The Advantages and Operation of Flying-Capacitor

Boosters
Viktor Antoni, Development Engineer - Electronic Design, Vincotech, Hungary

1 Introduction
High-efficiency solar inverters are getting more and more in demand in the recent years.
However, cost efficient solutions are also desirable. To achieve this, not only the inverter but
also the booster stage have to be low cost and high efficient. Two- and three- level boosters
are commonly used in solar inverters. The three-level solutions are able to decrease the
voltage stress on the semiconductors and the output voltage ripple, and therefore the inductor
size can be decreased. Due to three-level operation, switched voltage level is half of the DC-
link voltage. Thus, one can use semiconductors with lower blocking voltage, which are faster
and cheaper. For three-level operation an adequate DC-link capacitor (capacitive voltage
divider) has to be utilized, which can then split the two-level DC voltage into three voltage
levels. In this case the PWM signal needs to be corrected to ensure the symmetry of the
neutral point of the divider. Usually two inductors are used in the input in three-level boosters.
This article describes the Flying Capacitor Booster solution, which increases the efficiency while
being still cost efficient without enormous three-level DC-link capacitors and with only one
choke on the input.

2 The Flying-Capacitor Booster

In this topology the additional voltage levels are synthetized by a capacitor, so-called flying-
capacitor.

In a three-level case the voltage of the flying-capacitor is the half of the output voltage. The
𝑉𝑉𝐷𝐷𝐷𝐷
capacitor can offset the output voltage with in positive and negative direction. The three-
2

level flying-capacitor booster can be seen on Figure 1.

Page 1 of 14
 Figure 1: The three-level flying-capacitor booster

In the flying-capacitor booster due to the phase shift in the control of transistors, the input
frequency is p times the switching frequency (p is the number of stages described later).

3 The Commutation Loops and Operation Modes of the

Flying-Capacitor Booster
In a flying-capacitor booster the commutation loops include capacitors. A capacitor from the
commutation point of view can be considered as zero impedance. Its main role in the
commutation loop is to offset the two outer semiconductors from each other. With this offset
the three-level flying-capacitor booster can be considered as two standalone boosters, in which
the outer one’s commutation loop includes the DC-link capacitor, the outer diode, the flying-
capacitor and the outer switch. The inner commutation loop includes the flying-capacitor, the
inner diode and the inner switch. The two commutation loops can be seen on Figure 2.

Page 2 of 14
 Figure 2: The two communication loops

In general the number of voltage levels are theoretically endless, but in practice three, four
and five levels are used. The additional levels in n level solution can be realized by adding
extra outer commutation loops to the three-level converter. Every added booster’s
commutation loop will be similar to the blue loop on Figure 2. The number of voltage levels can
be calculated as the following:

𝑛𝑛𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 = 𝑝𝑝 + 1

where p is number of the commutation loops (boosters). The voltage of the capacitor can be
calculated:
𝑖𝑖 − 1
𝑉𝑉𝐹𝐹𝐹𝐹,𝑖𝑖 = 𝑉𝑉𝐷𝐷𝐷𝐷 ∙ �1 − �
𝑝𝑝

where i is number of the given commutation cells. The first loop refers always the most outer
loop.

This article describes the operation and behavior of the three-level flying-capacitor booster. All
other solutions can be realized based on this article.

In the operation of the three-level flying-capacitor booster, four different modes can be
derived. During normal operation the voltage of the flying-capacitor is the half of the output
voltage and the inductor current is perpetual. In the first mode both two switches are off, the
current goes through the two diodes, and they are working in bypass mode. In this mode the

Page 3 of 14
voltage of the flying-capacitor is not changing, the current of the choke is decreasing, and the
output voltage is increasing. In the second mode, the lower switch (T27) is turned on. The
current is charging the flying-capacitor resulting in its voltage to increase. In the third mode
the inner switch is turned on (the outer switch is turned off), the current goes through the
flying-capacitor, while its voltage is decreasing, and the output voltage will increase. In the last
mode, both two switches are turned on. The voltage of the flying-capacitor will be stationary,
while the current of the choke will be increasing. In the second and the third mode the
inductor current change is dependent on the duty cycle (D). The operations and their effects
can be seen on Figure 3 and Table 1.

