1 kV or 2 kV Surge Level: How Big Should TVS Ipp Be?

INTEK (www.intek.vip) is an authorized distributor of AMSEMI (Anhui Anmei Semiconductor) TVS devices; datasheets, samples and FAE review for the 5.0SMDJ standard and foldback -N families are available on request.

One-sentence conclusion: TVS Ipp is not chosen, it is derived from the short-circuit current of the surge generator. Start from the test level and source impedance to get the peak current the generator can deliver, subtract what the front stage shares, convert across waveform conventions by energy (8/20us and 10/1000us are not interchangeable), apply the three deratings for pulse width, temperature and repetition, and only then land on a grade with 1.5 to 2 times margin. The Ipp on a datasheet is the ceiling of the device, not your requirement - confusing the two produces two classic failures: a part rated 91.8 A destroyed during a 1 kV test, or a design over-specified by an order of magnitude because 500 A in 8/20us terms was treated as 500 A in 10/1000us terms.

flowchart TD
    A["Step 1: read the test level<br/>1 kV / 2 kV / 4 kV plus 2 ohm or 12 ohm"] --> B["Convert to generator short-circuit current<br/>1 kV at 2 ohm = 500 A in 8/20us terms"]
    B --> C{"Step 2: protection topology<br/>single stage or two stage"}
    C -->|TVS alone| D["TVS takes nearly all the current<br/>only cabling and decoupling subtracted"]
    C -->|GDT or MOV front stage| E["TVS takes ten to twenty percent<br/>the front stage absorbs the bulk"]
    D --> F["Step 3: energy equivalence<br/>8/20us versus 10/1000us, never a direct ratio"]
    E --> F
    F --> G["Step 4: three deratings<br/>pulse width, temperature, repetition"]
    G --> H["Step 5: 1.5 to 2 times margin<br/>then pick voltage and power grade"]
    H --> I{"Step 5 review: did the failure come from<br/>clamp overstress or from energy?"}
    I -->|Clamp too high| J["Use a foldback low-clamp part<br/>or a two-stage design to lower residual"]
    I -->|TVS destroyed by energy| K["Raise the power grade for more Ipp<br/>the clamp stays the same"]

1. Three convention mismatches that make copying Ipp fail

The conversation in the field usually runs like this: the customer says the board is qualified to 2 kV, purchasing opens a datasheet and finds Ipp 57.5 A, and asks whether 57.5 A is enough for 2000 V. None of the three numbers lives in the same coordinate system.

• Waveform mismatch: the test level is a voltage (1.2/50us open-circuit), the datasheet number is a current (10/1000us single pulse), and the only bridge between them is source impedance. Skipping that step is like weighing with a ruler.
• Position mismatch: the generator delivering 1000 A does not mean the TVS carries 1000 A. A GDT or MOV front stage, cable impedance and decoupling elements all share it - in a two-stage design the TVS often sees only ten to twenty percent.
• Repetition mismatch: datasheet Ipp is a single non-repetitive pulse. Qualification fires once; the field may fire a dozen times a year (lightning, hot plug, paralleled-cluster circulating current) and repeated pulses take another bite out of the margin.

Each mismatch on its own can move the answer by an order of magnitude, and stacked together they produce the familiar "all parameters check out, everything still blows up".

2. Step 1: convert the test level into current

A combination-wave generator per GB/T 17626.5 (identical adoption of IEC 61000-4-5) delivers a 1.2/50us voltage wave into an open circuit and an 8/20us current wave into a short circuit, with the peak ratio fixed by the source impedance. That is the starting point: the level is a voltage, the device needs a current, and the two are one division apart.

Test level (open-circuit voltage)	Source impedance and coupling	Generator short-circuit current (8/20us)	Typical use
1 kV	2 ohm, line-to-line	500 A	In-board DC bus, secondary-side ports
2 kV	2 ohm, line-to-line	1000 A	ESS and BMS bus, telecom power
4 kV	2 ohm, line-to-line	2000 A	Outdoor cabinets, long cable entries
1 kV	12 ohm, line-to-ground	about 83 A	Asymmetric lines, single-point ground
2 kV	12 ohm, line-to-ground	about 167 A	Common safety qualification level
4 kV	12 ohm, line-to-ground	about 333 A	Outdoor common-mode testing

Two readings to remember:

1. Same level, different source impedance, six times the current. 1 kV is 500 A at 2 ohm but only 83 A at 12 ohm. Internal specs often read "1.2/50us, 1 kV, 12 ohm", which corresponds to about 83 A - so many designs qualified "to 1 kV" were only checked against one sixth of the current.
2. That current is generator capability, not TVS burden. Cable inductance, fuses, common-mode chokes and decoupling all sit between the injection point and the die. Treat it as an upper bound and confirm on the bench how much the TVS branch actually takes.

