Does a 48V DC Bus Need a TVS? Surge Protection and Low-Clamping TVS Selection for Server and Telecom Power
One-line conclusion: yes, but in the right place - a 48V bus TVS is not a lightning protector; it is the last clamp for residual transients on the DC bus. The judgement comes down to two sentences: 1) do not pick the grade from the nominal 48V figure - VRWM must be derived from the maximum steady-state bus voltage (tolerance, float charge, plug-in overshoot) times 1.15 to 1.2; a 48V bus commonly reaches 52 to 54V steady state, so start at the 64V grade. 2) the grade is set by the downstream device rating - the clamping ceiling is 0.85 times that rating, and with a 48V bus feeding 100V-class MOSFETs or Schottky rectifiers the standard 5.0SMDJ 64V grade clamps at 103.0V, already past the 85V window, so only a foldback (snap-back / low-clamping) TVS - the AMSEMI -N series at 82.4V on the same grade - fits. In other words, the hard part on a 48V bus is not whether to fit a TVS, but that a standard TVS clamp window simply does not line up with 100V-class parts.
1. Two frequently asked questions
"It is only 48V, nowhere near high voltage - does it really need a TVS?"
Yes, but not because 48V is high. A 48V bus carries three kinds of transient that reach the downstream devices: plug-in and hot-swap - inserting a board or a load into a live bus produces an L di/dt kick from the loop and cable inductance, so the bus sees a spike far above steady state; cable-coupled surge - the AC/DC front end, remote-feed long runs and cabinet wiring all couple lightning-induced and switching energy onto the bus; and load-side disturbance - load dump, recovery after a short circuit and DCDC input capacitor charging all produce high-di/dt overshoot on the bus.
Equally important is the boundary: a bus TVS handles residual transients that have already arrived on the DC bus. The main lightning energy belongs to the AC input front stage (GDT / MOV coarse diversion), and an SMC-package TVS cannot substitute for that. The full staged logic is set out in Surge Protection for Energy Storage / BMS.
"The 48V bus looks like it has plenty of margin - a 100V MOSFET or rectifier should be comfortable?"
The opposite is true. This is the most under-estimated window conflict in 48V-class power supplies. The bus peaks near 54V, so a 1.2 factor lands you on the 64V grade; once VRWM is 64V, a standard TVS clamps at 103.0V maximum on the same grade (10/1000us convention), while a 100V-class part at 0.85 derating allows only 85V: 103.0V > 85V, and the first window fails outright. This is not a "pick a bigger one" problem - going up a grade only raises the clamp, and going down will not hold off the maximum bus voltage. That is exactly why 48V and -48V buses are a prime battleground for low-clamping foldback TVS: the same grade pulls Vc down by roughly 20 percent (64V grade 103.0V to 82.4V) and the window is valid again.
2. Where the bus transient comes from, and who owns it
Separate the responsibilities first, so AC-side and DC-side duties are not mixed up:
| Transient source | Typical trigger | Primary protection | Role of the bus TVS |
|---|---|---|---|
| Plug-in / hot-swap kick | Board or load insertion, connector bounce | Bus TVS plus soft-start or current limit | Primary (clamps the L di/dt peak) |
| Cable-coupled surge | Lightning induction, remote-feed long runs, cabinet wiring | Entry GDT / MOV plus decoupling | Secondary fine clamp |
| Load-side disturbance | Load dump, short-circuit recovery, DCDC input charging | Bus TVS plus local RC snubber | Primary |
The table is a division of labour, not a parts list: where energy is high or runs are long, stage it as coarse front stage, decoupling, then fine clamp. A single bus TVS suits in-board transients of controlled energy.
Why energy forces staging: a TVS power rating is a single-pulse convention and its continuous capability is only a few watts, further derated by pulse width and temperature - see What Does a 5000 W TVS Rating Actually Mean?. Handing system-level lightning energy to one SMC part destroys the part and leaves the downstream device exposed anyway.
