Table 1. Test house selection process for the 30 houses of Phase 1.

Polk County building department staff were unable to assist us at this time because of staff availability.

Several problems arose in the search for test houses. While houses with air handlers in the garage were easy to find, air handlers in the house were more difficult to find, and air handlers in the attic were much more difficult to find. In Phase 1, for example, of 186 houses in Orange, Brevard, and Lake Counties combined, only eight were located in the attic. Of those eight, we were able to gain access to only one for testing. In order to find a sufficient sample of attic air handlers, we targeted Polk County and Marion counties where air handlers are more commonly installed in that location. In addition, we directly contacted builders and air conditioning contractors in Polk County. Of 23 attic units, 17 were tested in Polk County and 4 in Leon County.

Another problem arose with the data obtained from the visit to the Orange County building inspections department, namely that the homeowner name was not available on the 600A forms. We did not feel that sending letters to homeowners addressed to "homeowner" would yield a successful response rate. After investigating several options, it was discovered that homeowner names could be obtained for most properties through tax appraiser records by means of the internet.

It was also intended that a total of 12 gas furnaces would be tested, 4 for each air handler location. However, furnaces were readily found in garages but seldom in attics or indoors. In the end, we tested a total of 9 furnaces; 6 in the garage, 2 indoors, and 1 in the attic. Since we found that furnaces were, on average, considerably more leaky than electric air handlers (heat pumps and straight cool), we decided to test no more than 6 units in the garage so that the sample of garage air handlers would not be too heavily weighted toward gas furnace leakage.

3.0 Training Field Testers

Staff training was carried out during this project. Chuck Withers headed up the field testing effort, and also the training of other project staff members, including Jeff Sonne (Senior Research Engineer), Janet McIlvaine (Research Analyst), and Matt Lombardi (student assistant). This training consisted of "on the job" training by means of demonstration and tutoring of various test methods.

4.0 Develop a Low Air Flow Measurement Device

A low air flow measurement device was designed and constructed to allow measurement of air handler airtightness below 30 Q25, since 30 cfm is the low-end threshold measurement accuracy of the Energy Conservatory Duct Blaster. The low-flow device consists of two Dwyer Series RM flow meters (model number RMC-123 with valve) operating in parallel. Each of these rotometers (column with floating metal ball) has a range from 0 to 30 cfm. Therefore, the low-flow device can measure air flow rates from 0 to 60 cfm. These meters are connected to a vacuum cleaner. The vacuum cleaner currently being used provides sufficient flow to measure up to 31 cfm. Valves on the flow meters allow regulation of flow. An adjustable slide opening on the vacuum hose also allows adjustment of the vacuum cleaner air flow rate.

The low flow measurement device was calibrated (Figure 1). The air flow rate through the low flow device agreed to within 2% compared to a TSI Wind Tunnel.

Figure 1. Calibration of Dwyer flow meters using TSI Wind Tunnel model 8390.	Figure 2. Calibration of Duct Blaster at flow < 30 cfm using TSI Wind Tunnel.

The Duct Blaster was also calibrated for air flows below 30 cfm. The accuracy of the Energy Conservatory Duct Blaster was assessed for air flows between 10 and 38 cfm by operating the Duct Blaster in series with the TSI Wind Tunnel (Figure 2). Throughout this range of air flows, the Duct Blaster reads very nearly 2 cfm lower than TSI Wind Tunnel for all air flows. Therefore, we have been able to use the Duct Blaster, with a 2 cfm offset, for measurements down to 10 cfm. Calibration results for the low air flow measurement device and the Duct Blaster are shown in Figures 3 and 4.

Figure 3. The average offset is 1.9 cfm for the Duct Blaster	Figure 4. Y= 1.007 * X + 0.085; R^2= 0.9998

Because the Duct Blaster could be used down to 10 cfm, nearly all testing could be performed without the need for the low air flow measurement device.