Automotive work follows a different convention altogether: ISO 7637-2 pulses 1, 2a and 5a have their own definitions (pulse 5, load dump, is far larger in energy than a combination wave) and cannot be interchanged with 61000-4-5 levels. For e-motorcycle and two-wheeler BMS, the two systems need separate evidence - one report does not cover the other.

3. Step 2: sharing - the TVS only sees part of the current

Protection topology	Share of current taken by the TVS (order of magnitude)	Suitable level	Key constraint
Single stage: one TVS	close to 100%, minus cabling and fuse	around 1 kV at 2 ohm	All energy lands on silicon; rely on power grade
Two stage: GDT or MOV plus decoupling plus TVS	ten to twenty percent	2 to 4 kV	Decoupling sets the split; front-stage residual must be checked too
Three stage: GDT, MOV, TVS	smaller still; TVS clamps only	4 kV and above, outdoor	Residual falls stage by stage; each stage needs an energy budget

Sharing is not "add a MOV and forget it" - two accounts must be settled together:

• The front-stage residual: an MOV clamps far more coarsely than a TVS, so after decoupling the TVS starts from a raised voltage. Its Vc must still contain the downstream withstand window from that raised starting point. Checking current sharing without checking residual is the most common two-stage mistake.
• The decoupling element: resistive decoupling is simple but drops voltage and dissipates in normal operation; inductive decoupling "arrives late" on fast edges - too much and the TVS is destroyed before the front stage conducts, too little and the front stage lifts the residual. That trade-off is settled on the bench with the real waveform, not on paper.

4. Step 3: 8/20us and 10/1000us are not convertible

This is the trap most often skipped, and it is where "pick a part for 500 A" comes from.

At the same peak current, a wider pulse pushes more charge and energy into the die. A 10/1000us pulse lasts about fifty times longer than 8/20us by waveform integration, which means a part rated 91.8 A in 10/1000us terms withstands far more than 91.8 A in 8/20us terms, and conversely, hunting for "Ipp at least 500 A at 10/1000us" because the event is 500 A at 8/20us over-specifies by an order of magnitude - over-designed, and usually too large to fit.

Order of preference:

1. Use the same-waveform curve when it exists. If the vendor publishes an 8/20us Ipp or energy curve, judge against it and do not convert. AMSEMI 5.0SMDJ series (datasheets 8060014 standard, 8060055 foldback -N) publish Vc and Ipp exclusively in 10/1000us terms with no 8/20us data; Littelfuse 8x20us figures are that vendor's own convention and must be attributed to Littelfuse - the two must never be mixed, converted or subtracted.
2. When no same-waveform data exists, check energy order of magnitude. Energy is the integral of v times i; with the clamp roughly constant, energy scales with charge, and charge scales with peak current times duration. This yields a magnitude verdict (enough, marginal, not enough), never a specification promise.
3. Always close on measurement: inject the real waveform and measure the voltage at the protected device pin plus the current in the TVS branch. The calculation is a design starting point only.

5. Step 4: three deratings, none optional

• Pulse-width derating: longer surges allow lower peak power. Datasheet Ipp refers to 10/1000us; consult the derating curve when the pulse is wider (see What Does the 5000 W on a TVS Datasheet Mean).
• Temperature derating: TVS junction temperature already runs near 150 C, and available peak power falls with ambient on the published curve. BMS boards sit against the cells inside a sealed pack at 60 to 70 C year round, so margin computed at room temperature shrinks.
• Repetition derating: qualification fires once, the field fires many times. Hot plug in AIDC servers and circulating current between paralleled ESS clusters are high-frequency repetitive; BMS bus surges are one-off incidents. Repetitive events need the average-power and derating cut, and single-shot capability is not repeat capability (see TVS Pulse Heating and Board-Level Thermal Design).

