This page covers the wear-out and the sudden failures of silicon power switches: the power MOSFET, the IGBT and the thyristor family (SCR, GTO and IGCT). The general models of avalanche, short circuit, latch-up and Wunsch-Bell heating are on the Electrical Stress page, the solder and wire-bond fatigue models on the Mechanical and Package Stress page, and single-event burnout on the Radiation page; this page gives the device-level lifetime models and limits used for these switches.

1. Why It Happens

Trigger: Load cycles that swing the junction temperature (power cycling), an inductive turn-off that drives the device into avalanche, a load short circuit, a fast current or voltage rise on a thyristor, a surge, and atmospheric neutrons while the device blocks a high voltage (Semikron AN 21-001; onsemi AN-7514; Mohsenzade et al.; Semikron AN 18-002; Semikron AN 17-003).
Mechanism, power cycling: Aluminium and silicon expand by a factor of 6 differently, so the bond wires crack at the heel and lift off, and the chip solder fatigues, from the centre in inverter tests and from the corners at high junction temperature (Kuenzi; Semikron AN 21-001).
Mechanism, avalanche: In older MOSFETs the avalanche current under the source turns on the parasitic bipolar, and in rugged modern devices the junction heats until the intrinsic carrier density reaches the drift doping (onsemi AN-7514).
Mechanism, short circuit: The IGBT carries several times its rated current at the full bus voltage for microseconds and fails by latch-up, gate oscillation and oxide breakdown, or by thermal runaway from the leakage current tens of microseconds after a successful turn-off (Mohsenzade et al.).
Mechanism, thyristor: A current that rises faster than the conducting area spreads from the gate makes hot spots, and a fast voltage rise drives a displacement current through the junction capacitance that fires the device by itself (Motorola AN-849; Semikron AN 18-002).
Mechanism, cosmic rays: A neutron makes a dense charge track in the high-field drift region of a blocking device and starts a destructive avalanche, with a rate that rises almost exponentially with the voltage (Semikron AN 17-003; Findeisen et al.).
What sets the rate: The swing and the mean or minimum junction temperature, the heating time, the current per bond and the wire diameter for power cycling (Alavi et al.); the avalanche current and time for avalanche (onsemi AN-7514); the bus voltage and the detection time for short circuits (Mohsenzade et al.); the gate drive and the temperature for thyristor triggering (Semikron AN 18-002); the blocking voltage, the temperature and the altitude for cosmic rays (Semikron AN 17-003).
Failure criterion: A rise of 20 per cent of the forward voltage, the thermal resistance or the temperature swing in power cycling (Semikron AN 21-001), or 5 to 20 per cent of $V_{C E}$ in other studies (Alavi et al.); a short between the main terminals, the most common thyristor failure (Littelfuse AN1009); a destroyed chip after avalanche, short circuit or a neutron event.

2. Where to Look

Location: The heel of the emitter bond wires and the chip solder layer for power cycling, and the baseplate solder for slow cycles (Semikron AN 21-001); the cell region under the source for avalanche (onsemi AN-7514); the region next to the gate of a thyristor for di/dt failure (Motorola AN-849); the drift region of the blocking device for neutron failures (Findeisen et al.).
Signature: A $V_{C E}$ that rises in steps as wires lift off (Kuenzi); a rising thermal resistance from solder voids (Alavi et al.); a melted spot in the active area after avalanche or short circuit; a burn mark at the gate edge of a thyristor (Motorola AN-849); a single short in the blocking state with no preceding drift for a neutron event (Semikron AN 17-003).
Confirm with: Power-cycling tests with the on-state voltage and the thermal resistance recorded during the test (Semikron AN 21-001); unclamped inductive switching tests at several inductances to separate constant-energy from constant $I^{2} t$ behaviour (onsemi AN-7514); beam tests with acceleration factors up to $1 \times 1 0^{10}$ or mountain tests for neutron rates (Semikron AN 17-003).
Look-alikes: A neutron failure looks like an overvoltage failure, since both are a short in the blocking state, but it comes at a voltage below the rating; a power-cycling $V_{C E}$ rise that is not followed by failure is not the end of life, since one device ran about 600k further cycles after a 5 per cent rise (Alavi et al.).

3. Prevention

Design: Reduce the swing first, since life falls as about the fourth power of $Δ T_{j}$ (Semikron AN 21-001); derate the DC voltage, since the neutron rate falls almost exponentially with it, and use three-level topologies, which have about a third of the rate (Semikron AN 17-003); limit the short-circuit current and detect it within a few microseconds (Mohsenzade et al.); add series inductance and an RC snubber to a thyristor and keep its steady DC below half of $V_{D R M}$ (Semikron AN 18-002).
Material and process: Sintered chips and modules without baseplate change the constants of the power-cycling model (Semikron AN 21-001); thicker or more bond wires lower the current per stitch (Alavi et al.); rugged MOSFET cells suppress the parasitic bipolar (onsemi AN-7514).
Operation: Remember that the neutron rate is highest in a cold device, since the temperature factor falls with heating, so a cold start at full voltage is the worst case, and expect about twice the rate per 1000 $m$ of altitude (Semikron AN 17-003); keep the current below 75 per cent of the rating (Littelfuse AN1009); drive the thyristor gate hard and fast (Semikron AN 18-002).
Screening and test: Power-cycling qualification with the 20 per cent criteria and Weibull statistics (Semikron AN 21-001); avalanche tests rated by $I^{2} t$ (onsemi AN-7514); for neutron rates, interpolate the data sheet only between the measured voltages (Semikron AN 17-003).

