This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2018.2826046, IEEE
Transactions on Power Electronics
1

Stray Impedance Measurement and Improvement of

high power IGCT Gate Driver Units
Zhengyu Chen, Zhanqing Yu, Xuan Liu, Jiapeng Liu and Rong Zeng, Senior Member, IEEE

Thyristor Transistor
Abstract—Turn-off capability of integrated gate commutated
thyristor (IGCT) is highly dependent on the commutation Anode iA
iA IGrr vAK
capability of its gate driver unit (GDU). To enhance it, stray vAK
+
P
impedance is the crucial limitation, which has not been obtained iG
accurately before. Thus in this paper, acquisition method of N-

impedance based on experimental measurements was carefully Gate

iG
analyzed, including electrolytic and ceramic capacitors, P 0
Qoff vGK
MOSFETs, PCB and GCT housing. And each part of stray vGK N+
TD
impedance was obtained based on a series of experiments with C1 +
6-inch GCTs and two types of conventional GDUs. To further vC
vC Cathode -20V
enhance commutation speed, an improved GDU was developed,
reducing the total stray inductance from 0.395 nH to 0.037 nH, T0 T1 T2 T3
stray resistance from 0.444 mΩ to 0.227 mΩ. An extremely high Fig. 1. IGCT turn-off operation schematic and waveforms.
current of 12.5 kA was successfully turned off, which is the
maximum turn-off current level so far. Besides, the maximum voltage vAK rises - the so-called hard-drive. In another word, the
commutation capability was tested as over 17 kA. Consequently, commutation time (T1) must be shorter than the delay time (TD =
no further optimization of GDU is necessary. However, there was T1+T2, T2 is the time of excess carrier recombination), which is
still about 1.0 nH existing inside the GCT housing we used. A determined by the design of GCT wafer, commonly about 1 μs.
novel housing with lower stray inductance is expected for higher Obviously, stray impedance (inductance LS and resistance RS)
commutation speed in the future. of turn-off circuit is the crucial limitation to reduce T1 and
enhance commutation capability [8][11]. To ensure the
Index Terms—IGCT, gate driver unit, turn-off, stray sufficiently low impedance, a large number of electrolytic
impedance. capacitors and MOSFETs are paralleled to reduce their
equivalent inductance and resistance, at the expense of a large
I. INTRODUCTION volume. Meanwhile, to minimize the circuit loop area which
will also yield high LS, six-layer printed circuit board (PCB) is
T he integrated gate commutated thyristor (IGCT) has been
developed since the mid 1990’s [1]-[4] and widely applied
to lots of power equipment, such as medium voltage converters
adopted for GDU and integrated to the GCT housing [1].
However, the typical LS is about 3 - 5 nH, meaning that only 4.0
- 6.6 kA can be commutated to GDU in 1 μs [2]. It meets the
and dc circuit breakers [5] -[7]. As a thyristor-based device, requirement of 4-inch GCTs, but no longer adapts to 6-inch
low on-state voltage, huge surge current capability, excellent GCTs. Recent years, some renewed efforts have been done. For
reliability and large utilization of silicon area are the example, by paralleling ceramic capacitors to electrolytic
remarkable advantages of IGCT [8]. capacitors, which has lower equivalent inductance and
When designing a high power, large size IGCT, high resistance, the commutation capability of GDU was enhanced to
turn-off capability is a major target, which is highly dependent 7 kA in 0.8 μs, and its turn-off ability was increased by up to
on the commutation capability of gate driver unit (GDU). As 25% [8][9]. And some discussions about the combination of
shown in Fig. 1, the common turn-off circuit of GDU consists of electrolytic and ceramic capacitors was presented, on the view
a precharged electrolytic capacitor bank C1 (usually 20 V) and a of recharging process and minimum time between two turn-off
MOSFET bank Qoff as the switching element [9]. During events[10]. Besides, according to the characteristics comparison
turn-off process, the total anode current iA must be commutated of different types of capacitors, GDU was improved for the low
from cathode of GCT wafer to GDU before the anode-cathode temperature applications by utilizing the combination of
different capacitors [11]. Until now, the maximum turn-off
current up to 9.5 kA at 2.8 kV have been realized by the 150 mm
Manuscript received September 16, 2017; revised December 17, 2017; reverse conducting GCT wafer and GDU with low inductance
accepted April 2, 2018. This work was supported by National Key Research and [12]. Furthermore, some new concepts have been proposed such
Development Program of China (2017YFB0903203) and the Science and
Technology Project of State Grid Corporation. (SGZJ0000BGJS1600326).
as internally commutated thyristor (ICT), which greatly reduced
(Corresponding author: Rong Zeng, Zhanqing Yu.) the stray inductance of turn-off circuit loop by integrating the
The authors are with the Department of Electrical Engineering, Tsinghua capacitors and MOSFETs into the housing [13][14].
University, Beijing 100084, China (e-mail: [email protected]. It is essential to figure out the accurate LS and RS of each part
edu.cn; [email protected]; liuxuan14@126. com; [email protected]; in turn-off circuit, not only for estimating the maximum
[email protected]).

0885-8993 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2018.2826046, IEEE
Transactions on Power Electronics
2

commutation capability, but also for guiding the path of future

optimization. However, there was only a rough estimated LS of
GCT housing and the whole GDU. Neither accurate value of
each part on GDU nor detailed acquisition method has been
provided before. Furthermore, RS was not considered.
Thus in this paper, combining turn-off circuit model, the
acquisition method of LS and RS based on experimental
measurement is carefully analyzed. With conventional GDUs
applied to 6-inch GCT, the value of each part in turn-off circuit
loop is obtained, which is an important reference for
improvement. Furthermore, a new GDU with extremely low LS
and RS is developed, and verified by experiments.

