Dimensionnement et récupération des diodes de limitation dans les circuits de sortie d'allumage plasma

Dimensionnement et récupération des diodes de limitation dans les circuits de sortie d'allumage plasma

Abstract

A clamp diode bounds the excursions that cable reflections, winding ringing and gap collapse send down the ignition output. It also absorbs the trapped energy of the output loop, has to turn off cleanly afterwards, and survives that duty thousands of times a day. This article works the trapped energy, the unclamped overshoot, the clamp absorption, the clamp capacitor rise, the recovery spike and the clamp-cable resonance.

1. Introduction

The output stage of a plasma igniter is the transformer secondary, the coupling capacitor and the torch cable. It sees excursions the design never intends: reflected pulses from a cable impedance mismatch, ringing between the transformer leakage and the stray capacitances, and the collapse transient when the gap breaks down or the arc extinguishes.

Without a clamp those excursions reach two to three times the nominal ignition voltage. That stresses the cable insulation, the coupling capacitor and the transformer windings, and the failures arrive months later and somewhere else in the machine.

The clamp is not a free component. It has to absorb the trapped energy of the output loop, turn off cleanly when the transient ends, and survive the absorption duty of a machine firing hundreds of times a second. Misapplied clamps fail in visible ways: one that rings on recovery becomes the transient source it was installed to remove, and one that absorbs more than its rating dies short and turns protection into a fault.

This article works the trapped energy, the overshoot it produces, the clamp absorption and voltage rise, and the recovery spike that decides whether the clamp stays a protective element.

2. Mécanisme de base

2.1 Trapped energy in the output loop

When the gap stops conducting, the current in the output inductance has nowhere to go and its energy has to be absorbed.

E_trapped = ½ × L_out × i_out²

where E_trapped is the trapped energy, L_out is the total output loop inductance and i_out is the current at the moment the gap stops conducting. With 1.2 mH of secondary leakage and cable inductance at 8 A the loop holds 38.4 mJ.

2.2 Overshoot without a clamp

With no clamp, that energy charges the parasitic capacitance of the output and the voltage rises until the parasitic path breaks down instead.

V_peak = V_nominal + i_out × √(L_out / C_par)

where V_peak is the unclamped excursion, V_nominal is the working output voltage, i_out is the current at extinction and C_par is the parasitic capacitance the loop charges. A 300 pF parasitic path with 1.2 mH presents 2000 Ω, so 8 A of current adds 16 kV to a 10 kV output. The swing reaches 26 kV, which is 2.6 times nominal and where cable and winding insulation failures start.

2.3 What the clamp has to absorb

Once the clamp conducts, the loop current decays into the clamp while the clamp holds its threshold voltage, and the extra time the source keeps driving current makes the clamp absorb more than the trapped energy.

E_clamp = ½ × L_out × i_out² × V_clamp / (V_clamp − V_nominal)

where E_clamp is the energy the clamp absorbs per event, V_clamp is the clamp threshold and V_nominal is the working output voltage. A 13 kV threshold against a 10 kV output gives a ratio of 4.33, so the 38.4 mJ of trapped energy becomes 166 mJ at the clamp. That ratio is the reason a clamp threshold set close to the working voltage is expensive: at 11 kV the same event hands the clamp 422 mJ.

2.4 Clamp capacitor voltage rise

The clamp capacitor takes that energy, and its voltage rise per event sets how stable the threshold stays.

ΔV_clamp = E_clamp / (C_clamp × V_clamp)

where ΔV_clamp is the rise on the clamp capacitor, E_clamp is the energy per event, C_clamp is the clamp capacitance and V_clamp is the clamp threshold. Holding the rise below 3% of 13 kV, which is 390 V, needs 33 nF. At that value the rise is 387 V, or 2.98%, and every clamp operation leaves the threshold within a few volts of where it started.

2.5 Recovery and the spike it produces

When the output falls back below the threshold, the diode turns off and its stored charge has to be removed first.

V_rr = L_par × (di/dt)_rr

where V_rr is the spike the recovery produces, L_par is the inductance of the clamp loop and (di/dt)_rr is the rate at which the reverse current falls. The clamp carries the full 8 A of loop current, so its stored charge reaches 4 µC and the reverse current peaks near 8 A.

A part that snaps off in 25 ns drives 3.2 × 10⁸ A/s through 1 µH of loop inductance and puts 320 V on top of the clamp threshold. A soft-recovery grade that falls over 1.5 µs drives 5.3 × 10⁶ A/s and produces 5.3 V.

2.6 Clamp resonance with the cable

The clamp capacitor and the cable inductance form a resonator that rings after every event unless it is damped.

Q = (1 / R) × √(L / C)

where Q is the quality factor of that resonance, R is the series damping resistance, L is the cable inductance and C is the clamp capacitance. A 33 nF clamp capacitor with 1.2 mH of cable inductance resonates at 25.3 kHz with a characteristic impedance of 191 Ω. The same 191 Ω of series resistance brings the Q to 1 and stops the ring, so a 200 Ω element in series with 100 nF across the clamp is enough to keep the diode from reconducting on its own tail.