5.0 Field Testing for Air Handler Leakage

Testing was performed on 69 air conditioning systems to determine the extent of air leakage from air handlers and adjacent connections. Testing and inspection was performed in the selected houses to obtain the following information:

• Q₂₅ in the air handler, Q₂₅ at the connection to the return plenum, and Q₂₅ at the connection to the supply plenum (Q₂₅ is the air flow rate through leak sites when exposed to a 25 pascal pressure differential).
• Operating pressure at four locations - at the return plenum connection, in the air handler before the coil, in the air handler after the coil, and at the supply plenum connection. An estimate was made of what portion of the AH is above and below the coil, based on visual inspection within the air handler cabinet.
• Return and supply air flows were measured by means of a flow hood. Air handler flow rates were measured by means of an air handler flow plate device (per ASHRAE Standard 152P methodology). The calibrated flow plate device (manufactured by the Energy Conservatory) is inserted into the filter access tray of the air handler (Figures 5 and 6). A manometer measures the pressure differential produced by the air velocity moving across the orifices of the flow device. This pressure differential is converted to an air flow rate for the specific flow plate used. Since the flow plate creates resistance to air flow and therefore modifies the normal air flow rate, a correction factor is used. This correction factor involves measuring the supply plenum pressure once with the plate in place and once with the air handler in its normal operating mode.
• Overall duct system and house airtightness was measured in 20 of the 69 homes.
• Cooling and heating system capacity was determined (ARI directory look-up) based on AH and outdoor unit model numbers.
• The location and type of filter was recorded. (15 filters were located in the air handler, one was located in the return duct, and 53 were at the return grill(s). Filters for 22 attic units were located at the grill; one was in the air handler.)
• The dimensions and the surface area of the air handler cabinet were measured and recorded. (Average air handler surface area was 30.7 square feet.) The average cooling system capacity was 38,000 Btu/hr. On average, there were 1.74 tons of cooling per 1000 square feet of floor area.
• The fraction of the air handler under negative pressure and under positive pressure was determined by visual observation within the air handler cabinet. (On average, 13.8% of the air handler cabinet was under positive pressure and 86.2% was under negative pressure. Nearly all of the positive pressure portions of air handlers were found in gas furnaces.)
• The types of sealants used at AH connections were recorded.
• An estimate was made of the fraction of the initial air handler leak area that are sealed "as found". (37 of the 69 air handlers had some measure of sealing in evidence. On average, 15.8% of the leakage in the 69 air handlers had been sealed, based on our visual observation. Eight air handlers had 50% or more sealed. In one case, 90% had been sealed.)

• A portion of the air distribution system (ADS) containing the air handler was isolated. Typically this involved cutting through the main supply plenum (Figure 7), placing an air barrier through the supply plenum (Figure 8), and then sealing this air barrier to the exterior surface of the supply plenum (Figure 9). In a number of homes, it was not possible to cut through the supply plenum either because there was insufficient length of plenum before branch ducts or because the supply plenum was a flex duct (Figure 10). In these homes, all of the supply registers were sealed, in effect testing the entire duct system.

  Figure 11. A calibrated blower and a manometer were attached to the isolated portion of the ADS.

• A calibrated blower and a manometer were attached to the isolated portion of the ADS (Figure 11).
• An airtightness test was performed on this portion of the ADS, obtaining Q₂₅.
• Leaks at the AH-to-supply plenum connection were then repaired.
• The Q₂₅ test was repeated.
• Leaks at the AH-to-return plenum connection were then repaired.
• The Q₂₅ test was repeated.
• Leaks in the AH cabinet were repaired.
• The Q₂₅ test was repeated.
• The leakage at each of the three indicated locations was calculated by subtraction from the preceding Q₂₅ value.

5.2 Air Handler Leakage Test Results

A total of 69 air handlers were tested in 64 houses. Nine were gas furnaces. One was electric strip heat. Two were hydronic heat. Fifty-seven were heat pumps.