II. ACQUISITION METHOD OF STRAY IMPEDANCE

A. Stray Impedance Distribution in Turn-off Circuit

Fig. 2 is the conventional 6-inch GDU and the simplified
physical structure of turn-off circuit. A considerable number of
electrolytic capacitors and MOSFETs are paralleled to achieve
as low impedance as possible. Besides, to investigate the Fig. 2. Conventional GDU applied to 6-inch GCT and physical structure of
improvement of ceramic capacitors, two types of GDUs (GDU1, turn-off circuit.
only with electrolytic capacitors C1, and GDU2, with both
electrolytic capacitors C1 and ceramic capacitors C2) are between gate and cathode electrodes of GCT wafer, but the
analyzed. measured one between gate and cathode measurement points on
In order to obtain stray impedances, some voltage and GDU, where a voltage probe is easier to be placed. Similarly, vC
current should be measured during experiments, including is the measured voltage of capacitors including their inner
gate-cathode voltage vGK, capacitor voltage vC, anode current iA, equivalent series impedance, but vC1 is the real voltage of
and GDU current iG. The positions of voltage and current probes electrolytic capacitors.
are shown in Fig. 2. The measurement point of vGK should be as
close to the GCT housing as possible. The direct measurement B. Calculation Method of Stray Impedance
of iG based on the available 6-inch GCT housing is impractical, As the flowing path of iA varies with working state of IGCT,
however, it can be obtained indirectly in proportion to single different stray impedances can be acquired respectively based
MOSFET current iMOS. Considering the current deviation of on the measurement results at different state (T0-T2 in Fig. 1).
different MOSFETs as well as capacitors, three positions of iMOS 1) On-state T0
(the gate signals of MOS1 and MOS3 is the furthest from the During on-state, the MOSFETs in turn-off circuit Qoff
public point, and that of MOS2 is the nearest) and two positions remains off-state (iG = 0). When iA keeps constant, RK can be
of vC are measured to compare with each other, as shown in calculated based on the measurement of vGK and iA, expressed
Fig.2. as:
According to Fig. 2, LS and RS distributed in turn-off circuit
mainly consists of five parts, as following. vGK vGK
RK  
(1) RH1 and LH1 are the impedance between gate contact on  iA  iG  (1)
iA
GCT wafer inside housing and gate measurement point on
2) Turn-off Transient During Commutation T1
GDU.
When Qoff is turned on, commutation begins forced by the
(2) RM+P and LM+P are the impedance between gate and
precharged capacitors. Fig. 3 is the corresponding schematic
cathode measurement point on GDU, except for capacitor bank.
model during commutation.
They are mainly from MOSFETs and PCB.
In the condition of using GDU1 in Fig. 3(a), based on the
(3) RC1 and LC1 are the equivalent series impedance of C1. If
Kirchhoff’s Voltage Law (KVL) and Kirchhoff’s Current Law
C2 are paralleled to C1, the equivalent series resistance RC2 of C2
(KCL), vGK and vC can be expressed as:
also need to be considered. Nevertheless, the equivalent series
inductance of C2 is so low that can be negligible.
(4) RH2 and LH2 are the impedance between cathode contact vGK   iA -iG   RK  RH  iG  ( LH  LK )  diG dt
(2)
on GCT wafer inside housing and cathode measurement point
on GDU. vGK  vC  RM+P  iG  LM+P  diC1 dt
(3)
(5) LK is the equivalent impedance of gate-cathode junction
of GCT wafer. And RK is its equivalent resistance which varies vC  vC1  RC1  iG  LC1  diG dt
(4)
to the cathode current, appearing like a diode.
For simplification, part (1) and (4) can be merged as RH and When utilizing GDU2, the equivalent series resistance RC2
LH. In fact, they are mainly the stray impedance of GCT housing. and voltage vC2 are added to the schematic, as shown in Fig. 3(b).
Ignoring the asymmetrical distribution, paralleled MOSFETs Considering that iG is divided by two types of capacitors, (4)
and capacitors are idealized as no current deviation. Moreover, should be replaced by the equations below:
it should be noted that vGK in Fig. 2 is not the real voltage

0885-8993 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2018.2826046, IEEE
Transactions on Power Electronics
3

iA iA

RM+P RH iG
RM+P LM+P RH LH iG

LK
vGK
vGK RC1 vC1
RC1 LC1 vC1 RK +
+

vC (a)
(a)
iA
iA

RM+P LM+P RH LH iG RM+P RH iG

RC2 vC2 LK iC1 RC2 vC2

iC1 +
+ vGK vGK
iC2 RC1 LC1 vC1 RK iC2 RC1 LC1 vC1
+ +
vC vC
(b) (b)
Fig. 3. Schematic model of turn-off circuit during commutation T1 using (a) Fig. 4. Schematic model of turn-off circuit at the beginning of T2 using (a)
GDU1 and (b) GDU2. GDU1 and (b) GDU2.