Mode 2
Mode 1

Mode 3 Mode 4

Figure 3: Operation modes of a flying-capacitor booster

Page 4 of 14
 Transistors Inductor current
Mode FC voltage DC-link voltage
T25 T27 D<0.5 D>0.5
Mode 1 OFF OFF decreasing - - increasing
Mode 2 OFF ON increasing decreasing increasing decreasing
Mode 3 ON OFF increasing decreasing decreasing increasing
Mode 4 ON ON - increasing - decreasing
Table 1: Output and FC voltage states

The transfer function (y) of the flying-capacitor booster is the following:

𝑉𝑉𝑂𝑂𝑂𝑂𝑂𝑂 1
𝑦𝑦 = =
𝑉𝑉𝐼𝐼𝐼𝐼 1 − 𝐷𝐷

where 𝐷𝐷 is the duty cycle.

The used modes are dependent on the duty cycle. If < 0.5 , then 𝑦𝑦 < 2. In this case Mode 4 is
not used and the operation will be the following:

…  Mode 1  Mode 2  Mode 1  Mode 3  …

If > 0.5 , then 𝑦𝑦 > 2 , and the operation will be:

…  Mode 4  Mode 2  Mode 4  Mode3  …

In case of 𝐷𝐷 = 0.5, 𝑦𝑦 = 2 ,the operation:

…  Mode 2  Mode 3  Mode 2  Mode3  …

The most commonly used operation is, when 𝐷𝐷 is less than 0.5.

4 The Operation of the Flying-Capacitor Booster

In the flying-capacitor booster topology the two transistors have to be controlled by 180°
phase shifting (Figure 4).

Figure 4: The reference and modulation signals for the PWM

Page 5 of 14
This results that in case of 𝐷𝐷 = 0.5, the operation modes will change between Mode 2 and Mode
3. The typical curves of the flying-capacitor booster can be seen on Figure 5 in case of 𝐷𝐷 = 0.2.

Page 6 of 14
 f

Figure 5: The typical curves

a, the gate signals

b, the output, the input and the FC+ voltage
c, the voltage of the Flying Capacitor and the inductor current
d, the Flying Capacitor current
e, the currents of the diodes
f, the current of the transistors

5 The Advantages of the Flying-Capacitor Booster

The flying-capacitor booster topology compared to the booster topology has the following
advantages:

• As the operation is three-leveled, the voltage stress on the semiconductor is decreased.

This results in lower EMI, lower current and voltage ripple.

The flying-capacitor booster topology compared to the symmetrical booster topology has the
following advantages:

• It has two-level input and output connection, while the third voltage level is synthetized
by the flying-capacitor. This way the large three-level capacitors can be eliminated on
the input and the output.
• Only one input choke is needed.

In both cases the input frequency is double of the switching frequency. This results in a lower
input ripple current or the inductance can be decreased. Because of the double frequency,
slower semiconductors can be used, which deceases the costs while the switching losses are
also lower. This means for optimal behavior SiC MOSFETs are not needed, but Si IGBTs can be
used.

Page 7 of 14
For more detailed comparison of the topologies and component selection, please see
Vincotech’s benchmark “Boost your 1500 V string inverter”.

6 The Flying-Capacitor

6.1 Sizing of the Flying-Capacitor

The voltage supplied by a flying-capacitor has a key role in this topology. To keep the voltage
ripple on the capacitor low, a suitable capacitor size is needed. To determine the needed
capacitance, the switching frequency and the maximum allowed voltage ripple need to be
considered. The size of the capacitance can be calculated as:
𝐼𝐼𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝
𝐶𝐶𝐹𝐹𝐹𝐹 =
∆𝑈𝑈𝐹𝐹𝐹𝐹 ∙ 2𝑓𝑓𝑆𝑆𝑆𝑆

, where ∆𝑈𝑈𝐹𝐹𝐹𝐹 is the maximum allowed voltage ripple, 𝐼𝐼𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 is the maximum current, and 𝑓𝑓𝑆𝑆𝑆𝑆 is
the switching frequency of the transistors.

6.2 The Balancing of the Capacitor Voltage

For the appropriate operation the flying-capacitor, voltage has to be half of the output voltage.
To achieve this, it must be regulated at all time. This can be done by changing the operation
modes. As it can be seen in Table 1, Mode 1 and Mode 4 have no effect for the flying-
capacitor, so for regulation Mode 2 and Mode 3 have to be used. The regulation state diagram
can be seen on Figure 6.