5.2.1 Air Leakage in the Air Handler and Adjacent Connections

The airtightness results from all 69 air handlers are as follows: 20.4 Q₂₅ in the air handlers, 3.9 Q₂₅ at the return connection, and 1.6 Q₂₅ at the supply connection. These measured leakage amounts were "as found", that is, we measured the leakage of the system without making any changes to the system with one exception. If the filter access door was off or ajar, then it was placed in its proper position. The filter access door was found to be removed or ajar in two homes, both interior air handlers. In one case, a missing filter access door represented 189 Q₂₅. In the other case, an ajar filter access door represented 37 Q₂₅.

Pressure differentials were measured at four locations;

• at the return to air handler connection
• in the lower portion of the air handler (typically between the cooling coil and the filter access location)
• in the upper portion of the air handler (typically between the cooling coil and the blower)
• at the supply to air handler connection

The lower portion of the air handler was always under negative pressure. The upper portion of the handler was under negative pressure in 59 cases (all electric plus one hydronic) and under positive pressure in 10 cases (all furnaces plus one hydronic). In those 10 cases, on average, 56% of the air handler unit was under positive pressure.

Average measured pressure differentials for gas furnaces (with one hydronic unit) and electric air handlers (with one hydronic unit) are shown in Table 2.

	Gas Furnaces	Electric
pressure return connection (pa)	-92.1	-85.5
pressure AH (-) region (pa)	-95.1	-153.4
pressure AH (+) region (pa)	114.2	78.5
pressure supply connection (pa)	96.8	50.8

Based on the measured operational pressures and the Q25 for each location, estimated air leakage has been calculated for both the negative pressure and the positive pressure zones of the air handler plus connections for the 69 systems. The negative pressure zone had an average (as operated) leakage of 58.8 cfm. This 58.8 cfm of estimated return leakage in the air handler cabinet and adjacent connection represents 4.9% of the average 1207 cfm of air handler air flow. The positive pressure zone had an average leakage of 9.3 cfm, or 0.8% of air handler flow.

5.2.2 Air Leakage Variations by Air Handler Type and Location

It was intended that 12 furnaces would be tested, 4 from each air handler location. However, only nine furnaces were tested, because relatively few furnace units were found in interior or attic locations. In addition to the 9 furnaces, 2 hydronic gas heating system were tested (this air handler was also located in the garage). The hydronic system uses a hydronic heating coil with hot water supplied by a gas water heater. The 9 gas furnaces were found to be considerably more leaky than the non-furnace units. Since there is a difference among types of furnaces, averages for non-gas furnaces are shown in Table 3 and Table 4 shows gas furnace averages.

The total cabinet leakage of positive and negative pressure regions are 28.3 Q₂₅ versus 19.2 Q₂₅ (found in Tables 3 and 4 by adding Q₂₅ AH (-) region to Q25 AH (+) region). The hydronic units had air handler leakage of 35.1 Q₂₅, even more than the gas furnace air handlers. When converted to actual leakage, the 9 furnace units experienced 25.9 cfm of return leakage and 41.5 cfm of supply leakage in the cabinet. By comparison, the 60 non-furnace air handlers experienced 56.4 cfm of return leakage and 0.4 cfm supply leakage in the cabinet.

When gas furnaces are excluded, there were only minor variations in cabinet airtightness by air handler location. Q₂₅ for the non-gas furnace units is essentially the same for each air handler location; 19.7 Q₂₅ for attic, 18.7 Q₂₅ for garage, and 19.0 for indoors (found in Table 3 by adding Q25 AH (-) region to Q₂₅ AH (+) region). However, there is a noticeable difference in operational leakage because return side pressure is substantially lower for attic installations. While air handler pressure for garage and interior units are -181 pascals and -160 pascals, respectively, it is -126 pascals for attic units. As a result, return side cabinet leakage for attic units is 17.1 % less compared to garage units and 13.1 % less compared to interior units. However, since a larger fraction of the return ductwork of attic units is "outdoors" (see discussion in Section 6.0), the actual air flow from "outdoors" is likely to be greater for attic units than for other air handler locations.