3. Choosing the grade: VRWM does not follow the nominal 48
This is the habit worth correcting: few engineers would deliberately pick a 48V grade for a 48V bus, but some do reason that way. The AMSEMI 8.0SMDJ foldback (8000 W) range covers VRWM 48V to 90V in 11 grades, and one of them is VRWM 48V - on a 48V bus that puts the standoff voltage level with the bus steady state, so any bus excursion enters breakdown; never use it. The rule is:
VRWM >= maximum steady-state bus voltage x 1.15 to 1.2
"Maximum steady-state bus voltage" comes from the equipment specification, not the nominal figure - it includes tolerance, the float-charge point and plug-in overshoot. For the -48V telecom convention, use the top of the float range (-48V systems commonly reach 55 to 57V steady state; confirm against the system specification), otherwise the grade ends up too low. The table below puts typical 48V-class situations next to the clamping figures for both versions:
| Max steady-state bus voltage | Suggested VRWM grade | Standard 5.0SMDJ: Vc max | AMSEMI foldback -N: Vc max |
|---|---|---|---|
| about 52-54V (48V bus, +10 to 12 percent) | 64V | 103.0V | 82.4V |
| about 54-57V (top of -48V telecom float range) | 75V | 121.0V | 96.8V |
| 60V and above (wide-input, battery or supercap direct bus) | 85V | 137.0V | 110.0V |
Convention warning: every Vc above is on the 10/1000us convention, taken from AMSEMI documents 8060014 (standard 5.0SMDJ) and 8060055 (foldback -N), and those documents contain no 8/20us data. Littelfuse publishes 8x20us figures for its own parts (for example 36.5 A at 10x1000us and 273.8 A at 8x20us for the same 5.0SMDJ85CA), and those must be attributed to Littelfuse explicitly - the two conventions must never be mixed or subtracted. Within one voltage grade, the 3000 W, 5000 W and 8000 W tiers share identical VBR and maximum Vc and differ only in Ipp, so a higher power tier cannot be used to lower the clamp. Grade choice, margin and derating remain subject to the latest manufacturer datasheet and to measurements on the assembled unit.
4. Back-calculation: the downstream rating decides which grade you can use
Choosing the grade is only half the job - whether you can use it depends on the downstream device rating. The engineering check is Vc <= 0.85 x downstream device rating (automotive-adjacent projects sometimes tighten to 0.8; follow the project specification):
| Downstream device rating (nominal minimum convention) | Vc ceiling = 0.85 x rating | Usable grade (AMSEMI foldback -N) | Can the standard 5.0SMDJ on that grade pass? |
|---|---|---|---|
| 80V | 68V | 54V grade 69.6V (slightly over; re-check at the real current point) | No (87.1V) |
| 100V | 85V | 64V grade 82.4V fits | No (103.0V) |
| 150V | 127.5V | 75V grade 96.8V | Marginal (121.0V, about 5 percent margin) |
| 200V | 170V | 85V grade 110.0V / 90V grade 116.8V | Yes (137.0V / 146.0V) |
Two points matter. First, ratings come in conventions: the nominal minimum and the measured typical for a MOSFET or rectifier differ, with the measured typical often about 10 percent higher, so 0.85 against the nominal minimum is the conservative route and a measured value buys more margin. Second, Vc is not a constant: the datasheet figure is one point at the calibrated Ipp, and clamping rises along the dynamic resistance with current, so check at the actual surge current rather than comparing datasheet Vc directly.
This table also answers a familiar confusion: "I swapped the 5000 W TVS for an 8000 W part - why does the downstream device still fail?" Because VBR and maximum Vc are identical within a voltage grade; at the 54V grade the three tiers carry 34.4 A / 57.5 A / 91.8 A of Ipp, so a higher power tier only makes the TVS harder to destroy - the residual voltage at the downstream device does not move. Only two routes lower the residual: a lower-clamping part, or a higher downstream rating. The voltage-budget version of this argument is in TVS Fitted, MOSFETs Still Blow: The 23S / 24S Voltage Budget, and the part-number version in AMSEMI Foldback TVS: Full Series and Cross-Reference.
5. The selection path for a 48V bus TVS
Chained together, the order of work looks like this:
flowchart TD
A["48V / -48V DC bus<br/>Does it need a TVS?"] --> B["Fix the max steady-state bus voltage<br/>tolerance + float + plug-in overshoot"]
B --> C["Pick the VRWM grade at x1.15~1.2<br/>e.g. 54V bus -> 64V grade"]
C --> D["Set the window from the downstream rating<br/>Vc ceiling = 0.85 x rating"]
D --> E{"Does the standard TVS clamp<br/>fit the window?"}
E -- "No" --> F["Move to foldback -N on the same grade<br/>about 20 percent lower Vc"]
E -- "Yes" --> G["Pick the power tier from surge current<br/>3 kW / 5 kW / 8 kW"]
F --> G
G --> H{"High transient energy or long cable runs?"}
H -- "Yes" --> I["Coarse GDT / MOV front stage<br/>+ decoupling + bus TVS fine clamp"]
H -- "No" --> J["Bus TVS alone"]
I --> K["Verify on the board: measure at the downstream<br/>device pins, not across the TVS"]
J --> KThe last step is the one most often done wrong: do not declare the design good after measuring the clamp across the TVS. The L di/dt drop along the trace adds on top of the TVS clamp, so the downstream device pin sees "TVS clamp plus trace-induced drop", and the two can differ by a meaningful fraction. The method and pass criteria are in How to Verify the TVS on a BMS Protection Board - that article uses BMS as its example, but "measure at the protected device" applies just as much to server and telecom power on a 48V bus.
