How Many Surges Can a TVS Take? Datasheets Never Say - So the Lifetime and Scrap Criteria Are Yours to Build

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

One-sentence conclusion: "How many surges can this TVS take" has no answer in the datasheet and never will - because surge damage is the accumulation of microscopic junction damage, not mechanical wear with a fixed cycle count. The rated Ipp guarantees that one 10/1000us pulse will not break the part, and that is all it guarantees. The working substitutes are three: keep the energy of each real-world pulse below the derated rating with margin; separate event-type from dense-type surge regimes and manage them differently; and run a three-parameter lifetime program on IR / VBR / Vc with scrap limits defined jointly by the end manufacturer and the component vendor - rather than waiting for the part to tell you.

I. Why the datasheet will never give a count

Start by taking the question apart. The peak pulse power a datasheet quotes (5000 W for the AMSEMI 5.0SMDJ series, for example) and the Ipp figure are defined for a 10/1000us standard waveform, single non-repetitive pulse - they answer "will one such pulse break it", and the answer is no. As for the second pulse, the hundredth, the ten-thousandth, the datasheet is silent.

The silence is not vendor caution; there is no universal number to give:

• Damage is cumulative and statistical. Every clamp, even a fully in-rating one, leaves microscopic damage in the junction: local hot spots, lattice defects, interface degradation. No single event is fatal; accumulation shows up first as parameter drift and only later as failure.
• The accumulation rate depends strongly on conditions. The same pulse damages faster at an 85C starting junction than at 25C; a 10/1000us long wave carries an order of magnitude more energy than an 8/20us short wave; spacing of one second versus one millisecond decides whether the heat ever leaves. Any guaranteed count would have to be locked to one full operating profile, and would be meaningless for every other customer.
• The failure criterion itself needs defining. How much drift counts as "failed"? Does a doubling of leakage count? That depends on how much protection margin the system still has - a question for the system builder that the component vendor cannot answer unilaterally.

So the useful form of the question is not "how many surges" but: what fraction of the rating is each surge in my application, which three parameters do I monitor, and where does the scrap line sit? The rest of this article gives the method.

II. The aging chain: from one in-rating clamp to a dead short

TVS aging is not sudden; it is a causal chain, and every link in it can be monitored:

flowchart LR
    A["Every surge clamped<br/>even fully within rating"] --> B["Microscopic junction damage<br/>hot spots and lattice defects"]
    B --> C["Reverse leakage IR creeps up<br/>the most sensitive early sign"]
    C --> D["Static leakage power rises<br/>self-heating raises starting Tj"]
    B --> E["Breakdown voltage VBR drifts slowly<br/>batch distribution widens"]
    B --> F["High-current clamp Vc rises<br/>protection margin shrinks"]
    D --> G{"Keep running?"}
    E --> G
    F --> G
    G -->|No monitoring, no derating| H["Next surge exceeds<br/>the degraded capability"]
    H --> I["Breakdown short<br/>failed part drags down the bus"]
    G -->|Monitored, scrap line set| J["Replace, system stays safe"]

Two links deserve emphasis. First, rising IR accelerates the aging itself: leakage is continuous power dissipation; as it climbs, self-heating raises the starting junction temperature, and the same surge then does more damage at higher temperature - aging is a positive feedback loop, not a linear wear-out. Second, a rising Vc eats the system protection margin: the margin you kept by selecting a TVS whose Vc sits below the MOS rating is consumed bit by bit as the part ages, until one large surge finds the clamp already past the MOS limit - the MOS is what visibly fails, but the root cause is the aged TVS.

III. Event-type versus dense-type: sort your surges first

The same TVS lives under two completely different lifetime regimes:

Regime	Typical scenarios	Energy signature	Lifetime approach
Event-type	Hot-plug, switching events, residual lightning, ESD	Low frequency (a few to a few dozen per day), each pulse inside the derated rating	Select per datasheet, verify the energy margin once, expect years of service
Dense-type	Frequent relay/contactor cycling, repeated reverse-polarity insertion, switching spikes every cycle	High frequency, energy accumulates into average power via duty cycle	Single-pulse ratings are not enough; send the pulse sequence to the vendor, measure temperature rise

Event-type deserves less anxiety than it gets: a few dozen pulses a day, each inside the derated rating, is exercise rather than wear; datasheet selection plus an energy-margin check is enough. The dangerous one is dense-type - a TVS can only dissipate a few watts continuously, and by P_avg ≈ P_peak × (pulse width ÷ period), a 5 kW pulse every few milliseconds pushes average power to the watt-scale limit. Dense regimes cannot be selected on single-pulse Ipp alone; assemble the pulse sequence (width, peak, frequency, duty cycle, starting junction temperature) and send it to the manufacturer for confirmation.

One in-between case is routinely mis-classified as event-type: the dielectric withstand test. A 1500 V AC test crosses zero every half cycle - about 3000 avalanche drives per minute - so what looks like a "test" is a dense-type regime for the device, and it is one reason reworked boards keep failing the retest.

