Backflow assembly test procedure field guide for testers

The annual backflow test, and what it confirms

A backflow assembly test is a field test that confirms a backflow preventer still does its one job: stop water from reversing out of a hazard and back into the potable supply. The assembly sits at a cross-connection, the point where the drinking water could meet something it must never meet, an irrigation line, a boiler, a chemical feed, a fire system full of stagnant water. The test proves the moving parts inside it still seal and still open when they are supposed to.

You prove that by reading pressure differential across the working parts with a calibrated gauge. On a reduced pressure assembly you read whether the relief valve opens at the right point and whether the two checks hold. On a double check you read both checks. On a pressure vacuum breaker you read the air inlet and the single check. The reading is the whole test. A backflow assembly looks fine from the outside while a fouled check or a stuck relief leaves the supply wide open.

The test is annual in most jurisdictions, and it is a legal requirement, not a courtesy. The water purveyor or the local cross-connection control program tracks every assembly on its system and sends a notice when the year is up. Miss it and the customer can lose water service. Confirm the exact cadence and the rules against the program for the address, because the authority having jurisdiction governs this end to end.

What does a backflow assembly test prove?

The test proves the assembly still holds pressure the way its standard requires, so the cross-connection is still protected on the day you test it. It is a functional check of the seals and the springs inside, read as pressure differential, and it gives a pass or fail against published criteria.

What it does not prove is worth naming so nobody oversells the report. It does not prove the assembly is the right type for the hazard, which is a survey and design question. It does not prove it was installed correctly, in the right orientation, with the right clearances and an air gap below any relief port. And it does not prove the assembly will still pass next month, because a main break the week after your test can drop a grain of scale on a seat and fail it overnight. The test is a snapshot of a single day.

The value of the snapshot is that it is the only routine check this device ever gets. A backflow preventer has no display, no alarm, and no way to tell you it has quietly stopped protecting the water. The annual test is how a fouled check or a weak relief spring gets caught before someone draws contaminated water from a tap. When the test fails, the work shifts from testing to repair, which is its own discipline covered in the failed-test repair guide.

The assemblies you test, and what the test checks on each

Four assemblies cover almost everything a tester sees, and the field test checks different things on each because they are built differently. The reduced pressure principle assembly, the RP or RPZ, is the high-hazard unit: two independent check valves with a relief valve in the zone between them that dumps to atmosphere the moment a check passes pressure backward. The test on an RP reads the relief opening point and both checks. The double check, the DC or DCVA, is two checks in series with no relief valve, used on lower hazards, and the test reads both checks. The pressure vacuum breaker, the PVB, has one check and a spring-loaded air inlet, protects against backsiphonage only, and the test reads the air inlet opening and the check. The spill-resistant vacuum breaker, the SVB or SPVB, is a PVB built so it does not spill across the gauge during the test, which is what lets it go indoors.

The standards behind the assemblies, which you verify against the edition your program adopts, map cleanly and are covered in depth in the failed-test repair guide. ASSE 1013 covers the RP, ASSE 1015 the double check, ASSE 1020 the PVB, and ASSE 1056 the spill-resistant version. The fire-line detector assemblies, the reduced pressure detector and double check detector, fall under ASSE 1047 and 1048 and carry a metered bypass that is tested too. Know the assembly in front of you before you hook a single test cock, because the procedure and the pass criteria change with it.

The differential pressure gauge and its calibration

The test gauge is a differential pressure gauge, and on the common kit it reads the pressure difference across one part at a time through three needle valves, a high side, a low side, and a bleed. Older and some current kits run five needle valves and a duplex layout, but the logic is the same: you connect the high and low hoses to two test cocks, open the needle valves to put the gauge across the part you are reading, and the needle shows the differential, not the line pressure. A reading near zero with the line at 70 psi is a part that is not holding. That is the whole skill of the instrument.

