Surge Protection on the BMS Main Circuit: Clamping Contactor Coil Flyback, Bus Transients, Precharge and Gate Supply

One sentence conclusion: a BMS main circuit is not a bus with one TVS across it, it is five transients with different mechanisms sharing one board - the L di/dt left across the charge and discharge MOSFETs, relay and contactor coil turn-off flyback, precharge inrush at kiloampere scale, negative excursions and noise pushed onto the gate and auxiliary supply, and contact bounce injecting repeated surges. The design order is always locate the path first, then select the device - get the path wrong and no amount of surge rating will help. Three sentences decide whether the board survives: coil flyback is dissipated at the coil itself, bus residual must be measured at the MOSFET drain and source, and closing the main relay with no precharge means charging the bus capacitance through milliohms.

1. Split the main circuit into five transients: locate by position, not by part

In most field failures on a main circuit the first reaction is that the TVS rating is too low, and the part is swapped from 5 kW to 8 kW. But at least five transient classes exist on a main circuit, and they differ in energy source, timescale and conduction path. Getting the clamping position wrong for just one of them can collapse the whole protection chain.

Transient class	Energy source	Typical shape	First choice of clamp location
Bus L di/dt	Busbar or harness inductance times current step	Microsecond spike, recurring	Across the bus, close to the MOSFET and busbar
Coil turn-off flyback	Relay or contactor coil energy 1/2 L I squared	Single triangular pulse, amplitude several times the supply	At the coil terminals
Precharge inrush energy	Bus capacitance 1/2 C U squared charged through milliohms	Millisecond high current pulse	Current limiting plus timeout in the precharge branch
Gate supply negative excursion	Turn-off dv/dt coupled through Cgs and Cgd, plus ground bounce	Short negative dip and overshoot	Gate loop and gate supply decoupling
Contact bounce surge	Contact chatter repeatedly making and breaking an inductive loop	Burst of small high frequency pulses	RC or MOV across the contacts, plus coil clamping

That table is the whole skeleton. Each section below answers only three questions: which parameter is being clamped, where the device goes, and what it is sized on.

Convention note: main circuit transient levels depend heavily on harness length, busbar layout, relay type and capacitance. Every figure here is an engineering order of magnitude or an algorithm, not a substitute for the original datasheet and system level testing.

2. Contactor and relay coil flyback: the device most often placed in the wrong spot

2.1 Why it is so often misplaced

The contacts of a relay or contactor sit in the main circuit, but its coil is usually driven from a separate driver board or from a BMS IO. Seeing the drive MOSFET destroyed, engineers place a TVS across that MOSFET drain and source. That puts the clamp downstream of the current path.

The coil energy lives in the coil, and at turn-off it must find a loop. The shortest and most effective loop is right at the coil terminals. The further the clamp sits, the larger the trace inductance and the sooner the spike lands somewhere else.

flowchart TD
    A["Relay or contactor turn-off command"] --> B{"Is there a clamp<br/>across the coil itself"}
    B -- "No" --> C["Coil energy 1/2 L I squared<br/>finds no low impedance loop"]
    C --> D["Coil terminal voltage pushed<br/>to several times the supply"]
    D --> E["Drive MOSFET or driver IC<br/>breaks down or back feeds the control side"]
    B -- "Yes" --> F{"Which clamping method"}
    F --> G["Flyback diode<br/>lowest voltage, slowest release"]
    F --> H["Diode plus Zener or TVS<br/>clamp level can be set"]
    F --> I["RC snubber<br/>also quenches contact arcing"]
    F --> J["Foldback TVS<br/>lowest clamp, faster release"]
    G --> K["Check drive device rating<br/>and release time"]
    H --> K
    I --> K
    J --> K

2.2 Four methods: decide how fast release must be, then how high the clamp can sit

Coil clamping is not a case of lower is better. The lower the clamp, the slower the coil current decays, and the slower the relay releases. So the first question is mechanical and safety related: how fast must this contactor open?

Clamping method	Clamp level	Release speed	Where it fits
Flyback diode only	About 0.7 V plus the supply	Slowest	Fast opening not required, drive device rating tight
Diode plus Zener in series	Settable, supply plus Vz	Medium	The usual compromise, most common
RC snubber	Set by the time constant	Medium to fast	Contact arcing must also be quenched
Foldback TVS	Lowest of the four at the same grade	Fairly fast	Fast opening required and drive rating margin is thin

This is where a foldback TVS first earns its place: for the same release time it holds the coil terminal voltage at a lower level, so the drive MOSFET rating does not have to move up a grade. Note however that coil flyback is a single, low energy event (millijoules). A high power tier is almost never needed here; a small package with adequate energy rating is enough, and the power headroom should be spent on the bus.