(iii) Then using GDU2, RK, RM+P and LM+P can be calculated
vC  vC1  RC1  iC1  LC1  diC1 dt again by (1), (3) and (9), as comparisons to the results of GDU1.
(5)
(iv) Still using GDU2, RC2 can be obtained utilizing
vC  vC2  RC2  iC2 calculated RC1 and LC1, based on (5) - (8).
(6)
(v) Referred to (2), RH and LH+LK can be calculated using
vC2  20   iC2 dt C2 GDU1. The repetitive calculated results of GDU2 can also be
(7) comparisons.
iG  iC1  iC2 It is important to pay attention to that RK in (2) is the
(8) dynamic value during commutation, other than the steady one
where vC1, vC2, iC1 and iC2 are the voltage and current of C1 during on-state. However, the dynamic RK is unavailable based
and C2. Considering the sufficient capacity of C1, vC1 can be on experiments. In fact, as cathode current reduces during
idealized as constant voltage sources of 20 V. commutation, the dynamic RK increases more steeply than the
3) Turn-off Transient After Commutation T2 steady one. It is caused by the deviation of segment turn-off
After commutation and the reverse recovery of gate-cathode storage time due to the distance from segment rings to the gate
junction, it is under reverse bias so that RK appears high contact, as well as by the capacitance effect of gate-cathode
resistance, meaning that iG equals to iA. Furthermore, during T2 junction. Limited by the unavailable dynamic RK, RH cannot be
after reverse recovery, the voltage drop caused by LH and LM+P obtained based on experiments neither. However, it could be
can be ignored, due to that iA maintains constant for a short simulated based on a three dimensional finite element model of
while before vAK rises. Similarly, LC1 can also be ignored when GCT housing.
using GDU1. However, it should be taken into account when Fortunately, benefited from the small values of RK and RH,
using GDU2, because of the current dynamic sharing by C1 and their fluctuation has few influence on the calculation of LH+LK,
C2. Consequently, the turn-off circuit schematic during T2 after referred to (2). To further avoid the effect of dynamic RK and
reverse recovery is shown in Fig. 4. inaccurate RH on the result of LH+LK, the period at the start of
When using the GDU1 in Fig. 4(a), vGK and vC can be commutation should be adopted to calculate, where iG is very
expressed as: low and RK changed barely.
vGK  vC  RM+P  iG (9)
III. EXPERIMENT AND CALCULATION RESULTS
vC  vC1  RC1  iG (10) To obtain stray parameter distribution in turn-off circuit, an
experimental platform was established. And a series of
When using GDU2 in Fig. 4(b), (10) should be replaced by
experiments using 6-inch GCT wafers with two types of GDUs
(5) - (8).
were carried out. Based on the measurements of relevant
Based on the experimental measurements, stray impedance
parameters, LS and RS distribution was carefully calculated and
can be calculated as followed.
analyzed.
(i) Using GDU1, during on-state, RK at different cathode
current can be calculated by (1). And during T2 when diG/dt is A. Experimental Platform
almost zero, RM+P and RC1 can be calculated by (9) and (10). The experimental circuit [7] is shown in Fig. 5. It consists of
(ii) Still using the results of GDU1, and already calculated a high-voltage DC source VDC, charging resistor RC, bus
RM+P, RC1 and RK, during commutation T1, it is easy to obtain capacitor CDC, inductor L to limit di/dt, freewheeling diode D,
LM+P and LC1, referred to (3) and (4). snubber RSnub and CSnub, and MOV to protect IGCT from

0885-8993 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2018.2826046, IEEE
Transactions on Power Electronics
VAK IAK VC IMOS VGK 4

K1 RC =5Ω K2 L=10 ~ 200 μH 6 i 24 24 6 60

A IGrr
5 20 20 5

CDC=5mF 4 16 16 4 40
K3 D MOV
VDC 3 i 12 12 3

vAK (kV), iA, iG (kA)

RSnub =0.5Ω G 20
2 8 8 2

vGK, vC (V)
CSn ub =15 μF vAK
1 4 4 1
0

IAK (kA)
UC (V)
IGCT 0 0 0 0
Fig. 5. Experimental circuit for testing IGCT turn-off process.
-1 vGK -13.8V -4 -4 -1 -20

-2 -12.8V -8 -8 -2
overvoltage. To test the commutation capability and obtain stray -3 -40
-3 -12 -12
parameters in a high current, CSnub with large capacitance is used
-4 -16 -16 -4
to reduce turn-off energy and protect the GCT wafer from it. -60
-5 vC -20 -20 -5
Before the test, CDC is charged by VDC. Then open switch K1 and
-6 -16.3V -24 -24 -6
close switch K2. IGCT is turned on for hundreds of -80

microseconds, and then turned off when anode current rises to 127.0 127.5 128.0 128.5 129.0 129.5 130.0
VAK IAK Vt C(μs) IMOS VGK
several kiloamperes. The turn-off current can be adjusted by
turn-on time, voltage of CDC or the inductance of L. (a)
During experiments, iA and iMOS were measured by 6 iA 24 24 6 60
IGrr
Rogowski coil. To shield interference at high di/dt, coaxial 5 20 20 5
40
cables were utilized for measuring vGK and vC. Besides, to figure 4 16 16 4

out the current deviation of different capacitors and MOSFETs, 3 i 12 12 3 20

vAK (kV), iA, iG (kA)

different measurement points of vC and iMOS were tested. 2 G 8 8 2
vAK

vGK, vC (V)
1 4 4 10
B. Experimental Results

IAK (kA)
UC (V)
0 0 0 0
-20
Fig. 6(a) and (b) show typical 5.0 kA turn-off waveforms of -1 -4 -4 -1
vGK -16.7V
a 6-inch GCT with GDU1 and GDU2 respectively. Therein, iG -2 -8 -8 -2 -40
-14.9V
was normalized in proportion to the measured iMOS of single -3 -12 -12 -3
MOSFET, but not the real measured current of GDUs (the same -4 -16 -16 -4 -60
below). In Fig. 6(a), at 127.2 μs, Qoff was turned on. Thereafter -5 -20 -20 -5
vC
commutation began, forced by precharged capacitors. The diG/dt -6 -24 -24 -6
-80