6. Step 5: is this grade actually enough?

Take the common 48 V and 54 V bus. The AMSEMI 5.0SMDJ54CA-N has VRWM 54 V, VBR 60.0 to 66.3 V and maximum clamp Vc 69.6 V; across the three power grades (3000 W, 5000 W, 8000 W) VBR and Vc are identical and only Ipp differs: 34.4 A, 57.5 A and 91.8 A (10/1000us terms).

Putting the steps together for a single-stage topology (all current into the TVS, clamp taken as 70 V, room temperature, single shot):

Injected event (8/20us current terms)	Generator peak current	Single-shot energy (clamp at 70 V)	Verdict for one 5000 W grade (about 5 J)
1 kV, 2 ohm	500 A	about 1 J	enough, roughly 5 times margin
2 kV, 2 ohm	1000 A	about 2 J	enough, roughly 3 times; re-check after temperature and repetition
4 kV, 2 ohm	2000 A	about 4 J	marginal: fails once hot, repeated or already drifted
2 kV, 12 ohm (line-to-ground)	about 167 A	about 0.3 J	ample; the limit is Vc, not energy

Three practical conclusions follow:

1. 1 to 2 kV on a single part is workable; 4 kV needs a front stage. The 2000 A of a 4 kV, 2 ohm event is not solved by moving to 8000 W - that grade still only carries 91.8 A, more than an order of magnitude short, and the clamp does not drop at all. Add a GDT or MOV front stage to remove the energy.
2. A bigger power grade fixes "destroyed", not "punched through". The three grades share one Vc; if the failure is Vc exceeding the downstream MOSFET or interface withstand, raising the grade is wasted money. Switch to a foldback low-clamp part (54 V grade Vc drops from 87.1 V standard to 69.6 V) or go two-stage. The diagnostic is simple: look at what died - the TVS itself, or the part behind it.
3. A bigger power grade lowers the voltage ceiling. In the foldback -N family the 5000 W grade reaches 100 V, while 3000 W and 8000 W stop at 90 V. Above 21S, raising the grade hits the voltage ceiling before it buys margin, so change the architecture rather than the part number (see TVS Architecture for 32S and 48S High-Voltage BMS).

Note also that Ipp and clamp are a pair: paralleling parts shares current but never lowers the clamp, and series connection raises voltage but brings grading and leakage with it (see TVS in Parallel and Series).

7. Bench verification: trust only the waveform at the protected pin

The derivation is paper until three things are done:

• Inject at the connector pin, not across the TVS - measuring across the TVS measures the device, not the system.
• Measure the protected device pin to ground, not the TVS terminals. What decides survival is the residual the chip sees, and the trace inductance in between lifts the peak further.
• Re-test afterwards: whether TVS leakage and VBR have drifted (a drift means the part has already been stressed), and whether the interface recovers. Criteria and fixture details are in Bench Surge Verification for BMS TVS.

8. Common pitfalls

• Comparing kV to A. Skipping source impedance is where the whole error chain starts; 2 ohm and 12 ohm differ by six times at the same level.
• Using an 8/20us peak as a 10/1000us Ipp. The requirement is either inflated by an order of magnitude or the criterion is relaxed by one - both end badly.
• Checking TVS current sharing without the front-stage residual. In a two-stage design the MOV lifts the starting voltage and the real clamp sits above the datasheet figure.
• Expecting a bigger power grade to fix everything. Same Vc within a grade, lower voltage ceiling overall - find the failure mode before spending.
• Declaring pass after one shot. Repetition and high temperature consume the margin computed at room temperature, especially in ESS and AIDC.

FAQ

Q1: For a 1 kV and a 2 kV surge level, what Ipp does the TVS need?

You cannot read Ipp off the level; convert to current first. Under GB/T 17626.5 and IEC 61000-4-5 the generator delivers 1.2/50us open-circuit voltage and 8/20us short-circuit current with the peak set by source impedance: 1 kV at 2 ohm is 500 A, 2 kV at 2 ohm is 1000 A, while 12 ohm line-to-ground gives about 83 A and 167 A. That is generator capability; front-stage impedance, cabling and decoupling share the current, and a two-stage design often leaves the TVS ten to twenty percent. Then the convention question: those amps are 8/20us while AMSEMI 5.0SMDJ Ipp is 10/1000us, so compare them only by energy order of magnitude and settle on the bench.

Q2: Can 8/20us and 10/1000us Ipp be converted into each other?

