1. A SOLAR LIQUID-DESICCANT AIR CONDITIONER
Andrew Lowenstein
AIL Research, Inc.
P.O. Box 3662
Princeton, NJ 08543
e-mail: ail@ailr.com
ABSTRACT
A new technology for cooling and dehumidifying buildings
that uses liquid desiccants is now entering the U.S. market.
This technology, which can efficiently serve large latent
loads, will greatly improve indoor air quality by both
allowing more ventilation as well as more tightly controlling
humidity. Furthermore, since a liquid-desiccant air
conditioner can be packaged as a roof-top air-handler, it will
compete directly with the most popular cooling system now
being used in the U.S.—electric DX roof-top systems.
Early applications for the liquid-desiccant air conditioner
will be as a thermally activated cooling system for
processing ventilation air to buildings in humid climates.
For systems that are gas-fired, the total ownership costs can
be reduced by adding solar thermal collectors. In roof-top
installations the collectors can be located near the air
conditioner to simplify installation and reduce costs.
Energy storage can be done with concentrated desiccant
rather than hot water, further reducing installation costs.
At 2002 U.S. energy prices and government subsidies for
solar technologies that are now available in some U.S.
states, a solar-powered liquid-desiccant air conditioner that
uses single-glazed flat plate collectors could pay back the
installation costs for its solar components in as little as eight
years. In the longer term, advances both in solar collectors
and liquid-desiccant regenerators will lead to a hightemperature system that could payback in less than eight
years without government subsidies.
Key Words: solar cooling, solar air conditioner
1. INTRODUCTION
For the second time in three winters prices for fossil fuels
have been extremely volatile. In February 2003, the U.S.
market price for natural gas exceeded $9.70 per million Btu.
Although this price is less than the $10.53 per million Btu
reached in December 2000, it is three times higher than the
February 2002 price. And, as noted by the U.S. Energy
Information Administration, “[a] major consideration for
energy markets through 2025 will be the availability of
adequate natural gas supplies at competitive prices to meet
growth in demand1”.
High prices and availability are not the only energy issue
now facing the world. Climate change brought on by the
burning of fossil fuels is a reality. Only the magnitude of
the change remains debated.
A cooling system powered by solar thermal energy could be
an important part of a renewable solution to the preceding
problems. With the availability of solar insolation
coinciding roughly with cooling loads, solar energy appears
to be an ideal energy source for an air conditioner.
However, the technical challenges of developing an efficient
heat pump that runs on low-temperature thermal energy and
can deliver cooling when the energy source is not available
have been formidable.
Most past work on solar cooling has used absorption chillers
to convert solar thermal energy into cooling. Work on
solar/absorption systems has progressed steadily since the
1970s. The earliest work used single-effect absorption
chillers. More recent demonstrations have adapted doubleeffect chillers to run on either solar energy or fossil fuel.
Unfortunately, solar-powered absorption chillers will
always have problems competing in the broad HVAC
market. About 95% of the commercial buildings in the U.S.
use packaged air conditioners, the majority of these systems
being roof-top units. Roof-top systems are relatively small
(most under 50 tons) and they match naturally to the air
distribution system found in most buildings. In the U.S.,
gas-fired absorption chillers already have trouble competing
in this market. Replacing natural gas with a solar energy
source will not improve their competitiveness.
A new generation of liquid-desiccant air conditioners
(LDACs) is now coming to the market that is designed to

2. compete against roof-top systems. This technology, which
is thermally activated, offers new opportunities for
commercializing a solar cooling system.
2. A LOW-FLOW LIQUID-DESICCANT AIR
CONDITIONER
Perhaps the single biggest obstacle to moving liquiddesiccant technology from its secure base in industrial
applications into the HVAC market has been the high
maintenance requirements for the technology. The most
common configuration for a liquid-desiccant conditioner
(i.e., the component that dries and cools the process air) is a
bed of porous contact media that is flooded with desiccant.
The process air moves through this bed and is dried and
cooled as it comes in contact with the desiccant.
The process air that flows through a flooded packed-bed
conditioner will entrain small desiccant droplets. These
droplets are removed by a “mist eliminator”, i.e., a filter for
liquid droplets. In industrial applications, these “mist
eliminators” will be maintained to insure that desiccant
droplets do not carryover downstream. However, it is
unlikely that “industrial” maintenance procedures will be
widely accepted in HVAC applications.