Air in the gauge or the hoses is the first thing that ruins a reading. A bubble in the line is a spring, it compresses and lets the needle drift, and it makes a good assembly look like it is leaking down. So you bleed the gauge and the hoses every time before you trust a number: open the bleed and the needle valves and let water run through until it runs solid with no spitting, the same discipline as bleeding air out of any hydraulic line. New testers skip the bleed, chase a phantom failure for twenty minutes, and learn the lesson once.

The gauge has to be calibrated, and the calibration is annual in common practice, with the gauge accurate to within about 0.2 psid. The calibration date and certificate ride with the test paperwork, because most programs reject a report from an out-of-cal gauge no matter what the needle showed. An uncalibrated gauge is the single most common reason a clean-looking report gets bounced. Treat the cal sticker the way an electrician treats the calibration on a torque wrench: if it is expired, the reading does not count. Confirm the calibration interval and the accuracy your program requires against the AHJ.

Test cocks, shutoffs, and isolating the assembly

Every testable assembly has shutoff valves at each end and a set of small test cocks tapped into the body, and the test runs off those test cocks. They are numbered in the direction of flow. Test cock #1 is on the inlet ahead of the first check, #2 is in the zone between the first check and the second part, #3 is downstream of the second check, and #4, on assemblies that have it, is past the second check or at the relief, depending on the assembly. Identify them on the assembly in front of you and against the procedure for that type, because the numbering and what each cock reaches is specific to the assembly.

Before you put a gauge on anything, tell the customer the water is going off. Testing means closing the downstream shutoff and isolating the assembly, so everything fed through it loses pressure and flow for the length of the test. On a building that is a few minutes of no water. On a process line, a clinic, a restaurant mid-service, or a data center cooling loop, an unannounced shutoff is a real problem, so you coordinate the window instead of surprising them. The two-minute phone call ahead of time is the difference between a routine test and an angry callback.

Then isolate and verify the isolation. Close the downstream shutoff, then the upstream shutoff, so the assembly is closed at both ends, and confirm the upstream valve is actually holding and not weeping by, because a shutoff that passes throws every reading off. Flush the test cocks briefly to clear grit before you connect, so a piece of debris in a cock does not read as a leak. Setup is most of the test. A sloppy setup produces a confident wrong answer.

How do you test an RP assembly?

You test an RP by reading three things in order: the relief valve opening point, the first check, and the second check. The relief valve must open at a differential of at least 2.0 psid, the first check must hold tight at a high differential, commonly at least 5.0 psid and at least 3.0 psid above the relief opening point, and the second check must hold drip-tight against backpressure. Those are the common minimums from the standard field procedure. Confirm the exact values against the procedure and edition your AHJ has adopted, because the numbers and the steps vary by program.

The sequence follows the assembly. With the gauge bled and the high and low hoses on the right test cocks, you isolate the second check and slowly bleed the zone, watching the differential fall until the relief valve just starts to discharge. The differential at that first discharge is the relief opening point, and it has to read 2.0 psid or higher. Then you read the first check, which is the static differential it holds across itself, and it has to hold tight at its minimum and well above the relief opening so the zone stays protected. Last you read the second check, confirming it holds against backpressure without leaking down.

The order matters because the readings build on each other. The relief opening tells you the zone will dump when it has to. The first check has to hold higher than that opening point, or the relief would chatter and weep in normal service. The second check is the backstop. An RP fails most often at the relief discharge, which is usually the first check passing pressure into the zone rather than the relief itself, and reading in order is how you tell which part is at fault. A failed reading is the start of the repair path in the failed-test repair guide, not the end of the job.

How do you test a double check assembly?

You test a double check by reading each of the two checks on its own, and each one has to hold drip-tight at a differential of at least 1.0 psid in the direction of flow. That is the common minimum from the standard procedure. A double check has no relief valve, so there is no opening point to read and no discharge to watch. The whole test is the two checks, read independently.

Reading them independently is the point that catches people. The protection on a DC is the redundancy of two checks in series, so a unit with one strong check and one that leaks has lost half its protection even though it might still hold overall on a rough look. You isolate and read the first check, confirm it holds its differential, then read the second check the same way. Each carries its own pass or fail. Reporting the assembly as a single pass without proving both checks held is how a half-failed DC stays in service.