Power Cycling of Modules

Name	Formula	Related Formulae	Variables/Constants
Power-cycling life of a module (CIPS08, Bayerer)	$N_{f} = K Δ T_{j}^{β_{1}} e^{β_{2} / (T_{j}^{*} + 273)} t_{o n}^{β_{3}} I^{β_{4}} V^{β_{5}} D^{β_{6}}$	$N_{f} = A Δ T_{j}^{α} e^{E_{a} / (k_{B} T_{j m})}$	$N_{f} : cy c l es t o f ai l u r e$ $K : t ec hn o l o g y f a c t or$ $Δ T_{j} : j u n c t i o n t e m p er a t u r e s w in g in K, T_{j}^{*} : r e f er e n ce j u n c t i o n t e m p er a t u r e in d e g r ees C e l s i u s (t h e minim u m in B a y er e r^{'} s f or m)$ $t_{o n} : h e a t in g t im e p er cy c l e in s$ $I : c u r r e n t p er b o n d - w i r e s t i t c h in A$ $V : b l oc k in g v o l t a g e c l a ss d i v i d e d b y 100$ $D : b o n d - w i r e d iam e t er in mi cr o m e t r es, β_{1} t o β_{6} : f i t e x p o n e n t s$ $A, α : C o f f in - M an so n - A r r h e ni u s co n s t an t s$ $E_{a} : a c t i v a t i o n e n er g y$ $k_{B} : B o l t z mann co n s t an t 8.617 \times 1 0^{- 5} eV/K$ $T_{j m} : m e an j u n c t i o n t e m p er a t u r e in K; A l a v i e t a l . : B a y er er β = - 3.483, 1917, - 0.438, - 0.717, - 0.751, - 0.564, an d f or I n f in eo n E a sy P A C K K = 2.03 \times 1 0^{14}, β = - 4.416, 1285, - 0.463, - 0.716, - 0.761, - 0.5; L E S I T ha s E_{a} = 0.808 eV f or A l b o n d - w i r e f ai l u r e$ $(co m p u t e d w i t h t h e E a sy P A C K se t an d a ss u m e d v a l u es Δ T_{j} = 60 K, t_{o n} = 2 s, I = 10 A, V = 12, D = 300 : N_{f} = 2.1 \times 1 0^{5} a t T_{j, min} = 40^{\circ} C an d 1.2 \times 1 0^{5} a t 90^{\circ} C; r ai s in g Δ T_{j} f r o m 60 t o 100 K c u t s t h e l i f e t o 0.105, a h o tt er minim u m f r o m 40 t o 90^{\circ} C t o 0.57)$
Power-cycling life with a swing-dependent exponent (Semikron)	$N_{f} = A_{0} A_{1}^{β} Δ T_{j}^{α - β} e^{E_{a} / (k_{B} T_{j m})} \frac{C + t _{o n}^{γ}}{C + 2 ^{γ}} k_{t hi c k n ess}$	$β = e^{- (Δ T_{j} - T_{0}) / λ}$	$A_{0}, A_{1} : sc a l e co n s t an t s$ $β : s w in g - d e p e n d e n t cor r ec t i o n$ $Δ T_{j} : j u n c t i o n t e m p er a t u r e s w in g in K$ $α : C o f f in - M an so n e x p o n e n t$ $E_{a} : a c t i v a t i o n e n er g y in J$ $k_{B} : B o l t z mann co n s t an t 1.38 \times 1 0^{- 23} J/K$ $T_{j m} : m e an j u n c t i o n t e m p er a t u r e in K$ $C : t im e coe f f i c i e n t$ $t_{o n} : h e a t in g t im e in s (n or ma l i se d t o 2 s)$ $γ : t im e e x p o n e n t$ $k_{t hi c k n ess} : 1 f or I GB T s t o 1200 V, 0.65 f or 1700 V I GB T s an d C A L d i o d es, 0.5 f or r ec t i f i er s, 0.33 f or S i C t o 1200 V$ $T_{0}, λ : co n s t