Why Do RS485 and CAN Ports Keep Failing? Signal-Port TVS Selection and the Capacitance Trade-Off
Bottom line: when a communication port keeps dying, changing the chip rarely helps; what needs changing is the protection structure. Signal-port protection logic is completely different from a power bus. Power ports are "large energy, few events"; signal ports are "low energy, many events, plus permanent common-mode." The selection order is: identify the threat source first (long-cable coupled surge, hot-plug, ground potential difference), then select the device in four steps — VRWM covers the common-mode swing, VC sits below the transceiver tolerance, junction capacitance matches the baud rate, and bidirectional or unidirectional follows the signal shape — and finally deliver the protection through layout (close to the connector) and front-stage coordination (GDT or PTC grading). The two most common failures in communication products are not "no TVS fitted" but too much junction capacitance wrecking the edges, or forgetting to check VC at the surge current.
flowchart TD
A["1. Identify the threat<br/>long-cable surge / hot-plug / ground offset"] --> B["2. Signal characteristics<br/>levels / speed / differential or single-ended"]
B --> C["3. Choose VRWM<br/>at or above max common-mode swing"]
C --> D["4. Check VC<br/>at or below transceiver bus pin rating"]
D --> E["5. Match junction capacitance Cj<br/>higher speed needs lower Cj"]
E --> F["6. Directionality<br/>differential bidirectional / grounded single-ended unidirectional"]
F --> G["7. Layout and front stage<br/>near connector + GDT / PTC grading"]
G --> H["8. Validate by measurement<br/>surge / common mode / edges and BER"]First, the two questions asked most often
Question: RS485 ports keep dying. Does a new transceiver help? Identify the threat source first. Port damage is usually three causes stacked together: common-mode surge and lightning induction coupled into long cables, live hot-plug transients, and ground potential differences between devices pushing common-mode voltage outside the chip range. A higher-voltage transceiver only lifts the tolerance ceiling slightly and cannot stop a surge an order of magnitude larger in energy. The correct approach is a signal-port TVS placed right at the port, with GDT or PTC grading in front when needed, plus a review of isolation and grounding.
Question: what junction capacitance suits a CAN bus TVS? Speed decides it. Classical CAN (250 kbps to 1 Mbps) is not capacitance sensitive and tens of picofarads usually works. CAN-FD at 2 to 5 Mbps has a short bit time, and the RC formed by the capacitance and the bus characteristic impedance slows the edges, so pick a device an order of magnitude lower. After selection, check the measured edges and bit error rate rather than the parameter table alone.
Step 1: work out where the threat comes from
The three threat classes need completely different handling, so separate them before selecting:
| Threat source | Typical scenario | Character | Protection approach |
|---|---|---|---|
| Long-cable coupled surge / lightning induction | Parallel RS485 bus in residential storage, outdoor CAN trunk | Large energy, few events, common-mode dominant | TVS plus GDT or PTC front-stage grading |
| Live hot-plug | Module insertion and removal, maintenance | Small energy, many events, steep spike | Local TVS clamping, with plug-cycle life in mind |
| Ground potential difference / permanent common mode | Distributed multi-device networks | Not a transient, a continuous over-range condition | Isolation plus single-point grounding, TVS as backstop |
The third class is the one most often missed: if a ground potential difference holds the common-mode voltage outside the chip range permanently, the TVS has not failed — the chip is being worn out by continuous over-range operation. That case needs galvanic isolation and equipotential bonding; piling on protection devices treats the symptom. System-level graded surge design (test levels, waveform convention, front and rear stage coordination) is covered separately in How to Design Surge Protection for Storage and BMS Products.
Step 2: four parameters set the device
Signal-port TVS uses the same parameter set as power TVS (VRWM, VBR, VC, IPP), but the checks differ:
- VRWM (reverse working peak voltage) at or above the maximum signal common-mode swing. Do not look at the differential level alone: RS485 transceivers typically tolerate a certain common-mode range (commonly on the order of minus 7 V to plus 12 V, per the specific chip datasheet), and the static common mode on a 24 V system CAN line is higher than on 12 V. Too low a VRWM leaves the TVS slightly conducting, heating and leaking; too high raises VC with it and weakens protection.
- VC (clamping voltage) at or below the transceiver bus pin tolerance, checked at the surge current. The VC in the parameter table corresponds to a specific test current; at a larger real surge current the residual voltage is higher, so VC must be checked against the expected surge current level from the VC curve or confirmed with the manufacturer. Missing this step is the number one cause of "TVS fitted and the chip still dies."