during commutation was about 14 kA/μs. Caused by LS of GDU, 127.0 127.5 128.0 128.5 129.0 129.5 130.0
vGK appeared a steep onset of negative voltage to -13.8 V instead t (μs)
of -20.0 V. Meanwhile, vC rose from -20.0 V to -16.3 V. With iG (b)
increasing, vGK rose from -13.8 V to -12.8 V attributed by RS. At Fig. 6. Experimental results of 5.0 kA turn-off waveforms with a 6-inch GCT,
127.5 μs, iA was commutated completely from cathode to gate using (a) GDU1 and (b) GDU2.
within 300 ns, after which a recovery current spike IGrr was
observed. It was believed that the amount of IGrr and its shape two kinds of GDUs. Fig. 7 is the typical turn-off waveforms
was a good monitor of driver commutation capability. It proved with GDU2 at 2.5 kA, 5.0 kA, 7.5 kA and 10.0 kA. It was
the hard turn-off and indicated potential for switching off even measured that the commutation time was 190 ns, 360 ns, 540 ns
at higher current [9]. Besides, the steep drop of vGK and vC at and 730 ns respectively, increasing linearly with iA. On the
127.6 μs also indicated the end of commutation. After contrary, the delay time TD (before iA decreased) was almost
commutation, iA was totally conducted by GDU1. And GCT fixed at different iA, about 1.5 μs. Besides, the stray resistance
transformed from thyristor mode to transistor mode. After a and poor capacitance of C2 led to a bigger increment of vGK and
delay time of excess carrier recombination, vAK rose and iAK vC at higher iA. During the commutation process of 10.0 kA, vGK
reduced. Affected by RS, vGK did not reduce to -20 V rose from -16.7 V to -12.5 V, which would reduce the
immediately until iA decreased to zero, as well as vC. commutation diG/dt at a certain degree.
Additionally, the difference before vGK and vC was caused by Furthermore, in order to figure out the maximum
RM+P and LM+P. commutation capability of GDU2, a very high surge current of
Fig. 6(b) shows the 5.0 kA turn-off waveforms using GDU2. over 50 kA was generated by shortening the inductor L in Fig. 5.
iA was similar to Fig. 6(a). But during commutation, vGK rose The comparison of 10 kA and 50 kA (maximum value) turn-off
from -16.7 V to -14.9 V, lower than that in Fig. 6(a). Meanwhile, experimental results are shown in Fig. 8. The 50 kA of iA is
vC rose slowly towards a final value of -17.6 V, instead of a steep beyond the range of Fig. 8 and not shown. The IGCT failed
voltage rise under high diG/dt. That confirmed the extremely low short when turning off 50 kA because of unsuccessful
inductance of C2. Benefited from the lower impedance, the commutation. However, it could be found that about 17 kA was
recovery spike occurred 40 ns earlier than Fig. 6(a), indicating a forced to GDU2 within 1.5 μs, which could be thought as the
faster commutation speed of 16 kA/μs. Considering the maximum commutation capability of GDU2.
negligible inductance but poor capacitance of C2, it mainly
conducted iG at first, leading vC2 and vC to rise in a certain degree. C. Stray Impedance Calculation Results and Analysis
As iC1 increased gradually, C1 conducted all of iG, and charge C2 Based on the experiments at different turn-off current with
back to -20 V in the end. two kinds of GDUs, the calculated results of stray resistances
To obtain accurate stray impedance distribution, a series of and inductances are provided in Fig. 9. Considering that the
turn-off experiments at different current were conducted with transient commutation period at low iA is too short, the

0885-8993 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2018.2826046, IEEE
Transactions on Power Electronics RM+P2 RC1 RK RP1+3
RM+P2 RC2 RK RP1+3 5

2.5 kA 5.0 kA 7.5 kA 10.0 kA 0.7 0.7 0.7 0.7

with GDU1
10.0 0.6 0.6 with GDU2 0.6 0.6
iA RK
7.5 0.5 0.5 0.5 0.5
iA, iG (kA)

RH
5.0 0.4 0.4 0.4 0.4
RM+P

RG+P13
R (mΩ)
2.5 0.3 0.3 0.3 0.3

RK
RC1
iG 0.2 0.2 0.2 0.2
0.0
0 0.1 0.1 0.1 0.1

RC1 RC2
0.0 0.0 0.0 0.0
-5 -12.5V
-16.7V
vGK (V)

-0.1 -0.1 -0.1 -0.1

-10 1 2 3 4 L5P2+M 6L 7 L +L 8 9 10
i (kA)C1 P1+3 K

LP2+MA LP1+3+LK
-15 (a)
1.4
1.4 with GDU1 1.4 1.4 1.
-20 1.2
LH+LK with GDU2
0 1.2 1.2 1.2 1.
1.0
1.0 1.0 1.0 1.
-5 0.8

LG+P13+LK
vC (V)

0.8 0.8 0.8 0.

LP2+M2
L (nH)
-10 LC1
0.6
0.6 LC1 0.6 0.6 0.

-15 0.4
0.4 LM+P 0.4 0.4 0.
0.2
-20
0.2 0.2 0.2 0.
0.0
127.0 127.5 128.0 128.5 129.0 129.5 130.0
t (μs) 0.0 0.0 0.0 0.

