Дипломная работа - Проектирование 10 кВт преобразователя

Подождите немного. Документ загружается.

datasheet; however it may require some interpolation between values in graphs to

determine the appropriate value for the specific application, such as estimated junction

temperature.

•

Total power loss in the FWD (P

losstot

)

The total power dissipation can be calculated as in equation 3.2.2.2 [13]:

satoffonlosstot

PPPP +

= (3.2.2.2)

3.3. Lowest loss estimation of IGBT and FWD

In appendix D it can be found a MOSFET in comparison with various IGBT with almost

the same voltages and currents and as predicted the power loss in the MOSFET is greatly

exceeding any IGBT regardless of model. This is easily explained by the high voltage

ratings.

On the other hand, in other applications in example as an inverter with a DC-bus supplied

from a regular car battery (12V) which is consuming 50A an off-the-shelf IGBT with

CEsat

= 2.5V has the efficiency of 79%. An of-the-shelf MOSFET with R

DS(on)

= 0.007 Ω

has an efficiency of 97% and if we add switching losses the benefit of the MOSFET is

increased further more. The conclusion looks like choosing the IGBT with the lowest

power losses are the best choice without a doubt but there are other concerns as well.

By viewing the chart of suitable transistors/diodes again and removing all those which

may have a problem with either too high peak currents or too high peak voltages

a few

good transistor solutions remain, in this case with respect to power loss per IGBT with

FWD. Calculations are done using the equation 55 and 57 with the parameters cosφ=1,

=15 A and the model specific parameters can be found in appendix D where all the

parameters have been normalized to fit the actual values. The switching frequency is

assumed to be 10kHz giving it a m

= 15.

Manufacturer Model P

LOSSTOT

Technology

CSJC

θθ

IRF IRG4PSH71UD 26,3W SI IGBT, SI FWD 0.36+0.24

IRF/CREE IRGPSH71U/

C2D05120

22,8W SI IGBT, 2 x SiC FWD 0.36+0.24,1.1+0.24

CREE 2 x CID100512 40W 2 x SI IGBT, 2 x SiC FWD 1.15+0.24

Table 11 : Suitable discrete IGBTs

Although even if the discrete transistors CO-packed with free-wheeling-diodes above

seems to outperform the modules below in respect to loss the cabling has to be considered

since it will contribute with some capacitance and possibly some inductance. In some

cases the effects of this is large enough to be ineligible in favor of the big modules.

The discrete transistors also have the possibility to be mounted on a separate heatsink or

to displace the dissipated heat over a larger area making cooling easier in some

The rated voltage and current should be at least twice the nominal to handle overshoots [7].

Conduction loss and switching loss for one IGBT with FWD

applications. However, the modules are much easier to handle, one big bulky brick with

optimized internal wiring prepared for mounting a PCB on top along with built-in brake

IGBT and a high performing rectifier-bridge results in a lot less wiring and complexity.

On the other hand a module requires a larger heatsink or equal device capable of

removing a lot of heat from a relatively small area.

Manufacturer Model P

LOSSTOT

Technology

CSJC

θθ

Infineon

FS25R12W1T4

29,6W SI IGBT, SI FWD 0.66 + 0.88

FUJI

7MBR35UB120

33W SI IGBT, SI FWD (0.76,1.07)+0.05

Infineon

FP25R12KT3

30,7W SI IGBT, SI FWD (0.8,1.35) + 0.3

Table 12: Suitable IGBT modules

In the end a more detailed approach than only the power loss has to be considered for

choosing the most suitable solution, mostly depending on the choice of cooling but also

on how much the transistor-performance can be increased with the help of snubber

circuits.

3.4. Rectifier

The total conduction loss can be derived from the average current running through a leg

times the forward voltage drop of two diodes since the current always runs through two.

This voltage drop is somewhat depending in some grade on the current and the

temperature of the junction and the proper value can be found in the datasheet. The

average current is simply the output current divided by three, notice that this value will

not set the voltage drop since there is momentarily a higher current running through the

diode when actually conducting, they are just not conducting more than a half period at a

time.

×=

avgSATLOSSCOND

IVP

(3.4.1)

The “2” since the current is running through two diodes and the “3” since there are three

legs in the rectifier bridge as illustrated in figure 13. Mainly three rectifiers from three

different suppliers have been considered, FUJI, International Rectifier and Semikron.

They have all been chosen from a wide range of models and only the best suiting rectifier

for this application has been considered. Their main properties are described in the table

below.

The module is equipped with a built-in rectifier and brake function

Manufacturer Model

SAT

V @14,5A

125° C

LOSSTOT

Weight Size(packet)

CSJC

θθ

Fuji 6RI30E 0,87 35,7 W 100g 52 x 22 x 26

0.80+ 0.10

IRF 36MT80 0,9 36,9 W 20g 29 x 26 x 21 1.16 + 0.2

Semikron SKD 33 1,21 49,6 W 30g 64 x 30 x 21 0.417

Table 13 : Suitable rectifiers

As seen in the table above a tradeoff has to be made, the dissipated energy is lower in

some of the cases but it is also harder to remove the heat due to the larger thermal

resistance. For instance the dissipated energy in the Semikron diode bridge is

approximately 39% higher compared to the Fuji, however the Semikron has only about

46% of the thermal resistance making it a lot easier to cool. If an external efficient

cooling system is available then the Fuji would be the best choice to maximize efficiency

but with worse conditions for cooling the Semikron would be the best choice neglecting

the slightly lower efficiency.

