 1996; Parker et al. 1997b) has lead us to use
slightly lower air handler flows.
� Base ease - This case describes an average new house in California.
Duct Leakage numbers are based on Walker et al. (1997) and Modera and
Wilcox (1995), refrigerant
charge and airflow are based on average numbers from Blasnik et al. (1996)
and Proctor (1997), and air conditioner sizing is based on field surveys
and a contractors rule of thumb of 1 ton of air conditioner for every 500
ft2 of floor area (Brown et al. 1994; Proctor et. al. 1995; Proctor &
Albright 1996; Vieira et al. 1996).
� Poor case - this case describes a below-average typical house in
California. Duct Leakage numbers represent the poorest quartile of
measured leakage in California houses by Jump et al. (1996), refrigerant
charge and airflow are based on measurements from Blasnik (1996) and
Proctor (1997), and air conditioner sizing is the same as the base case.
This case is bad, but certainly not the worst that exists in the region.
� Best case - this case describes an average new house in California
that has been improved by duct sealing, refrigerant charge addition, and
correction of reduced air flow. The air conditioner has not been changed
as this would not typically been done in a retrofit situation. This is the
best that can be reasonably achieved with a retrofit.
� Best resized - this is the best case with a properly sized, according
to Manual J and Manual S (ACCA 1986, 1992) air conditioner.
These four cases were examined for a conventional attic with an insulated
floor and typical venting (1:300 vent area to roof area) and a
cathedralized (unvented and insulated at the roofline) attic. Each of the
cases were run for two different thermostat operation conditions:
� Pulldown - The air conditioner is turned off at 9 am and turned on at
4pm to a set point of24�C (75�F). This simulates a common condition in
California in which occupants are not home during the day, the air
conditioner is turned off during that period, and is turned on to cool
down the house near the time when they return home. The air conditioner
actuation is often done with a programmable thermostat.
� Continuous Cycling - One set point of24�C (75�F) (as used in ACCA
1996). The air
conditioner cycles throughout the day to meet this set point, regardless
of whether the homeowner is present. No internal gains were assumed for
The weather that was used for this study were the design day conditions
Sacramento, CA. The input weather data is more completely described in
Walker et al. (1998).
Energy Use Parameters
The parameters that serve as the basis for the comparisons between the
described above are:
� Pulldown Time - is how long the air conditioner takes to cool down a
house that has been heating up all day to the set point temperature of24�C
(75�F). Homeowners typically like a fast pulldown time (Kempton et al.
� Tons at the register (TAR) -represents the amount of useful cooling
delivered to the house through the supply registers. Walker et al. (1998)
report that this number is always much lower than the nameplate capacity
of the air conditioner.
� Energy Consumption - is the total energy used by the air conditioner
over the course of a day. It includes the energy consumption of the
compressor, air handler blower and the condenser fan.
� Peak Power - is the 15 minute peak power use of the air conditioning
of little interest to homeowners (because they usually do not pay demand
charges), it is of interest to utilities, particularly those which have to
deal with heat storms and other "loads from hell" caused by
residential air conditioners (Brown et al. 1994; Triedler and Modem 1992).
The simulations were
performed using REGCAP and a 1-minute time step to capture the transient
response of the system. House and attic temperatures for the Base pulldown
case (both conventional and cathedralized attics) are shown below in
Figure 1. In the conventional case the attic gets quite hot, 47�C (117�F).
This has a dramatic impact on air conditioner performance because it
increases conduction losses. On peak temperature days, this would be even
higher. In addition, any air entering return leaks is at this high
temperature. Also note that in both cases, the attic temperature is
reduced by supply leaks when the air conditioner is on. Much of this
energy is lost in the conventional attic, but is recovered in the
Figure 1: Temperatures for Conventional and Cathedralized Base Case
Pulldown Operation Results
To examine the impact on
performance, Table 2 shows the pulldown times and the tons at the
register, ranked by increasing pulldown time. The pulldown time ranged
from about 100 minutes for the Best cathedralized attic case to almost six
hours for the poor conventional case.
Pull down times are fastest for the two improved houses with oversized (4
ton) air conditioners (Best cases) and the cathedralized Best case is an
hour faster than the conventional Best case. For the remaining three
comparisons, including the two houses with properly sized 3 ton air
conditioners (Best Resized cases), all of the cathedralized cases pulldown
at least an hour faster than their conventional counterparts. This gap
increases to two and a half hours for the Poor cases. Pull down times are
grouped quite tightly for the cathedralized cases, but are widely spaced
for the conventionally vented and insulated attics.