an t s o f t h e cor r ec t i o n; S e mik r o n A N 21 - 001 : so l d er e d c hi p s w i t h ba se pl a t e A_{0} = 2.9 \times 1 0^{9}, A_{1} = 60, T_{0} = 40 K, λ = 17 K, α = - 4.3, E_{a} = 4.5 \times 1 0^{- 20} J, C = 1, γ = - 0.75 (w i t h o u t ba se pl a t e C = 0.38, γ = - 0.7; s in t er e d c hi p s A_{0} = 2.05 \times 1 0^{11}, T_{0} = 38 K, E_{a} = 2.54 \times 1 0^{- 20} J, C = 1.44, γ = - 1.21), p u b l i s h e d a t 15 p er ce n t f ai l u r e p r o babi l i t y, ba se pl a t e so l d er f a t i g u e e x c l u d e d$ $(co m p u t e d : f or so l d er e d c hi p s w i t h Δ T_{j} = 60 K an d t_{o n} = 2 s, N_{f} = 8.74 \times 1 0^{5} f or 40 t o 100^{\circ} C an d 2.61 \times 1 0^{5} f or 90 t o 150^{\circ} C, w hi c h r e p r o d u ces t h e ab o u t 880 k an d 260 k o f t h e n o t e; t_{o n} = 1 an d 10 s g i v e 1.10 \times 1 0^{6} an d 6.45 \times 1 0^{5}; t h e s in t er e d se t g i v es 9.86 \times 1 0^{5} f or 40 t o 100^{\circ} C; E_{a} = 4.5 \times 1 0^{- 20} an d 2.54 \times 1 0^{- 20} J a r e 0.281 an d 0.159 eV)$
End of life in a power-cycling test	$\frac{V _{C E} ( N ) - V _{C E} ( 0 )}{V _{C E} ( 0 )} \geq x_{V} or \frac{R _{t h} ( N ) - R _{t h} ( 0 )}{R _{t h} ( 0 )} \geq x_{R}$	$D = \sum_{i} \frac{n _{i}}{N _{f, i}} = 1$	$V_{C E} (N) : o n - s t a t e v o l t a g e a t t h e s am e c u r r e n t an d t e m p er a t u r e a f t er N cy c l es (V_{f} or V_{D S} f or d i o d es an d M O S F E T s)$ $x_{V} : t o l er a t e d r i se$ $R_{t h} : j u n c t i o n - t o - c a se t h er ma l r es i s t an ce$ $x_{R} : t o l er a t e d r i se$ $D : a cc u m u l a t e d d ama g e$ $n_{i} : cy c l es o f a s w in g c l a ss in a mi ss i o n p r o f i l e$ $N_{f, i} : cy c l es t o f ai l u r e f or t ha t c l a ss (M in e r^{'} s r u l e : t h e d ama g es o f t h e s w in g c l a sses a dd, an d t h e m o d u l e f ai l s w h e n t h e s u m r e a c h es 1, P l e x im r ain f l o w d e m o); S e mik r o n : 20 p er ce n t r i se o f t h e f or w a r d v o l t a g e, o f R_{t h} or o f t h e s w in g; A l a v i e t a l . : 20 p er ce n t a s a g e n er a l cr i t er i o n, 5 t o 20 p er ce n t r i se o f V_{C E} in o t h er s t u d i es, o n e d e v i ce r an ab o u t 600 k cy c l es a f t er a 5 p er ce n t r i se, an d a 20 p er ce n t R_{t h} r i se p o in t s t o so l d er f a t i g u e; a r i s in g V_{C E} p o in t s t o b o n d w i r es (K u e n z i)$ $(n o n e w n u mb er s; A l a v i e t a l . s a w t h e s u dd e n r i se a t ab o u t cy c l e 26, 000 f or Δ T_{j} = 105 K, w i t h T_{j, ma x} u p 20 p er ce n t an d V_{C E} u p 10.2 p er ce n t)$

