How a Branch Steals From the Branch Next to It

Why this matters

The complaint is always about one room. The cause is almost never in that room. Parallel branches, whether they are radiant loops off a manifold, duct runs off a trunk, or fixture branches off a riser, are coupled to each other through a shared pressure difference, and that coupling means a change anywhere changes everything. A tech who does not know this spends the call on the cold room, finds nothing wrong with it, and eventually replaces something in it. The tech who does know it walks past the cold room and goes looking for what got easier somewhere else.

Before you start opening and closing things

A manifold or header on a heating loop is above scald temperature, and the action is to isolate and let it cool before you loosen a flow meter cartridge or a balancing insert, not to work around it with a rag. If the job will not wait for cooling, stand out of the plane of the port, use a face shield rather than glasses, and crack the fitting a fraction of a turn at a time so any release is a weep rather than a jet.

If you are pulling a circulator or opening the loop rather than adjusting a setting, that is mechanical isolation and stored energy under 29 CFR 1910.147, which means lock and tag the disconnecting means rather than switching it off, and relieve the trapped pressure before the first joint comes apart. Work inside the appliance panel or at the disconnect itself is electrical, under 29 CFR 1910.333(b)(2), with the live-dead-live proving sequence taken from NFPA 70E-2021, 120.5.

The call

A three-loop radiant manifold, commissioned five years earlier with each loop metered at 2.0 gpm and documented on the panel door. The complaint: two rooms slow to come up on cold mornings, one room too warm, and a system that "was fine until the kitchen was redone."

The homeowner had already had two visits. The first replaced a thermostat in one of the cold rooms. The second bled the loops. Neither helped, and neither was unreasonable given what was in front of them.

The rule that explains all of it

Every loop off that manifold sees the same pressure difference, because they all start at the same supply header and end at the same return header. They do not get assigned a flow. They take whatever flow that shared pressure difference drives through their own resistance.

For turbulent flow, resistance behaves as a square law: head loss equals some constant times flow squared. Turn that around and flow through a branch varies as the square root of the available differential divided by that branch's own constant. Two things follow, and both are counter-intuitive until you have seen them once:

• Make one branch easier and it takes more flow, which drags the shared differential down, which starves every other branch. Nobody restricted the other branches. Their conditions did not change at all.
• The total can go up while every individual branch except one goes down. A pump on a constant speed sits where its own curve crosses the combined system, so a lower combined resistance means more total flow at less head.

What was eliminated, and on what evidence

• Air. The previous visit had bled the loops and the symptom persisted, and a genuinely air-bound loop shows an unstable, gulping flow meter rather than a steady low reading. The meters on the two cold loops were steady.
• A failing circulator. Total flow at the manifold had gone up rather than down, which is the opposite of a tired pump.
• A thermostat or zone valve. Both cold rooms were affected identically and neither shared a thermostat or a valve with the other, so a single control fault could not explain both.
• A blockage in the cold loops. A blockage raises that loop's resistance and its flow falls, which fits. But a blockage in one loop pushes flow into the others and raises the shared differential, so the third loop would have gained a little and the total would have fallen. The total had risen. That is the discriminator, and it is the reason the loop reading high was the suspect rather than the victim.

The kitchen remodel had shortened one loop substantially and rerun it in a larger tube. Nobody re-balanced afterwards.

The numbers, worked through

Treat each loop as a square-law resistance and treat the differential available at the manifold as falling with total flow, which is what a constant-speed circulator plus fixed distribution piping produces.

As commissioned: three loops of roughly equal length, each metering 2.0 gpm at a manifold differential of 4.0 ft. So each loop's constant is 4.0 divided by 2.0 squared, which is 1.0 ft per gpm squared. Total 6.0 gpm.

After the remodel: the shortened, larger-tube loop has roughly a quarter of its original resistance constant, 0.25. The other two are untouched at 1.0.