Fig. 7. Experimental results of turn-off waveforms at 2.5 kA, 5.0 kA, 7.5 kA -0.2
1 2 3 4 5 6 7 8 9 10
and 10.0 kA with a 6-inch GCT, using GDU2.
iA (kA)
300
30 24 24 30 30 24 (b)400 24
Turn off 10 kA
25 20 Fig. 9.
20 Stray impedance
25 25 distribution
20 of 350
GDU1 and
20 GDU2.
17kA Turn off 50 kA 200 300
20 16 250
16 20 20 16 16

15 GDU2, 12
RC2 15needs more
12
15
attention
12
during12commutation, because
iA 200 100
vAK (kV), iA, iG (kA)

10 iG is8 mainly 10
conducted
810 8
by C 2
150
with8 extremely low stray
vGK, vC (V)

iG 0 inductance. RH is 5 obtained based on 4 simulation of a three

5 vAK 4 100
UC (V)

4 5 4
UAK (kV)

dimensional finite element model50of GCT housing, as 0.3 mΩ at

UAK2 (kV)

IMOS2 (A)

UGK2 (V)
IAK2 (kA)
IMOS (A)

UGK (V)

UC2 (V)

0 0 0 00 0 0
-1001MHz. Additionally, RK is the 0 equivalent resistance of
-5 -4 -5 -4
-5 -4 -4
gate-cathode junction at steady-50 turn-on current iA, whose
-10 -8 -10 -8
-10 -8
-200resistance becomes lower as i increase,
-8

vGK A -100 similar to a diode.

-15 -12 -15 -12
-15 -12 -12
It can be found from Fig. 9(b) -150 that the major stray inductance
-20 -300 -16 -20 -16
-20 -16
comes from LH+LK. That means -200
-16
the structure of housing is the
-25 vC -20 -25 -20
-25 -20 -250 -20
main limitation to commutation during turn-off process. Besides,
-30 -400 -24 -30 -24
-30 -24 -300
LC1 and LM+P take nearly equal share of -24 the stray inductance on
127.0 127.5
128.0 128.5 129.0 129.5 130.0
GDU1. However, when C1 is paralleled with C2, LC1 can almost
t (μs) be ignored. LM+P consists of MOSFETs and PCB between vGK
Fig. 8. Experimental results of 10 kA and 50 kA turn-off waveforms using measurement points on GDU except for capacitors. According
GDU2. UAK toIAK UC typical
datasheet, IMOS internal
UGK inductance of paralleled MOSFETs
UAK2 IAK2 UC2 IMOS2 UGK2
bank is 0.12 nH. In another word, the stray inductance of PCB
calculation results at iA higher than 2.5 kA are supposed to be
between vGK measurement points is very low, only about 40 pH.
more accurate. Besides, part of stray impedances is presented
Detecting high di/dt over 10 kA/μs, high voltage such as 4
repeatedly based on the experimental measurements with GDU1
kV, and low voltage below 1 V at the same time was quite
and GDU2, but some inconformity is observed. That is because
difficult. Besides, current deviation of different capacitors and
of the variation in the different commutation speed (different
MOSFETs was not considered in the calculation above. These
equivalent frequency) with two kinds of GDUs.
factors might have a certain effect on the calculation results
Fig. 9(a) is the stray resistance distribution of each part. RM+P
above. Therefore, various measurement points of vC and iMOS
is mainly the equivalent resistance of MOSFETs and PCB. It
(shown in Fig. 2) were performed to verify the results. Fig. 10 is
appears a slightly upward trend from 0.2 mΩ to 0.3 mΩ as iA
the comparison of different measurement points on GDU2
increases, which is accordant with the characteristics of
during turn-off 5.0 kA experiments. At the start of commutation,
MOSFETs. Besides, RC2 is slightly lower than RC1. When using

0885-8993 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
 右下 IMOS1 IMOS2 IMOS3
This article has been accepted forU31
publication
G31in a future
I31 issueImos
of this but左上
journal, Uc has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2018.2826046, IEEE
TransactionsUonC1Power Electronics
UC2
6

80 TABLE I
70 iMOS2 0.0 60 0.0
Characteristic Comparison Between Single MOSFET and DirectFET
60 iMOS1 -2.5 40
Type 60 -2.5
MOSFET DirectFET
50 -5.0 40 Code IRFR3505
-5.0 IRF7749
Breakdown voltage 55 V 60 V
40 iMOS3 -7.5 20
Continuous drain40 current -7.5
30 A 200 A
(VGS = 10 V, 25 ℃)
iMOS (A)

vC (V)
30 -10.0 20 -10.0

Imos左 上

Imos右 上

Imos右 下

Uc左 上
Pulsed drain curernt
280 A 800 A
20 -12.5 (VGS0 = 10 V, 20
25 ℃) -12.5
Static on resistance 11 mΩ 1.1 mΩ
10 vC2 -15.0 0
Internal inductance
-15.0
12 nH Not provided
0
0 -17.5 -20 -17.5

-10 -20.0 mainly

-20
determined by its capacitance,
-20.0
which was not in highly
-20
vC1 demand. The voltage variation △vC after turn-off is expressed
-20 -22.5 as:
-40 -22.5