Avalanche of Power MOSFETs

Name	Formula	Related Formulae	Variables/Constants
Unclamped inductive switching energy of a power MOSFET	$E_{A S} = \frac{L I _{A S}^{2}}{2 ( 1 - V _{D D} / V _{D S X (S U S)} )}$	$E_{A S} = \frac{L I _{A S} V _{D S X (S U S)}}{R} [1 - K ln (1 + \frac{1}{K})], K = \frac{V _{D S X (S U S)} - V _{D D}}{I _{A S} R}$ $t_{A V} = \frac{L I _{A S}}{V _{D S X (S U S)} - V _{D D}}$	$E_{A S} : s in g l e - p u l se a v a l an c h e e n er g y$ $L : in d u c t an ce$ $I_{A S} : a v a l an c h e c u r r e n t a t t h e s t a r t o f t h e a v a l an c h e$ $V_{D D} : s u ppl y v o l t a g e$ $V_{D S X (S U S)} : s u s t ain e d a v a l an c h e v o l t a g e, ab o u t 1.3 B V_{D S S}$ $R : ser i es r es i s t an ce o f t h e c i r c u i t$ $t_{A V} : a v a l an c h e t im e (f or R = 0, d er i v e d f r o m L d i / d t = - (V_{D S X (S U S)} - V_{D D})); o n se mi A N - 7514 : o l d er d e v i ces f ai l e d b y t u r n - o n o f t h e p a r a s i t i c bi p o l a r, w hi c h d e p e n d s o n c u r r e n t an d t e m p er a t u r e an d i s n o t e n er g y r e l a t e d, w hi l e r ug g e d m o d er n d e v i ces f ai l t h er ma l l y w h e n t h e in t r in s i c c a r r i er s b eco m e co m p a r ab l e w i t h t h e d r i f t - l a y er d o p in g; t h e f ai l u r e c u r r e n t f a l l s w i t h a v a l an c h e t im e a t a s l o p e o f ab o u t - 1/2 o n a l o g - l o g pl o t, so t h e l imi t i s a co n s t an t I_{A S}^{2} t_{A V}, n o t a co n s t an t e n er g y$ $(co m p u t e d f or t h e F D B 8444 e x am pl e, L = 1 mH, V_{D D} = 0, B V_{D S S} = 40 V : 24.8 A g i v es E_{A S} = 0.308 J (307 m J in t h e n o t e) an d t_{A V} = 477 μ s; t h e co n s t an t I^{2} t v a l u es a r e 0.654 A^{2} s a t 25^{\circ} C (32.4 A, 623 μ s) an d 0.250 a t 150^{\circ} C (23.5 A, 452 μ s); t h e n o t e^{'} s 0.381 a t 100^{\circ} C b e l o n g s t o 25.2 A a t 600 μ s)$
Bipolar turn-on during avalanche	$I_{A V, b} R_{B} \geq V_{B E, o n} (T)$	$\frac{d V _{B E, o n}}{d T} < 0, \frac{d R _{B}}{d T} > 0$	$I_{A V, b} : a v a l an c h e c u r r e n t t ha t f l o w s l a t er a l l y u n d er t h e so u r ce t h r o ug h t h e p - b o d y$ $R_{B} : r es i s t an ce o f t ha t p - b o d y p a t h$ $V_{B E, o n} : v o l t a g e t ha t f or w a r d - bia ses t h e b o d y - so u r ce j u n c t i o n (t h e in e q u a l i t y w r i t es t h e co n d i t i o n t ha t o n se mi A N - 7514 g i v es in w or d s); o n se mi A N - 7514 : t h e d r o p r i ses w i t h t e m p er a t u r e t h r o ug h t h e p os i t i v e coe f f i c i e n t o f t h e r es i s t an ce w hi l e t h e t u r n - o n v o l t a g e f a l l s, so h e a t in g p u s h es t h e d e v i ce t o w a r d t h e bi p o l a r m o d e, an d w h e n f or w a r d bia s i s r e a c h e d t h e d e v i ce f ai l s; m o d er n d es i g n s s u pp r ess i t$ $(n o n e w n u mb er s; w i t h t h e o n se mi s l o p e o f - 104 mA /^{\circ} C a t 0.6 ms t h e f ai l u r e c u r r e n t f a l l s f r o m 33 A a t 25^{\circ} C t o 20 A a t 150^{\circ} C)$

Short Circuit of IGBTs

Name	Formula	Related Formulae	Variables/Constants
Short-circuit energy and withstand time of an IGBT	$E_{S C} = \int_{0}^{t_{S C}} v_{C E} i_{C} d t \approx V_{b u s} [I_{S C, L} t_{D} + I_{S C, r e f} (t_{S C} - t_{D})]$	$Δ T_{j} \approx \frac{E _{S C}}{ρ c A d _{e f f}}$	$E_{S C} : e n er g y d i ss i p a t e d in t h e f a u l t$ $t_{S C} : d u r a t i o n o f t h e s h or t c i r c u i t$ $v_{C E}, i_{C} : co l l ec t or v o l t a g e an d c u r r e n t$ $V_{b u s} : D C - l ink v o l t a g e$ $I_{S C, L} : c u r r e n t b e f or e t h e l imi t in g a c t s$ $t_{D} : d e l a y o f t h e l imi t in g$ $I_{S C, r e f} : l imi t e d c u r r e n t (M o h se n z a d e e t a l ., E q . 10)$ $Δ T_{j} : a d iaba t i c t e m p er a t u r e r i se$ $ρ, c : d e n s i t y an d s p ec i f i c h e a t o f s i l i co n$ $A : a c t i v e a r e a$ $d_{e f f} : d e pt h t ha t i s h e a t e d in t h e p u l se (a d iaba t i c f or m, d er i v e d f r o m t h e h e a t c a p a c i t y ρ c A d_{e f f} o f t h e h e a t e d l a y er); M o h se n z a d e e t a l . : I GB T s s u r v i v e ab o u t 10 μ s o f s h or t c i r c u i t b e f or e t h er ma l r u na w a y, f ai l u r e m o d es a r e l a t c h - u p, g a t e osc i l l a t i o n an d o x i d e b r e ak d o w n, se l f - t u r n - o f f, t h er ma l r u na w a y a f t er t u r n - o f f f r o m t h e l e ak a g e c u r r e n t (50 t o 100 μ s l a t er), an d d es a t u r a t i o n d e t ec t i o n t ak es 1 t o 5 μ s; w i t h o u t l imi t in g, 96 kW f or 7 μ s t oo k t h e a v er a g e j u n c t i o n t o ab o u t 198^{\circ} C in t h e i r s im u l a t i o n, w i t h l imi t in g ab o u t 116^{\circ} C, f r o m 75^{\circ} C$ $(co m p u t e d w i t h t h e i r T ab l e 5 v a l u es, 600 V, 95 A, 45 A, 200 ns an d 7 μ s : E_{S C} = 0.195 J; t h e u n l imi t e d 96 kW f or 7 μ s i s 0.672 J)$