- Junction capacitance Cj matched to speed. The TVS sits across the bus, and its capacitance with the bus characteristic impedance forms an RC low pass that eats edge steepness directly. The higher the speed, the lower the allowed capacitance; see the table below for orders of magnitude.
- Directionality follows signal shape. For differential pairs such as RS485 A and B or CAN H and L, the signal can swing either way with respect to ground or carry an uncertain common-mode offset, so bidirectional devices are usual (part numbers ending in C). Single-ended IO known to stay above ground, such as a 0 to 5 V enable or feedback line, can use a unidirectional device whose forward conduction at one PN junction drop clamps negative spikes lower. Both approaches are in use; the key is to confirm that the negative swing stays inside the absolute maximum rating of the transceiver.
Voltage grade orders for common interfaces
| Interface | Signal shape | TVS VRWM grade order | Notes |
|---|---|---|---|
| RS485 (A/B) | Symmetrical differential | Cover the transceiver common-mode tolerance (commonly within a little over ten volts) | Bidirectional; watch front-stage diversion on long trunks |
| CAN / CAN-FD (H/L) | Symmetrical differential | Set by system voltage; 12 V and 24 V systems differ | Bidirectional; check Cj carefully for CAN-FD |
| 12 V / 24 V industrial IO | Single-ended | 12 V or 24 V grade | Unidirectional acceptable for above-ground signals |
| On-board low-voltage digital | Single-ended | A few volts | Closer to ESD protection; pick a low-capacitance device |
The table gives grade orders only. Real voltage values must follow the transceiver and the actual system levels, per the manufacturer datasheet.
Junction capacitance versus speed
| Interface speed | Capacitance sensitivity | Cj order of magnitude | How to verify |
|---|---|---|---|
| RS485 up to 115.2 kbps | Low | Tens of picofarads usually fine | Normally no dedicated check needed |
| RS485 above 1 Mbps | Medium | Lower order recommended | Measure edges and bit error rate |
| CAN 250 kbps to 1 Mbps | Low to medium | Tens of picofarads usually fine | Watch bit sampling point margin |
| CAN-FD 2 to 5 Mbps | High | Push lower; use a low-capacitance series | Check rising edge, bit time and BER |
These are engineering orders of magnitude. The allowable capacitance ceiling depends on bus length, node count, characteristic impedance and transceiver drive strength, so there is no single number; the final answer is the measured edge and bit error rate. For speed-sensitive interfaces (CAN-FD, high-speed RS485, Profibus class) filter directly on low-capacitance series.
Step 3: layout and front-stage coordination decide whether protection is delivered
Correct parts with wrong layout still fail. Three hard rules:
- Put the TVS next to the connector. Trace inductance between the TVS and the connector adds an L times di/dt residual on a steep surge edge; placing the TVS beside the transceiver amounts to no protection. The TVS belongs at the connector entry, with a short and wide ground path.
- Grade where large surges are expected. At long cable entries and outdoor runs, one signal-port TVS alone (typically a few hundred watts under the 10/1000 microsecond convention, per the device specification) cannot absorb lightning-induced energy. Add a GDT or PTC in front so most energy diverts upstream and the TVS only does fine clamping. Decoupling for graded coordination is covered in Surge Protection.
- Protect both lines of a differential pair, and keep the common-mode path open. Fitting a single TVS on line A or line H only is the most common shortcut. Also confirm the TVS ground path is genuinely low impedance — without a clear downstream path for common-mode surge, the protection device is decorative.
Failure debug order (when a port keeps killing parts): look at what fails. If the TVS fails, the energy exceeded it (add front-stage diversion). If the transceiver fails, residual voltage or common mode exceeded limits (check VC, grounding and isolation). Then confirm the TVS is not reversed or the wrong part. Finally capture the surge and communication edges on a scope and check the common-mode voltage range and edge quality.