127.0 127.5 128.0 128.5 129.0 129.5 130.0 vC   iG dt C1 (11)

t (μs)
Fig. 10. Comparison of different measurement points on GDU2 during As a result, the number and capacitance of C1 on GDU3 was
turn-off 5.0 kA experiments. reduced to less than 35% of that on GDU2 (from 151.8 mF to
52.8 mF). And the size was shrunk by 16%. In this way, after
there is little deviation of different iMOS and vC measurement turning off 10 kA, vC would firstly rise to a high level, and then
points. After commutation and recovery current, all the three return closed to -20 V in about 10 μs. According to (11), △vC
iMOS are close to 50 A, whose ratio to the total current 5.0 kA is would only be less than 0.75 V. So it could be turned off again
equal to the number of MOSFETs. Despite a certain immediately in dozens of microseconds, which had little
nonuniformity among different iMOS and vC measurement points influence on the maximum turn-off capability. Furthermore, vC
is observed at recovery current spike (127.5 μs) and during could be recharged to -20 V in milliseconds, which depended on
current reduction process (after 128.5 μs), the measurement the charging power of driving circuit.
results of these duration were not utilized to calculate the stray After optimization, the less number and lower on-state
parameters. Consequently, the current diviation of parralleled losses of DirectFETs on GDU3 would lead to a lower
capacitors and MOSFETs has little influence on the calculation temperature rise than GDU1 / GDU2, implying a longer lifetime
result of stray parameters. operation. The decrease of C1 would increase the power of each
According to Fig. 9, the total stray inductance of turn-off electrolytic capacitor when charging C2 during turn-off process,
circuit including GDU1 and GCT housing is 1.44 nH, and that which would increase its temperature rise. However, it was
with GDU2 is 1.14 nH (ignoring LC1). Obviously, compared believed to be more reliable due to the less number of capacitors.
with RS, LS will yield most of the voltage drop under high diG/dt Turn-off experiments of a 6-inch GCT with GDU3 were
such as 14 kA/μs and 16 kA/μs. In another word, LS is the major carried out at different current up to 12.5 kA, as shown in Fig.
limitation to enhance commutation capability. It consists of both 12. Limited by the thin structure of DirectFETs, its current was
GDU and GCT housing. And an improved GDU with lower infeasible to monitor by Rogowski coil. However, the
inductance is presented in next section. commutation time could be inferred by the steep drop of vGK at
the end of commutation. Compared to the turn-off waveforms at
IV. IMPROVEMENT OF GATE DRIVER UNIT 10.0 kA with GDU2 in Fig. 6 and Fig. 7, the commutation time
Based on the stray impedance distribution analysis above, to with GDU3 in Fig. 12(a) was slightly accelerated from 730 ns to
further reduce stray impedance of GDU2, MOSFETs with lower 710 ns. Besides, during the commutation, vGK rose from -18.9 V
impedance should be selected carefully. Besides, both C1 with to -15.7 V, lower than that of GDU2, and the difference between
high capacitance and C2 with low stray inductance are essential. vGK and vC was tiny. At the end of commutation, the steep drop
However, GDU2 is bulky. The capacitance and paralleled of vGK was very small. All those features proved an effective
number of C1 and C2 should be optimized. optimization of GDU3 that LM+P and RM+P were substantially
In view of these two points, an improved GDU3 was reduced. By calculation, they were 0.037 nH and 0.027 mΩ. It
developed as shown in Fig. 11, to be compared with the was believed that the maximum commutation capability of
previous GDU2. DirectFETs with lower equivalent inductance GDU3 was higher than 17 kA within 1.5 μs. Additionally,
and resistance [15][16] were adopted to replace previous LH+LK was calculated as 1.0 nH, conforming to that of GDU1
MOSFETs. Table I is the characteristic comparison between and GDU2.
single MOSFET used in GDU1/GDU2 and DirectFET used in All the experiments above were conducted at about 20 ℃. In
GDU3 [17][18]. Although the internal inductance of DirectFET a wide temperature range from -60 ℃ to 60 ℃, the capacitance
is not provided in the datasheet, its eliminated conduction path of C2 will only reduce by up to 5%. And the impedance
and metal package obviously indicate an extremely low fluctuation of DirectFET and C2 in this temperature range is also
inductance. Besides, due to the stronger current capability of negligible, compared with the total impedance of commutation
DirectFET, its parallel quantity is reduced by 60%. circuit. Therefore, the temperature will have a very small
Considering that during the commutation C2 conducted iG influence on the commutation capability of GDU3.
first, and then was charged by C1, to ensure the sufficiently low Table II summarized the stray impedances distribution of
inductance, the number and capacitance on GDU3 is the same as three types of GDU. Considering the upward trend of RM+P as iA
GDU2 (1.68 mF for 190 capacitors). However, the demand of increases, its values of different GDUs in Table II are compared
LC1 was no longer harsh. Therefore, the number of C1 was based on the calculation of 5.0 kA turn-off current. By

0885-8993 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2018.2826046, IEEE
Transactions on Power Electronics
7

TABLE II
Stray Impedances Comparison of Different GDUs
GDU LM+P LC1 RM+P RC1 RC2 LS-All RS-All
type nH nH mΩ mΩ mΩ nH mΩ
GDU1 0.164 0.231 0.266 0.178 / 0.395 0.444
GDU2 0.145 Bypassed 0.255 Bypassed 0.193 0.145 0.448
GDU3 0.037 Bypassed 0.027 Bypassed 0.200 0.037 0.227

nor further optimization of GDU is necessary.