Thyristor Limits

Name	Formula	Related Formulae	Variables/Constants
Rate-of-rise limits of a thyristor	$L_{R} \geq \frac{V _{D}}{( d i / d t ) _{cr}}, L_{R} \geq \frac{2 V _{V}}{( d i / d t ) _{cr}}$	$i_{d v / d t} = C_{j} \frac{d v}{d t} \geq I_{GT} \Rightarrow false turn-on$	$L_{R} : ser i es in d u c t an ce t ha t k ee p s t h e c u r r e n t r i se b e l o w t h e cr i t i c a l v a l u e$ $V_{D} : o f f - s t a t e v o l t a g e b e f or e t u r n - o n, (d i / d t)_{cr} : cr i t i c a l r a t e o f r i se o f t h e o n - s t a t e c u r r e n t$ $V_{V} : R M S l in e - t o - l in e v o l t a g e$ $i_{d v / d t} : d i s pl a ce m e n t c u r r e n t t h r o ug h t h e j u n c t i o n c a p a c i t an ce C_{j}, d v / d t : r a t e o f r i se o f t h e o f f - s t a t e v o l t a g e$ $I_{GT} : g a t e t r i g g er c u r r e n t; S e mik r o n A N 18 - 002 : (d i / d t)_{cr} i s t es t e d a t 3 I_{T (A V)}, 85^{\circ} C c a se an d I_{G} = 5 I_{GT}, f a l l s w i t h t e m p er a t u r e an d v o l t a g e an d r i ses w i t h a s t r o n g er an d f a s t er g a t e p u l se, an d t h e s n u bb er d i sc ha r g e c u r r e n t a t t u r n - o n m u s t s t a y b e l o w 50 A; (d v / d t)_{cr} i s t es t e d t o 66 p er ce n t o f V_{D R M} a t T_{v j, ma x} w i t h an o p e n g a t e; M o t or o l a A N - 849 : a f a s t c u r r e n t r i se co n ce n t r a t es t h e p o w er n e a r t h e g a t e in h o t s p o t s, an d d v / d t t r i g g er in g g e t s e a s i er a t hi g h t e m p er a t u r e$ $(co m p u t e d w i t h a ss u m e d v a l u es : 400 V R M S an d 100 A/ μ s n ee d L_{R} o f a t l e a s t 5.7 μ H; 20 pF a t 500 V/ μ s d r i v es 10 mA, e n o ug h t o f i r e a se n s i t i v e g a t e)$
Surge current and the fusing integral	$\int i^{2} d t \approx \frac{I _{T S M}^{2} t}{2}$	$V_{D C} \leq 0.5 V_{D R M}$ $\frac{Δ V _{B R}}{V _{B R} Δ T} \approx 0.11%/ K$	$\int i^{2} d t : l o a d in t e g r a l o f a ha l f - s in e s u r g e$ $I_{T S M} : p e ak n o n - r e p e t i t i v e s u r g e c u r r e n t$ $t : ha l f - cy c l e d u r a t i o n (8.33 ms a t 60 H z, 10 ms a t 50 H z)$ $V_{D C} : s t e a d y D C b l oc k in g v o l t a g e$ $V_{D R M} : r e p e t i t i v e p e ak o f f - s t a t e v o l t a g e$ $V_{B R} : b l oc k in g v o l t a g e$ $T : j u n c t i o n t e m p er a t u r e; M o t or o l a A N - 849 : t h e a pp r o x ima t i o n an d ab o u t 100 n o n - r e p e t i t i v e s u r g es o v er t h e l i f e; S e mik r o n A N 18 - 002 : s t e a d y D C n o t ab o v e 50 p er ce n t o f V_{D R M}, a s a f e t y f a c t or o f 1.5 t o 2.5 f or in d u s t r ia l s u ppl i es, an d t h e b l oc k in g v o l t a g e c han g es b y ab o u t 0.11 p er ce n t p er K; L i tt e l f u se A N 1009 : k ee p t h e c u r r e n t b e l o w 75 p er ce n t o f t h e r a t in g, an d t h e m os t co mm o n f ai l u r e i s a s h or t b e tw ee n t h e main t er mina l s$ $(co m p u t e d w i t h an a ss u m e d I_{T S M} = 1000 A : 4165 A^{2} s a t 60 H z an d 5000 A^{2} s a t 50 H z; a c han g e o f 100 K m o v es t h e b l oc k in g v o l t a g e b y ab o u t 11 p er ce n t)$