	22 Attic	17 Garage	21 Indoors	60 Total
pressure return connection (pa)	-69.1	-114.4	-79.4	-85.5
pressure AH (-) region (pa)	-125.5	-181.2	-160.1	-153.4
pressure AH (+) region (pa)	121.0	36.0	na	78.5
pressure supply connection (pa)	52.8	51.2	48.3	50.8
Q25 return connection	2.1	2.4	3.8	2.8
Q25 AH (-) region	19.1	18.5	19.0	19.0
Q25 AH (+) region	0.6	0.2	0.0	0.2
Q25 supply connection	1.7	1.3	0.7	1.2
Q25 sum	21.7	21.1	22.8	21.9
Q return connection (cfm)	3.9	6.0	7.6	5.9
Q AH (-) region (cfm)	50.3	60.7	57.9	56.4
Q AH (+) region (cfm)	1.5	0.3	0.0	0.4
Q supply connection (cfm)	2.7	2.0	1.0	1.8
Q AH+connections (cfm)	58.4	68.7	66.5	62.6

	1 Attic	6 Garage	2 Indoors	9 Total
pressure return connection (pa)	-50.0	-98.3	-94.5	-92.1
pressure AH (-) region (pa)	-66.5	-100.6	-93.0	-95.1
pressure AH (+) region (pa)	75.0	120.9	113.8	114.2
pressure supply connection (pa)	74.0	100.7	96.5	96.8
Q25 return connection	1.2	15.8	3.8	11.4
Q25 AH (-) region	0.0	17.2	3.7	11.6
Q25 AH (+) region	0.0	21.8	6.8	16.7
Q25 supply connection	0.0	4.7	4.7	4.3
Q25 sum	1.2	59.5	18.9	44.0
Q return connection (cfm)	1.8	35.9	8.4	24.9
Q AH (-) region (cfm)	0.0	39.7	8.1	25.9
Q AH (+) region (cfm)	0.0	56.1	16.9	41.5
Q supply connection (cfm)	0.0	10.8	10.6	9.7
Q AH+connections (cfm)	1.8	142.5	44.0	102.0

5.3 Air Flow Rates

Air flow rates were measured for each system. Using a flow hood, air flow was measured at each supply and each return. Using a Flow Plate from the Energy Conservatory, the air handler flow rate was measured in 56 of 69 air handlers. In 46 of the 56 cases, the flow plate was installed in the filter tray, the intended location for the flow plate. In 10 of the 56 cases, the flow plate was installed at the single return grill because it would not fit into the air handler (most often because plate was too large or too small). In the 13 cases where the flow plate measurement was not performed, either there was no access to the filter tray or the flow plate would not fit the air handler dimensions.

Our best estimate of total air handler air flow is based on the flow hood at the return(s) and addition of estimated return leakage (based on Q25 and operating pressure, including the air handler and connections). Our best estimate of total air handler air flow was 1207 cfm per system. With nominal cooling capacity of 38,000 Btu/hr per system, this converts to 381 cfm per ton, or nearly on target with the nominal 400 cfm per ton normally indicated by most manufacturers as design flow.

Table 5.  Operating pressures, Q25, and Q (calculated operational leakage) for 69 tested systems.

	23 Attic	23 Garage	23 Indoors	69 Total
pressure return connection (pa)	-68.3	-110.2	-80.7	-86.4
pressure AH (-) region (pa)	-122.9	-160.2	-154.3	-145.5
pressure AH (+) region (pa)	98.0	108.8	113.8	107.7
pressure supply connection (pa)	53.7	64.1	52.5	56.8
Q25 return connection	2.0	5.9	3.8	3.9
Q25 AH (-) region	17.4	18.3	16.6	17.6
Q25 AH (+) region	1.4	5.7	1.7	2.8
Q25 supply connection	1.7	2.2	1.0	1.6
Q25 sum	22.5	32.1	23.1	25.9
Q return connection (cfm)	3.7	14.4	7.7	8.2
Q AH (-) region (cfm)	45.2	55.8	49.5	50.6
Q AH (+) region (cfm)	3.2	13.8	4.2	6.7
Q supply connection (cfm)	2.7	3.9	1.6	2.6
Q AH+connections (cfm)	54.8	87.9	63.0	68.1