This is shown graphically in Figure 2, which has the measured house
temperatures for each of the four conventional cases and the base
cathedralized case starting from a few minutes before the pulldown
occurred and going until the house temperature reaches the set point,
The remaining cathedralized cases have similar pulldown times to the base
case and are not shown on Figure 2 for clarity. Note that the individual
cases all start at different temperatures because the parameters that
differentiate the cases also affect the house temperatures during the
first 16 hours of the day. All of the conventional cases show a spike in
the house temperature right at the beginning of the pulldown. The spike is
caused by hot air in the attic and ducts which overwhelms the air
conditioner capacity for the first minute or so of the cycle. Also, the
thermostats were set to cycle once the house reached the set point of 24�C
(75�F). To avoid cluttering Figure 2, only the temperature data until
pulldown is achieved is shown.
The tons at the register
(TAR) data correspond to a fifteen minute average value from
an hour after the pulldown begins to avoid initial transients. The TAR
data shows similar patterns to the pulldown times. The Best cathedralized
case delivers about 60 more power to the house through the registers than
the Base conventional case and the Poor conventional case delivers about 5
less power as the Base conventional case. The Poor conventional
case delivers the same amount of energy as the Best Resized conventional
case even though the Poor case has a much larger air conditioner. The
ranking of cases in descending TAR is different than the pulldown times
because the two smaller air conditioners (the Best Resized case for both
types of attics) deliver less energy to the house than the comparison Best
cases because the air conditioner has less capacity (3 tons vs. 4 tons).
Also, in addition to the effects of refrigerant charge and system airflow,
the air conditioner capacity is a function of the inside and outside
conditions. In particular, the indoor conditions vary considerably among
the cases, which in turn affects the TAR.
The cathedralized attics are considerably cooler than the conventional
attics (see Figure 1) which means that duct leaks and conduction losses
have a much smaller effect on diminishing the tons at the register values.
Even though the TAR values are typically higher for the cathedralized
cases, they are an underestimate of the useful cooling delivered, because
much of the energy that leaks from ducts in a cathedralized attic is
recovered to the main part of the house. This energy regain, although it
contributes to the cooling of the house, is not directly reflected in the
TAR values, because they represent only the energy actually delivered
through the registers.
The last column of Table 2 shows the ratio of tons at the register to the
capacity. The results from these simulations confirm earlier work (Siegel
1999; Walker et al. 1998) that nameplate capacity is not a good indication
of how much cooling an air conditioner will deliver and that duct leaks,
poor refrigerant charge, and reduced air flow will lead to considerably
less delivered cooling.
The pulldown times and tons at the register performance results discussed
so far are also similar to the results of earlier work (Siegel 1999;
Walker et al. 1998). To supplement that work, the addition of the
equipment model now allows for energy consumption and peak power to be
simulated. Energy consumption and peak power use for the entire system
(compressor, air handler blower and condenser fan) are shown in Table 3
for the pulldown cases, and Table 4 for the continuously cycling cases.
The energy consumption data for the pulldown cases is displayed
graphically in Figure 3.
Daily energy consumption
values in Table 3 span a smaller range than the TAR and
pulldown numbers in Table 2. Specifically, compared to the energy used by
the Base conventional case, the Best cathedralized case consumes 45 less
whereas the Poor conventional case consumes 31 more. The cathedralized
cases all use less energy than their conventional counterparts and the
range of energy consumption is also much smaller between the cathedralized
cases, as shown in Figure 3. The energy consumed by each of the
cathedralized cases varies less than between the conventional cases,
because energy lost from the ducts to the attic is mostly recovered to the
There is one seemingly paradoxical result in Table 3. The Best Resized
consume about 10 more energy than the Best cases even though the only
difference is that the Best Resized cases have smaller air conditioner
than the Best cases (3 tons vs. 4 tons). The 4 ton air conditioner has a
higher flow rate through the same duct system as the resized unit (see
Table 1). This leads to proportionally less conduction loss for the 4 ton
unit. This effect explains almost all of the difference between the Best
and the Best Resized results with very minor additional contributions from
increased fan efficiency and a slightly higher energy efficiency ratio (EER)
for the Best case. This result should not be interpreted to mean that a
bigger air conditioner is better. Oversizing an air conditioner can have
significant negative impacts on the ability of a system to control
The energy consumption data
for the conventional simulations show that leaky ducts
in a hot attic, reduced airflow and low refrigerant charge have a profound
impact on energy use. Correcting these problems from the Poor conventional
case to the Best conventional case leads to a 42 decrease in energy use.