No - only checked by energy order of magnitude. At the same peak, 10/1000us lasts about fifty times longer than 8/20us, pushing more than an order of magnitude more charge into the die. Looking for Ipp above 500 A at 10/1000us because the event is 500 A at 8/20us over-specifies by an order of magnitude, and judging a 91.8 A part as failing a 500 A 8/20us event is the same error reversed. Prefer vendor curves in the same waveform: AMSEMI 5.0SMDJ publishes 10/1000us only with no 8/20us data, Littelfuse 8x20us figures belong to Littelfuse, and the two conventions must never be mixed or subtracted.

Q3: Why does moving from a 5000 W to an 8000 W TVS still fail the surge test?

Because within one voltage grade a higher power grade raises Ipp without lowering the clamp. The three AMSEMI 54 V grades share VBR and Vc (69.6 V) and differ only in Ipp: 34.4 A, 57.5 A, 91.8 A. If the clamp is punching through the downstream withstand window, the bigger grade changes nothing and you need a foldback low-clamp part or a two-stage design; raising the grade only helps when the TVS itself is dying of energy. Check what failed first - the TVS or the device behind it.

Q4: Can a single TVS absorb a 2 kV combination wave on its own?

Usually yes, provided three things are accounted for. A 2 kV, 2 ohm event is 1000 A in 8/20us terms; with a single-stage TVS taking all of it at a 70 V clamp the single-shot energy is on the order of 2 J, against roughly 5 J for a 5000 W grade 10/1000us part - about three times margin at room temperature. Temperature derating and repeated pulses consume that margin, and 4 kV at 2 ohm (2000 A, about 4 J) is already marginal. So 1 to 2 kV single-stage is workable, while 4 kV and above needs a front stage. All figures are order-of-magnitude estimates, not device specifications.

Contact us

The value of working backward is that the requirement gets computed before the part is opened - start from the datasheet Ipp as if it were the requirement, and every conclusion downstream inherits the error. Two reworks show up most often: a design specified for 500 A in 10/1000us terms that doubles the cost, will not fit the package and uses less than half the capability on the bench; and a 4 kV application with a single TVS that fails on every attempt, where swapping part numbers is pointless because the problem is the topology, not the device.

Send us your test level and source impedance, injection point, downstream device withstand and part numbers, current TVS grade and protection topology, and we will run the five steps in one pass: level to current, sharing ratio, waveform energy equivalence, the three deratings, and whether you should raise the power grade or change to two stages. AMSEMI 5.0SMDJ standard and foldback -N datasheets, samples and FAE review are available - please contact us.

Shenzhen Intek Technology Co., Ltd - authorized distributor of AMSEMI TVS and HOPERF CMT digital isolators Tel / WeChat: 136-3264-8484 Address: Room 4 A11, Building 2, No. 309 Ping'an Avenue, Pinghu Street, Longgang District, Shenzhen Website: www.intek.vip

---

Disclaimer: This article is provided for technical education and industry exchange only and does not constitute selection, purchasing or other commercial advice. The conversions of 1 kV at 2 ohm to 500 A, 2 kV at 2 ohm to 1000 A and 1 kV at 12 ohm to about 83 A follow the source-impedance definition of combination-wave generators in GB/T 17626.5 and IEC 61000-4-5; actual generator output and waveform parameters follow the equipment specification. The energy figures of about 1 J, 2 J, 4 J and 5 J are order-of-magnitude estimates from waveform integration assuming all current enters the TVS, a constant 70 V clamp and a single room-temperature shot; they are not energy promises for any device or design, and actual capability must follow the manufacturer energy and derating curves plus system measurement. AMSEMI 5.0SMDJ series Vc and Ipp (datasheets 8060014 standard and 8060055 foldback -N) follow the 10/1000us convention with no 8/20us data; Littelfuse 8x20us figures are that vendor's own convention and the two must not be converted or subtracted. The AMSEMI -N foldback family comes in 3000 W, 5000 W and 8000 W grades that share VBR and Vc at a given voltage and differ only in Ipp; the 8000 W grade is 8.0SMDJ**CA-N (VRWM 48 to 90 V, eleven grades, AEC-Q101, automotive suffix -H). Two-stage sharing ratios and temperature and repetition derating factors are engineering order-of-magnitude descriptions; actual values must follow manufacturer derating curves, system thermal design and bench measurement. High-voltage battery work should be performed by qualified professionals only.