IV. Three monitored parameters: incoming inspection, post-surge review, scrap decisions

On the production floor, lifetime management is a closed loop of three parameters measured at three points:

Parameter	Sensitivity and role	Incoming inspection	After surge exposure
Reverse leakage IR	Most sensitive early sign: junction damage shows first as rising leakage	Measure at 25C, limit per the datasheet (order of magnitude 5 uA for the 54-95 V grades), record batch distribution	Compare against the incoming distribution: a single part clearly off the batch level freezes the batch
Breakdown voltage VBR	Drifts slowly; widening distribution appears before single-point shift	Measure VBR against the datasheet window (upper limit 66.3 V for the 54 V grade), keep the batch distribution	Compare drift and spread against the same batch; out-of-line drift or spread triggers disposition
Clamp voltage Vc	Needs surge equipment; the real high-current protection capability	Not 100%-tested; sampling only	Re-measure Vc after surge sampling to confirm the protection margin survived

Read them in this order: IR is the thermometer, VBR is the medical history, Vc is the verdict. A normal-looking VBR does not mean a healthy part - IR may already have moved; and IR alone cannot substitute for Vc, which needs surge equipment. Three checkpoints close the loop:

1. Incoming inspection: measure IR and VBR at room temperature, set the window (datasheet limits plus measured batch distribution), archive the data as the baseline.
2. Surge sampling: sample-verify against the system surge profile, number and archive every sampled part, and never return them to the line.
3. Drift review: re-measure IR / VBR / Vc on the sampled parts, compare against the incoming baseline, and check drift magnitude and direction against the scrap line.
4. Periodic in-service review: for permanently biased nodes (a bus TVS sits across the pack for its whole life), track the IR trend with standby leakage monitoring, distinguishing temperature-driven drift from sudden leakage events.

V. The four most common production-line mistakes

• Treating bench passes as a lifetime rating. "We hit it with 200 surges and nothing failed" proves only that those 200 pulses under those conditions did not fail it. It predicts nothing about the thousandth pulse, nor about in-vehicle temperatures and waveforms. The bench is a sampling tool, not a lifetime report.
• Copying the datasheet min/max as the acceptance window. The VBR window in the datasheet (upper and lower limits for the 54 V grade, say) is the permitted range across all production batches, not an incoming-inspection line - copying it discards all visibility into batch distribution. Set the window from measured batch data: a distribution that suddenly widens or shifts as a whole flags a batch problem earlier than any single out-of-limit part.
• Returning surge-sampled parts to the line. A surged part has a stress history; passing the parameter re-test does not make it incoming stock again. Returning it mixes aged parts into new batches with no traceability. Number the samples, archive them, scrap them after destructive tests, and write it into the process card.
• Measuring VBR without IR. VBR is the breakdown-point voltage and is insensitive to early junction damage; IR moves first. An incoming-inspection flow that only checks VBR is a physical exam that measures height but never temperature.

VI. The energy margin: damage less before you monitor more

The first step of lifetime management is not monitoring - it is reducing the damage each surge does. The calculation order matches the TVS power rating article; two points matter specifically for lifetime:

• Deratings multiply. If the real waveform is longer than 10/1000us, derate for pulse width first; if the starting junction temperature is above 25C, derate for temperature next; the product is the energy this part can actually take in your application. Actual surge energy divided by that number is your per-event wear rate - the smaller the ratio, the slower the accumulation.
• There is no universal margin percentage. "Is 50% margin enough" depends on frequency, waveform, temperature and the system safety class; send the complete pulse sequence to the manufacturer for confirmation (the distributor can arrange the FAE link). The vendor may return an experience-based range, recommend a higher power grade, or ask for temperature-rise measurement - all of which beat any rule of thumb.

The foldback -N family has a structural advantage here: the 3000 W, 5000 W and 8000 W grades share VBR and Vc at a given voltage and differ only in Ipp (about 34.4 / 57.5 / 91.8 A at the 54 V grade, 10/1000us). When the energy margin runs short, step up a power grade at the same voltage: the clamping system and downstream coordination stay untouched, and the lifetime margin upgrade does not force a full re-verification of the clamp design.

Related solutions

• How to select TVS for a BMS protection board - the main selection flow and voltage window
• What the 5000 W on a TVS means - peak power, pulse-width derating and average power
• How to verify a BMS protection-board TVS - what a surge bench test should cover
• Does a failed TVS short or open - failure modes and collateral damage
• TVS leakage and static power - standby drain and the temperature behaviour of IR
• Why a BMS keeps failing its dielectric test - the dense damage a withstand test inflicts
• The AMSEMI -N foldback family - three power grades, 47 voltage grades and cross-references

FAQ

Q1: How many surges can a TVS withstand, and why do datasheets not give a count?