2.3 Two hard constraints that get skipped

• The clamp voltage must stay within the drive device rating, with margin. Coil clamp level is repetitive stress on the drive MOSFET drain and source, applied at every turn-off. It cannot be judged as a transient that passes and is gone.
• The contacts need protection too, not just the coil. Breaking an inductive loop arcs at the contacts as well, which over time shows up as oxidation and rising contact resistance. Where long life matters, add an RC or MOV at the contacts. One coil clamp device cannot fix both ends.

3. Bus L di/dt: why measuring across the TVS gives a false pass

3.1 Residual is set by the TVS plus the trace, not by the TVS alone

The datasheet clamping voltage is a reading at a calibrated point. What actually appears at the downstream MOSFET drain and source is:

V_DS(peak) is approximately Vc(TVS) plus L_trace times di/dt

Main circuit di/dt is very steep when current steps (MOSFET turn-off, relay switching, short circuit recovery), and busbar, harness and pad inductance commonly runs from tens to hundreds of nanohenries. Even at 50 nH and 100 A per microsecond that term is 5 V, and real main circuits switch faster still.

Conclusion: a clean waveform at the TVS terminals does not mean the MOSFET survived. Move the probe to the pins of the protected device.

3.2 Which parameter the bus is clamped on

The first constraint on the bus is whether the downstream device gets punched through, so what matters is the downstream withstand window, not the nominal surge rating:

Item to work out	How to take it	Pass criterion
Downstream MOSFET or IC rating	Nominal minimum, or measured typical about 10 percent higher	Take the more conservative of the two
Allowed residual ceiling	Rating times 0.85, 0.8 for automotive	Engineering practice, not a mandatory standard
Actual TVS residual	Vc at the target surge current	Not the single calibrated Ipp point
Trace contribution	L_trace times di/dt	Set by layout, must be measured

The full derivation of that chain is in the voltage window for 23S and 24S buses. This article stresses one point: raising the power tier does not change Vc.

3.3 A higher power tier does not lower the clamp

Across the 3000 W / 5000 W / 8000 W tiers of the AMSEMI -N foldback series, VBR range and maximum Vc at the same voltage grade are identical. What changes is surge current capability Ipp:

Power tier	54 V grade part	VBR min to max (V)	Vc max (V)	Ipp (A)
3000 W	SMDJ54CA-N	60.0 to 66.3	69.6	34.4
5000 W	5.0SMDJ54CA-N	60.0 to 66.3	69.6	57.5
8000 W	8.0SMDJ54CA-N	60.0 to 66.3	69.6	91.8

(Datasheet convention: AMSEMI Doc. 8060084 / 8060055 / 8060057, all 10/1000 us. The 8000 W tier figures follow the manufacturer published data; confirm against the latest datasheet.)

So if the field problem is residual high enough to punch the MOSFET, moving to a higher power tier does nothing at all to the residual. There are only two levers: move to a lower clamping part (the foldback version is about 20 percent lower than the standard part at the same grade), or raise the protected device rating by one grade. The first saves cost, the second adds it. Full model cross reference is in the AMSEMI foldback TVS model and cross reference page.

Convention rule when quoting: AMSEMI 5.0SMDJ datasheets (8060014 standard, 8060055 -N foldback) carry only the 10/1000 us convention and contain no 8/20 us data at all. Some imported brands publish both (for example the Littelfuse 5.0SMDJ85CA public table lists 36.5 A at 10x1000 us and 273.8 A at 8x20 us). That figure is Littelfuse own convention and must be attributed, and the two waveforms must never be converted into one another or subtracted.

4. The precharge branch: the most skipped and most dangerous section

There is one invisible high current device on the main circuit: the bus capacitance. Closing the main relay, or the main MOSFET, with the pack connected straight across milliohm class loop impedance produces an inrush that in theory equals pack voltage divided by total loop impedance. On a 48 V platform with a few tens of milliohms that reaches the kiloampere region.

Three typical outcomes: the main relay contact welds and cannot open again, the main MOSFET exceeds its single pulse avalanche or surge rating, and the bus collapses far enough to trip undervoltage or reset the controller.

