Can You Parallel TVS Diodes? How to Scale Surge Capability and Raise Standoff Voltage
One-line takeaway: paralleling several TVS parts does not split surge current by count — the I-V slope in the breakdown region is extremely steep, so the lowest-VBR die conducts first and swallows almost all of the current. A spread of a few percent in VBR is enough to destroy the assumption that "two in parallel equals twice the capability", so do not use paralleling to scale up. To raise the surge energy margin, move up a power tier at the same voltage grade: in the AMSEMI -N foldback series the 3000 W, 5000 W and 8000 W tiers have identical VBR and identical maximum clamping voltage Vc at any given grade and differ only in Ipp, so an upgrade needs no board change and no re-calculation of downstream margin. A series string does raise the voltage, but the clamp rises with it and you must additionally solve voltage sharing and leakage matching — in most cases it is better to pick a part with a higher VRWM grade.
1. Two frequently asked questions
"If I put two TVS parts in parallel, do I get double the surge capability?" No. The real capability after paralleling is decided by the lowest-VBR part, not by the count. As the surge edge arrives, the voltage reaches the breakdown point of the lower-VBR die, which conducts first and clamps the voltage; the other die, with a higher VBR, has not entered breakdown at all. Almost all the energy lands on the first die. After one hit the usual outcome is "one shorts first, the other follows".
"If I put two TVS parts in series, does the withstand voltage double?" The voltage does add up: two identical parts in series give roughly twice the breakdown voltage, so you can in principle build a high-voltage grade from two low-voltage parts. But the cost has to be counted too: the clamping voltage roughly doubles as well, so the residual voltage seen by the downstream MOSFET rises, and there is a steady-state voltage-sharing and leakage-matching problem between the two parts, which can leave one of them permanently over-biased and ageing early. The industry practice is therefore to select a part with a higher working-voltage grade rather than to build one by series connection.
2. Why paralleling does not scale
The defining property of a TVS is that once it breaks down, its I-V curve is extremely steep — the dynamic resistance in avalanche is very low, so a small increase in voltage produces a large increase in current. That is what decides how current divides between paralleled parts:
| Source of imbalance | Mechanism | Consequence |
|---|---|---|
| Initial VBR spread | Production tolerance within one part number and grade (typically a few percent, see the datasheet) | The lowest-VBR die breaks down first and takes almost the whole surge edge |
| Mismatched dynamic resistance | Datasheets normally give Vc only at the calibrated Ipp point, not the dynamic resistance Rd | Current sharing at high current cannot be predicted beforehand |
| Asymmetric traces and joints | Different copper length and via count to the protected node and to ground means different parasitic inductance | The lower-inductance die acts first, runs hotter and fails first |
On "thermal compensation": avalanche-type devices have a positive breakdown-voltage temperature coefficient, so a die that heats up after conduction sees its VBR rise slightly, which gives it a little self-balancing tendency. But heat spreading and junction heating are millisecond-scale processes while a surge lasts microseconds — a single fast surge is over long before that correction can act.
Conclusion: paralleling doubles the part count, not the current path. Adding a part only raises the nominal power sum; what the pair can really take is still set by the weakest die.
3. When paralleling is borderline acceptable
Three situations can be discussed, but none of them is a normal design method:
| Situation | Explanation | Boundary |
|---|---|---|
| The manufacturer explicitly supports it | The datasheet documents a parallel arrangement, or the device is a multi-die part matched at the factory | Manufacturer-endorsed use is not the same thing as a customer bolting two parts together |
| Same part number, same lot, measured pairing | Pair parts whose measured VBR values are close, with symmetric layout | It only improves matters; it does not guarantee them, and re-pairing is needed after lot changes |
| Multi-channel protection | One multi-channel TVS or ESD array protects several ports independently | That is one part covering several lines, not several parts on one node |
If you are already paralleling, do at least three things: use one part number from one lot; pair by measured VBR (the smaller the difference, the better); keep the traces and pads of the two parts left-right symmetric with equal length and equal via count. Then derate the nominal capability to a level clearly below the sum of the two.