However, on the other hand, the stray inductance inside
GCT housing (LH+LK) is about 1.0 nH, which is urgent to
improve. Recently, a new gate bushing for the hermetic GCT
housing has been developed by ABB, which is estimated to
minimize LP1+3+LK to 0.2 nH at 1 MHz [19][20]. The integrated
housing with turn-off circuit [14] may also reach the same
purpose. Expectantly, combining GDU3 and a novel GCT
housing, the total LS of turn-off circuit may be optimized to
under 0.3 nH. And an extremely high commutation speed of
over 50 kA/μs will be achieved in the near future.

V. CONCLUSION

Fig. 11. Comparison

VAK of Iprevious
VC GDU2
IMOS(above)
VGKand improved GDU3 In this paper, combining turn-off circuit model, the
AK
(below). acquisition method of each part of LS and RS of IGCT based on
experimental
60 12
measurements was carefully analyzed, including
12 24 24
electrolytic and ceramic capacitors, MOSFETs, PCB, and GCT
10 20 20 10
housing.
iA 40
8 16 16
A 8
series of relevant experiments were conducted with 6-inch
6 12 12
GCTs
20 6
and two types of conventional GDUs. According to the
vAK (kV), iA (kA)

4 8 8 4
experimental measurements, each part of stray impedance was
vGK, vC (V)

0
2 vAK 4 4 2
calculated, which was a crucial reference for IGCT
IMOS (A)
IAK (kA)
UC (V)

0 0 0 0
improvement. LS was the major limitation to commutation rather
-20
-2 -4 -4 -2
than RS. The total LS on GDU1 was 0.395 nH. 58% of it came
-4 vGK -18.9V -8 -8
-15.7V -40 -4 electrolytic capacitors LC1, and the rest mainly came from
from
-6 -12 -12
MOSFETs-6
LM+P. The stray inductance of PCB was only about
-8 -16 -16 -60 -8
40 pH, which can be ignored. From GDU2, it was proved that
-10 -20 -20
L-80C1-10can be bypassed by ceramic capacitors. Besides, the total RS
vC
-12 -24 -24
on -12 both GDU1 and GDU2 was about 0.45 mΩ, which would
127.0 127.5 128.0 128.5 129.0 129.5 130.0 also yield a large voltage at high current such as over 10 kA.
VAK IAK t C(μs)
V IMOS VGK
60% of it came from capacitors, and the rest came from
(a) MOSFETs.
60 12
12 i 24 24 Based on stray impedance calculation results, an improved
A
10 20 20 GDU3
40
10 was developed. Previous MOSFETs were replaced by
8 16 16 DirectFETs
8 with lower impedance. The number of electrolytic
6 12 12 capacitors
20 6 was reduced by 50%. During the experiments, an
vAK (kV), iA (kA)

4 8 8 extremely4 high current of 12.5 kA was successfully turned off,

vGK, vC (V)

2 vAK 4 4 0
which 2 is the maximum turn-off current level so far. And the
IMOS (A)
IAK (kA)
UC (V)

0 0 0 maximum 0 commutation capability of GDU3 was over 17 kA.

-20
-2 -4 -4 Experiments indicated that LS and RS was significantly
-2
vGK -18.9V
-4 -14.7V -8 -8 optimized
-40 -4 to 0.037 nH and 0.227 mΩ. No further optimization
-6 -12 -12 of GDU -6 is necessary.
-8 -16 -16 -60 -8
However, there was still about 1.0 nH existing inside GCT
-10 -20 -20
housing.
-10
In order to further enhance commutation capability and
vC -80
-12 -24 -24
increase
-12
turn-off current, it is necessary to utilize a compact
127.0 127.5 128.0 128.5 129.0 129.5 130.0 housing structure with lower stray inductance. Hopefully,
t (μs) combining our improved GDU and the novel housing of 0.2 nH,
(b) an extremely high commutation speed of over 50 kA/μs will be
Fig. 12. Experimental results of (a) 10.0 kA and (b) 12.5 kA turn-off achieved in the near future.
waveforms with a 6-inch GCT, using improved GDU3.

paralleling ceramic capacitors and replacing MOSFETs with ACKNOWLEDGMENT

DirectFETs, the stray inductance on GDUs is remarkably Thank Zhuzhou CRRC Times Electric Co. Ltd. for
reduced from 0.395 nH to 0.037 nH, and the stray resistance on providing 6-inch GCT wafers for our GDU experiments. Their
GDUs is optimized from 0.444 mΩ to 0.227 mΩ. Consequently, continuous support and help is gratefully acknowledged.

0885-8993 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2018.2826046, IEEE
Transactions on Power Electronics
8

REFERENCES Zhengyu Chen was born in Tianjin, China in 1992.