Do this	Why	Common mistake
Keep the precharge branch separate (precharge relay plus limiting resistor)	Limits inrush to the few ampere region	Sharing the main relay, which does neither job well
Close the main path only at 90 to 95 percent of pack voltage	Leaves only 5 to 10 percent differential, so no inrush burden	Threshold set too low, so inrush remains
Fit precharge timeout protection	With a welded relay the resistor sees P = V squared / R continuously, about 153 W at 48 V and 15 ohm	No timeout, resistor burns through or catches fire

Resistor selection, the dual check of pulse energy 1/2 C U squared against steady state V squared / R, and the three way precharge failure diagnosis are covered in how to design a BMS precharge circuit. Two points that get missed:

1. The precharge branch needs its own clamping protection. The resistor carries usable power during inrush and the precharge relay, or precharge MOSFET, still stands off the full bus voltage at the instant of closing. With a MOSFET the single pulse SOA must be checked; steady state current is not a valid basis.
2. Precharge failure is often misdiagnosed as a failed TVS. A welded contact producing sustained overcurrent overheats the bus TVS, so the symptom is a failed TVS while the root cause is in the sequence. See section 7 for the order of diagnosis.

5. Gate supply and gate loop: negative dips, ground bounce, and failures where the part is fine

Not all main circuit MOSFET failures are energy problems. Three transient classes are unrelated to energy and just as destructive:

Symptom	Mechanism	Countermeasure
Gate-source overvoltage or negative breakdown	Turn-off dv/dt coupled through Cgd lifting the gate, or ground bounce pushing Vgs negative	Separate on and off gate resistors, Miller clamp, gate-source clamp or ESD part close to the device
Driver malfunction or shoot through	False turn-on during the Miller plateau, insufficient dead time or drive strength	Check Qg against peak drive current, check Cgd times dv/dt against threshold
Gate supply sagging or resetting	Peak drive current back feeding the supply, inadequate decoupling	Local decoupling at the gate supply, separate return for the drive loop

The full derivation is in how to design a MOSFET gate drive circuit and why a BMS burns its MOSFETs, which attributes failure layer by layer; this article's classification is by position and complements it. One criterion is worth repeating:

A narrow, tall spike far above the rating means an energy path problem: look at clamping.A plateau lasting tens to hundreds of nanoseconds means a gate or drive problem: look at the gate loop.Rising case temperature before anything else means a thermal problem: look at current sharing and heatsinking.

6. The six step main circuit protection workflow

flowchart LR
    A["1 Split transients<br/>list five paths by position"] --> B["2 Draw boundaries<br/>AC entry, DC bus, load side"]
    B --> C["3 Capture waveforms<br/>measure before selecting"]
    C --> D["4 Work back the window<br/>rating times 0.85 sets residual"]
    D --> E["5 Place devices<br/>coil local, bus at MOSFET, precharge limiting"]
    E --> F["6 Bench verification<br/>probe the protected device pins"]
    F --> G["7 System retest<br/>ambient temperature and repeated surge"]
    G --> H["8 Parameter drift check<br/>leakage, VBR, clamp"]
    H --> C

Steps 3 and 6 are the ones most often skipped, and both are measurement rather than selection. They are also exactly what separates a design that looks compliant from one that is.

7. Bench verification: four questions and the step everyone skips

Main circuit surge practice follows the BMS bus surge bench closely (full criteria in how to verify the TVS on a BMS protection board). Three points are specific to the main circuit:

1. Test coil flyback and bus surge together. Passing each alone and failing in combination is common, because contact bounce superimposes coil flyback onto bus L di/dt.
2. Probe the pins of the protected device. Trace inductance times di/dt is not in the datasheet and only appears on the board.
3. Recheck parameter drift after repeated surges. Every surge leaves microscopic damage in the junction, showing up over life as a higher leakage limit, slowly drifting VBR and a higher clamp at high current. Continuously biased nodes deserve a look at TVS leakage and standby current.

Common pitfalls

Pitfall	Symptom	Correct approach
Coil clamp placed at the drive MOSFET	MOSFET repeatedly destroyed, coil flyback unchanged	Put the clamp across the coil
Reading only the nominal surge rating	Moving to 8 kW leaves the residual untouched	Compare Vc at the target current
Judging pass at the TVS terminals	Clean waveform, MOSFET still fails	Move the probe to the protected device pins
Sharing the precharge relay with the main relay	Either no limiting or a small contact carrying high current	Keep the precharge branch separate
Using a flyback diode as the only coil clamp	Contactor opens slowly, safety loop responds late	Choose the clamp level from the allowed release time
Mixing waveform conventions	Using an 8/20 us Ipp against a 10/1000 us Vc	Lock one convention before comparing

Related solutions

• How to select the TVS on a BMS protection board (the foldback anti-burn approach) -> BMS board burn and protection TVS selection
• The three voltage window checks on a 23S or 24S bus -> TVS fitted and the MOSFET still fails
• Layer by layer attribution of bus TVS failure -> Why does a BMS burn its MOSFETs
• Precharge resistor, relay and power-up sequence -> How to design a BMS precharge circuit
• Comparison of clamping methods for inductive turn-off spikes -> How to clamp a relay coil turn-off spike
• System level surge protection staging (MOV, GDT, TVS) -> How to design surge protection
• The three foldback TVS power tiers and representative models -> Foldback TVS family overview
• Selecting and paralleling the main circuit charge and discharge MOSFETs -> How to select BMS charge and discharge MOSFETs

FAQ

Q1: Which protection devices must a BMS main circuit have, and where do they go?

