Welding Inverter Design Calculator

This is a continuous-conduction, two-switch forward sizing model. Erickson and Maksimović, Fundamentals of Power Electronics, 3rd ed. (2020), provides the steady-state volt-second balance, RMS-current, loss and magnetic-design framework. The equations below are derived here; ST AN4638, pp.3–4 and Appendix A, supplies the welding topology and device-stress context.

With $n=N_s/N_p$, $V_p=V_{bus}-2V_{CE}$ and $R_t=R_s+n^2R_p$, inductor volt-second balance gives

The editable arc line is $U_{arc}=U_0+rI$. Defaults reproduce the MMA voltage endpoints in ST AN4638 Table 2, p.6; this is a design load characteristic, not a measured plasma model. The chart compares it with maximum-duty voltage capability using fixed turns and winding resistances. It does not predict a constant-current controller characteristic.

Primary turns are rounded upward from $N_p\ge V_{p,hi}D_{lim}/(fA_e\Delta B_{lim})$. The secondary is rounded upward to meet both approximate loaded duty and no-load pulse headroom at minimum bus. Copper voltage drop is then included in the actual duty check. Failure of that check requires revising the inputs. Forward flux is unipolar excursion, unlike the balanced bridge peak-flux convention in Transformer Design. Its current-density, bare-copper window budget, IACS resistivity, hot-resistance and skin-depth equations are reused and adapted here. Copper areas include a 10% allowance; the RMS triangle inequality bounds the primary sizing current including magnetizing current.

For ideal reset diodes, $t_r/T=DV_p/V_{bus}$. Primary current rises linearly from $n(I-\Delta I/2)$ to $n(I+\Delta I/2)+V_pD/(fL_m)$; a pulse with endpoints $a,b$ has $I_{rms}^2=D(a^2+ab+b^2)/3$. Secondary load current flows only during on-time. The primary winding also carries the decreasing magnetizing reset pulse, adding $I_{m,pk}^2(t_r/T)/3$ to its mean-square current. That reset current flows through the clamp diodes, not through the IGBTs; their RMS current includes only the on-time pulse. The primary copper sizing bound allows both on and reset magnetizing intervals. Parallel switch devices share current equally.

Secondary diode stress is checked by interval: the forward rectifier blocks the reset plateau $nV_{bus,hi}$, while the freewheel diode blocks the on-state pulse $n(V_{bus,hi}-2V_{CE})$. Their maximum is the ideal reverse-voltage baseline, consistent with the bus-based stress in onsemi AND8373/D, equation 31. Reset-diode drops and transient leakage/recovery overshoot need separate allowance; this plateau estimate is not a selected device rating.

Choke ripple is $\Delta I=[U_{arc}+I(R_{ch}+R_{lead})+V_{fw}](1-D)/(fL)$. The high-bus corner sizes inductance and peak energy. Choke turns follow $N\ge LI_{pk}/(B_{lim}A_e)$; $g_{eq}=\mu_0N^2A_e/L$ is an equivalent total reluctance gap. A physical gap requires core-reluctance and fringing corrections; this value is not a machining instruction.

Losses are evaluated at both DC-bus endpoints. IGBT conduction uses constant hot $V_{CE}$; switching energy uses equal on/off weighting and linear current/voltage scaling of the entered per-device $E_{on}+E_{off}$. Transformer copper, output diode conduction, choke DC resistance, leads, transformer core and other allowances are included. Enter choke core loss and excess AC copper loss in the other-loss allowance. Lead heating conservatively uses high-bus ripple at both corners. The larger endpoint budget feeds mains power through the entered front-end efficiency and true power factor.

Welding duty $X$ changes time-averaged arc power and the heating-equivalent mains estimate $\sqrt{XI_{weld}^2+(1-X)I_{idle}^2}$. It never reduces switch, magnetic or instantaneous winding ratings.

• One isolated forward secondary, one forward rectifier and one freewheel diode; steady positive choke current and fixed switching frequency. Both primary switch positions conduct together.
• The minimum and maximum DC-bus voltages are independent engineering inputs. Determine loaded valleys, high-line peaks or PFC limits separately; mains voltage alone does not establish them.
• Custom magnetic geometries are illustrative, not catalogue parts. Flux ceilings, copper current density, window allowance, loss factors and energy inputs must be chosen for the actual materials and temperature.
• Primary RMS current includes the ideal magnetizing reset pulse but not a separately resolved core-loss excitation component. Core loss is an entered power allowance; its small waveform contribution must be included in a detailed design. The transformer starts each cycle at a reset baseline. Remanence, flux walking, leakage, startup imbalance and saturation curvature are excluded. Maximum-duty high-bus excitation is checked even though operating duty is lower.
• The resistance-only duty model uses mean load current and excludes magnetizing-current voltage drop and within-cycle resistive distortion. Ripple assumes a constant output terminal voltage over one period.
• Choke hot DC resistance is supplied independently. Suggested turns, copper area and equivalent gap do not verify a realizable core, winding layout or resistance.
• Switching energy is a local scaling estimate. Equal on/off weights, linear scaling and equal sharing must be checked against device curves or measurements. Recovery energy must be counted once. Clamp losses, gate drives and other losses belong in the separate allowance.
• The waveform player is a slowed-down illustration of the low-bus corner. Current and flux remain calculated at the selected electrical operating point; no feedback dynamics are simulated.

Use for preliminary engineering comparisons of CCM IGBT two-switch forward welding power stages. The default case is approximately 4.2 kW at 160 A. Endpoint loss budgets do not guarantee that an interior bus voltage cannot produce a larger loss. Magnetic loss, switching energies, hot resistances and cooling require independent confirmation.

The no-load target only sets necessary secondary pulse headroom. At no load the CCM gain law fails, and an output capacitor can charge toward the secondary pulse peak. This tool does not establish actual open-circuit voltage, safe touch voltage or voltage-reduction-device behavior. Ignition, arc extinction, short circuits, anti-stick/hot-start functions, current-loop stability, output-capacitor dynamics, leakage spikes, transient SOA, inrush, bulk capacitance and EMC are outside scope.

Mains voltage, stored DC-link energy and welding current can cause fatal shock, burns and fire. This is a design calculation, not a build guide. IEC 60974-1:2021, edition 6, specifies safety and performance requirements for welding power sources; check against the standard in force and its applicable national adoption. No standards tables are reproduced and no compliance or nameplate rating is certified here. Isolation, creepage, clearance, protective earthing, discharge provisions, output-voltage limits, protective devices and thermal tests require separate qualified assessment. Mains installation work belongs to a qualified person. The cycle RMS estimate does not select a supply cable, plug, breaker or fuse.