6.0 Extended Testing

Figure 12

For 20 of the 69 systems additional (extended) testing was performed. This extended testing included measuring the overall duct system airtightness and house airtightness. The duct system airtightness testing followed the duct airtightness test method of ASHRAE Standard 152P (ASHRAE, 2001), obtaining both total leakage and leakage to out. First, the air handler was turned off and masking material was placed over supplies and returns. Second, the ductwork was split (sealed or blocked) at the air handler, either by placing masking over the blower intakes (preferred method) or inserting a barrier into the filter rack in the bottom of the air handler.

Two Duct Blasters were installed, one at a return register and one at a supply register (Figure 12.). With all other registers masked off, both sides of the system were taken to -25 pascals at the same time. ASHRAE 152P allows both sides of the system to be tested separately. When testing Q_25,total, there is benefit to running the test with both sides depressurized simultaneously. With both sides at essentially the same pressure, any leakage which might exist across the seal would be very small, and therefore would have essentially no impact on the test results. Furthermore, any leakage past the seal in the air handler would only "steal" from the leakage on one side of the system and "give" it to the other side of the system. By contrast, if each side is tested separately, then leakage past the seal is added to each side of the system. This test yielded Q_25s,total and Q_25r,total where "s" refers to supply and "r" refers to return.

On the other hand, when measuring Q_25,out and only one side of the system is tested at a time, then there will be relatively little pressure differential across the seal in the air handler, and leakage past the seal will have less impact. This test yielded Q_25s,out and Q_25r,out. In this test, the house is depressurized to the same -25 pascal pressure as the duct system. Duct system leaks to and from the conditioned space "disappear" during this test because the pressure differential between inside the ducts and the room go to zero. This yields duct leakage to outdoors. For the 20 houses in which extended testing was done, duct leakage test results are shown in Table 6.

house no.	Q25r,total	Q25s,total	Q25total	Q25r,out	Q25s,out	Q25out	ACH50
1	30	98	128	30	48	78	5.2
3	143	308	451	64	154	218	8.7
4	45	116	161	34	81	115	5.8
5	41	115	156	33	57	89	4.4
10	61	189	250	41	129	170	3.6
11	80	141	221	73	73	147	4.4
12	79	161	240	47	67	114	4.7
17	36	146	182	15	1	16	4.6
18	64	93	157	19	42	61	9.7
19	42	41	83	4	29	34	6.0
27	20	45	65	17	21	38	4.2
33	8	130	138	5	57	61	5.3
34	98	180	278	65	94	158	5.9
35	77	247	324	68	127	195	5.9
37	91	98	189	4	84	88	13.3
38	20	102	133	4	27	20	4.5
39	28	196	223	13	101	114	5.3
41	39	33	71	30	26	56	8.4
45	14	81	95	13	43	55	na
46	20	166	186	18	77	95	6.3
AVE.	53	134	187	30	67	97	6.1

6.1 "Total" Duct Leakage

Some important observations can be made from the extended test data in 20 houses. Total leakage on the return side of the system (including the air handler, return connection) was 53 Q_25r,total. With weighted operating pressure on the return side of about -100 pascals (including the air handler), operating return leakage is calculated to be 122 cfm, or 9.7% of rated system air flow.

Total leakage on the supply side of the system (Q_25s,total) was a very large 134. ASHRAE 152P method suggests using half of the supply plenum pressure as an estimate of overall supply ductwork operating pressure, if actual duct pressures are not known. For the 20 systems in which extended testing was performed, supply plenum pressure was 73.3 pascals. Based on a pressure of 37 pascals, actual leakage would be 167 cfm, or about 13.3% of rated air flow. In order to test the divide by two method, supply duct operating pressure measurements were taken for 14 systems at locations which were felt to be representative of the overall supply ductwork. These average 35.9 pascals, compared to 65.7 pascals for the supply plenums for those same 14 systems. For these systems, duct pressure was 55% of supply plenum pressure. Therefore, it seems that divide by two method is reasonable for central Florida homes on the supply side of the system.