Datasheets genuinely do not answer this. The peak pulse power and Ipp ratings are single, non-repetitive 10/1000us figures that guarantee one such pulse will not damage the part. Surge damage is the statistical accumulation of microscopic junction damage - every clamp, even a fully in-rating one, leaves some - and the accumulation rate depends strongly on starting junction temperature, waveform and pulse spacing, so no universal count can be guaranteed. The working substitutes are: derate the actual pulse for width and temperature, keep the energy below the rating with margin, and manage lifetime by monitoring drift in IR, VBR and Vc.

Q2: How do I tell whether my surges are event-type or dense-type?

Look at frequency and energy accumulation. Event-type surges are infrequent occurrences - hot-plugging, switching events, residual lightning - a few to a few dozen per day with each pulse inside the derated rating; parts selected per the datasheet serve for years. Dense-type surges come from relay or contactor cycling, repeated reverse-polarity insertion, or switching spikes fed into the TVS every cycle, where energy accumulates into average power through duty cycle and a pulse every few milliseconds can push average power past the few-watt continuous limit. For dense regimes send the full pulse sequence to the manufacturer and verify by temperature-rise measurement if needed. Note that a 1500 V AC dielectric test - about 3000 avalanche drives per minute - is a dense-type regime as far as the device is concerned.

Q3: Which parameters drift as a TVS ages, and how do I monitor them on the line?

Three parameters, ranked by sensitivity. Reverse leakage IR is the earliest warning: junction damage first shows as leakage creeping up; at 25C the upper-limit order of magnitude for the 54 to 95 V grades is 5 uA (per the datasheet), and both incoming inspection and post-surge checks should measure IR at room temperature. Breakdown voltage VBR drifts slowly, and batch distribution comparison is more informative than checking single parts against min/max. High-current clamp voltage Vc needs surge equipment and usually belongs in sampling. Read them together: a normal VBR does not prove the part is healthy - IR may already have moved.

Q4: Can surge-sampled parts go back to the production line, and how should scrap limits be set?

No, and it should be written into the process card. A part that has been through surge sampling has a stress history; even if parameters re-test in spec, it is not equivalent to untouched incoming stock, and returning it mixes aged parts into new batches. Sample parts should be numbered, archived and scrapped after destructive tests. There is no universal scrap number to copy: the VBR and IR acceptance window should combine the datasheet limits with measured batch distributions, and the allowable post-surge drift depends on how much protection margin the system retains, so it must be defined jointly by the end manufacturer and the component vendor as part of the system safety assessment.

Q5: What surge convention does the AMSEMI 5.0SMDJ series use, and can it be compared directly with Littelfuse data?

The AMSEMI 5.0SMDJ series (datasheets 8060014 standard and 8060055 foldback -N) gives 10/1000us figures only, with no 8/20us data. The 8x20us Ipp figures Littelfuse publishes are that vendor's own convention; the two waveform families must not be converted or subtracted, and cross-convention comparison should use relative quantities such as percentage reduction with test conditions labelled on every row. The foldback -N 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 - at the 54 V grade roughly 34.4 / 57.5 / 91.8 A under the 10/1000us convention.

Contact us

The productive way to handle "how many surges can a TVS take" is not to chase a number that does not exist, but to nail down three things: whether your surges are event-type or dense-type, what fraction of the derated rating each pulse carries, and where the scrap lines for the three monitored parameters sit. Two passive situations show up most often in the field: a TVS whose standby leakage has visibly risen after two years in service, with no incoming baseline archived, so nobody can tell device aging from a system fault; and an incoming flow that only checks whether VBR falls inside the datasheet window, so batch-level aging signals - a widening or shifting distribution - are missed entirely.

Send us your surge profile (waveform, peak, frequency, duty cycle, temperature), the current TVS part numbers, and your incoming-inspection and sampling flow, and we will help you close the lifetime-management loop in one pass: the energy-margin calculation, window recommendations for incoming inspection, a scrap-line discussion for post-sampling re-tests (with the FAE link arranged), and the event-type versus dense-type classification of your application. AMSEMI 5.0SMDJ standard and foldback -N (3000 W / 5000 W / 8000 W grades) datasheets, samples and FAE review are available - please contact us.

Shenzhen Intek Technology Co., Ltd - authorized distributor of AMSEMI for South China, under a certificate renewed through 2028-12-01 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. No universal standard or universal scrap threshold exists for surge-withstand counts; the energy-margin, acceptance-window and re-test flows in this article are a method framework, and concrete thresholds must be defined jointly by the end manufacturer and the component vendor as part of the system safety assessment - this article promises no specific percentage or count. 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 54 V grade VBR window upper limit of 66.3 V, the 5 uA leakage (54-95 V grades, 25C) and the roughly 34.4 / 57.5 / 91.8 A Ipp of the three -N grades at 54 V are quoted from the manufacturer datasheet convention; the latest manufacturer datasheet governs. Statements about the aging feedback loop, leakage rising with temperature, and about 3000 breakdown events per minute in a 1500 V AC test are order-of-magnitude descriptions and theoretical counts; actual values follow the manufacturer reliability reports and system measurement.