4. Three accounts to settle when connecting in series
| Account | Where the problem is | What to do |
|---|---|---|
| Clamping | Total Vc is roughly the sum of the two, so the downstream residual voltage doubles | After series connection, re-check the clamping gate of the three-stage voltage window; the MOSFET rating may need to go up a grade |
| Voltage sharing | The part with lower leakage sees a higher steady-state share and may sit over-biased or even enter breakdown | Add sharing resistors to pin the division down, but account for the continuous leakage and dissipation they bring |
| Failure | If one part shorts, the other must withstand the whole voltage alone | Keep the string as short as possible; two parts is already the practical limit |
One exception worth noting: junction capacitance falls when parts are in series (series-capacitance formula), so there is a historical practice of putting two low-capacitance TVS parts in series to cut Cj — at the price of a higher clamp and a weaker protection level. Also note that a bidirectional TVS is already two junctions back to back, so do not mistake "one bidirectional part" for "a part that still needs another in series".
5. The correct paths for scaling and raising voltage
Take "not enough" apart and the answer depends entirely on which item is short:
| What you want | Common wrong move | Recommended move |
|---|---|---|
| More surge current capability | Parallel another part of the same type | Move up a power tier at the same voltage grade (3000 W → 5000 W → 8000 W), or go to a larger package |
| Higher working voltage VRWM | Put two lower-grade parts in series | Pick a part with a higher VRWM grade directly; on a DC rail also evaluate a metal-oxide varistor (MOV) |
| Lower clamping voltage | Parallel more parts | Move to a foldback (snap-back) low-clamping part; about 20 percent lower Vc at the same grade |
| Withstand a larger system-level surge | Let a single TVS stage take it all | Staged protection: a coarse front stage (GDT / MOV) + decoupling + a fine-clamping TVS in the rear |
| Protect several signal lines | Parallel several parts onto one node | Use a multi-channel TVS or ESD array, one channel per line |
flowchart TD
A["Protection not enough: energy, standoff voltage or clamp"] --> B{"Which item is short?"}
B -- "Surge energy" --> C["Do not parallel: move up a power tier\n3kW to 5kW to 8kW, or a larger package"]
B -- "Working voltage VRWM" --> D["Do not use series: pick a higher VRWM grade\nOn a DC rail evaluate an MOV"]
B -- "Clamping voltage too high" --> E["Move to a foldback low-clamping part\nabout 20 percent lower Vc at the same grade"]
B -- "System-level energy too large" --> F["Staged protection: coarse GDT / MOV front stage\nthen decoupling, then a fine-clamping TVS"]
C --> G["One part per node\nscaling never by adding parallel parts"]
D --> G
E --> H["Re-check the three-stage window\nVRWM / VBR / Vc"]
F --> H6. Scaling without paralleling: the three AMSEMI -N power tiers
The AMSEMI -N shallow-foldback (also called snap-back or low-clamping) series is available in 3000 W, 5000 W and 8000 W tiers, and the key rule is that at any given voltage grade the three tiers share exactly the same VBR range and the same maximum clamping voltage Vc — only Ipp rises. Take the 54 V grade, the most common in BMS work:
| Tier and 54 V part | VBR min–max (V) | Vc max (V) | Ipp (A, 10/1000us) |
|---|---|---|---|
| 3000 W · SMDJ54CA-N | 60.0–66.3 | 69.6 | 34.4 |
| 5000 W · 5.0SMDJ54CA-N | 60.0–66.3 | 69.6 | 57.5 |
| 8000 W · 8.0SMDJ54CA-N | 60.0–66.3 | 69.6 | 91.8 |
What this means: going from 5000 W to 8000 W leaves the package (DO-214AB/SMC), the VRWM, the VBR range and the clamping voltage completely unchanged (69.6 V in both cases) and only makes the device harder to destroy. So for "not enough energy", moving up a tier is simpler and far more predictable than paralleling — no board change, and no need to re-check the downstream MOSFET margin.