The divide by two method is not reasonable for central Florida homes for the return side of the system. For these 20 systems, 38% of the Q_25r,total is in the air handler and 62% of the Q_25r,total is in the return ductwork. Given air handler pressure of -133 pascals, a return plenum pressure of -81.5 pascals, and return duct pressure of approximately -70 pascals, the weighted return side pressure is approximately -95 pascals. By contrast, the divide by two method would predict approximately -41 pascals. Clearly, in systems such as those commonly found in Florida that have only a single, short return duct (plenum), the actual operating pressure should be greater than the return plenum, perhaps 1.2 times the plenum pressure.

In addition to the extended testing results from these 20 units, return side leakage (Q_25r,total; including the return duct, return connection, and air handler) is available for 38 other systems. This additional return airtightness data results from the test method used to characterize air handler airtightness. The supply side of the system was isolated and excluded by cutting through and blocking the supply plenum. What remained was the air handler and the return ductwork. After leaks in the air handler cabinet and at the return and supply plenum connections were repaired, the remaining leakage was for the return ductwork.

Therefore, return side leakage is available for 58 of the 69 systems. Return leak air flow (Q_r,total) combined for the air handler, return connection, and the return ductwork was found to be 152.4 cfm, or 11.8% of total rated system air flow for this group of 58, based on Q_25r,total and operational pressures. For this larger sample, Q_r,total is considerably greater than for the 20 houses of extended testing. The results are rather alarming, that in newly constructed homes on the order of 12% of return air and 13% of supply air are leaking. But how much of this leakage is to "out" or unconditioned space?

6.2 Duct Leakage to "Out"

In 20 homes, duct leakage to "out" was measured (Table 6). On average, 56% of the leakage of the return ductwork (including air handler) and supply ductwork was to "out" ("out" defined as outside the conditioned space, including buffer spaces such as attic or garage). The fraction of the leakage that was to "out" varies by air handler location (Table 7). For return ductwork (including air handler), the proportion of total leakage that is to "out" is 81.4% for attic, 67.6% for garage, and 28.0% for indoors. For supply ductwork, the proportion of total leakage that is to "out" is in the range of 52% to 56% for all three locations.

Air handler location	Return	Supply	Entire duct system
Attic	81.4%	56.5%	63.2%
Garage	67.6%	51.7%	56.0%
Indoors	28.0%	52.6%	37.1%

The biggest variable which seems to predict leakage fraction to "out" is the fraction of the return ductwork that is in the attic, since nearly all of the supply ductwork is in the attic. All of the return ductwork of attic units was in the attic, whereas much of the return ductwork for other units was in the house. As a consequence, the energy penalty associated with locating the air handler in the attic is greater than indicated in the computer modeling results in Section 7.0 (compared to other air handler locations) because the modeling only considers the leakage of the air handler cabinet and the adjacent connections, and not the return ductwork itself. If the modeling were to take into account the fraction of the return duct leakage that comes from "out" (specifically the attic), then the energy penalties associated with the attic air handler location would be greater than indicated by the modeling.

Estimates can be made for actual operating duct leakage to "out", based on the fraction of the leakage to "out" from Table 7. These are presented in Table 8. For the supply duct leakage, the Q_25s,total and Q_25s,out are measured data from 20 houses. For the return duct leakage, Q_25r,total is measured data from 58 systems. Q_25r,out is calculated from the 58 system sample by using the "out"/"total" fraction obtained from the 20 system sample.