Common mistakes
| Mistake | Real consequence | Correct handling |
|---|---|---|
| Checking VRWM only, not VC at surge current | Residual voltage exceeds chip tolerance; chip dies with a TVS fitted | Check the VC curve at the expected surge current level |
| Chasing "smaller is better" on capacitance | Low-capacitance devices are often a grade weaker in power and clamping, and cost more | Match capacitance to speed, good enough is enough, confirm by edge measurement |
| Protecting only one line of a differential pair | The other line still admits surge, no common-mode path | Fit A/B or H/L as a pair, short wide ground path |
| Reusing a power-port TVS on a signal port | Large capacitance wrecks edges, communication unstable | Use a low-capacitance device intended for signal ports |
Related reading
- System-level surge grading (test levels, waveform convention, GDT / MOV / TVS coordination): How to Design Surge Protection for Storage and BMS Products
- Power-bus TVS selection (the three-stage VRWM / VBR / VC voltage window): How to Select a Low-Clamping TVS for a BMS Protection Board (Chinese)
- TVS fundamentals (principle, unidirectional versus bidirectional, parameter meanings): What Is a TVS Diode
- MCU resources and interfaces for communication and control: Which MCU for a Storage BMS? GD32 Selection Guide (Chinese)
- Part number quick reference (power-port series, VRWM 24 to 100 V): AMSEMI Foldback TVS Full Series and Cross-Reference
FAQ
Q1: RS485 ports keep getting destroyed. What protection works, and does changing the transceiver help?
Identify the threat source before changing anything. RS485 port failure usually comes from three causes: common-mode surge and lightning induction coupled into long cables, transients from live hot-plugging, and ground potential differences pushing common-mode voltage outside the chip range. A higher-voltage transceiver only raises the tolerance ceiling slightly and cannot stop a surge an order of magnitude larger in energy. The correct approach is a signal-port TVS placed close to the A and B lines, with GDT or PTC grading when needed; VRWM must cover the common-mode swing, VC must sit below the transceiver bus pin tolerance, and junction capacitance must match the baud rate. Check isolation and grounding as well. Device parameters follow the manufacturer datasheet.
Q2: How do I select a TVS for CAN, and what junction capacitance is appropriate?
Four parameters decide it. VRWM must cover the maximum common-mode voltage on CAN H and CAN L, set by system voltage and transceiver specification, with different grades for 12 V and 24 V systems. VC must sit below the absolute maximum rating of the transceiver bus pins. Junction capacitance must match speed: classical CAN from 250 kbps to 1 Mbps usually works with tens of picofarads, while CAN-FD at 2 to 5 Mbps needs a lower order of magnitude, otherwise the RC slows the rising edge and bit-time margin shrinks. Symmetrical differential lines normally use a bidirectional TVS. Final judgement follows measured waveforms and the manufacturer datasheet.
Q3: Why does the port still fail with a TVS fitted, and what happens if VC exceeds the transceiver rating?
Four common causes: the TVS clamping voltage under surge current sits above the transceiver bus pin tolerance and the residual voltage kills the chip, so VC must be checked at the surge current rather than VRWM alone; the TVS sits too far from the connector and trace inductance adds an L times di/dt spike to the real residual; there is a TVS but no front-stage diversion, so a large surge punches through the TVS; or ground potential difference and common-mode voltage are permanently out of range so the TVS is fine while the chip works outside its limits. Place the TVS near the connector, add GDT or PTC grading where needed, check isolation and single-point grounding, and validate by surge measurement.
Q4: Bidirectional or unidirectional TVS for a signal port, and what power rating?
Directionality follows signal shape. For differential pairs such as RS485 A and B or CAN H and L, the signal can swing either way with respect to ground or carry an uncertain common-mode offset, so a bidirectional TVS is normally used. Single-ended IO known to stay above ground, such as a 0 to 5 V enable or feedback line, can use a unidirectional part whose forward conduction at roughly one PN junction drop clamps negative spikes lower. Power rating follows expected surge energy: signal ports commonly use a few hundred watts under the 10/1000 microsecond convention, and near long cable entries or where large surges are expected they should work with a GDT or PTC front stage so most energy is handled upstream. Final choice follows real waveforms and the manufacturer datasheet.
Contact us
For signal-port TVS and ESD device selection review and sample testing, communication port protection assessment (RS485, CAN, industrial IO), and matching AMSEMI protection devices plus CR Micro power devices, contact us.
Shenzhen Intek Technology Co., Ltd — electronic component distributor and system solution provider Phone / WeChat: 136-3264-8484 Address: 4F, Building 2, Jingwei Center, No. 309 Ping an Avenue, Pinghu Street, Longgang District, Shenzhen Website: www.intek.vip
Disclaimer: This article is technical education and industry exchange for reference only and is not selection, procurement or other commercial decision advice. Interface levels, voltage grades, junction capacitance orders and power orders quoted here are typical industry ranges or order-of-magnitude estimates; real values vary significantly with transceiver, system voltage, bus length and device specification. Final design must follow the ratings in the transceiver and TVS manufacturer datasheets, real waveforms, and measurement on the complete product.