In work funded by the Gas Research Institute, Lowenstein
and Gabruk2 studied a novel technology that could move
liquid-desiccant systems from their industrial base into the
broader field of comfort conditioning. The most dramatic
departure from the prior art is the use of very low flow rates
of liquid desiccant on the contact surfaces of the
conditioner: flow rates that are one-tenth to one-fiftieth the
levels in industrial systems.
Low desiccant flow rates can be effective only if the
desiccant is continually cooled as it absorbs water vapor
from the process air. Thus, a low-flow conditioner must be
internally cooled. Lowenstein and Gabruk solved this
problem conceptually, by configuring the conditioner as a
parallel-plate indirect evaporative cooler.
In work funded by the National Renewable Energy
Laboratory, AILR developed and tested a water-cooled
liquid-desiccant conditioner that uses low-flow technology3.
This conditioner is built as a water-cooled heat exchanger.
The plates of the heat exchanger are plastic extrusions that
are 48 inches (1.22 m) long, 12 inches (0.30 m) wide and
0.1 inches (2.5 mm) thick with 132 cooling passages
running their length. Process air flows through the 0.1inch
(2.5 mm) spaces between plates. Liquid desiccant is
delivered to the top edges of each plate. Wicks on the outer
surfaces of the plates create thin, uniform desiccant films.
The distribution and collection of desiccant is incorporated
into the design of the heat exchanger and occurs without
sprays or drip pans that would create droplets. A 6,000-cfm
(2.83 m3/s) conditioner has 200 plates.
In addition to a conditioner, a liquid-desiccant cooling
system will have a regenerator and an interchange heat
exchanger. This configuration is shown in Figure 1. The
regenerator uses heat to remove the water that the desiccant
absorbed in the conditioner. The interchange heat
exchanger improves the efficiency of the air conditioner by
using the hot concentrated desiccant that leaves the
regenerator to preheat the entering weak desiccant.
Conditioner
Regenerator
Interchange
Heat Exchanger
Fig.1: Generic Liquid-Desiccant Air Conditioner
The regenerator also uses low-flow liquid-desiccant
technology. It is again a liquid-to-air heat exchanger, but
one that has a hot fluid flowing within the plates. This
thermal energy desorbs water from the desiccant films that
flow down on the outer surfaces of the plates. A scavenging
air flow between the plates carries away the water vapor.
The preceding scavenging-air regenerator will have a
Coefficient of Performance (COP) that is close to 0.65. A
much more efficient regenerator with a 1.25 COP can be
made by adding an atmospheric-pressure desiccant boiler
“upstream” of the scavenging-air regenerator. The steam
that leaves the boiler with a 212 F (100 C) saturation
temperature provides the thermal energy for the scavengingair regenerator. This advanced system is commonly referred
to as a 1½-effect regenerator (similar to the double-effect
generators used in absorption chillers). The boiler, which
must operate at about 300 F (150 C), will be available in
about two years.
A 1,200-cfm (0.57 m3/s) LDAC that uses a scavenging-air
regenerator began operating in AILR’s laboratory in
November 2002. As of March 2003, the conditioner in this
system has operated effectively with no signs of carryover
of desiccant droplets. There have been material
compatibility problems in the higher temperature

3. regenerator. Changes in both the selection of materials and
in quality control procedures are being implemented that
will eliminate these problems.
Two 6,000-cfm (2.83 m3/s) LDACs will be tested in the
field during the 2003 summer. Both systems will use
scavenging-air regenerators and will process outdoor air.
Solar collectors will be added to one site in 2004.
4. THE COMMERCIALIZATION STRATEGY FOR A
SOLAR LDAC
A LDAC that uses “low flow” technology creates new
opportunities for solar cooling. Two characteristics of this
technology are most important:
•
•
water heater
cooling tower
conditioner
A LDAC can be configured as a small to mid-sized
roof-top air conditioner, and
A LDAC can serve very high latent loads*.
The cost of a solar cooling system can be significantly
reduced if the engineering time and contractor labor needed
to design and install a new system are reduced. Our
approach to achieving this is to target a very popular
packaged cooling system—roof-top air conditioners
between 10 and 30 tons (35 and 105 kW). By making the
“feel and function” of the liquid-desiccant system close to
that of a conventional system, more contractors will accept
the technology, and their learning curve to becoming
proficient installers will be shortened. Furthermore, by
keeping the solar collectors and cooling system near each
other on the roof, installation costs can be reduced.