Solve the manifold for the new balance point and the differential settles at about 2.88 ft. At that differential:

Nobody touched loops A or C. They lost 15 percent of their flow because the differential they share fell from 4.0 ft to about 2.88 ft, and their flow follows the square root of that: the square root of 2.88 divided by 4.0 is 0.85.

Why the total looked fine, and what it cost

At the boiler, total flow was up 13 percent and the pump sounded no different. Anyone checking system-level numbers would have signed it off.

Read the loop temperatures instead. At a 140 F supply, loops A and C measured an 18.0 F drop before and a 19.5 F drop after. Using the sensible heat relationship for plain water near 60 F, roughly 500 times gpm times the temperature difference in degrees F, loop A carried about 18,000 Btu/h before, at 2.00 gpm and 18.0 F, and about 16,575 Btu/h after, at 1.70 gpm and 19.5 F. That is about 92 percent of the original, an 8 percent loss of output, delivered with a larger temperature drop, not a smaller one. A tech reading only the drop would conclude the loop was working harder.

Loop B, at 3.39 gpm, measured a 10.5 F drop, so about 17,800 Btu/h into a room that had gotten smaller. Hence the overheating complaint.

Now mix the returns, all three at the same 140 F supply so the comparison is like for like. Before: three loops at 2.00 gpm each returning at 122.0 F, so the mixed return is 122.0 F. After: two loops at 1.70 gpm returning at 120.5 F and one at 3.39 gpm returning at 129.5 F, which mixes to about 125.0 F.

That 3 F rise in return temperature is the part that costs money on a condensing appliance, because condensing depends on return water temperature being below the flue gas dew point, which for natural gas at typical excess air sits in the region of 130 F. The system went from about 8 F of margin below that region to about 5 F. It was still condensing, with less room, and on the coldest days when supply temperature is reset upward it would stop.

The failure mode of missing this: the shop raises the supply temperature to fix the two cold rooms. That does warm them. It also pushes the return past the condensing region on every cycle, overheats loop B further, and converts a distribution problem into a permanent efficiency loss that nobody will ever connect back to a kitchen remodel.

Three ways people make it worse

• Speeding the pump up. More differential does lift the starved branches, but it lifts the greedy one too, and by more in absolute terms because it starts from a bigger flow. You end up with the same imbalance at higher velocity, more noise and more pump power.
• Restricting the starved branch. It sounds absurd written down and it happens constantly, because a tech adjusts the valve on the branch that is complaining. Restricting a starved branch starves it further.
• Chasing the symptom room. Replacing emitters, thermostats or registers in the cold room cannot change a differential set somewhere else.

The correct move is to restrict the branch that is taking too much, which is the subject of the balancing card in this group.

How to confirm a theft rather than a blockage

Both present as one branch reading low, and they call for opposite actions, so separate them before you touch a valve.

1. Read every parallel branch, not just the one complained about. A theft shows one branch high and the others low together. A blockage shows one branch low and the others slightly high.
2. Read the total. Theft raises total flow because combined resistance fell. A blockage lowers it.
3. Read the shared differential at the header, on the same gauge and reference for both readings. Theft drops it. A blockage raises it.
4. Check what changed physically. A branch whose resistance fell got shorter, got larger, lost a fitting, or had a restrictor removed. Ask what work was done, not what broke.
5. Re-read all branches after any correction. Because they are coupled, correcting one moves every other one, so a single-branch verification after a single-branch adjustment is not a verification.

References

• 29 CFR 1910.147, control of hazardous energy, for lock and tag plus pressure relief before opening a circulator or a loop
• 29 CFR 1910.333(b)(2) for work inside an appliance panel or at the disconnect, with NFPA 70E-2021, 120.5 for live-dead-live proving
• Manufacturer documentation for the appliance's minimum return water temperature and reset schedule
• ASHRAE guidance on hydronic distribution and flow balancing in parallel circuits
• See related: Why Balancing Is Not Optional; Why Throttling a Valve Changes More Than Flow; Why Velocity Matters as Much as Volume