Key Takeaway#

In a non-vacuum load tap changer, acetylene and hydrogen are normal. They are the by-product of contact arcing under oil on every tap operation — the device doing its job, not a fault. The single most common field error is reading an LTC oil sample against main-tank expectations and raising a false arcing alarm. Interpreting an LTC sample correctly means using a different framework entirely: population-derived limits instead of a fixed threshold table, acetylene-normalised ratios, Duval Triangle 2 instead of Triangle 1 — and knowing which compartment and which tap-changer type the sample came from before you say anything about condition.

The Trap#

Dissolved gas analysis of a main transformer tank rests on a simple premise: combustible gases mean trouble. Hydrogen points to partial discharge, the heating gases to thermal faults, and acetylene to arcing. IEEE C57.104 and IEC 60599 codify that reading through threshold tables, ratios, and the Duval triangles. The premise is sound — for the main tank.

The load tap changer inverts it. An on-load tap changer (OLTC) regulates voltage by switching between winding taps while the transformer stays energised, and in a non-vacuum design that switching happens by drawing an arc through the oil. Every tap operation produces gas. IEEE C57.139 — the IEEE guide for DGA in transformer load tap changers — is explicit that combustible gases, especially the arcing gases hydrogen and acetylene, are expected as a normal product of operation, and that the quantity of gas is roughly proportional to how much tap-changing the unit has done.

That last point has a consequence people rarely anticipate: high absolute gas levels can be entirely normal. An arc-furnace transformer's LTC, switching hundreds of thousands of times a year, will read far higher than an urban distribution LTC of the same type — and both can be healthy. Run that arc-furnace sample through a main-tank threshold table and it screams arcing fault. It is doing exactly what it was built to do.

So the first discipline is to stop treating the LTC sample as a small main tank. It is a different device, gassing for a different reason, and it needs a different rulebook.

Why the Hardware Gasses the Way It Does#

To read LTC gas you have to know how the switching hardware is designed, because the design envelope sets what "normal" looks like. Two tap-changer standards define that envelope: IEC 60214-1 on the IEC side and IEEE C57.131 on the IEEE side. They are deliberately harmonised — on the number that matters most for gas interpretation, they are identical.

A resistor-type OLTC bridges between taps through transition resistors that carry current for a fraction of a switching cycle. Those resistors get hot, and hot metal in oil produces the heating gases — methane, ethane, and especially ethylene. Both standards bound how hot the resistor is allowed to get: under a test at 1.5 times maximum rated through-current, the transition-resistor temperature rise must not exceed 400 K for compartment-type tap-changers or 350 K for in-tank tap-changers. (The bare "350 K" figure that circulates in the literature is the in-tank value; the compartment value is 400 K.)

Here is where field judgment enters. CIGRE work referenced by C57.139 found that some arcing-resistor-type LTCs generate large amounts of heating gas — dominant ethylene — without any fault, simply because of how their transition resistors heat during normal switching. The ethylene/acetylene ratio in these units can drift upward with age without a problem developing. Knowing that the resistors are designed to a bounded, deliberate temperature rise is what lets you recognise heavy ethylene in a healthy resistor-type LTC as the exceptional-but-normal pattern it is, rather than chasing a coking fault that isn't there. The hardware standard is not a fluid-analysis document, but it is the context that keeps you from over-reacting to the gas.

The False Positives That Live in the Application#

Some of the most stubborn false alarms don't originate in the diverter switch at all. They come from the way the tap changer is applied, and they are catalogued in the application-guidelines standard, IEC/IEEE 60214-2. Its field-service guidance states plainly that DGA of OLTC liquid is completely different from transformer evaluation. Three traps recur.

Change-over selector discharges. When a coarse/fine or reversing change-over selector operates, it momentarily disconnects part of the winding, and the resulting discharge generates gas — in particular acetylene and hydrogen. The volume is small, but if that selector operates within the main transformer tank liquid, the gas shows up in the main-tank DGA and, in the words of the standard, "can be falsely interpreted as a defect." When a main-tank sample shows unexplained acetylene and the unit has an in-tank change-over selector, this is the first thing to check.

Acetylene from a resistance measurement. Measuring winding resistance with DC current present while operating the tap changer can pit the contacts and generate acetylene. That acetylene is a procedural artefact, not an in-service arcing fault — but it sits in the next oil sample looking exactly like one. Knowing the maintenance history disambiguates it.

Pyrolytic carbon on stuck contacts. Contacts left in one tap position for months, or running with low contact pressure or high oil temperature, can grow carbon deposits that eventually generate free gas. This one is a developing problem — and the standard's recommendation is that regular DGA and corrosivity testing catch it early.

The common thread: an LTC gas reading is only interpretable in the context of the tap changer's design, application, and recent maintenance. The sample alone is not enough.

When LTC Gas Shows Up in the Main Tank#

Everything above is about reading the LTC sample correctly. The same physics runs the other way, and it is worth stating on its own, because it produces one of the more convincing false positives in main-tank DGA: an LTC that gasses normally can raise the main tank's acetylene and ethylene, mimicking an internal arcing fault in the transformer itself. IEEE C57.104 acknowledges this directly — it is part of why the standard formally excludes LTC-coupled units from its scope and defers tap-changer DGA to C57.139.

There are two routes for the gas to travel. The first is an in-tank change-over selector, which — as noted above — can generate acetylene and hydrogen during normal switching that ends up in the main-tank liquid. The IEC/IEEE 60214 family treats this as expected behaviour, not a defect. The second is a hardware leak: a degraded barrier between the LTC compartment and the main tank lets diverter-switch gas migrate where it does not belong.