Fig. 1 — The clamp diode holds the output at 13 kV while a 33 nF capacitor absorbs 166 mJ per event.

3. Points faibles en matière d'ingénierie

Pain Point 1: Clamp threshold drift

If the reference behind the clamp threshold moves with temperature, the threshold moves with it. An upward drift lets the excursion through, and a downward drift clips the useful ignition pulse and spends energy that the gap never sees. A ±30 ppm/°C capacitor in the reference filter holds its value across a 50 °C swing inside 0.15%, which keeps the threshold where the design put it.

Pain Point 2: Energy absorption duty

The clamp absorbs 166 mJ per activation. At 400 ignition events per second a clamp firing on 5% of them takes 3.3 W, and one firing on half of them takes 33 W. A resistor sized for the first number fails on the second, and the diode fails with it, usually short, which turns the protection into a fault.

Pain Point 3: Hard recovery injecting noise

A fast-recovery part with a sharp snap-off produces the 320 V spike of the recovery equation, and that spike couples into the ignition transformer and into the control plane. The clamp that was installed to suppress transients becomes a transient source, and the fault shows up on the control board, not at the clamp.

Pain Point 4: Clamp-capacitor resonance with the cable

The clamp capacitor rings against the cable inductance at 25.3 kHz with a Q of 190 when nothing damps it. The ring prolongs the transient, drives the diode back into conduction, and adds a second and third absorption event to every excursion. The diode then absorbs three times the energy the calculation predicted.

4. Étapes de conception

Strategy 1: Size the clamp capacitor from the event energy

Compute the energy per event at the highest output current and the highest loop inductance the machine will ever present, then choose the capacitance so the voltage rise stays under 3% of the clamp threshold. On a 13 kV threshold and 166 mJ per event that is 33 nF, and the rise is 387 V.

Strategy 2: Choose a soft-recovery clamp diode

Specify a part with controlled reverse recovery and verify the recovery waveform on the bench. A soft-recovery grade lowers the reverse current slope from 3.2 × 10⁸ A/s to 5.3 × 10⁶ A/s and the spike from 320 V to 5.3 V. Use a diode à récupération rapide only where the circuit needs the speed, and add series inductance to control the slope when it does.

Strategy 3: Build the threshold from stable parts

Implement the clamp threshold with a precision reference and a stable capacitor. An N4700 part with a dissipation factor below 0.1% and a ±30 ppm/°C coefficient keeps both the clamp capacitor and the reference filter in place across the operating temperature, so the protection margin measured cold is the margin delivered hot.

Strategy 4: Damp the clamp-cable resonance

Add 200 Ω in series with 100 nF across the clamp capacitor. The branch brings the Q of the 25.3 kHz resonance from 190 to 1, which keeps the diode out of conduction after the main event and stops the repeat absorptions that the energy calculation never counted.

Strategy 5: Monitor the clamp duty in the controller

Count clamp activations per ignition event and trend the count over weeks. A rising count means the output circuit is degrading, through rising leakage inductance or cable damage, and it flags maintenance before the clamp is destroyed. The count is also the cheapest indicator of a cable that has been re-routed or crushed.

Strategy 6: Validate against the worst-case output faults

Fire into a dead short, into an open circuit and into a marginal gap, and confirm in each case that the clamp holds the peak inside the design margin and that the diode junction stays below 75% of its rating. The short and the open circuit are the two cases that a bench test into a healthy gap never exercises.

Paramètre de composant	Valeur typique	Condition de test
Tension inverse de crête répétitive	10 kV VRRM, two in series	Série 2CL, 25 °C
Courant direct moyen	0.8 A SI(AV)	Onde rectangulaire, Tc ≤ 85 °C
Courant de surtension direct	100 A à 8.3 ms	surtension unique, IFSM
Temps de récupération inverse	en dessous de 1 µs trr	niveau de récupération souple
capacité de jonction	25 pF typique	VR = 4 kV, 1 MHz
Clamp capacitance	33 nF, rise below 3%	166 mJ per event
Température de fonctionnement	−55 °C à +125 °C	Tj max, 75% derated

5. HVC N4700 (T3M) capacitors for the clamp circuit

The clamp holds its threshold only while its reference holds its value. HVC builds the N4700 (T3M) Class I family for that role, with the résistances de gradation that share the clamp voltage across a series diode string, from a 6,000 m² owned factory in Dongguan across a 1 kV to 100 kV range.

Six paramètres importants ici

• Stable clamp capacitance. A ±30 ppm/°C coefficient keeps the clamp threshold and the absorption behaviour constant across temperature, so the 387 V rise measured on the bench stays the rise delivered in a hot cabinet.
• An absorption path that stays cool. A dissipation factor below 0.1% leaves the absorbed energy in the resistor where it is meant to land, instead of heating the capacitor and moving its value.
• Coverage of the whole output range. Ratings from 1 kV to 100 kV match the clamp voltages of pilot and main-arc stages with one qualified dielectric, and a series string of 10 kV parts covers the 13 kV threshold with margin.
• Discharge below 5 pC. The clamp capacitor sits at the full threshold voltage on every event, and a discharge-free part does not add its own noise to a node that exists to be quiet.
• Low loop inductance. A self-inductance below 8 nH keeps the clamp loop short, which is what holds the recovery spike that the loop inductance produces.
• Proven in high-voltage transient service. The 40 kV 441-series condensateurs poignée de porte replaced first-tier brand parts in high-voltage field-service programs, and a références croisées contre un condensateur à film shows the drift a clamp reference would otherwise have to absorb.

