Forniture per pompe laser a fibra: tracciamento della frequenza di risonanza

Forniture per pompe laser a fibra: tracciamento della frequenza di risonanza

Astratto

Resonant LLC and CLLC supplies keep zero-voltage switching only while the switching frequency tracks the tank. This article gives the drift sources, the loop stability criterion, the ESR phase term, four field pain points at 10,000 h, and five steps that hold frequency tracking from 10% to 100% load.

1. introduzione

A pump source that uses an LLC, LCC, or CLLC tank gets soft switching, low emission, and high efficiency from the resonance itself. Every one of those benefits depends on the switching frequency sitting close to the tank frequency.

The tank frequency moves with temperature, load, component aging, and production tolerance. A fiber laser controller also steps the diode current between 5% and 100% of full load as the process changes, so the tank sees fast load steps as well as slow drift.

When the switching frequency leaves the tank, the converter loses zero-voltage switching. Switching loss rises 3× to 5×, emission rises 20 dB to 30 dB, and junction temperature jumps 30 °C to 50 °C.

This article covers the drift sources, the tracking loop equations, four field pain points, and five design steps built on N4700 Class I ceramic capacitors.

2. Meccanismo fondamentale

2.1 Resonant frequency and its drift sources

The tank frequency follows the resonant product:

f_r = 1 / (2π × √(L_r × C_r))

where f_r is the resonant frequency, L_r the resonant inductance, and C_r the resonant capacitance.

Small changes propagate with a half factor for each element:

∆f_r / f_r = −0.5 × (∆L_r / L_r + ∆C_r / C_r)

where ∆f_r / f_r is the fractional frequency shift, ∆L_r / L_r the fractional inductance change, and ∆C_r / C_r the fractional capacitance change. Capacitance drift enters from three sources: temperature, DC bias, and aging.

An X7R part combines ±15% over temperature, −20% to −50% at rated bias, and −2% to −5% per decade hour, for a total of −60% to +15% and a frequency shift of −18% to +35%. N4700 Class I holds ±0.35%, so the shift stays at ±0.175%.

2.2 ESR phase error in the phase detector

The tank capacitor ESR delays the current zero crossing, which shifts the phase the detector reads:

φ_ESR = arctan(2π × f_r × C_r × ESR)

where φ_ESR is the phase error, f_r the tank frequency, C_r the capacitance, and ESR the equivalent series resistance. A 2.2 nF capacitor at 250 kHz with 100 mΩ of X7R ESR gives 0.02 degrees, and 5 mΩ of N4700 ESR gives 0.001 degrees. Both are small enough that phase detection does not need ESR compensation.

2.3 Loop crossover and phase margin

A tracking loop stays stable while its gain crossover sits below the tank bandwidth:

f_c < f_r / (2Q)

where f_c is the crossover frequency, f_r the resonant frequency, and Q the loaded quality factor of the tank. At 250 kHz and Q = 3 the crossover must stay below 41.7 kHz, so a 20 kHz loop keeps margin.

Phase margin at that crossover follows the loop filter poles:

PM = 180 − arctan(f_c / f_pole1) − arctan(f_c / f_pole2)

where PM is the phase margin, f_c the crossover frequency, and f_pole1 and f_pole2 the loop filter poles. Correct pole placement reaches 60 degrees, which leaves 15 degrees of margin above the 45 degree limit.

2.4 Load changes the tank Q

The quality factor tracks the load resistance:

Q = √(L_r / C_r) / R_load

where Q is the tank quality factor, L_r and C_r the resonant elements, and R_load the reflected load resistance. At 10% load the tank Q is about 10× the full-load value, so the phase slope through resonance is steeper and the loop has less phase margin at the moment it needs to move fastest.

Fig. 1 — A 100 °C rise moves an X7R tank 7.5% and takes it past the 5% loop range, while an N4700 tank shifts 0.15%.

3. Punti critici di ingegneria

Pain Point 1: X7R drift leaves the loop range

Temperature and DC bias act on the same capacitor, so 20% to 40% of capacitance can disappear over the operating range. That moves f_r by 10% to 20%, past the 5% to 10% range the loop can follow.

The loop then loses lock and the converter runs at the wrong frequency without zero-voltage switching. In a 5 kW pump source the junction temperature can climb from 100 °C to 150 °C inside 100 ms and trip thermal shutdown.

Pain Point 2: Aging moves the tank too slowly to see

Class II ceramic loses 2% to 5% of capacitance per decade hour. After 10,000 h the loss reaches 8% to 20%, which shifts f_r by 4% to 10%.

The drift is too slow to show up as a transient and too large to ignore, so the converter sits at the edge of its switching boundary without a fault. Field data from fiber laser installations shows 18% of resonant converter failures inside two years come from this drift.

Pain Point 3: Light load sharpens the phase slope

Tank Q varies with load. At 10% load the phase transition through resonance is about ten times steeper than at full load, so the same loop gain produces a much smaller phase margin.

A loop tuned at full load can then limit-cycle between two frequency points, which shows up as audible noise and sub-harmonic emission. Lowering the bandwidth fixes the light-load case and slows the response to a load step.