Cosmic-Ray Failures

Name	Formula	Related Formulae	Variables/Constants
Cosmic-ray failure rate against voltage, temperature and altitude	$λ (V, T_{j}, h) = λ_{1} (\frac{λ _{2}}{λ _{1}})^{\frac{V - V _{1}}{V _{2} - V _{1}}} F (T_{j}) A (h)$	$F (T_{j}) = e^{- (T_{j} - 25) /47.6}, A (h) = 2^{h /1000}$ $λ_{m o d u l e} = \sum_{i} λ_{c hi p, i} A_{c hi p, i}$	$λ : f ai l u r e r a t e in F I T (1 F I T = o n e f ai l u r e in 1 \times 1 0^{9} d e v i ce h o u r s)$ $λ_{1}, λ_{2} : r a t es a t t h e v o l t a g es V_{1} an d V_{2} g i v e n b y t h e mak er$ $V : D C v o l t a g e a cr oss t h e b l oc k in g d e v i ce$ $F : t e m p er a t u r e f a c t or w i t h T_{j} in d e g r ees C e l s i u s (S e mik r o n a f t er A B B)$ $A : a l t i t u d e f a c t or w i t h h in m e t r es$ $λ_{c hi p} : r a t e p er a r e a o f e a c h c hi p in FIT/c m^{2}$ $A_{c hi p} : c hi p a r e a; S e mik r o n A N 17 - 003 : in t er p o l a t e o n l y, s in ce e x t r a p o l a t i o n er r s b y m or e t han t h r ee or d er s o f ma g ni t u d e; A (h) er r s b y l ess t han 10 p er ce n t t o 3000 m an d o v er es t ima t es b y 2 a t 5000 m; t h e f l ux v a r i es ab o u t 20 p er ce n t w i t h t h e so l a r cy c l e an d T o k y o ha s ab o u t 40 p er ce n t l ess t han N e w Y or k; an I GB T a t 900 V ha s ab o u t 6 FIT/c m^{2} an d a f r ee - w h ee l in g d i o d e b e l o w 0.02; F in d e i se n e t a l . : t h e J u n g f r a u j oc h t o N e w Y or k f l ux r a t i o i s 13, an d t h e r a t e f a l l s s ha r pl y b e l o w a c ha r a c t er i s t i c bia s, w hi c h t h e m o d e l d oes n o t p r e d i c t$ $(co m p u t e d : F = 0.35, 0.12 an d 0.072 a t 75, 125 an d 150^{\circ} C; A = 2 a t 1000 m, 4 a t 2000 m an d 12.0 a t 3580 m, a g ain s t t h e m e a s u r e d r a t i o o f 13 f or J u n g f r a u j oc h; ha l f w a y b e tw ee n p o in t s o f 1 an d 100 F I T t h e in t er p o l a t i o n g i v es 10 F I T; s i x m o d u l es o f 2 s w i t c h es a t 41 F I T g i v e 492 F I T, a s in t h e n o t e)$

Important Questions

Ques: Which power-cycling model should be used: LESIT, CIPS08 or a maker’s model?

The models disagree with test data by large factors. Alavi et al. report that CIPS08 predicted about 2 and 2.4 times the real life in two studies and 2.3 times in a third, and that LESIT predicted about 12 times; their own discrete IGBTs failed near 26,000 cycles at a swing of 105 $K$ . The Semikron model is fitted to the maker’s own product lines and reproduces its published examples, $8.74 \times 1 0^{5}$ and $2.61 \times 1 0^{5}$ cycles for a 60 $K$ swing (computed against about 880k and 260k). A maker’s model for the actual module is therefore the more likely to be right, with CIPS08 as a cross-check; LESIT, built on older modules, should not be used alone. Any model is valid only inside its test range of swing, heating time and temperature. Reference: O. Alavi, W. De Ceuninck and M. Daenen, ‘Optimizing Insulated-Gate Bipolar Transistors’ Lifetime Estimation: A Critical Evaluation of Lifetime Model Adjustments Based on Power Cycling Tests’, Energies 17, 2616 (2024), DOI 10.3390/en17112616 (fetch-tool summary); Semikron Danfoss, ‘Power Cycle Model for IGBT Product Lines’, Application Note AN 21-001, Rev. 02, 23 July 2024 (fetch-tool summary).

Ques: Does the minimum or the mean junction temperature set the power-cycling life?

The two models use different temperatures and give different sensitivities. CIPS08 uses the minimum junction temperature in Bayerer’s form, and with the EasyPACK constants raising it from 40 to 90 $^{\circ} C$ at a 60 $K$ swing cuts the life to 0.57 (computed). The Semikron model uses the mean temperature and an activation energy of 0.281 $eV$ , and the same shift, from 40 to 100 $^{\circ} C$ to 90 to 150 $^{\circ} C$ , cuts the life to 0.30 (computed, $8.74 \times 1 0^{5}$ to $2.61 \times 1 0^{5}$ ). The difference comes from the fitted exponents, not from physics. For a module of the maker’s own line, the Semikron value is the more likely, since it is fitted to those parts; for others the spread should be taken as the uncertainty (my reading). Reference: Semikron Danfoss, ‘Power Cycle Model for IGBT Product Lines’, Application Note AN 21-001, Rev. 02, 23 July 2024 (fetch-tool summary); O. Alavi, W. De Ceuninck and M. Daenen, ‘Optimizing Insulated-Gate Bipolar Transistors’ Lifetime Estimation: A Critical Evaluation of Lifetime Model Adjustments Based on Power Cycling Tests’, Energies 17, 2616 (2024), DOI 10.3390/en17112616 (fetch-tool summary).