Livraison et support

Standard parts ship in 4 to 8 weeks and matched clamp sets in 6 to 10 weeks. HVC applications engineering supports trapped-energy calculation, clamp capacitor sizing, damping branch design and worst-case fault testing. Send the output voltage, the loop inductance, the working current and the clamp threshold, and HVC returns a clamp selection sheet with absorption data and PD reports within five working days.

6. Questions fréquentes

Q1: How much energy does the output loop trap?

Half the loop inductance times the square of the current at the moment the gap stops conducting. With 1.2 mH of secondary leakage and cable inductance at 8 A that is 38.4 mJ, and the unclamped voltage then swings 16 kV above the working level, to 26 kV on a 10 kV output.

Q2: Why does the clamp absorb four times the trapped energy?

Because the loop current decays slowly while the clamp holds its threshold, and the source keeps driving current into the clamp during that decay. The energy scales with V_clamp divided by V_clamp minus V_nominal, which is 4.33 for a 13 kV clamp on a 10 kV output. Setting the clamp closer to the working voltage makes it worse: 11 kV gives 11.

Q3: How big does the clamp capacitor have to be?

Big enough that the voltage rise per event stays under 3% of the threshold. At 166 mJ per event and a 13 kV threshold that is 33 nF, which gives a 387 V rise, or 2.98%. A smaller capacitor lets the threshold walk upward with every event.

Q4: What makes a recovery spike large?

Loop inductance and the reverse current slope together. The clamp carries the full 8 A of loop current, so its stored charge reaches 4 µC. A part that snaps off in 25 ns drives 3.2 × 10⁸ A/s through 1 µH and produces 320 V. A soft-recovery grade that falls over 1.5 µs produces 5.3 V.

Q5: Why does the clamp ring after the event?

Because the clamp capacitor and the cable inductance form a resonator, and nothing in the clamp damps it. A 33 nF capacitor with 1.2 mH rings at 25.3 kHz with a Q of 190, which drives the diode back into conduction and adds repeat absorptions. A 200 Ω resistor in series with 100 nF brings the Q to 1.

Q6: What tests prove the clamp works?

Fire into a dead short, into an open circuit and into a marginal gap, and confirm the peak stays inside the design margin and the diode junction stays below 75% of its rating in each case. Add a controller counter for clamp activations per event and trend it, because a rising count catches cable damage weeks before the diode fails.

7. Références

1. E. Kuffel, WS Zaengl, J. Kuffel, Principes fondamentaux de l'ingénierie haute tension, 2e éd., Newnes, 2000.
2. Mohan, N., Undeland, TM et Robbins, WP, Électronique de puissance : convertisseurs, applications et conception, 3e éd. Hoboken, NJ : John Wiley & Sons, 2003.
3. Avis MIL-HDBK-217F 2, Prédiction de la fiabilité des équipements électroniques 1995.
4. CEI 60384-1:2021, Condensateurs fixes pour équipements électroniques - Partie 1 : Spécifications générales, 6e éd. Genève : CEI, 2021.
5. CEI 61071:2017, Condensateurs pour l'électronique de puissance, Éd. 2.0. Genève : Commission électrotechnique internationale, 2017.
6. Y. Zhou, B. Chen, C. Tian, ​​J. Yuan et Y. Chen, Convertisseur linéaire à diodes parallèles pour essais de tenue diélectrique et mesure des décharges partielles d'équipements électriques, Haute Tension, vol. 8, no. 1, pp. 171-182, 2023. https://doi.org/10.1049/hve2.12312.
7. V. Belyaev, Modèle de contournement superficiel d'un isolateur sous vide, dans les actes du 18e Symposium international sur les décharges et l'isolation électrique sous vide (ISDEIV), pp. 151-154, 1998. https://doi.org/10.1109/deiv.1998.740596.
8. AND9800/D — Sélection du condensateur d'amortissement pour les applications d'entraînement de moteur (Tableau 15 Données de test), onsemi, AND9800-D, https://m.onsemi.jp/pub/Collateral/AND9800-D.PDF .
9. États-Unis 8,570,711, Composant électronique en céramique multicouche (Suppression des décharges d'arc pour les MLCC de 1 à 5 kV), Samsung Electro-Mechanics Co., Ltd., accordé en 2013, https://www.freepatentsonline.com/8570711.html .
10. ГОСТ IEC 61071-2014, Condensateurs électrolytiques (Condensateurs pour l'électronique de puissance). Moscou : Standard, 2014.
11. Équipe d'ingénierie des applications HVC, Condensateur céramique haute tension Ressources techniques, https://www.hv-caps.com