A DC fails quietly, which is exactly why the annual test earns its keep on this assembly. With no relief valve there is no weep, no puddle, no warning. A check just stops holding, and the only thing that finds it is the gauge on the yearly test. By the time a DC reads failed, it has often been failing for a while with nobody the wiser, which is the strongest argument there is for not letting the annual test slide on a low-hazard line that looks like it is fine.

How do you test a PVB or SVB?

You test a pressure vacuum breaker by reading two things: the air inlet has to open at a differential of at least 1.0 psid, and the single check has to hold drip-tight at at least 1.0 psid in the direction of flow. Those are the common minimums. The SVB, the spill-resistant version, reads the same air inlet and check, and its whole design difference is that it does not spill water across the gauge during the test, which is what allows it to be installed indoors where a spilling PVB would flood the room.

To read the air inlet you isolate and slowly bleed the supply pressure down, watching the gauge as the pressure in the body falls, and the differential at which the air inlet just cracks open is the opening point. It has to be 1.0 psid or higher, and the inlet has to open fully as the body drains, not stick partway. Then you read the single check, confirming it holds its differential without leaking down. A PVB is the simplest of the testable assemblies, so the test is short, but the air inlet is exposed and moves every time the system depressurizes, so it is the part that fouls and sticks first.

Remember what a PVB cannot do, because it shapes whether the assembly was even the right choice. A PVB protects against backsiphonage only, never backpressure, and it has to be installed at least 12 in above the highest downstream outlet, a height to confirm against the local code. If the test fails because the air inlet will not open or the check leaks, the move is the same as any failure: diagnose, flush, repair with the matching kit, and retest. The irrigation PVB that comes up failed every spring is almost always a freeze casualty, covered later.

The standardized field test procedure

Backflow testing is not improvised. There is a written, standardized field test procedure, and the tester follows it step for step, the same way every time, because consistency is what makes one tester's report mean the same thing as another's. The procedure most programs adopt is the one published by the USC Foundation for Cross-Connection Control and Hydraulic Research, in its Manual of Cross-Connection Control, with separate step procedures for the RP, the DC, the PVB, and the SVB.

What the procedure pins down is the order of operations and the pass criteria, so the result does not depend on which tester showed up or which way they decided to bleed the assembly. Open this cock, connect this hose here, bleed in this order, read this differential, record it. The common pass values, relief opening at 2.0 psid or greater, checks holding at their minimums, air inlet opening at 1.0 psid or greater, all come from that framework. The numbers in this guide are the common ones, not a universal law, because the exact criteria and the accepted gauge format vary.

The variation by jurisdiction is real and it is the thing to get right. Some programs adopt the USC procedure outright, some adopt a state or regional procedure that differs in the details, and some accept an approved digital gauge with its own guided sequence. The procedure the AHJ requires is the one you follow on a given assembly, and a report run by a procedure the program does not accept can be rejected even when the assembly is fine. Confirm which procedure and which gauge the program accepts before you test, and never cite a step or section number you have not verified against the adopted edition.

Who is allowed to test: the certified tester

Not everyone can test a backflow assembly and submit the report. The work is done by a certified backflow assembly tester, often called a BAT, who has passed a written and a hands-on practical exam and holds a current certification recognized by the water purveyor or the AHJ. A plumbing license alone usually is not enough. The program wants the specific tester certification, and it tracks the certification number against every report it accepts.

The certification is not permanent. It carries a renewal cycle, commonly every two to three years, often with continuing education or a re-test on a calibrated station, and the rules differ by certifying body and jurisdiction. Let the certification lapse and the reports stop being accepted, even if the testing is sound, because the program verifies the tester's standing when the report comes in. Confirm your renewal cycle and the requirements against your certifying body and the AHJ, since this is local.

The certification number and a current gauge calibration date are the two credentials a program checks on every form, and a report missing either one gets bounced. The qualification standards for testers fall under the ASSE Series 5000 professional qualification standards in many programs, but which certification a given purveyor recognizes is the purveyor's call. Find out which certification the program for that address accepts before you bid the work, because a cert that is valid in the next county over may not be on this program's accepted list.

