How to Avoid a Dissimilar Metal Joint You Will Regret

Why this matters

You rarely get to pick both halves of a joint. One half is already in the building, chosen decades ago by someone who is not available, and your job is to attach something to it that will still be sound when the warranty runs out. The transition is where the callbacks live: the rotted nipple above a finished ceiling, the seized fastener that shears when the next tech touches it, the flashing that lifts. This is the procedure for making that transition once.

Before you open anything: kill the energy at the joint

A transition joint is almost always on a system that stores energy, so the sequence comes before the diagnosis.

• On a pressurized water or hydronic line: close the isolating valve, open a downstream tap or vent to atmosphere, and confirm zero pressure at a gauge or a fully open drain rather than by feel. Hot water systems hold enough heat to scald after the pressure drops, so let the line cool before you drain it. Where somebody else could reopen that valve while you are in the joint, lock or tag it and verify the zero-energy state before loosening the first fastener, per 29 CFR 1910.147.
• On anything with a gas connection: close the appliance gas shutoff and the branch shutoff before touching a union. If you smell gas at any point, everybody leaves the building immediately, nothing is switched on or off, no lights, no phone used inside, and the call to the gas utility is made from outside.
• On anything with an electrical connection or a bonding conductor you might disturb: de-energize at the disconnect, lock or tag it, and prove the conductors dead with the live-dead-live sequence in NFPA 70E-2021, 120.5, before you treat any conductor as safe. The de-energizing and lockout duty for electrical work is 29 CFR 1910.333(b)(2), not the general lockout standard.
• On anything held up by the joint: support the load independently before you loosen it. A corroded fastener does not yield gradually, it lets go.

Step 1: Identify the metal that is already there

You cannot pick a partner for a metal you have not named. Guessing costs you the whole exercise, because the two most common mistakes are calling galvanized steel "steel" and calling any bright fastener "stainless."

• Magnet. Sticks to carbon steel, cast iron, galvanized steel, and 400-series stainless. Does not stick to copper, brass, aluminum, or properly annealed 300-series stainless. A weak pull on a 300-series part usually means it was cold-worked, not that it is carbon steel.
• File a hidden edge. Galvanized steel shows bright grey zinc, then dull steel underneath. Brass shows yellow through. A plated steel fastener shows steel one or two strokes in, and once you know it is plated you know the plating is the only thing standing between the steel and the joint.
• Weight and sound. Aluminum is obviously light for its size. Cast iron rings dull, steel rings bright.
• Markings. Fastener head markings, pipe stampings, and fitting castings carry alloy information more often than people check.

Skip this and: you specify a "compatible" fitting against an assumed base metal, and the assumption is what fails, not the fitting.

Step 2: Choose the second metal from the same neighborhood, if you get the choice

When both halves are yours, the cheapest corrosion control in the trade is picking two metals that sit close together electrochemically. Copper with copper alloys. Steel with steel. Aluminum with aluminum or with a coating system rated for aluminum contact.

When the choice is constrained by pressure rating, code, or availability, close the gap rather than ignoring it. A silicon bronze fastener into copper is a much smaller couple than a stainless fastener into copper, and a much smaller one than plain steel.

Skip this and: you spend the rest of the procedure managing a problem you could have deleted at the counter.

Step 3: Decide which half is allowed to be sacrificial, and make it the big one

You will not always avoid a couple. What you can always control is which half gives up metal and whether it is small enough to matter. The active (less noble) metal corrodes, and corrosion current concentrates when the active metal is small. So:

• Fasteners, rivets, screws, and clips are the noble half. Always. They are small, they carry load, and they are the last thing you want thinning.
• Panels, brackets, and pipe bodies are where a sacrificial role belongs. They are large, the loss spreads out, and they are visible when it starts.
• Where a system genuinely needs a sacrificial part, make it a designed one on a replacement schedule rather than a structural member. A water heater anode rod is exactly this: a deliberately active metal, sized to be consumed, listed as a maintenance item.

Skip this and: you build the joint from case two of every galvanic article ever written, a small active fastener holding a large noble panel, and it fails at the fasteners while the panel still looks new.

Step 4: Break the electrical path, then prove you broke it

Isolating hardware only works if the circuit is actually open. Install the sleeve, isolating washers, nonconductive gasket, or dielectric fitting, then put a meter on continuity across the joint and confirm it reads open.

The reason this step exists separately from the install step is that in a building the two halves are frequently tied together somewhere else. Metal piping systems are commonly bonded for electrical safety under NEC Article 250, so a dielectric union in the middle of a bonded run breaks nothing. Where a bond is required, it stays. You do not solve a corrosion problem by removing a bonding conductor; you solve it by removing the electrolyte or by changing the metals.