A second important part of our commercialization strategy
is to first target applications that have high latent loads.
LDACs will have a strong advantage in these applications.
Unlike vapor-compression technology, which dries air by
cooling it below the dewpoint, a LDAC can dry air without
over-cooling. Large latent loads can be served without
reheating, and temperature and humidity can be
independently controlled within the building.
An effective latent cooling system also addresses the need
for better air quality within buildings. ASHRAE Standard
62 recommends much higher ventilation rates, specifically
to improve indoor air quality. However, in the more humid
eastern half of the U.S., the humidity that accompanies this
higher ventilation can push indoor relative humidity to
uncomfortable and unhealthy levels. A solar LDAC that
*
processes 100% ventilation air will purge indoor pollutants
while keeping indoor relative humidity low.
Our commercialization strategy for a solar cooling system
starts with the premise that gas-fired LDACs will soon
become the preferred technology for processing ventilation
air in humid climates. To demonstrate this we modeled the
performance of the 6,000-cfm (2.83 m3/s) roof-top LDAC
that is shown in Figure 2. In addition to the three desiccant
components shown in Figure 2, this packaged system has a
hot-water boiler as its heat source and a cooling tower for
rejecting heat to ambient.
The term “latent” is interchangeable with “humidity”. In
thermodynamics, the energy needed evaporate water to
create humidity is the Latent Heat of Evaporation.
interchange
heat exchanger
regenerator
Fig. 2: 6,000-cfm Liquid Desiccant Air Conditioner
for Building Make-up Air
The target application for this study is a 46,000 square foot
(4,275 m2) building in Miami, FL. The main air handlers
for the building circulate 40,000 cfm (18.9 m3/s). The
building is continuously ventilated from 6 AM through
7 PM, seven days per week at a rate of 12,000 cfm
(5.66 m3/s). (At 0.26 cfm per square foot [0.0013 m3/s per
square meter], the ventilation rate in this example would be
representative of a retail store.)
Two 6,000-cfm (2.83 m3/s) packaged LDACs process the
ventilation air to the building. This conditioned ventilation
air is supplied directly to the building. Approximately
12,000 cfm (5.66 m3/s) of return air is exhausted, and the
remaining 28,000 cfm (13.2 m3/s) is conditioned by 84 tons
of conventional DX roof-top units.
The energy use for the preceding system is compared to a
conventional design that uses 120 tons (422 kW) of DX

4. roof-top units to process a 70/30 mix of return air and
outdoor air. The DX capacity for this system has been sized
so that most of the latent load on the building is met. When
necessary, reheat is provided by an 80% efficient boiler to
prevent overcooling the building.
All non-ventilation loads for the building are calculated
using a TRNSYS† building simulation that was developed in
ASHRAE research project 1120-TRP4. Both latent and
sensible loads are calculated for each hour of the year.
The operation of the DX roof-top units is modeled using
manufacturer’s data for a 15-ton, 11.5 EER model that
operates at its minimum design air flow to maximize its
latent performance.
The LDAC is modeled using computer simulations that
were developed at AILR. The performance predicted by
these models compares favorably with laboratory data
collected both at AILR and at an independent laboratory.
The operating costs for the conventional and liquiddesiccant systems are calculated for gas and electricity
prices that are the 2002 averages for commercial customers
in Florida as reported by the U.S. Energy Information
Administration: $0.065 per kWh and $0.785 per therm
($7.44 per GJ).
Including parasitic power for fans and pumps, the
conventional DX system has slightly lower operating costs
than the system that preconditions the ventilation air with
the liquid-desiccant system and uses a simple scavenging-air
regenerator: $45,553 versus $47,223 for one year of
operation. However, when the scavenging-air regenerator
is replaced with a more efficient 1½-effect regenerator, the
liquid-desiccant system gains a 30% advantage: $32,816
versus $45,553. This operating cost advantage provides a
strong incentive for the advanced gas-fired LDAC to
displace the DX system in this application.
35% rebate for solar cooling and heating technologies as the
subsidy that the solar LDAC will receive in the near term.
In the longer term, we allow for improvements in the solar
and desiccant technology, but eliminate subsidies.