From the perspective of a homeowner who operates her system in a pulldown
manner it is clear that the Best cathedralized case is the most desirable.
Compared to the Best Resized cathedralized case, it uses slightly less
energy and the pulldown is considerably shorter. The pulldown is faster
for the Best case because it has a larger air conditioner.
However there are two disincentives to having an oversized air
conditioner. The first is that the capital of an extra ton of air
conditioning is approximately $500, which is ultimately paid by the
homeowner. The second is that the additional peak power load is a cost to
the electric utility. The smaller air conditioner has a 20-25 reduced
load. Furthermore when operated in a continuous cycling manner (such as on
the weekends) in a hot humid climate, the significantly oversized air
conditioner may not provide sufficient dehumidification for occupant
Continuous Cycling Results
Table 4 shows the energy consumption for the continuously cycling cases.
energy consumption for these cases is larger than the energy consumption
for the pulldown operation cases, because the house is kept cooler for
more of the day than for the pulldown case, particularly when it is hot
outside. With continuous cycling, the houses with cathedralized attics all
consume less energy than their counterparts with conventional attics. They
also consume less energy relative to the base case than for the pulldown
operation cases discussed above.
Although the data in Table 4 indicate the energy consumption benefits of
cathedralized attics there is also an important comfort benefit. The air
conditioning systems in the cathedralized cases are better able to
maintain a constant and comfortable air temperature in the house because
they ultimately lose less cooling energy. The Poor conventional case went
above 25 �C (77�F) twice over the course of day because the load on the
house was greater than the cooling delivered to the house.
Our results indicate that cathedralizing (sealing the attic and insulating
the roof) is a practical way to improve air conditioner performance in the
hot dry climate typical of California and the southwestern United States.
Although it is not as efficient as actually having ducts fully in the
interior space, it overcomes some of the aesthetic and construction
concerns associated with interior ducts. This improvement is even apparent
for "poor" systems because most of the energy losses to a
cathedralized attic are ultimately recovered. The energy performance for
systems in cathedralized attics is a weaker function of cooling system
installation problems (excess duct leakage, insufficient insulation, low
coil air flow and refrigerant charge) that make installations in
conventional attics consume more energy. Compared to conventional attics,
cathedralized attics also require lower energy consumption for both
continuously cycling and pulldown operating conditions. When operated in a
continuous cycling mode (i.e. with a constant thermostat set point of 24�C
(75�)), a poor system in a cathedralized attic will use slightly less
energy than a fully retrofitted and properly operating system in a
For conventional attics, pulldown time, energy consumption and peak demand
improve dramatically when refrigerant charge, reduced air flow, and duct
leakage are corrected. A larger capacity air conditioner will, of course,
cool down a house more quickly than a properly sized air conditioner.
However, bigger air conditioners have higher peak energy use, as well as
increased first cost. Also, when using oversized air-conditioners, the
performance improvements are quite modest, especially when compared to
those that result from moving the duct system into a cathedralized attic.
Although energy consumption is moderately improved by sealing excessive
leakage in a vented attic, real energy benefits come from also correcting
refrigerant charge and improving air flow. These results hold for
continuous cycling operation as well as pulldown operation.
The ultimate decision about whether to achieve energy savings and
performance improvements in homes by cathedralizing the attic or improving
the air conditioner and ducts will likely come down to one of cost. It is
hoped that this study will aid decision making by presenting the
quantitative benefits of different improvement strategies.
The research reported here was supported by the California Institute for
Energy Efficiency (CIEE), a research unit of the University of California,
through the U.S. Department of Energy under Contract No.
DE-AC03-76SF00098. Publication of research results does not imply CIEE
endorsement of or agreement with these findings, nor that of any CIEE
ACCA. 1986. Manual J - Load Calculations for Residential Winter and Summer
Conditioning. Washington D.C.; Air Conditioning Contractors of America (H.
Rutkowski, Technical Director).
ACCA. 1992. Manual S - Residential Equipment Selection. Washington D.C.:
Conditioning Contractors of America (H. Rutkowski, Technical Director)
Blasnik, M., T. Downey, J. Proctor, and G. Peterson. 1996. Assessment of
Installations in New Homes in APS Service Territory. Proctor Engineering
Report for Arizona Public Service Company.