Ques: Is the end of life a 5 or a 20 per cent rise of the on-state voltage?

Both are used. Semikron defines the end of life as a 20 per cent rise of the forward voltage, of the thermal resistance or of the temperature swing, and Alavi et al. call 20 per cent the general criterion, but note that other studies use 5 to 20 per cent of $V_{C E}$ . They also report that one device ran about 600k more cycles after a 5 per cent rise, so 5 per cent marks the first lifted wire rather than the end. In their own test the on-state voltage rose by 10.2 per cent and the peak temperature by 20 per cent at failure. The 20 per cent criterion is therefore the more likely to mark real end of life, and 5 per cent is a useful early warning, especially when a design cannot tolerate the extra loss (my reading). Reference: Semikron Danfoss, ‘Power Cycle Model for IGBT Product Lines’, Application Note AN 21-001, Rev. 02, 23 July 2024 (fetch-tool summary); O. Alavi, W. De Ceuninck and M. Daenen, ‘Optimizing Insulated-Gate Bipolar Transistors’ Lifetime Estimation: A Critical Evaluation of Lifetime Model Adjustments Based on Power Cycling Tests’, Energies 17, 2616 (2024), DOI 10.3390/en17112616 (fetch-tool summary).

Ques: Is the avalanche rating of a power MOSFET an energy?

Not for modern devices. The onsemi note says that the failure current of rugged MOSFETs falls with the avalanche time at a slope of about -1/2 on a log-log plot, so the limit is a constant $I_{A S}^{2} t_{A V}$ , and that older devices failed by turn-on of the parasitic bipolar, which is not energy related either. For the FDB8444 at 1 $mH$ , the constant-energy point of 0.308 $J$ is 24.8 $A$ for 477 $μ s$ , while the constant $I^{2} t$ line at 25 $^{\circ} C$ , 0.654 $A^{2} s$ , allows 32.4 $A$ for 623 $μ s$ (computed). The data-sheet $E_{A S}$ is therefore a test point, not a limit at other inductances. The note’s 0.381 $A^{2} s$ at 100 $^{\circ} C$ is printed next to 27.1 $A$ , but 0.381 belongs to 25.2 $A$ at 600 $μ s$ (computed), which is its own derived value. Reference: onsemi, ‘Single-Pulse Unclamped Inductive Switching: A Rating System’, Application Note AN-7514/D, Rev. 2, December 2023 (fetch-tool summary).

Ques: How long can an IGBT survive a short circuit, and when does it fail?

Mohsenzade et al. state that IGBTs survive about 10 $μ s$ of short circuit before thermal runaway, so the desaturation detection, with a blanking time of 1 to 5 $μ s$ , must act within that time. Failure can come during the pulse by latch-up or gate oscillation, at turn-off, or 50 to 100 $μ s$ after a successful turn-off, when the hot chip runs away through its leakage current. In their simulation an unlimited fault of 96 $kW$ for 7 $μ s$ , 0.672 $J$ , took the junction from 75 to about 198 $^{\circ} C$ , above the 175 $^{\circ} C$ rating, while limiting the current to 45 $A$ kept it near 116 $^{\circ} C$ with 0.195 $J$ (computed). Lowering the current is therefore as effective as faster detection, since the energy scales with both. Reference: S. Mohsenzade, J. Naghibi and K. Mehran, ‘Reliability Enhancement of Power IGBTs under Short-Circuit Fault Condition Using Short-Circuit Current Limiting-Based Technique’, Energies 14, 7397 (2021) (author copy, fetch-tool summary); Electrical Stress page of this vault.

Ques: Why does the cosmic-ray failure rate of a power device rise when it is cold?

Because the avalanche that a neutron starts multiplies more easily at low temperature, where the ionisation coefficients are higher (Ayalew, section 3.5.3: the coefficients fall as the temperature rises, through the phonon term, stated there for SiC). Semikron uses the ABB temperature factor $F = e^{- (T_{j} - 25) /47.6}$ , so the rate at 125 $^{\circ} C$ is 0.12 of that at 25 $^{\circ} C$ (computed), and a cold device at full voltage is the worst case. The voltage matters far more: the rate rises almost exponentially with the blocking voltage, and Semikron warns that extrapolation can err by more than three orders of magnitude. Altitude doubles the rate about every 1000 $m$ , which gives 12.0 at 3580 $m$ (computed) against the flux ratio of 13 that ABB measured between Jungfraujoch and New York. A doubling rule is therefore close enough to that height. Reference: U. Schilling, Semikron Danfoss, ‘Cosmic Ray Failures in Power Electronics’, Application Note AN 17-003, Rev. 01, 6 September 2024 (fetch-tool summary); Ch. Findeisen, E. Herr, Th. Stiasny and H. R. Zeller (ABB Semiconductors), ‘Cosmic rays interacting with biased high power semiconductor devices’, HFSJG Activity Report 1999/2000 (fetch-tool summary; exponents garbled); T. Ayalew, ‘SiC Semiconductor Devices Technology, Modeling, and Simulation’, PhD thesis, TU Wien, section 3.5.3 ‘Impact Ionization’.