5.1 Near-Term Competition between a Gas-Fired and Solar
Cooling System in Miami
Having demonstrated that a gas-fired LDAC that cools and
dries a building’s make-up air can be an attractive
alternative to an electric DX system, we next show that the
same LDAC can further improve its competitiveness if a
significant fraction of its thermal input is solar energy.
A 46,000 square foot (4,275 m2) building in Miami, FL on
which the liquid-desiccant system cools and dries
12,000 cfm (5.66 m3/s) of ventilation air again serves as the
target application for comparing the two systems. This
application produces a large number of operating hours—a
building in Miami requires cooling for almost the entire year
and the building is occupied seven days per week—which
leads to faster paybacks for the solar components.
The near-term competition assumes that the LDAC uses the
less efficient scavenging-air regenerator operating with
190 F (88 C) hot-water provided by either a single-glazed
flat-plate collector or an 80% efficient boiler. There is no
significant storage of hot water. Energy is stored as
concentrated lithium chloride solution at 43%. The
regenerator runs whenever the collector can deliver 190 F
(88 C) hot water and there is available storage for desiccant.
The desiccant storage tank is stratified with weak desiccant
drawn from the top of the tank and concentrated desiccant
returned to the bottom.
The LDAC processes 12,000 cfm (5.66 m3/s) outdoor air
whenever the building is occupied and outdoor wet-bulb
temperatures are above 55 F (13 C). The regenerator is run
on natural gas whenever the concentrated desiccant in
storage is depleted.
5. THE PERFORMANCE OF A SOLAR LDAC
Our ultimate goal is a solar cooling system that can capture
a significant fraction of the cooling market without relying
on government or utility subsidies or unrealistically high
energy prices. However, we recognize that this goal will
not be reached in a single step. If the long-term benefits
offered by an emerging technology are sufficiently great,
state and federal agencies in the U.S. have been willing to
provide financial incentives to encourage its adoption. In
the following analysis, we use North Carolina’s existing
A simple payback is used to assess the competitiveness of
the solar cooling system. The key cost assumptions in this
assessment are:
•
•
•
•
†
TRNSYS is a computer program developed at the
University of Wisconsin that simulates the time-dependent
operation of coupled energy systems.
The installed cost of flat-plate collectors is $18 per
square foot ($194 per square meter)
Desiccant storage costs $2.50 per pound of
anhydrous lithium chloride ($5.50 per kg)
The owner receives a rebate equal to 35% of the
installed costs of the solar components
The price for natural gas is $7.85 per million Btu
($7.44 per GJ), which is the average price paid by
commercial customers in Florida for the first ten
months of 2002.

5. The flat plate collectors are oriented facing directly south at
an angle from the horizontal that equals the latitude of the
site. No attempt was made to either optimize the installation
angle or seasonally adjust the angle.
Desiccant is stored in an uninsulated plastic tank. The cost
of the tank is relatively minor compared to the cost for the
desiccant. Large plastic tanks cost on the order of $1 per
gallon ($0.27 per liter). A gallon of weak lithium chloride
solution (i.e., 37% weight concentration), weighs contains
3.75 pounds (1.7 kg) of anhydrous salt. Lithium chloride is
now selling for about $2.00 per pound ($4.40 per kg) in
large quantities. Thus, the cost for the salt will be about 7.5
times the cost for the tank. The $2.50 per pound ($5.50 per
kg) cost for storage includes the cost of the salt, the storage
tank and a small markup by the contractor.
Figure 3 presents the total ownership costs over eight years
for the solar LDAC as a function of the size of the collector
array. The capital component to the ownership costs only
includes the incremental cost for adding the solar subsystem
to a gas-fired LDAC. The $240,000 cost for a system with
zero array area is the eight-year fuel cost for the gas-fired
version of the air conditioner.
For array areas below 5,000 square feet (465 m2), the total
costs stay constant at about $240,000. As shown in
Figure 4, arrays up to this size are essentially fully utilized
(i.e., over 90% of the energy that they collect is used to
regenerate desiccant). The flat portion of the curve in
Figure 3 occurs because the value of energy collected over
eight years equals the array’s installed cost whenever an
array is fully utilized.