What does a failed reading tell you?

A failed reading points at a specific part before you ever open the assembly, and learning to read it is what separates a tester from someone who just writes down numbers. A relief valve that weeps or discharges on an RP is almost always the first check passing pressure backward into the zone, not the relief itself. A check that will not hold its differential is debris on the seat or worn rubber on the disc. An air inlet that will not open on a PVB is a fouled, corroded, or stuck poppet or a broken spring.

The reading tells you which part, but it does not always tell you why, and debris imitates almost every failure. A single grain of scale on a seat holds a check open just enough to fail the differential, and it looks exactly like worn rubber on the gauge. That is why a failed test is the start of a diagnosis, not a verdict on a part. The move after a failure is to isolate, flush the line and the assembly to clear debris, and then retest before condemning anything, because a real share of failures pass on a clean flush alone.

When the failure survives a flush, the part is genuinely worn and the work shifts to repair: find the failed component the gauge accused, rebuild it with the kit matched to the make, model, and size, and prove it with a passing retest. That full diagnosis-to-repair path, the failure modes by assembly, the kit selection, and the safe isolation and rebuild, is the subject of the failed-test repair guide. As a tester your job at the gauge is to read the failure correctly and record it, so the repair starts from the right part.

The retest after repair

A backflow assembly that has been repaired has to pass a retest before it goes back in service, with no exception, because the repair is not proven until the gauge says so. A rebuild can fail its own retest from a diaphragm in backward, a pinched o-ring, debris that landed on the fresh seat, or simply the wrong kit. The retest catches your own work before the customer does.

Run the full field test again, every reading, the same procedure as the original, and record every value, not just the word pass. The retest readings are what the program accepts as proof the assembly works.

Whether the same person can perform the repair and the retest, and whether it can happen the same day, varies by jurisdiction. Some programs allow it, some restrict who signs, some want a separate report. Confirm the rules with the AHJ before you assume the closeout is done, and see the failed-test repair guide for the full retest and submittal sequence.

The test report and the submittal

The test ends in a report submitted to a specific authority, usually the water purveyor or a cross-connection control program, and an incomplete report is a test that does not count. The form ties the readings to one physical assembly and to a qualified tester. It is the only record that the cross-connection was proven this year, and it is what defends the customer and you if the assembly is ever questioned.

What the form carries is consistent even when the layout differs by program. The assembly identity, the make, model, size, and serial number, so the report is provably about that unit. The location and the service it protects. The test cock readings and the result for each part, the relief opening differential and the check holds on an RP, the check holds on a DC, the air inlet opening and check on a PVB. The pass or fail. The tester's name and certification number. The gauge's calibration date. Many programs reject a report missing the certification number or the gauge calibration, so those two are not optional.

The submittal runs on a deadline, and the deadline is local. A completed test typically has to be submitted within a window the program sets, often counted in days from the test date, and a failed assembly carries its own clock to repair and resubmit a passing report. Miss the window and the customer can face a notice, a fine, or a shutoff even though the assembly is fine, because the program tracks the dates, not just the outcome. Confirm the form, the deadline, who may sign, and where it goes with the AHJ or purveyor for the address, because two towns on the same water system can run different programs with different clocks.

How often must a backflow assembly be tested?

A testable backflow assembly must be tested at installation and then at least annually in most jurisdictions. The annual test is the common requirement, set by the water purveyor or the local cross-connection control program, and on some high-hazard or critical services a program can require more frequent testing. Confirm the cadence for the assembly against the program for the address, because the AHJ sets it and a few programs run on a different interval.

The customer rarely tracks the date. The water purveyor does, and it sends a test notice when the assembly's year is up, giving a deadline to get a passing test submitted. That notice is the trigger most testing runs on. The program keeps a database of every assembly on its system, the hazard it protects, and the date it was last proven, and the notice goes out on that clock.