Ipp for the two documented tiers across the common voltage grades (Vc and Ipp are both on a 10/1000us convention, taken from AMSEMI documents 8060084 and 8060055):
| VRWM grade (V) | Vc max (V) | Ipp: 3000 W tier (A) | Ipp: 5000 W tier (A) |
|---|---|---|---|
| 54 | 69.6 | 34.4 | 57.5 |
| 58 | 74.8 | 32.1 | 53.5 |
| 64 | 82.4 | 29.1 | 48.6 |
| 75 | 96.8 | 24.8 | 41.4 |
| 85 | 110.0 | 21.9 | 36.5 |
| 90 | 116.8 | 20.5 | 34.3 |
Convention reminder: for the AMSEMI 5.0SMDJ series (document 8060014 standard, 8060055 for the -N foldback) the Vc and Ipp figures are exclusively 10/1000us; there is no 8/20us data in the datasheet. Littelfuse's published 8x20us figures are Littelfuse's own convention for Littelfuse parts and must be attributed to Littelfuse whenever quoted. The same part shows completely different numbers under the two waveforms, so subtracting one from the other or converting between them is forbidden.
Data note: the 3000 W and 5000 W rows above come from documents 8060084 and 8060055. The 8000 W tier (8.0SMDJ**CA-N, VRWM 48–90 V, 11 bidirectional parts, AEC-Q101, automotive order codes with the -H suffix) figures are taken from the manufacturer's published material and should be confirmed against the latest original datasheet. Full models and grades: AMSEMI Foldback TVS: Full Series and Cross-Reference.
7. Common pitfalls
| Pitfall | Symptom | Correct approach |
|---|---|---|
| Treating two paralleled parts as "double capability" | After one surge one part shorts and the other soon follows | Move up a power tier or use a larger package instead of paralleling |
| Paralleling different voltage grades | The high-grade part never acts and the low-grade part takes all the energy | One part per node, with the grade chosen from VRWM |
| Raising voltage in series and forgetting the window | The clamp doubles and the downstream MOSFET blows again | After a series string, re-check both the VBR and the Vc gates |
| Series string with no voltage sharing | The lower-leakage part sits over-biased and ages early | Add sharing resistors and account for their continuous leakage and dissipation |
| Asymmetric traces on a parallel pair | The part with the shorter copper blows first | If paralleling is unavoidable, keep the pair symmetric, equal length, equal vias and matched by lot |
| Using "a bigger part" instead of doing the numbers | Assuming a higher power tier clamps lower | At one grade the three tiers share identical VBR and Vc, so the clamp still has to be checked |
Related articles
- How to read the power rating and derating (what 5000 W really means): What Does a 5000 W TVS Rating Actually Mean?
- Full series models and grades: AMSEMI Foldback TVS: Full Series and Cross-Reference
- Foldback versus standard TVS: Foldback TVS Cross-Reference: AMSEMI vs Littelfuse vs Vishay
- Bus-level selection overview: How to Select TVS for BMS Protection Boards
- How to check the three gates: TVS Fitted, MOSFETs Still Blow: The 23S / 24S Voltage Budget
- Diverting system-level energy: Surge Protection for Energy Storage / BMS
- How to verify on the board: How Do You Verify the TVS on a BMS Protection Board?
- Telling a genuinely low-clamping part apart: Which Chinese Brands Make Low-Clamping TVS?
FAQ
Q1: Can TVS diodes be used in parallel? Do two parts share the surge current?
Not recommended. The I-V curve of a TVS in breakdown is extremely steep, so the die with the lowest VBR conducts first and clamps the voltage while the higher-VBR part may not have entered breakdown at all, leaving almost all the energy on the first die. In other words, the capability of a parallel pair is set by the lowest-VBR part, not by the count. To raise the energy margin, move up a power tier at the same voltage grade (3000 W to 5000 W to 8000 W) or use a larger package.