	ATTIC		GARAGE		INDOORS		COMBINED
	total	out	total	out	total	out	total	out
Q25,r [58]	61.9	50.4	93.3	63.1	67.8	19.0	75.7	44.9
Q25,s [20]	109.1	61.6	170.6	88.2	119.5	62.9	134.3	71.4
Qr [58]	118.1	96.1	194.4	131.4	134.6	37.7	152.4	90.4
Qs [20]	135.6	76.6	212.0	109.6	148.5	78.1	166.9	88.7

From Table 8 actual operating supply leakage to "out" is large for all three air handler locations, averaging 89 cfm. Actual operating return leakage to "out" is slightly larger, on average, at 90 cfm. However, there is a large difference depending upon air handler location; 96 cfm for attic systems, 131 cfm for garage systems, but only 38 cfm for indoor systems. From an energy point of view, the attic systems may experience the greatest "real" energy penalties, because all of the return ductwork and air handler are located in the attic. By contrast, a majority of the return leakage for the garage system likely comes from the garage (which is considerably cooler than the attic). For indoor systems, the return leakage to "out" probably originates primarily from the attic. However, since the return leakage is so much smaller, the energy impact is likely to be considerably less than that of both the attic and garage systems.

6.3 Correlation of Supply Duct Leaks with Number of Registers

When analyzing the supply leakage in the extended test data, a surprising correlation was observed. This correlation indicates that there is a systematic and consistent duct fabrication problem that exists across a wide range of air conditioning contractors.

Figure 13 illustrates this correlation. This plot shows that each supply duct has a remarkably predictable amount of total duct leakage associated with it. The coefficient of determination is 0.86, indicating that 86% of the variability in total supply duct leakage is explainable by the number of supply registers. Figure 14 shows a similar relationship between supply leakage to "out" and the number of supply registers. In this case the coefficient of determination is 0.69, indicating that 69% of the variability in total supply duct leakage is explainable by the number of supply registers. Note that one of the two houses with 13 registers shows considerably less leakage than expected. In this case, supply ducts were located in the interstitial space between floors. When the house was taken to -25 pascals, it is probable (though not measured) that the interstitial spaces were substantially depressurized as well, so leaks in those supply ducts would show less air flow (i.e., less pressure differential = less leakage air flow) and therefore be under-represented.

Figure 13. Supply CFM25total leakage vs. the number of supply registers.	Figure 14. Supply CFM25out leakage vs. the number of supply registers

 

Figure 15. Register to ceiling leakage.

Figure 16. Connection of flexible duct to metal collar.

These two figures suggest that a duct leakage problem is recurring in nearly all new homes. We have identified what we believe are three issues that are creating this leakage. The first is the connection of the supply register or return grill to the sheet rock as seen in Figure 15. This leak shows up as supply leakage in the "total" test but not much in the leakage to "out" test. This leakage occurs when the register or grill does not "snug" tightly to the ceiling or wall board. In many cases, the register frame bends when the screws are tightened, often because it is fabricated from light gauge metal. This bending of the frame causes gaps between the register and the dry wall. While this leakage has relatively little impact on infiltration, energy use, or space humidity, since the leakage is to and from the conditioned space, it will still show up in the "total". This can be the cause of not meeting the 10% of system flow requirement of the Florida Energy Code.

The second and third issues are the 1) boot (supply box) to sheet rock connection and 2) flex duct to collar connection. These typically show up as leakage to both "out" and "total". Figure 16 shows how the flexible duct connections are typically made. In some cases metal tape is used, but wrinkles when applied to the complex angles and over bumps associated with this type of connection. Although small in size, the many wrinkles in the tape at each connection allow air to pass through. Best practice would use mastic from one air barrier to the next. For instance, mastic applied from the foil of the duct board over the metal tabs of collar and then onto the collar where the air duct is fastened down with the plastic strap.

The high correlation between the number of registers (typically representing two flex-to-collar connections, one connection at the main trunk or a supply junction box and the other connection at the supply boot) and the amount of supply leakage is high indicates that this installation practice failure is nearly universal. For "total" leakage, there are about 12 Q₂₅ per register. For leakage to "out", there are about 6 Q₂₅ per register (or 3 Q₂₅ per flex duct connection).

6.4 House Airtightness

Figure 17

House airtightness was also measured using a blower door (Figure 17.). House airtightness test results are shown in Table 6. On average, house airtightness was found to be 6.1 ACH50, which is in line with test results on other samples of homes built in the past decade in Florida.