The consequence of skipping it has teeth, which is why this is a legal requirement and not a recommendation. A program that does not receive a passing test by the deadline can issue a second notice, a fine, and ultimately shut off water service to the property until a passing test is on file. The cross-connection is a public-health matter to the purveyor, so the enforcement is real. The customer who treats the annual notice as junk mail finds out the day the water goes off.

New install versus annual test

A backflow assembly gets tested twice in its first year of life: once at installation, before it is accepted into service, and then on the annual cycle after that. The installation test proves the assembly was installed right and works from day one, before the program will sign off on the new cross-connection. The annual test after that proves it has not drifted out of compliance since.

The two tests read the same way, but the installation test catches a different class of problem. A new assembly can fail its first test from debris flushed in during the install, a check installed backward, an assembly plumbed in the wrong direction of flow, or shipping damage. Flushing the line before the first test matters even more here, because fresh construction puts pipe dope, solder, and cuttings into the line, and that debris lands on the new seats.

From there it is the annual cadence and the purveyor's notice. The assembly that passed clean at install can fail a year later from a fouled seat, a worn spring, or a freeze, which is exactly why the test repeats every year instead of being a one-time acceptance. A device with no alarm and no display only gets checked when someone puts a gauge on it.

Common test errors that fail a good assembly

More clean assemblies get reported as failures from tester error than most new testers believe, and the errors are a short, repeatable list. Air not bled from the gauge and hoses is the big one. A bubble compresses, the needle drifts, and a perfect check reads like it is leaking down. Bleed until the kit runs solid water before you trust a single number.

The rest cluster around setup and the gauge. Test cocks not fully open, or fouled with grit, throttle the reading and make a part look weak. A gauge out of calibration reads off across the board, so a real failure passes or a good assembly fails, and either way the report is wrong and the program may bounce it. Running the wrong procedure for the assembly, reading an RP sequence on a DC or skipping the relief opening, produces a confident wrong answer. And reading the wrong direction, getting the high and low hoses or the test cock order reversed, inverts the differential and reads garbage.

The pattern under all of them is the same: the assembly is fine and the test is wrong. When a reading surprises you, suspect the setup before you condemn the assembly. Re-bleed the gauge, confirm the test cocks are full open and clear, check the calibration date, and confirm you are running the right procedure for the assembly in front of you. A second clean run settles most surprises, and it is far cheaper than sending a customer to repair a part that never failed.

Cold weather, freeze protection, and the winter test

Backflow assemblies fail on a seasonal clock, and freeze is the leading reason, so cold weather changes both how an assembly survives and how you test it. Water left in an above-grade assembly through a hard freeze expands as it turns to ice and cracks the body, splits the cover, distorts seats, and tears rubber. Irrigation PVBs and outdoor RPs are the classic casualties, and they come up failed or visibly cracked in spring because nobody drained them for winter. A cracked body is not a repair, it is a replacement, and you note it on the report.

The winterizing is the prevention. Above-grade assemblies on seasonal lines get drained down and the test cocks left open, or get an insulated enclosure with heat, before the first freeze. The irrigation contractor who blows out the system in fall and leaves the PVB cocks cracked hands the spring tester a working assembly instead of a cracked one. Confirm the winterizing approach against the assembly type and the local climate, because what works in a mild winter is not enough where the line freezes solid.

Testing in winter has its own hazards. You do not want to leave an assembly full and isolated in freezing conditions any longer than the test takes, and water spilled from a relief port or a PVB during the test freezes on the ground and on you. Schedule outdoor tests for the warm part of a winter day where you can, drain promptly after, and on a spill-prone PVB outdoors plan for where the water goes so you are not standing on an ice rink by the end of the test.

What to document

A test you cannot reconstruct from the record is a test the program cannot accept and you cannot defend. The record is what proves the assembly was proven this year, and it is what answers the question a year out when the assembly is doubted and someone asks whether it was ever really tested.

Capture the assembly identity, every reading, and the credentials, tied together on one form. Record the assembly identity by serial, the type, each test cock reading, the relief valve opening differential on an RP and the check holds, the air inlet opening on a PVB, the pass or fail, your tester certification number, and the gauge calibration date. If it failed, record the failure mode and tie it to the repair and the passing retest. The form is only as good as the worst-recorded field on it.