At least three positions. 1. A clamp across the bus, close to the main circuit MOSFET and busbar, covering L di/dt and residual coupled in from outside. 2. A clamp across the relay or contactor coil terminals, covering turn-off flyback, and note this is not at the drive MOSFET. 3. Current limiting plus timeout protection in the precharge branch, covering closing inrush. Add interface level protection if there is an external communication port, a sampling connector or a long harness. The placement rule is that the clamp belongs upstream in the energy dissipation path - as close to the energy source as possible and as close to the protected device as possible, and when those conflict, favour the energy source.

Q2: For relay coil turn-off flyback, should I use a flyback diode or a TVS?

Start from how fast the contactor must open. A flyback diode holds the coil at the lowest voltage, roughly 0.7 V forward drop plus the supply, but the coil current decays slowest and release is slowest. A diode in series with a Zener lets you set the clamp level and is the usual compromise. A foldback TVS gives the lowest clamp voltage at the same grade, so the drive MOSFET rating need not move up a grade while still meeting the release time. Coil flyback is single event, millijoule class energy, so a high power tier is unnecessary; size a small package for the energy and spend the power headroom on the bus.

Q3: My main circuit MOSFETs keep failing. Is the TVS power tier too low?

Probably not. Do two things first. Move the probe from the TVS terminals to the MOSFET drain and source to see the real residual, since trace inductance times di/dt is not in the datasheet. Then confirm the TVS clamping figure is taken at the target surge current rather than the single calibrated Ipp point. If the residual genuinely exceeds the downstream window, a higher power tier will not help - the AMSEMI -N series 3000 W / 5000 W / 8000 W tiers share identical VBR and maximum Vc at the same voltage grade and differ only in Ipp, which is 34.4 / 57.5 / 91.8 A at the 54 V grade. To lower the residual either move to a lower clamping part, about 20 percent lower for the foldback version at the same grade, or raise the downstream rating one grade.

Q4: If the main circuit already has surge protection, is precharge still needed?

Yes. The two handle entirely different transients. The bus TVS handles microsecond spikes with millijoule to joule class energy, while precharge handles the millisecond capacitor charging inrush with joule to tens of joules of energy, several orders of magnitude apart. Using a TVS to absorb precharge inrush does not work at the power level: charging a 1000 uF bus capacitance on a 48 V platform takes about 1.2 J, while one calibrated 10/1000 us pulse is millijoule class. On a BMS or inverter with a large bus capacitance, precharge is effectively mandatory.

Contact us

Shenzhen Intek Technology Co., Ltd is an authorised distributor for AMSEMI TVS and HOPERF CMT digital isolators, focused on low clamping and shallow foldback TVS protection for BMS boards, energy storage and telecom power. If you need a transient path breakdown for your main circuit topology, a bus clamping window calculation, the I-V curve and V_H / I_H data for a particular grade, or a combined coil flyback and bus surge bench plan, please contact us. Send the main circuit schematic, the relay or contactor model, the bus capacitance and the surge test standard, and we will check it against your actual operating conditions.

Shenzhen Intek Technology Co., Ltd - electronic component distributor and system solution provider Tel / WeChat: +86 136-3264-8484 Address: 4F, Building 2, Jingwei Center, 309 Ping'an Avenue, Pinghu Street, Longgang District, Shenzhen, China Website: www.intek.vip

---

Disclaimer: This article is technical and engineering exchange material covering how to protect the various transient classes on a BMS main circuit by path. It is not selection, procurement or business advice. The V_RWM / V_BR / V_c / Ipp conventions quoted here come from AMSEMI published datasheets (standard 5.0SMDJ series Doc. 8060014, -N foldback 5.0SMDJ series Doc. 8060055, -N foldback SMDJ series Doc. 8060084, -N foldback 8.0SMDJ series Doc. 8060057) and are all on the 10/1000 us convention with no 8/20 us data in those documents. Littelfuse 8x20 us figures are Littelfuse own published convention and are attributed as such; the two waveforms must never be converted into one another or subtracted. Coil flyback energy, trace inductance magnitudes, precharge inrush and temperature multipliers are engineering order of magnitude estimates and empirical rules rather than datasheet guarantees. Actual design must follow the datasheet of the chosen part, measured waveforms on the assembled unit, and the applicable system test standard.