Figure 6: The flying-capacitor regulation

Page 8 of 14
The needed modes are depending on the duty cycle. In case of ≤ 0.5 , the operation will be the
following:

…  Mode 1  Mode 2  Mode 1  Mode 3  …

If the flying-capacitor voltage exceeds the set point, the operation can be modified to decrease
the voltage:

…  Mode 1  Mode 3  Mode 1  Mode 3  …

If the flying-capacitor voltage is less than the set point:

…  Mode 1  Mode 2  Mode 1  Mode 2  …

In case of ≥ 0.5 , the needed modification will be the same, only Mode 4 will be used instead of
Mode 1:

…  Mode 4  Mode 3  Mode 4  Mode 3  … to decrease the voltage

…  Mode 4  Mode 2  Mode 4  Mode 2  … to increase the voltage

6.3 The Pre-Charge of the Flying Capacitor

This section describes the details of the method proposed by Mitsubishi Electric Corporation [1]
to protect the flying capacitor booster, when there are no control signals (e.g.: during startup).
In case when all the control signals of the transistors are low, the flying-capacitor voltage
cannot be regulated. In that operation an extra effort is needed to keep the flying-capacitor
voltage on the safe side. Failing to eliminate the overvoltage on the semiconductors may cause
a fatal error in the system. There are two operation modes, when all the transistors are OFF:
1) When the input is applied and the output is equal with the input (e.g.: startup), and 2)
when the input is zero and the output is not. This happens, for example, when one string is not
connected to the circuit and other boosters are working. In both two cases the voltage of the
flying-capacitor is zero, and the voltage sharing of the two transistors is not defined. To keep
the voltage level of the semiconductors below the breakdown voltage, additional balancing has
to be used.

During startup the current flows through the two diodes and charge the output capacitance. In
this case the output voltage is equal to the input voltage, while the flying-capacitor voltage is
zero. This is dangerous for the lower switch. To eliminate this problem, another current path
has to be added, in which the current can charge also the flying-capacitor. For this a diode can
be used, a cathode of which has to be connected to a capacitive voltage divider, where the
lower point of the flying-capacitor is clamped at the half of the DC-link voltage. This can be
seen on Figure 7

Page 9 of 14
 Figure 7: The additional current path during startup

𝑄𝑄
As the voltage of the capacitor can be calculated from the following expression: = , and the
𝐶𝐶

charge will be the same for 𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜1 and 𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜2 + 𝐶𝐶𝐹𝐹𝐹𝐹 , the flying-capacitor voltage will be the
following:
𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜1
𝑉𝑉𝐹𝐹𝐹𝐹 = 𝑉𝑉𝑂𝑂𝑂𝑂𝑂𝑂
𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜1 + 𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜2 + 𝐶𝐶𝐹𝐹𝐹𝐹

where 𝑉𝑉𝑂𝑂𝑂𝑂𝑂𝑂 is equal with 𝑉𝑉𝐼𝐼𝐼𝐼 (if the forward voltage of the diodes are not considered).

If the capacitance of 𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜1 and 𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜2 is equal and the capacitance of 𝐶𝐶𝐹𝐹𝐹𝐹 is significantly smaller
than the capacitance of 𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜1 and 𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜2 , the voltage of the flying-capacitor is half of the output.
𝑉𝑉𝐼𝐼𝐼𝐼
𝐶𝐶 = 𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜1 = 𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜2 , 𝐶𝐶𝐹𝐹𝐹𝐹 ≪ 𝐶𝐶, 𝑉𝑉𝐹𝐹𝐹𝐹 ≈
2
When the string is not used and other boosters are working, the input voltage is zero, while
the output voltage is not. In this case another diode has to be added to charge the flying-
capacitor.