Brown, K., C. Blumstein, L. Lutzenhiser, B. Hackett, and Y.J. Huang. 1996.
"Does the Air-
Conditioning Engineering Rubric Work in Residences?" In Proceedings
of the 1996
ACEEE Summer Study on Energy Efficiency in Buildings, 8:11-20. Washington
American Council for an Energy-Efficient Economy.
California Department of Finance, Demographic Research Unit. 1997.
Population and Housing Estimates, January 1st 1997.
Carlson, J.D., J.E. Christian, and T.L. Smith. 1992. "In Situ Thermal
Modified Bitumen Roof Membranes Coated with reflective
ASHRAE/DOE/BTECC/CIBSE Thermal Performance of the Exterior Envelopes of
Buildings V, Clearwater Beach. Florida. December 1992,420-428.
Jump, D.A., I.S. Walker, and M.P. Modera. 1996. "Field Measurements
of Efficiency and
Duct Retrofit Effectiveness in Residential Forced Air Distribution
Proceedings of the 1996 ACEEE Summer Study on Energy Efficiency in
1:147-156. ACEEE, Washington D.C.: American Council for an
Kempton, W., D. Feuermann, and A.E. McGarity. 1992. "'I always turn
it on super': User
Decisions About When and How to Operate Room Air Conditioners."
Buildings, 18: 177-191.
Modera, M.P., and B. Wilcox. 1995. "Treatment of residential duct
leakage in Title 24
Energy Efficiency Standards", California Energy Commission Contract
Parker, D.S., J.R. Sherwin, and L. Gu. 1997a. Monitored Peak Attic Air
Florida Residences: Analysis in Support ofASHRAE Standard 152P.
97, Florida Solar Energy Center, Cocoa, Fl.
Parker, D., J. Sherwin, R. Raustadt and D. Shirey. 1997b. "Impact of
Evaporator Coil Air
Flow in Residential Air Conditioning Systems." ASHRAE Transactions.
405 ASHRAE, Atlanta, GA.
Proctor, J. 1999. Personal Communication about Air Conditioner Model.
Proctor, J. 1998. "Monitored In-Situ Performance of residential
ASHRAE Transactions,. 104(1B): 1833-1840. ASHRAE, Atlanta, GA.
Proctor, J. 1997. "Field Measurements of new residential air
conditioners in Phoenix,
Arizona." ASHRAE Transactions, 103(1): 406-415. ASHRAE, Atlanta, GA.
Proctor, J. and P. Albright. 1996. "Sizing air Conditioners: If
Bigger is not Better, What
Is?". Home Energy, 13(5): 13-18.
Proctor, J., Z. Katsnelson and B. Wilson. 1995. "Bigger is Not
Better: Sizing Air
Conditioners Properly." Home Energy, 12(3): 19-26.
Rudd, A F. and J. W. Lstiburek. 1998. "Vented and Sealed Attics in
Hot Climates." ASHRAE
Transactions, 104(2): 1199-1210. ASHRAE, Atlanta, GA.
Siegel, J. 1999. The REGCAP Simulation: Predicting Performance in New
Homes. Unpublished Masters Thesis. Department of Mechanical Engineering,
University of California, Berkeley.
Treidler, B. and M.P. Modera. 1994. "Peak Demand Impacts of
Residential Air Conditioning
Conservation Measures." Proceedings of the 1994 ACEEE Summer Study on
Efficiency in Buildings, 7: 249-266. Washington D.C.: American Council for
Vieira, R.K., D.S. Parker, J.F Klongerbo., J.K. Stone, J. Cummings. 1996.
Really Size Air Conditioning Systems. Proceedings of the 1996 ACEEE Summer
Study on Energy Efficiency in Buildings, 8:207-213. Washington D.C.;
Council for an Energy Efficient Economy.
Walker, I.S., J. Siegel, K. Brown, M.H. Sherman. 1998. "Saving Tons
at the Register." In
Proceedings of the 1998 ACEEE Summer Study on Energy Efficiency in
1:367-385. Washington D.C.: American Council for an Energy-Efficient
Wenzel, T.P., J.G. Koomey, G.J. Rosenquist, M. Sanchez, and J.W Hanford.
data Sourcebook for the U.S. Residential Sector", Lawrence Berkeley
Laboratory, Report # 40297