Pain Point 4: The inductor heats the capacitor

In a compact module the resonant inductor dissipates 2 W to 5 W and raises the temperature of the nearby capacitor by 10 °C to 20 °C. The two parts then drift in opposite directions when their coefficients differ.

An inductor with +200 ppm/°C beside an X7R capacitor at −1500 ppm/°C gives a 2.5% frequency shift from a 15 °C gradient. The loop has to track a drift that the layout created.

4. Fasi di progettazione

Strategy 1: Use a Class I capacitor in the tank

N4700 holds capacitance within ±0.35% from −55 °C to +125 °C, with no DC bias loss and no aging loss, so the frequency shift stays at ±0.175% against a loop range of ±5%.

Usa il N4700 parts at the resonant node and the DC blocking position. The self-heating term also falls: at 3 A rms of resonant current the 0.1% dissipation factor holds loss under 0.5 W per capacitor against 3 W to 8 W for X7R.

Strategy 2: Make the loop bandwidth follow the load

Program the loop filter so the bandwidth scales with load current: 30 kHz at full load for fast tracking and 5 kHz at 10% load to stop the limit cycling of Pain Point 3.

Scale the transition, do not step it. A square-root law in load current gives a smooth bandwidth curve and avoids a discontinuity that would excite the tank itself.

Strategy 3: Compensate the frequency reference with temperature

Fit an NTC thermistor on the resonant capacitor pad and trim the reference by the measured coefficient: f_ref(T) = f_ref_25C × (1 − 0.5 × TCR_C × (T − 25)).

With N4700 at ±30 ppm/°C the correction is 0.015% per 10 °C, so the term is small but it removes the residual error. With an X7R part at −1500 ppm/°C the same term runs 1.5% per 10 °C and becomes mandatory.

Strategy 4: Match the tank capacitors within ±2%

Tolerance sets the initial frequency error, ahead of drift. Parts matched within ±2% put the initial f_r inside ±1% of the design target, which on a 250 kHz tank is ±2.5 kHz.

Use a matched set from one production lot and mount the parts on the same pad group so they see one temperature. Add resistori di gradazione where the tank uses series capacitors, so the voltage splits evenly and each part runs at the same bias point.

Strategy 5: Split the thermal paths and add open-loop feedforward

Separate the resonant inductor and capacitor by at least 20 mm with a grounded copper pour between them, or place the capacitor on the cooler side of the heatsink. Hold the temperature difference below 5 °C, which at ±30 ppm/°C is 0.015% of capacitance.

During a load step faster than 100 µs, drive the frequency open loop from the sensed current and return to closed loop once the transient settles inside 1 ms to 5 ms. That keeps the loop from losing lock while it waits for the tank to respond. Where the DC link feeds the tank, a condensatore a film damps the step that the HVCT8G bank cannot absorb.

Parametro del componente	Valore tipico	Condizione di test
Tensione nominale (serie HVCT8G)	10 kV DC	Hi-pot 1.5× per 60 s
Capacità	4700 pF ±20% (472M)	1 MHz, 1 V rms
ESR a 1 MHz	inferiore a 50 mΩ	-
Induttanza di anello (accordo E40)	inferiore a 8 nH	Distanza tra i terminali: 40 mm
resistenza di isolamento	superiore a 10 GΩ a 500 V CC	25 °C, 60% UR
Range di temperatura	da -40 °C a +125 °C	Serie IEC 60068-2
Capacità di misurazione di impulsi dv/dt	oltre 50 kV/µs	a tensione nominale

5. HVC N4700 capacitors for resonant tanks

Three requirements drive this node: a capacitance that does not move with temperature or bias, no aging term that the loop cannot see, and matched pairs across the tank. HVC builds N4700 (T3M) Class I parts on a SrTiO3 base for that set, in a 6,000 m² owned factory in Dongguan, from 1 kV to 100 kV.

Sei parametri importanti in questo caso

• Temperature coefficient of ±30 ppm/°C. Capacitance stays within ±0.35% from −55 °C to +125 °C, which limits the tank frequency shift to ±0.175% against ±7.5% for an X7R part.
• No DC bias capacitance loss. Tank capacitance holds at any bus voltage, so a 400 V LLC does not lose the 5% to 15% of frequency margin that an X7R part gives up under bias.
• No ferroelectric aging. Class I dielectric has no aging mechanism, so the 4% to 10% frequency drift that an X7R tank accumulates in 10,000 h never appears.
• tan δ of 0.1% or lower. Loss stays under 0.5 W per capacitor at 3 A rms of resonant current against 3 W to 8 W for X7R, so the capacitor adds 3 °C to 5 °C of self-heating instead of driving its own drift.
• No long-term drift. Capacitance stays within 2% after 50,000 h at 85 °C, with intrinsic life beyond 100,000 h, so the loop keeps the same tracking margin over a 10-year service life. Coverage runs from 1 kV to 100 kV, with matched sets inside ±2%.
• Qualified as a drop-in for first-tier brands in field-repair programmes and specified by laser OEMs in new resonant platforms. HVC parts match the electrical parameters and package dimensions of Murata Componenti ad alta tensione con terminali a vite e radiali.