Page 10 of
14
 Figure 8: The additional diode when VIn is zero

As it can be seen on Figure 8, the current path is the following:

𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜2 → 𝐷𝐷ℎ → 𝐶𝐶𝐹𝐹𝐹𝐹 → 𝐷𝐷45 → 𝐿𝐿 → 𝐶𝐶𝑖𝑖𝑖𝑖

In this case the sum of the flying-capacitor voltage and the input voltage can be calculated as
the following:
𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜2
𝑉𝑉𝑓𝑓𝑓𝑓𝑓𝑓 = 𝑉𝑉𝑂𝑂𝑂𝑂𝑂𝑂
𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜1 + 𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜2 + 𝐶𝐶𝐹𝐹𝐹𝐹 × 𝐶𝐶𝑖𝑖𝑖𝑖

As in the last expression, if 𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜1 and 𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜2 is equal and 𝐶𝐶𝐹𝐹𝐹𝐹 × 𝐶𝐶𝑖𝑖𝑖𝑖 is negligible compared to 𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜1
and 𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜1 , the voltage is half of the output voltage.
𝑉𝑉𝑂𝑂𝑂𝑂𝑂𝑂
𝐶𝐶 = 𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜1 = 𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜2 , 𝐶𝐶𝐹𝐹𝐶𝐶 × 𝐶𝐶𝑖𝑖𝑖𝑖 ≪ 𝐶𝐶, 𝑉𝑉𝑓𝑓𝑓𝑓𝑓𝑓 ≈
2
This voltage is divided on the two capacitors. If the capacitance of the 𝐶𝐶𝐹𝐹𝐹𝐹 is much smaller than
the 𝐶𝐶𝑖𝑖𝑖𝑖 capacitor, then the voltage of the 𝐶𝐶𝑖𝑖𝑖𝑖 capacitor is as low as it can be considered zero,
and the voltage of the flying-capacitor is near to the half of the output voltage.

This method can be improved by closing the T27 switch. In this case the voltage is not divided
by 𝐶𝐶𝑖𝑖𝑖𝑖 and 𝐶𝐶𝐹𝐹𝐹𝐹 capacitors and the current path will be the following:

𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜2 → 𝐷𝐷ℎ → 𝐶𝐶𝐹𝐹𝐶𝐶 → 𝑇𝑇27

And the flying-capacitor voltage:

𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜2
𝑉𝑉𝐹𝐹𝐹𝐹 = 𝑉𝑉𝑂𝑂𝑂𝑂𝑂𝑂
𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜1 + 𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜2 + 𝐶𝐶𝐹𝐹𝐹𝐹

Page 11 of
14
6.3.1 Design considerations

During normal operation T27 creates an overvoltage spike at turn-off. If Df turns on to clamp
this spike, T27 switch will be loaded with the reverse recovery of Df. To avoid Df clamping this
overvoltage spike, an additional Zener diode (Dz) can be added in series with Df. The Zener
voltage of the Zener diode should be higher than the spike of turn-off. This can be seen on
Figure 9.

Figure 9: The additional Zener diode

𝑉𝑉𝑂𝑂𝑂𝑂𝑂𝑂
If the voltage of flying-capacitor is extremely higher than , an additional current ripple will
2

appear on the inductor. This ripple causes increased losses and noise. This ripple can be also
moderated with this Zener diode.

If the voltage of 𝐶𝐶𝐹𝐹𝐹𝐹 is less than the voltage of 𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜2 , an equalization current will flow between
𝐶𝐶𝐹𝐹𝐹𝐹 and 𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜2 resulting in an unbalance between 𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜1 and 𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜2 . This unbalance can be
decreased with a current limiting resistor (Rf), which can be seen on Figure 10.

Page 12 of
14
 Figure 10: The current limiting resistor

7 Conclusion
The flying-capacitor booster is a high-efficient, low cost solution for solar inverter applications.
The main advantages are the frequency multiplication, the lower semiconductor voltage, the
lower voltage and current ripple, the lower switching losses, and the low EMI emission. The
flying-capacitor size is significantly smaller than the required DC-link capacitor used in
traditional booster topologies with the same power rating. The challenge is to regulate the
voltage of the flying-capacitor especially when all the transistors are in OFF state (i.e.: before
the converter is turned on). The balancing can be achieved by state regulation. Also the pre-
charge challenge can be solved easily by additional diodes based on Mitsubishi Electric
Corporation’s patent. With these diodes the flying-capacitor booster is a cost efficient
alternative for the other booster solutions with higher efficiency.

Page 13 of
14
8 References

[1] M. Tauer, "Boost your 1500 V string inverter," Unterhaching.

[2] T. Okuda and H. Ito, "DC/DC POWER CONVERSION APPARATUS". United States Patent US
2013/0021011 A1, 24 01 2013.

Page 14 of
14