$400,000
$380,000
Cost for Planning Period ($)
The performance of the single-glazed flat-plate collectors is
simulated using TRNSYS. The TRNSYS modules for the
flat-plate collectors have been modified so that they
simulate equipment manufactured AET.
requirements of the regenerator for some hours. Although
desiccant can be regenerated, stored and used at a later hour,
$340,000
$320,000
$300,000
$280,000
$260,000
$240,000
$200,000
0
5000
7500 10000 12500 15000
Fig. 3: Eight-Year Costs for a Near-Term Solar Cooling
System that uses Flat Plate Collectors
the cost for storage will lead to higher costs for the eightyear planning period. If storage is not used, costs again
increase since the array in excess of 5,000 square feet
(465 m2) is not fully utilized.
In the near-term application that we are studying here,
storage will lead to lower total costs only if the cost of gas is
higher or the planning period is lengthened. If the planning
period is increased from 8 years to 12 years, the system with
the lowest total costs now uses 14,000 square feet
1.0
0.8
0.6
0.4
flat plate collectors
no storage
8-year planning period
0.2
For arrays larger than 5,000 square feet (465 m2), the
availability of solar energy will start to exceed the thermal
2500
Collector Area (s.f.)
2
With a 5,000 square-foot (465 m ) array, the solar collectors
are providing about 25% of the total thermal input to the
liquid-desiccant regenerator. At this array size, the hourly
thermal output of the array almost never exceeds the hourly
thermal requirements of the regenerator.
flat plate collectors
no storage
8-year planning period
$360,000
$220,000
Solar Utilization
The $18-per-square-foot ($194 per m2) cost for installed
flat-plate collectors assumes a relatively simple installation,
which is made possible by locating the collectors and
LDAC close to each other on the roof of the building. There
is no significant hot water storage and no piping runs within
the building. There are no “issues” regarding either the
integrity of the roof or structural support.
0.0
0
2500
5000
7500
10000
12500
15000
Collector Area (s.f.)
Fig. 4: Utilization of Solar Collectors for a Near-Term
Solar Cooling System that uses Flat Plate Collectors

6. (1,300 m2) of array. With this much larger array, desiccant
can be regenerated, stored and used at a later time. The
lowest total costs will occur when 10,000 lbs (4,545 kg) of
desiccant storage is used. The 12-year total costs with
storage will be $304,000 versus $359,000 for the gas-fired
alternative. The utilization of the arrays drops slightly to
88%, but solar energy now provides 66% of the total
thermal input for regenerating desiccant.
5.2 Long-Term Competition between a Gas-Fired and Solar
Cooling System in Miami
A new technology cannot indefinitely rely on subsidies to
promote its sale. If a solar air conditioner is to become a
meaningful alternative to electric and gas cooling systems, a
mature product must compete without subsidies.
Although not now available, mature LDACs will use the
more efficient 1½-effect regenerator that was described in
an earlier section. The COP for this regenerator will be
almost twice that of the simpler scavenging-air unit.
However, to gain this improvement, the regenerator must
have a 300 F to 320 F (150 C to 160 C) source of heat. This
high temperature will require either a tracking,
concentrating collector or an evacuated-tube collector.
For the study presented here, we’ve selected a tracking
parabolic-trough collector to supply heat to the regenerator.
Consistent with the assumption that the solar cooling system
is mature with fairly high annual sales, we’ve assumed that
the installed costs for the collectors are $17.50 per square
foot ($188 per m2). As with the flat-plate collectors, this
installed costs assumes a relatively simple installation.
The performance of the tracking parabolic-trough collectors
is again simulated using TRNSYS. The TRNSYS modules
for parabolic-trough collectors have been modified so that
they simulate the performance of the equipment
manufactured by Industrial Solar Technology.
The parabolic-trough collectors were oriented with their
axes aligned north/south (which will collect more solar
energy in the summer months than an east/west orientation).
The axes of the collectors were in a horizontal plane.
Given the extreme volatility present in the natural gas
market at the time this paper was prepared in February
2003, it is difficult to forecast a long-term price for natural
gas. For 2001, the average price paid by commercial
customers in Florida was $10.61 per million Btu. This price
is used in the long-term analysis.
For the long-term analysis there will also be a planning
period for which the total costs for the solar system are
almost constant as the array size is increased (as is the case
in Figure 3 for the flat-plate collectors). This planning
period is 7.5 years, for which all systems with less than
2,500 square feet (230 m2) of array (including the non-solar
system) have a total cost close to $156,000. With 2,500
square feet (230 m2) of parabolic-trough collectors, about
30% of the thermal input for the regenerator is supplied by
solar energy.