Skip this and: you install an isolating fitting, bill for it, and the couple keeps running through a jumper eight feet away.

Step 5: Coat the noble half, or coat both, never only the active half

This is the step that gets done backwards most often, and backwards is worse than not doing it.

If you paint or coat only the active metal, every pinhole, scratch, and holiday in that coating becomes a tiny exposed anode facing the entire uncoated noble surface. You have not reduced the couple, you have concentrated it into a pit. The same joint that would have shed a slow, spread-out haze now drills a hole.

Coat the noble half and you shrink the cathode, which reduces the driving current on the active half even where the active half is bare. Coating both is better still. Coating only the active half is the one option to strike off the list.

Skip this and: the joint pits through faster than an untreated one would have, and it does it under paint where nobody sees it.

Step 6: Get the flow direction and the fitting sequence right

In a water system, dissolved metal travels. Copper released upstream deposits on zinc and steel downstream and pits it, so where a system mixes copper with galvanized steel, put the copper downstream of the galvanized where the layout allows. Reversing that is a slow, invisible failure that shows up as pinholes in the galvanized run rather than at the transition itself.

Also treat the transition fitting as a wear item with a known weakness. A dielectric union with a steel nipple inside it rots at the nipple, hidden, and the union looks fine from outside until it does not. Choose the transition hardware your fitting supplier lists for that pipe pair, install it where it can be reached, and note in the job record that it is above a finished ceiling if that is where it ended up.

Worked example: a heater swap into existing galvanized

The existing supply and return are galvanized steel. The replacement heater comes with copper-alloy nipples in the tappings. Someone is going to thread copper alloy into galvanized steel in a permanently wet, permanently bonded system.

Run the four decisions.

Pairing: copper alloy against galvanized zinc is a wide couple, and the zinc is the active half. That is condition one, and it is not negotiable because both halves came with the job.

Area ratio: the zinc on the galvanized pipe is the large half, so the couple works in the favorable direction. That is why these installations survive at all. The risk is not the pipe body, it is the short threaded nipple at the transition, where the zinc is thinnest because threading cuts through the galvanizing.

Electrical path: present and required. The piping is bonded, so a dielectric fitting alone does not open the circuit. Install it anyway, because it also breaks the direct metal contact and moves the corrosion front out of the threads, but do not record the job as "isolated."

Electrolyte: permanent, hot, and continuously replenished. Nothing to remove.

The specification that comes out of that: a listed dielectric transition at each tapping, accessible, with the transition placed where it can be replaced without opening a finished surface, and a note in the record naming both metals and the fitting type.

Now the time arithmetic, which is the only argument that survives contact with a customer. Fitting the two isolating transitions and documenting them adds roughly 0.3 hours to the install. The callback it prevents is a rotted transition nipple in a wet ceiling: isolate and drain, open the ceiling, cut and replace the transition, then patch. Call it 3.0 hours of trade time and a separate drywall visit. That is a 10-to-1 ratio of prevented hours to spent hours on the trade labor alone, before the finish work, and the prevented hours are unbillable warranty hours while the 0.3 is billable install time. That is the comparison to make: 0.3 billable hours now against 3.0 unbillable hours later, not hours against hours as if the two were the same currency.

What would flip this: a soft-water or highly aggressive water supply changes which half you worry about, and a system that is being fully repiped in one material removes the couple entirely and makes the transition hardware unnecessary. Buying isolating fittings for a joint you are about to delete is waste, not diligence.

Step 7: Write down what you joined

The record needs four fields and it takes under a minute: both metals by name, the isolating method used, whether continuity reads open or closed across the joint, and where the transition physically is. That last field is the one that saves the next tech an hour of ceiling.

How to verify you got this right

• Continuity across any joint you claim is isolated reads open on a meter, not "should be."
• The smaller piece in the joint is the noble one. If the small piece is the active one, go back to step 3.
• No coating exists on the active half without at least as much coating on the noble half.
• Any bonding conductor you disturbed is reconnected and verified before the system is re-energized.
• The system is back to normal operating pressure with no weep at the transition after a full heat cycle, checked with a dry hand or a paper towel rather than by eye, because a slow weep at a threaded transition evaporates and leaves only a stain.

References

• 29 CFR 1910.147 for lockout/tagout of stored pressure and mechanical energy, and 29 CFR 1910.333(b)(2) with NFPA 70E-2021, 120.5 for de-energizing and proving electrical conductors dead
• NFPA 70 (National Electrical Code), Article 250, for bonding of metal piping systems and why a required bond is not removed to solve a corrosion problem
• Manufacturer documentation for listed dielectric and transition fittings, including approved base metals and orientation
• See related: Galvanic Corrosion and the Metals That Fight; The Galvanic Series as a Field Tool