Common mistakes

• Reading the gauge without bleeding the air out of the kit and hoses first, so a good assembly reads as a slow leak.
• Running the test on an out-of-calibration gauge, which voids the reading no matter what the needle showed.
• Running the wrong procedure for the assembly, such as reading a DC like an RP or skipping the relief opening point.
• Not confirming the upstream shutoff actually holds, so a valve that passes by throws off every reading.
• Reporting an RP or DC as a pass without proving both checks held on their own.
• Condemning a part on the first failed reading without flushing and retesting to rule out debris.
• Returning a repaired assembly to service without a passing retest with all readings recorded.
• Submitting a report missing the certification number or the gauge calibration date, so the program rejects it.
• Missing the purveyor's submittal deadline, so the customer is non-compliant even with a passing test in hand.

Field checklist

Standards and references

The field test procedure most programs adopt is the one from the USC Foundation for Cross-Connection Control and Hydraulic Research, published in its Manual of Cross-Connection Control, with separate step procedures for the RP, DC, PVB, and SVB. The common pass values, the relief opening at 2.0 psid or greater, check #1 holding tight and above the relief opening, each check on a DC at 1.0 psid or greater, and the air inlet on a PVB at 1.0 psid or greater, come from that framework. Confirm the exact criteria and the procedure edition against what your AHJ has adopted, and never cite a step number you have not verified.

The assembly standards come from ASSE International and map to the assemblies you test: ASSE 1013 for the RP, ASSE 1015 for the double check, ASSE 1020 for the PVB, ASSE 1056 for the spill-resistant vacuum breaker, and ASSE 1047 and 1048 for the fire-line detector assemblies. The tester qualification falls under the ASSE Series 5000 standards in many programs. AWWA's M14 manual, Backflow Prevention and Cross-Connection Control, is the water-industry reference for program practice. Confirm each standard and its edition against your jurisdiction's approval list, because the standards are revised on a cycle.

The plumbing code that adopts all of this is local, the IPC or the UPC as amended, and the water purveyor's cross-connection control program sets the testing cadence, the accepted procedure, the report form, and the submittal deadline. The standard that controls any given call is the one the AHJ has actually adopted and enforces. Verify the cadence, the criteria, the gauge calibration requirement, and the deadline against that program before you cite a number on a report.

Units and terms

Backflow testing carries its own shorthand, and the same part and the same reading go by a few names across forms and manufacturers.

The reading is pressure differential in psid, pounds per square inch differential, the difference the gauge sees across a part, not the line pressure behind it. The assemblies are abbreviated by type. The internal parts and the credential of the person testing have their own terms. The framework procedure is usually called by its source, the USC FCCCHR field test procedure, and the person allowed to run it is the certified backflow assembly tester.

RP / RPZ

Reduced pressure principle assembly, the high-hazard unit with two checks and a relief valve (ASSE 1013)

DC / DCVA

Double check valve assembly, two checks in series for low-hazard protection (ASSE 1015)

PVB / SVB

Pressure vacuum breaker and its spill-resistant version, a check plus an air inlet for backsiphonage only (ASSE 1020 / 1056)

psid

Pounds per square inch differential, the pressure difference the gauge reads across a part

Differential gauge

The test instrument that reads pressure across one part at a time through high, low, and bleed needle valves

Test cock

A small numbered valve tapped into the assembly body that the gauge hoses connect to

BAT

Backflow assembly tester, the person certified to test and submit, often on a two to three year renewal

USC FCCCHR

The USC Foundation for Cross-Connection Control and Hydraulic Research, source of the common field test procedure

Relief valve opening point

The differential at which an RP's relief valve starts to discharge the zone, commonly 2.0 psid or greater

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• Why did my RPZ fail the test?
• How do you water-test a DWV system?
• What is the difference between an RPZ and a double check?
• What is graywater?
• What is a cross-connection?