With 2,500 square feet (230 m2) of parabolic-trough
collectors, essentially all of the collected solar energy can be
immediately used to regenerate desiccant. Storage does not
improve the utilization of the arrays, a situation that is
identical to that which occurred in the near-term analysis
with 5,000 square feet (460 m2) of flat-plate collectors.
And, similar to the near-term analysis, storage will lead to a
lower total cost only if either the planning period is
lengthened or gas prices are increased.
For the long-term analysis, we again increase the planning
period to 12 years to study the benefits offered by storage.
Keeping gas at $10.61 per million Btus, the lowest total
costs for the planning period occur when the array is
increased to 6,000 square feet (560 m2) and 5,000 pounds
(2,270 kg) of lithium chloride are used for storage. This
system will have total costs of $217,000—a $32,000 savings
over the non-solar system.
6. CONCLUSIONS
A new thermally activated cooling technology is now
coming to the market that will significantly expand the
possibilities for solar cooling. This technology—low-flow
liquid-desiccant air conditioners—will first be used as a
packaged gas-fired roof-top unit that provides better control
of indoor humidity when latent loads are high.
Many factors affect the competition between a solar cooling
system and one that primarily uses either gas or electricity.
These include (1) hardware costs, (2) energy prices, (3) the
length of the cooling season as well as the hourly variation
of the cooling loads, and (4) the buyer’s criteria for
purchasing decisions (e.g., payback, return on investment,
etc.). The study presented here is not a thorough
investigation of how these factors impact the
competitiveness of a solar cooling. Instead, we’ve
presented a fairly simple analysis. i.e., only one application
was studied under one set of economic assumptions.
However, this one application does reflect conditions in a
very important market for solar cooling: humid climates
with long cooling seasons. In this market, assuming fairly
low energy prices (i.e., energy costs for 2001 and 2002 in
Florida), a solar liquid-desiccant cooling system will
payback in the near term with government subsidies in eight
years. In the longer term, without subsidies but with more

7. advance technology for the LDAC and the solar collectors,
the payback will be between seven and eight years. These
paybacks should lead to a growing solar cooling industry.
Obviously, a significant drop in energy prices will
discourage sales of a solar cooling system. However, for
the past five years energy prices have been at historically
low levels. A more likely scenario is that energy prices will
rise from these levels as the world emerges from a recession
that is still depressing economic activity at the start of 2003.
And, the need to limit greenhouse gas emissions will, in
effect, further increase the price of fossil fuels.
The solar LDAC is an economically viable technology for
cooling and dehumidifying some of the world’s buildings.
Whether it displaces 5% or 50% of the fossil energy use for
air conditioning will depend on both the maturation of the
technology, global environmental issues and future energy
prices.
7. ACKNOWLEDGEMENTS
The work presented here is based in part on research funded
by the U.S. Department of Energy under SBIR contract DEFG02-01ER83141. This support does not constitute an
endorsement by the U.S. DOE of the view expressed in this
paper. The author also acknowledges the financial support
of the National Renewable Energy Laboratory for the
development of the low-flow conditioner and regenerator,
and the Oak Ridge National Laboratory for the development
of the 6,000-cfm roof-top liquid-desiccant air conditioner.
Finally, the author thanks Mr. Ken May of Industrial Solar
Technologies for help in modeling parabolic trough
collectors.
8. REFERENCES
(1) Annual Energy Outlook 2003 With Projections to 2025,
DOE/EIA-0383(2003), January 2003
(2) Lowenstein and Gabruk, “The Effect of Absorber
Design on the Performance of a Liquid-Desiccant Air
Conditioner,” ASHRAE Transactions, pt. 1, vol. 98, AN-923-3, 1992
(3) Lowenstein, “Advance Liquid Desiccant Technology,”
DOE Peer Review, Nashville, TN, 2002 (downloadable as
PDF file from www.chpb.net/pdfs/020430IES/Lowenstein.pdf
(4) “Development of Equivalent Full Load Heating and
Cooling Hours for GCHPs Applied in Various Building
Types and Locations,” ASHRAE 1120-TRP, December
2000