The screen for this is the single most quotable diagnostic in the whole subject, because two independent standards land on the same number. When the main-tank C₂H₂/H₂ ratio runs above roughly 2 to 3, suspect OLTC contamination rather than a transformer arcing fault. IEC 60599 and IEEE C57.104 reach that threshold by different routes and agree on it — and that concordance is what makes it trustworthy as a first filter.

Confirming it is straightforward, and it is the step too often skipped: sample all three compartments on the same date — the main tank, the OLTC compartment, and the conservator — and compare. A genuine OLTC source shows the diverter compartment leading, with a gas signature that tracks back to the tap changer rather than the windings. IEC 60599 describes exactly this comparison sampling as the way to settle the question.

The TriboTech diagnostic tool supports this case directly. Our Duval Triangle 1 tool has an OLTC mode that overlays an oil-contamination (OC) zone — the high-acetylene, low-methane-and-ethylene corner of the triangle where main-tank samples contaminated by tap-changer gas tend to land. The OC-zone concept follows Bustamante et al. (2024), and the practical value is that a main-tank point falling in that corner is a flag to run the three-compartment comparison rather than to mobilise for an internal fault. Plot the gases in the live tool — the boundaries are implemented there — instead of working from memorised cut-offs.

How to Interpret an LTC Sample#

Once you accept that the LTC gasses normally, the obvious question is: against what do you judge "abnormal"? IEEE C57.139's answer is the part that surprises people most — there is no universal threshold table. The guide deliberately does not provide a fixed list of acceptance limits analogous to the main-tank tables. Instead it tells you to derive statistical limits from a population of similar LTCs operating under similar conditions, and to interpret each sample against those limits and the unit's own trend.

That sounds demanding, and where a client has enough comparable units it is the right approach. Where they don't, the guide offers a fallback for ratios — an ethylene/acetylene caution of about 0.5 and warning of about 1.0 — but only as a starting point to be adjusted to experience, never as a fixed rule. We say so explicitly when we use it.

The workhorses of LTC interpretation are the acetylene-normalised ratios: ethylene/acetylene and the total-heating-gas-to-acetylene ratio. They are robust precisely because they are largely independent of how many times the unit has switched and of gas lost to a breather — the two factors that make absolute concentrations so hard to read in an LTC. The guide instructs that these ratios be defined so the value gets worse by increasing, which is why ethylene/acetylene (not its inverse) is the one to track.

Two of the guide's case examples make the single most useful teaching point. In one, a unit with badly coked arcing contacts showed very low gas concentrations but ratios far above the warning level — the ratios caught a fault the concentrations missed. In another, a unit unable to operate showed high concentrations but perfectly normal ratios — the fault was mechanical, driving excess switching, and only the concentration limits plus the operations count caught it. The lesson: concentrations and ratios detect different fault classes, so use both. The number of operations, where you can get it, is a valuable third axis.

For graphical fault-type classification, the LTC-specific tool is Duval Triangle 2, not the main-tank Triangle 1. It uses the same three gases — methane, ethylene, acetylene — but with zone boundaries set for the way an LTC gasses. Using Triangle 1 on an LTC sample is one of the misreadings this whole article is about.

Try it yourself: the TriboTech Duval Triangle tool includes the Triangle 2 variant for tap-changer diagnostics. Plot your LTC gases there rather than reconstructing zone boundaries by hand — and remember the plotted position is unreliable if any of the three gases is near its detection limit.

The Pure-Vacuum Inversion#

There is one LTC type where the logic flips back. In a true vacuum tap changer, the current switching happens inside sealed vacuum bottles, not in the oil. The oil-immersed contacts — if any — do almost no arcing, so they form very little gas. For a pure vacuum LTC with no oil-immersed arcing selector or bypass switches, any significant gas is abnormal, and C57.139 itself says to use the main-tank Duval Triangle 1 for these units.

This is the exception that proves the rule. The reason a non-vacuum LTC needs its own framework is that it arcs in oil by design; remove the oil-immersed arcing and you remove the reason, and the main-tank tools become appropriate again. Which is exactly why you must know the switch architecture before you choose the tool. Acetylene in a vacuum LTC means something very different from acetylene in an arcing-type one.

Practical Close#

The discipline behind all of this is short.

1. Interpret the LTC sample separately from the main tank. Different device, different gas mechanism, different rulebook. A main-tank sample mislabelled as an LTC sample is a recognised data-quality failure — confirm what you are actually looking at.

2. Establish the compartment and the LTC type first. Diverter versus selector compartment, arcing versus vacuum, resistor versus reactor. The same gas reading means different things across these, and "normal" is defined per type.

3. Lead with ratios, support with concentrations, and trust trends over single samples. The acetylene-normalised ratios are the most reliable early indicators; concentrations and operation counts catch the fault classes ratios miss; and a developing fault shows up as a trend across samples, not as one alarming number.

4. Watch the nitrogen-to-oxygen ratio as a breather check. On an air-breathing LTC, an N₂/O₂ ratio drifting well above its normal range is a cheap early warning that the breather is clogged and combustibles are accumulating for a reason that has nothing to do with a fault.

None of this is exotic. It is the ordinary discipline of reading the instrument you actually have in front of you — a load tap changer, doing its job, gassing because that is what it does.

Talk to TriboTech#

If your monthly DGA programme covers transformers with on-load tap changers, the LTC samples deserve their own interpretation, not a main-tank threshold table applied by reflex. We build LTC condition assessments on the right framework and tell you, in plain language, whether the gas is the device working or a fault developing. Get in touch to discuss your fleet.