6. Unidirectional or bidirectional?
The criterion is still just one: can the protected node ever see a voltage of the reverse polarity for the TVS?
| Bus form | Common choice | Reason |
|---|---|---|
| +48V single-polarity bus to ground | Unidirectional (A) can be evaluated | The node never reverse-biases, saving the roughly 1V forward-conduction clamp window |
| -48V telecom convention (positive ground) | Bidirectional (CA) preferred | Wiring polarity is easily misread and float plus plug-in events can disturb both directions |
| Reverse connection, bidirectional transients, AC content | Bidirectional (CA) required | A unidirectional part conducts like a forward diode in reverse, clipping or shorting |
One practical note: the main AMSEMI -N foldback part numbers that Intek supplies are CA bidirectional types, which is why CA parts dominate on BMS and storage buses - not because "bidirectional is better", but because the foldback low-clamping route is built on bidirectional dies. The decision logic and structural differences are in Unidirectional or Bidirectional TVS? (Chinese).
7. Six common pitfalls
| Pitfall | Symptom | Correct approach |
|---|---|---|
| Picking the 48V grade for a 48V bus | Any bus excursion enters breakdown; the part runs hot or fails | Derive the grade from max steady state x 1.15 to 1.2 |
| Protecting only the AC entry | Bus transients still kill the downstream DCDC | Add a DC bus TVS and stage it with the front end |
| Assuming "48V leaves plenty of margin" | Standard 64V grade clamps at 103.0V, past a 100V device window | Compute Vc <= 0.85 x rating before choosing a grade |
| Expecting a power-tier upgrade to lower the clamp | An 8000 W part changes nothing at the downstream device | VBR and Vc are grade-determined; lower the clamp with a low-clamping part |
| Measuring the clamp across the TVS | Trace inductance is missed and the real peak is higher | Measure at the downstream device pins or drain-source |
| Ignoring leakage on a continuously biased node | Leakage rises with temperature and breaks the standby or thermal budget | Check the IR limit and HTRB behaviour, see the leakage article |
That last row matters especially in server and telecom power: the bus TVS is energised 8760 hours a year. The AMSEMI 5.0SMDJ standard and -N foldback series are both specified at 5 uA maximum leakage at the 54 to 95 V grades (25 C, at VRWM, maximum-value convention), negligible at room temperature, but leakage roughly doubles per 10 C, so a hot cabinet needs its own standby and thermal accounting - and on the same grade the foldback and standard versions share the same IR limit, because low clamping does not mean low leakage. The full arithmetic is in Does the TVS on a BMS Leak Enough to Drain the Battery?.
8. Four hard conclusions for 48V bus protection
- Derive the grade, do not assume it: VRWM from maximum steady-state bus voltage x 1.15 to 1.2.
- The window is set by the downstream part: Vc <= 0.85 x its rating (nominal minimum is conservative, measured typical gives more room). Without that number, do not freeze the part.
- 48V with 100V-class parts is the classic foldback win: the 64V grade moves from 103.0V to 82.4V and the window goes from invalid to valid, so the downstream part need not be pushed to 150V just to suit the TVS.
- High energy means staging: long cable runs, remote feeds and cabinet environments want coarse GDT / MOV, decoupling, then a bus TVS fine clamp - the bus TVS is not the lightning path.
Related solutions
- Bus-level selection overview: How to Select TVS for BMS Protection Boards
- Foldback principle and imported benchmarks: Foldback TVS vs Standard TVS
- Full series and 54 to 100V grade cross-reference: AMSEMI Foldback TVS: Full Series and Cross-Reference
- Which Chinese suppliers make low-clamping parts: Which Chinese Brands Make Low-Clamping TVS?
- How to work the voltage window: TVS Fitted, MOSFETs Still Blow: The 23S / 24S Voltage Budget
- Reading peak pulse power and derating: What Does a 5000 W TVS Rating Actually Mean?
