MARELLI M16FA655A Analogue AVR - Analogue Automatic Voltage Regulator for MARELLI Alternators | OEM Replacement

Short version. The M16FA655A is a Analogue Automatic Voltage Regulator, Mark X platform. It supplies the exciter field, senses the alternator output and corrects the difference several times a second. Its rating is 8 Adc continuous, its field resistance window 3-20 ohm, its accuracy +/-0.5 per cent, and it is intended for machines whose field winding sits inside that window. Read the rest of this page as the detail behind those three figures.

Excitation architecture. PMG fed, designed for Marelli MJB alternators with a permanent magnet pilot exciter. That single sentence decides most of the regulator's behaviour in a fault: whether the field survives a close-up short circuit, whether the machine builds up from residual voltage on its own, and how much the terminal voltage dips when a large motor starts on the same bus. The trim potentiometer is not a calibration device. If rated voltage is not reachable near the middle of its range, the sensing tap is wrong.

Sensing. Single-phase, 170-480 Vac, 50 / 60 Hz, +/-10 per cent tolerance. The sensing circuit is the regulator's only view of the outside world, so its wiring matters as much as its specification. A voltage error introduced by a loose terminal or a long run of untwisted cable is indistinguishable, from the regulator's point of view, from a genuine voltage error on the alternator. Field resistance is a temperature-dependent number. A reading taken on a cold morning and a reading taken after an hour at load are both correct and both different.

Power stage and field drive. Power input Auxiliary winding or main terminals, 170-277 Vac at 100 / 120 Hz, +/-10 per cent, output 100 Vdc maximum. Field current rating 8 Adc continuous, forcing capability 15 Adc for 1 minute, field resistance window 3-20 ohm. These are not independent numbers: the output voltage ceiling and the field resistance together set the maximum current the board can actually deliver, and the continuous rating is only reachable at the low end of the resistance window. If the machine will not hold voltage at all, disconnect the field and measure the sensing voltage first. That one reading separates a sensing fault from a field fault.

Accuracy and response. +/-0.5 per cent steady-state accuracy, One cycle, thermal drift +/-0.5 per cent for a 50 degC ambient change after a 10 minute warm-up. In practice the accuracy figure is the least interesting of the three. What an operator notices is recovery after a load step, which is governed by response and by available forcing current, and what a maintenance engineer notices after two years is drift, which is governed by the reference circuit and the ambient temperature inside the terminal box. Field resistance is a temperature-dependent number. A reading taken on a cold morning and a reading taken after an hour at load are both correct and both different.

Frequency behaviour. Rated for 50 / 60 Hz, with Under-frequency limiter with an adjustable knee. The frequency-sensitive function exists to protect the prime mover, not the alternator: as engine speed falls the regulator reduces excitation so the alternator stops demanding torque the engine cannot supply. Without it, an overloaded engine is dragged down until it stalls, and the alternator is blamed for a fault that belongs to the fuel or governor system. Field resistance is a temperature-dependent number. A reading taken on a cold morning and a reading taken after an hour at load are both correct and both different.

Paralleling. Reactive droop compensation, +/-20 per cent. On a single set this circuit does nothing. On two or more sets it is the only thing that stops the machines from exchanging reactive power in a slow oscillation, and it is the first place to look when a parallel installation shows a circulating current that nobody can explain. Static accuracy is measured on a resistive load at constant frequency. Real loads and real governors are neither, so treat the figure as a comparison between boards rather than a promise.

Protection built into the regulator. Under-frequency limiter, over-excitation limiter, internally replaceable fuse. These functions protect the alternator and the regulator, not the installation. Downstream protection, loss of mains and reverse power all belong to the controller, and a site that relies on the AVR for them is a site that will eventually trip in a way nobody can explain. Never flash a field through the regulator. Disconnect the board, flash the field directly, then reconnect. Flashing through the AVR is how a working board becomes a failed board.

What it fits. Marelli MJB 225-500 frame alternators. Lifecycle status: In production. Both halves of that sentence matter at order time. A board that fits perfectly but has been discontinued is a different commercial decision from one that fits and is in current production, and the price difference rarely reflects the difference in risk. Where a board has been superseded, the replacement usually fits the same footprint but not the same terminal numbering. Read the drawing, do not read the old wiring.

Field notes on this model. The Mark X is fully resin-encapsulated, which is what makes it survive humidity, conductive dust, salt spray and the vibration of a marine or offshore installation. It is designed around a PMG, so excitation holds up through a short circuit. Marelli also publishes the external control interface: a 100 kohm potentiometer gives +/-5 per cent voltage trim, and a -3 / +3 V DC signal lets a power factor controller drive it directly.

What the customer is actually buying. When a regulator is quoted as a replacement, the buyer is not purchasing a circuit board. They are purchasing the certainty that the set will hold voltage on the next load step. On the M16FA655A that certainty comes from the combination of 8 Adc continuous continuous field current and the One cycle response, and it disappears the moment the field resistance drifts outside the window the board was set up for.

Fusing. Fit the fuse the manual specifies and nothing larger. A bigger fuse does not protect the regulator better, it protects it later. On a M16FA655A the fuse is there to clear a field short before the thyristor is destroyed, and a 20 A fuse in place of an 8 A one turns a cheap board replacement into a rewind.

Stability before accuracy. Set the stability adjustment until the voltage stops hunting on a load step, then leave it alone. Chasing the last tenth of a per cent of static accuracy by tightening stability is how a set ends up oscillating at 2 Hz under a resistive load and looking like a governor fault. The M16FA655A is specified at +/-0.5 per cent from no load to full load at 0.8 power factor on balanced undistorted load at constant frequency. Steady-state accuracy +/-0.1 per cent, or +/-1 per cent with 4 per cent engine governing; a real machine on a real load will not do better than that no matter how the trimmer is set.

Measure, do not assume. Nameplate field resistance figures are nominal at 20 degC and are frequently wrong after a rewind. Take the measurement with a four-wire meter if one is available, because a two-wire reading on a low resistance winding includes the lead resistance and can be off by twenty per cent.

Voltage high and unresponsive. If the machine runs to maximum excitation regardless of the trim, disconnect the field and see whether the voltage collapses. If it does, the regulator has failed shorted; if it does not, the field is being fed from somewhere else, usually a shorted thyristor or a mis-wired paralleling circuit. Do not run the machine at full excitation for more than a few seconds while diagnosing, because the winding will not survive it.

New, old stock or refurbished. Discontinued regulators are available as new old stock, which is genuine and works, and as refurbished, which is a gamble on how the board was stored. Ask which one is being quoted. A board that has spent ten years in a damp warehouse has aged electrolytics even though it has never been installed.

Standby versus prime power. A standby set runs a few hundred hours a year and its regulator spends most of its life cold; a prime power set runs continuously and its regulator spends its life at operating temperature. The M16FA655A is rated for -30 to +70 degC, which is fine in both cases provided the terminal box has any ventilation at all. What changes between the two duties is the capacitor ageing inside the board, not the semiconductor, so a prime power machine should have the regulator inspected on a schedule rather than on failure.

Documentation that survives the job. Photograph the terminal block before disconnecting anything, and write the sensing tap and the field resistance on the inside of the terminal box lid. Ten years later that note is the only thing that will let the next engineer choose a replacement without a full survey.