Ques: Does the Microchip example of a neutron failure rate add up?

Not as printed. The note starts from 2500 FIT for one device, multiplies by a die-volume factor of 1.6, a latitude factor of 0.35 and a flux ratio of 500/13, and reports 5384 FIT. The product is 53,800 FIT (computed), ten times larger, so either a factor of 10 is missing from the stated inputs or the result has lost a digit. The calendar-time conversion by 8/24 that follows is right. Until the original is checked, the larger value is the more likely to follow from the stated inputs, and the example should not be used as a reference number. The note also says that its altitude equations are valid only to about 15 $km$ , and that the value for 300 $m$ should be used below 300 $m$ . Reference: Microchip Technology, ‘Excess Failure Rate Calculation Due to Neutron Flux with SiC MOSFETs and Schottky Diodes’, AN4589 (DS00004589B), 2024 (fetch-tool summary; equations garbled).

Ques: Why does a thyristor fail at a fast current rise, and how is it prevented?

At turn-on only the region next to the gate conducts, and the conducting area spreads outward with time, so a current that rises too fast flows through a small area and makes hot spots that destroy or degrade the blocking junction (Motorola AN-849). Semikron specifies $(d i / d t)_{cr}$ at three times the average current, 85 $^{\circ} C$ case and a gate current five times the trigger current, and the capability falls with temperature and voltage and rises with a stronger, faster gate pulse. A series inductance of at least $2 V_{V} / (d i / d t)_{cr}$ limits the rise: 5.7 $μ H$ for 400 $V$ and 100 $A/ μ s$ (computed, assumed values). The snubber must not discharge more than 50 $A$ into the device at turn-on. Reference: Motorola Semiconductor Products, ‘Guide to Thyristor Applications’, Application Note AN-849 (year not stated) (fetch-tool summary); Semikron Danfoss, ‘Thyristor Triggering and Protection of Diodes and Thyristors’, Application Note AN 18-002, Rev. 03, 11 November 2024 (fetch-tool summary).

References

O. Alavi, W. De Ceuninck and M. Daenen, ‘Optimizing Insulated-Gate Bipolar Transistors’ Lifetime Estimation: A Critical Evaluation of Lifetime Model Adjustments Based on Power Cycling Tests’, Energies 17, 2616 (2024), DOI 10.3390/en17112616 (fetch-tool summary).
Semikron Danfoss, ‘Power Cycle Model for IGBT Product Lines’, Application Note AN 21-001, Rev. 02, 23 July 2024 (fetch-tool summary).
onsemi, ‘Single-Pulse Unclamped Inductive Switching: A Rating System’, Application Note AN-7514/D, Rev. 2, December 2023 (fetch-tool summary).
S. Mohsenzade, J. Naghibi and K. Mehran, ‘Reliability Enhancement of Power IGBTs under Short-Circuit Fault Condition Using Short-Circuit Current Limiting-Based Technique’, Energies 14, 7397 (2021) (author copy, fetch-tool summary).
Electrical Stress page of this vault.
U. Schilling, Semikron Danfoss, ‘Cosmic Ray Failures in Power Electronics’, Application Note AN 17-003, Rev. 01, 6 September 2024 (fetch-tool summary).
Ch. Findeisen, E. Herr, Th. Stiasny and H. R. Zeller (ABB Semiconductors), ‘Cosmic rays interacting with biased high power semiconductor devices’, HFSJG Activity Report 1999/2000 (fetch-tool summary; exponents garbled).
T. Ayalew, ‘SiC Semiconductor Devices Technology, Modeling, and Simulation’, PhD thesis, TU Wien, section 3.5.3 ‘Impact Ionization’. Available: https://iue.tuwien.ac.at/phd/ayalew/node74.html.
Microchip Technology, ‘Excess Failure Rate Calculation Due to Neutron Flux with SiC MOSFETs and Schottky Diodes’, AN4589 (DS00004589B), 2024 (fetch-tool summary; equations garbled).
Motorola Semiconductor Products, ‘Guide to Thyristor Applications’, Application Note AN-849 (year not stated) (fetch-tool summary).
Semikron Danfoss, ‘Thyristor Triggering and Protection of Diodes and Thyristors’, Application Note AN 18-002, Rev. 03, 11 November 2024 (fetch-tool summary).
R. Kuenzi, Paul Scherrer Institut, arXiv:1607.01578 (fetch-tool summary).
Littelfuse (Teccor brand thyristors), ‘Miscellaneous Design Tips and Facts’, Application Note AN1009, 2013 (fetch-tool summary).
Plexim GmbH, ‘Rainflow Counting and Lifetime Prediction’, PLECS demo model. Available: https://docs.plexim.com/plecs-demos/motor-drives/rainflow-counting/.

Quartz 5

Explorer

Silicon Power MOSFET, IGBT and Thyristors