- Bench verification method: How to Verify the TVS on a BMS Protection Board
- Leakage and thermal behaviour on biased nodes: Does the TVS on a BMS Leak Enough to Drain the Battery?
- Unidirectional versus bidirectional: Unidirectional or Bidirectional TVS? (Chinese)
- Staged protection and test levels: Surge Protection for Energy Storage / BMS
- TVS fundamentals: What Is a TVS Diode?
- Protecting the comms port beside the bus: Why Do RS485 and CAN Ports Keep Failing?
- Two-stage method with isolated interfaces: Do You Still Need a TVS Next to an Isolator?
FAQ
Q1: Does a 48V DC bus need a TVS, and is AC-side surge protection enough on its own?
No. A 48V bus needs a TVS not because the voltage is high but because three transients reach it: plug-in and hot-swap kicks from board or load insertion (L di/dt), cable-coupled surge from the AC/DC front end, remote feeds and cabinet wiring, and load-side disturbance such as load dump or recovery after a short circuit. AC-side protection (GDT / MOV) handles the main lightning energy as an entry-level coarse stage; the bus TVS handles transients already present on the DC bus and is the last clamp before the downstream devices. The two are staged and not interchangeable, and long cable runs or remote feeds need both.
Q2: Which voltage grade should a 48V bus use? Can I just pick the 48V grade?
No. The AMSEMI 8.0SMDJ foldback range does include a VRWM 48V grade, but on a 48V bus it puts the standoff voltage level with the bus steady state, so any excursion enters breakdown. Derive the grade from the maximum steady-state bus voltage (tolerance, float charge point, plug-in overshoot) times 1.15 to 1.2: with a 48V bus peaking near 52 to 54V, use the 64V grade; for a -48V telecom supply whose steady state reaches 55 to 57V, a 70V or 75V grade is the usual choice. Confirm against the system specification.
Q3: A TVS is fitted on the 48V bus and the 100V MOSFETs still fail - why?
Check the window first, then the part. The standard 5.0SMDJ 64V grade clamps at 103.0V maximum (10/1000us convention), while a 100V-class part at 0.85 derating allows only 85V, so the first window fails and an over-voltage at the downstream device is inevitable. The AMSEMI -N foldback on the same grade clamps at 82.4V and the window is valid again. Also remember Vc is not a constant - it rises along the dynamic resistance with current, so check at the real surge current and measure at the downstream device pins rather than across the TVS.
Q4: If I upgrade from 5000 W to 8000 W, does the clamp drop?
No. Within one voltage grade the AMSEMI -N foldback 3000 W, 5000 W and 8000 W tiers share identical VBR range and maximum Vc, differing only in surge current: at the 54V grade Ipp is 34.4 A / 57.5 A / 91.8 A. A higher tier makes the TVS harder to destroy without changing the residual seen by the downstream device, and needs no board change or margin re-calculation. To lower the residual, move to a lower-clamping part (for example from a standard 5.0SMDJ to the -N foldback, about 20 percent lower on the same grade) or raise the downstream rating.
Contact us
48V bus protection looks simple, but the real sticking point is the conflict between the voltage grade and the downstream rating: choose the grade from the nominal voltage and the part sits near breakdown, choose the clamp from the downstream rating and the standard TVS window does not close. Answering both needs three inputs: the maximum steady-state bus voltage (including tolerance and the float point), the downstream device type and rating, and the order of magnitude of transient energy at that node (plug-in? long cable run? lightning induction?).
Send those three and we will work the voltage window, the power tier and the staging through with you, and give a definite answer on which grade to use, whether to move from a standard TVS to a foldback low-clamping part, and whether an entry-level coarse stage is needed. Datasheet pages, samples and FAE review for the AMSEMI 5.0SMDJ standard series and the -N foldback (snap-back) series 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 Vc, breakdown voltage VBR, peak pulse current Ipp, reverse leakage IR and so on) and waveform conventions are subject to the latest datasheet from each manufacturer and to whole-system measurements. The AMSEMI 5.0SMDJ series (documents 8060014 standard and 8060055 foldback -N) quotes Vc and Ipp on a 10/1000us convention with no 8/20us data; Littelfuse 8x20us figures are Littelfuse own convention for its own parts. Maximum steady-state bus voltage, the 0.85 / 0.8 derating factors and topology or power-tier choices are engineering practice and orders of magnitude rather than mandatory standards; real designs must be based on the system specification, manufacturer data, your own circuit calculations and measurements on the assembled unit. Work involving high voltage or mains must be carried out by suitably qualified personnel.