Q2: Does a series string of two TVS parts double the withstand voltage, and what must be checked?
The voltage does add up: two identical parts in series give roughly twice the breakdown voltage, effectively building a high-voltage grade from two low-voltage parts. But three accounts must be settled at the same time — clamping (total Vc is roughly the sum, so the residual voltage doubles and the voltage window must be re-checked), voltage sharing (the lower-leakage part carries a higher steady-state share, needing sharing resistors that bring continuous leakage and dissipation), and failure (if one shorts, the other must take the whole voltage). In most cases it is better to simply choose a part with a higher VRWM grade.
Q3: If the surge energy is not enough, should I parallel more parts or move to a bigger power rating?
Move up a power tier. In the AMSEMI -N foldback series the 3000 W, 5000 W and 8000 W tiers have identical VBR and identical maximum clamping voltage Vc at any given grade and differ only in Ipp (54 V grade: 34.4 A / 57.5 A / 91.8 A, 10/1000us convention). A higher tier means "harder to destroy" — no board change and no margin re-calculation — whereas paralleling neither shares current by count nor removes the spread, and multiplies part count, layout difficulty and risk. If the system-level energy is simply too large, the right path is staged protection: a coarse GDT / MOV front stage, decoupling, then a fine-clamping TVS.
Q4: What is the difference between the 3000 W, 5000 W and 8000 W tiers at the same voltage grade?
Only the surge current capability. All three share the same VBR range and the same maximum clamping voltage Vc at a given grade, in the same DO-214AB/SMC package; only Ipp differs — 54 V grade 34.4 A / 57.5 A / 91.8 A, 85 V grade 21.9 A / 36.5 A (10/1000us convention; the 3000 W and 5000 W figures come from documents 8060084 and 8060055). The order is to fix the voltage grade from the three-stage voltage window first, then pick the power tier from the surge current you need to block. If the window itself does not close, a higher power tier will not help — change to a lower-clamping foldback -N part or raise the downstream MOSFET rating first.
Contact us
Parallel or upgrade, series or re-grade — it is the same question underneath: work out the voltage window and the energy budget first, then decide which part and how many. Adding an extra part by feel tends to cost more material while turning "one part blows" into a chain failure.
Send us your cell count and maximum working voltage, the surge waveform and current class you need to block (or your specification test conditions), and the downstream MOSFET part and rating. We will work the window and the energy budget through with you and give a definite answer on which tier to use, whether staged protection is needed and whether a board change is required; we can also run bench measurements to your specification waveform and issue a surge test report. Datasheet pages and samples for the AMSEMI 5.0SMDJ standard series and the -N foldback (snap-back) series, plus FAE review, are available on request. Contact us.
Shenzhen Intek Technology Co., Ltd — authorized distributor of AMSEMI (Anhui Anmei Semiconductor) TVS and HOPERF CMT series digital isolators Tel / WeChat: +86 136-3264-8484 Address: Room 4F, Building 2, Jingwei Center, No. 309 Ping'an Avenue, Pinghu Street, Longgang District, Shenzhen, China Website: www.intek.vip
Disclaimer: This article is technical and industry information for reference only and does not constitute selection, procurement or any other business advice. Device parameters (clamping voltage, breakdown voltage, peak pulse current, leakage current and so on), waveform conventions and derating ratios are subject to the latest datasheet from each manufacturer and to whole-system measurements. The AMSEMI 5.0SMDJ series (documents 8060014 / 8060055) quotes Vc and Ipp on a 10/1000us convention with no 8/20us data; Littelfuse 8x20us figures are Littelfuse's own convention; the 8000 W tier (8.0SMDJ**CA-N) figures come from the manufacturer's published material and should be confirmed against the latest original datasheet. Whether a parallel or series arrangement is usable must be judged against the original documentation and measurements for the specific part, and work involving high voltage and surge testing must be carried out by suitably qualified personnel.