He received the B.S. degree from the Department of
[1] Yamamoto, et. al., “GCT (Gate Commutated Turnoff) Thyristor and Gate Electrical Engineering, Tsinghua University, Beijing,
Drive Circuit”, PESC 1998.
China, in 2014, where he is currently pursuing a Ph.D.
[2] H. Gruening, T. Tsuchiya, K. Sotoh, et. al., “6 kV 5 kA RCGCT with degree.
Advanced Gate Drive Unit”, Proceedings of 2001 International His current research interests include power
Symposium on Power Semiconductor Devices & ICs, Osaka, 2001, pp: semiconductor devices and their gate unit drivers,
133-136. HVDC system and direct current circuit breaker.
[3] Satoh, et. al., “6kV/4kA Gate Commutated Turn-off Thyristor with
Operation DC Voltage at 3.6kV, ISPSD 1998.
[4] Takata, et. al., “Snubberless Turn-off Capability of Four-inch 4.5kV GCT
Thyristor”, ISPSD 1998.
[5] N.Zargari, S.Riuo, Y.Xiao, H.lwamoto, K.Sato, J.Donlon, "A New Zhanqing Yu was born in Inner Mongolia, China, in
Current Source Converter Using A Symmetric Gate Cornmutated 1981. He received his B.S. and Ph.D. Degree from the
Thyristor (SGCT)", IEEE- IAS, Rome 2000. Department of Electrical Engineering, Tsinghua
[6] J.-M. Meyer and A. Rufer, “A DC hybrid circuit breaker with ultrafast University in Beijing in July 2003, July 2008,
contact opening and integrated gate-commutated thyristors (IGCT),” respectively.
IEEE Trans. Power Del., vol. 21, no. 2, pp. 646–651, Apr. 2006. After graduation, he became a post doctor and
lecturer at the Department of Electrical Engineering,
[7] Zhengyu Chen, Zhangqing Yu, Gang Lv, “Researches on 10 kV DC
Tsinghua University in Beijing in July 2008 and July
Hybrid Circuit Breaker Based on IGCT Series”, Proceedings of the CSEE,
2016, Vol. 36, No. 2, pp:317-326. 2010 respectively, an associate professor in the same
department in December 2012.
[8] P. K. Steimer, O. Apeldoorn, B. Odegard, S. Bernet, and T. Bruckner, He has participated in several projects sponsored by High-Tech R&D
“Very high power IGCT PEBB technology,” in Proc. IEEE 36th PESC, Program (863 Program), National Basic Research Program of China (973
Jun. 2005, pp. 1–7.
Program), National Natural Science Foundation of China. His research interests
[9] H. Gruening and K. Koyanagi, “A new compact high dI/dt gate drive unit include DC grid, DC breaker, electromagnetic environment and
for 6-inch GCTs,” in Proc. IEEE International Symposium on Power electromagnetic compatibility, lightning protection.
Semiconductor Devices & ICs, Kitakyushu, Japan, 2004, pp. 336–340.
[10] H. Gruening, K. Koyanagi, “A modern low loss, high turn-off capability
GCT gate drive concept”, EPE, 2005, pp: P.1 – P.10. Xuan Liu was born in Lanzhou, Gansu, China in 1996.
[11] B. Backlund, M. Luscher, “Recent developments in IGCT gate units,” in He is currently pursuing a B.S. degree from the
2007 European Conference on Power Electronics and Applications, Department of Electrical Engineering, Tsinghua
Sept.2007 , pp.1 – 7. University, Beijing, China. His current research
[12] U. Vemulapati, M. Rahimo, M. Arnold, et. al. “Recent Advancements in interests are GCT gate driver and its power supply.
IGCT Technologies for High Power Electronics Applications”, EPE’ 15
ECCE-Europe, 2015 17th European Conference, Sep., 2015.
[13] P. Köllensperger, R. W. De Doncker, “The Internaly Commutated
Thyristor – A new GCT with integrated turn-off unit”, in Proc. Conf.
Integr. Power Electron. Syst., Naples, Italy, Jun. 2006.
[14] P. Köllensperger, R. W. De Doncker, “Optimized Gate Drivers for
Internally Commutated Thyristors (ICTs)”, IEEE Transactions on
Industry Applications, Vol. 45, No. 2, March/April 2009. Jiapeng Liu was born in Liaoning, China in 1994. He
[15] G. Engelmann, R. W. De Doncker, “A Systematic Comparison of Various received the B.S. degree from the Department of
Thermal Interface Materials for Applications with Surface-Mounted Electrical Engineering, Tsinghua University, Beijing,
(DirectFETTM) MOSFETs”, IEEE PEDS 2015, Sydney, Australia, June China, in 2016, where he is currently working towards
2015, pp. 112-117. Ph.D. degree in the research group of Prof.Zeng. His
[16] Sawle, Andrew, Blake, et. al, “Novel Power MOSFET Packaging current research interests include power
Technology Doubles Power Density in Synchronous Buck Converter for semiconductor device manufacture, modeling and
Next Generation Microprocessors”, APEC, pp. 106-111, 2002. development.
[17] “IRF7749L2TRPbF, DirectFETTM Power MOSFET”, Datasheet, PD -
97434, International Rectifier, Online: www.irf.com.
[18] “IRFR3505 HEXFET Power MOSFET”, Datasheet, PD - 94506A,
International Rectifier, Online: www.irf.com.
Rong Zeng (M’02-SM’06) was born in Shaanxi,
[19] Didier C., Thomas S., Tobia, W., “Semiconductor switching device with
gate connection”, US Patent 2010/0270584 A1. China, in 1971. He received the B.Eng., M.Eng., and
Ph.D. degrees from the Department of Electrical
[20] Wikström T., Arnold M., Stiasny T., et. al, “The 150 mm RC-IGCT: a Engineering, Tsinghua University, Beijing, China, in
device for the highest power requirements”, Proc. ISPSD, Waikoloa, HI, 1995, 1997, and 1999, respectively.
2014, pp. 91-94.
He was a lecturer with the Department of Electrical
Engineering, Tsinghua University in 1999, an
associate professor and professor with the same
department in 2002 and 2007, respectively. Currently,
he is the Dean of the Electrical Engineering
Department, Tsinghua University. He is currently working in the fields of air
gap discharge, lightning protection, and electromagnetic compatibility in power
systems, electric and magnetic field measurement by integrated electro-optical
sensors, power semiconductor, HVDC system and direct current circuit breaker.

0885-8993 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.