Consegna e supporto

Standard parts ship in 4–8 weeks, and matched custom sets in 6–10 weeks. HVC applications engineering supports tank simulation, loop stability analysis, matched-set selection for resonant networks, and custom voltage and capacitance combinations.

Send the tank inductance, resonant capacitance, switching range, load profile, and the loop crossover you plan. HVC returns a matched capacitor set with capacitance distribution, ESR data, and a two-sample evaluation kit within five working days.

6. Domande frequenti

Q1: How far can the tank frequency drift before the loop loses lock?

A typical tracking loop follows 5% to 10% of frequency shift. An X7R capacitor that loses 20% to 40% of its value to bias and temperature moves f_r by 10% to 20%, which is outside that range. N4700 Class I holds ±0.35% and limits the shift to ±0.175%.

Q2: How fast does the tank capacitor heat up during a load step?

Its thermal time constant runs 0.5 s to 2 s, so a 10% to 100% load step raises it 5 °C to 15 °C inside 1 s to 5 s. At −1500 ppm/°C an X7R part then changes 1.2% in capacitance and shifts f_r 0.6%, while N4700 shifts 0.015%.

Q3: What crossover frequency keeps the tracking loop stable?

Below f_c = f_r / (2Q). At 250 kHz with a loaded Q of 3 the limit is 41.7 kHz, so a 20 kHz loop keeps margin. Hold phase margin above 45 degrees with PM = 180 − arctan(f_c / f_pole1) − arctan(f_c / f_pole2); 60 degrees is reachable.

Q4: Why does the loop misbehave at light load?

Because tank Q rises with load resistance. At 10% load the phase slope through resonance is about 10× steeper, so the loop can limit-cycle between two frequencies and produce audible noise plus sub-harmonic emission. Drop the bandwidth from 30 kHz to 5 kHz as load falls.

Q5: Does capacitor ESR disturb the phase detection?

Barely. φ_ESR = arctan(2π × f_r × C_r × ESR), so a 2.2 nF capacitor at 250 kHz with 100 mΩ gives 0.02 degrees and with 5 mΩ gives 0.001 degrees. The phase detector does not need ESR compensation; the capacitance drift of §2.1 is the term that matters.

Q6: How much does the resonant inductor disturb the capacitor?

An inductor dissipating 2 W to 5 W raises a nearby capacitor 10 °C to 20 °C. With +200 ppm/°C on the inductor and −1500 ppm/°C on an X7R capacitor, a 15 °C gradient produces a 2.5% frequency shift. Keep 20 mm of separation and hold the difference under 5 °C.

7. Riferimenti

1. E. Kuffel, WS Zaengl, J. Kuffel, Fondamenti di ingegneria dell'alta tensione, 2ª ed., Newnes, 2000.
2. N. Mohan, TM Undeland, WP Robbins, Elettronica di potenza: convertitori, applicazioni e progettazione, 3a ed., Wiley, 2003.
3. Avviso MIL-HDBK-217F 2, Previsione dell'affidabilità delle apparecchiature elettroniche 1995.
4. IEC 60384-1:2018, Condensatori fissi per impiego in apparecchiature elettroniche — Parte 1: Specifiche generiche.
5. H. Nakao, Y. Yonezawa, T. Sugawara, Y. Nakashima e F. Kurokawa, Metodo di valutazione online del degrado dei condensatori elettrolitici per la previsione dei guasti degli alimentatori switching a controllo digitale, IEEE Transactions on Power Electronics, vol. 33, n. 3, pp. 2552-2558, 2018. https://doi.org/10.1109/tpel.2017.2691048.
6. M. Marchesoni e altri, Minimizzazione dell'ondulazione della tensione del condensatore nei convertitori di tensione per applicazioni HVDC, in AEIT HVDC International Conference (AEIT HVDC) 2019, pp. 1-6, 2019. https://doi.org/10.1109/aeit-hvdc.2019.8740647
7. Manuale applicativo Fuji IGBT, Capitolo 5 — Progettazione del circuito di protezione (REH984)Fuji Electric, REH984, https://www.fujielectric.com/device/semi/technical/application/pdf/REH984/REH984_05.pdf .
8. GOST IEC 61071-2014, Condensatori silos elettronici (Condensatori per elettronica di potenza). Mosca: Standard, 2014.
9. C. Oria, A. Ortiz, D. Ferreño, I. Carrascal e I. Fernández, Analisi dello stato dell'arte sulle prestazioni dei materiali dielettrici cellulosici nei trasformatori di potenza: risposta meccanica e invecchiamento, IEEE Transactions on Dielectrics and Electrical Insulation, vol. 26, n. 3, pp. 939-954, 2019. https://doi.org/10.1109/tdei.2019.007896.
10. Team di ingegneria applicativa HVC, Condensatore ceramico ad alta tensione Risorse tecniche, https://www.hv-caps.com