3.5. Filters

3.5.1. Input filter

As the current and voltage harmonics will not change in frequency with increasing rated

power of the inverter the components values can be scaled proportional as long as the

main voltage is considered fixed. Along with this proportional increase comes the power

loss. The factor that would have largest effect on the input filter components and loss

would be to change rectifier complexity. Stepping up from the present 6-pulse rectifier

bridge to at least a 12-pulse bridge would reduce initial harmonics significantly and

therefore almost completely removing the need for additional filter except for EMI.

However the EMI filter does barely represent any significant loss except in some

additional weight.

However if the 10kW input to a 6-pulse rectifier needs to filtered to an acceptable level

including a simple EMI filter it requires filters worth of 131g and 37 W of loss per phase,

which is not negligible. It adds up to a total of 393g and 111W.

3.5.2. DC-bus filter

As the losses already have been derived and explained some estimated values for

different output capacities and switching frequencies has been collected. A 3% ripple

voltage of the maximum DC-bus voltage is allowed in all scenarios.

Output power/switching

frequency @ 230/400VAC

6kHz 10kHz 14kHz

2kW 2.6 3.2 3.8

10kw 27 30 41

20kW 54 67 81

Table 14 : DC-bus filter losses when using 230/400 VAC input

Output power/switching

frequency @ 115/200VAC

6kHz 10kHz 14kHz

2kW 5 42 80

10kw 6.3 52 100

20kW 8.2 68 120

Table 15 : DC-bus filter losses when using 115/200 VAC input

The loss due to the surge limiting circuit is not included since it is only active during

start-up. A 12-pulse rectifying bridge would also here prove to be favorable as it

decreases the voltage ripple significantly doubling its frequency, relieving the capacitor

from much ripple currents which will require a smaller capacitor bank.

3.5.3. Output filter

The power loss in the output filter is depending on several factors where the largest is the

DC-bus voltage V

, f

, I

peak

and the maximum allowable dV/dt in the motor which can

be calculated using the simplified equation 3.5.3.1 from chapter 2.4.3.

fIV

swpeakDC

lossfilteroutput

×××

(3.5.3.1)

Considering the maximum dV/dt for the motor rather fixed would make this model only

consist of fundamental parameters and straight to calculate.

Using the parameters for the model examined in this paper a power loss per phase is

estimated to be 10 W, giving a total of 30 W.

3.6. Controller

The controller is responsible for an insignificant power loss in comparison to the high

power components. But as it is trivial to estimate the maximum loss there are no real

reason to neglect it. This loss can also be considered fixed regardless of the output power

or the switching frequency as it mainly depends on the internal clock frequency.

As it can be found in the datasheet the maximum loss is 1.2W@3.3V but as explained

before a voltage regulator is needed, further contributing to an additional 0.8 W loss.

3.7. Transistor drives

The power loss in the transistor drives consists of one rather fixed component and one

minor depending on the switching frequency. However if the transistor configuration

needs to be changed due to higher power it would alter the dynamic power loss as well.

Suppose the same transistor would be used in parallel to obtain higher power rating then

the dynamic power loss would increase with a factor corresponding to the number of

parallel transistors. Otherwise it would increase proportional to the switching frequency

as explained in chapter 2.7.

The 10kW inverter examined in this paper would accumulate about 5W of power loss due

to the three drives (

5.13× W) and the dynamic 0.5 W assuming the switching frequency

is 10 kHz. Even if the switching frequency would be changed to 6 kHz or 14 kHz it

would not change the total significantly, (4.8 W, 5 W, 5.2 W).

A parallel coupled transistor would double this loss as well as requiring a stronger driver.

3.8. Internal power supply

The power loss of the internal power supply is dependant on the devices it provides with

power as it usually has a rather fix percentage of loss. As the transistor drives have a high

peak current while low RMS current a buffering capacitor is favorable to use instead of

over sizing the power supply. Since the efficiency in the power supply is estimated to

80% the other 20% of the input power is pure loss dissipated in the supply.

As stated in chapter 3.7 and 3.8 the power loss from the controller and drives sums up to

7 W giving the power supply a loss of 1.75 W adding up to the total of 8.75 W put in to

the power supply.

4. Results

4.1. Power density

The power density is defined as the ratio between the rated output power and the weight

of the device, the higher the better. The power density is very much related to the

efficiency of the device as thoroughly explained in this report. The weight of three

different power levels, 2000W, 10kW and 20kW will be estimated using the same

method as used in the paper for 10kW. Two different input voltages are used, the 400V

three phase input used in this paper and a 200V three phase input which is common in

aircrafts. The weight is divided in the respective segment of the inverter to pinpoint the

specific weight. As can be seen some parts grow with rated output while some parts

remain constant. The switching frequency is considered constant at 10kHz for this

evaluation.

Weight (g) @ 230/400VAC / 540VDC

Rated effect/component 2000W 10kW 20kW

Input filter 72 393 720

Rectifier 13 30 97

Dc-bus filter w. surge limiter 240 780 1872

Power supply 70 70 70

Controller and driver 60 60 60

IGBT and FWD 54 54 54

Output filter 39 219 403

Internal wiring 130 240 370

Cooling 1200 3880 7230

Housing 630 940 1463

Sum: 2525 6666 12339

Table 16 : Approximate separated weight for 230/400VAC applications

Weight (g) @ 115/200VAC / 270VDC

Rated effect/component 2000W 10kW 20kW

Input filter 98 480 1230

Rectifier 30 97 165

Dc-bus filter w. surge limiter 90 130 585

Power supply 70 70 70

Controller and driver 60 60 60

IGBT and FWD 54 54 54

Output filter 53 365 623

Internal wiring 173 360 470

Cooling 1330 4160 8440

Housing 630 940 1463

Sum: 2588 6716 12360

Table 17 : Approximate separated weight for 115/200VAC applications

4.2. Scalability

The ability and simplicity to up-scale or down-scale a solution to fit the actual needs

without any unnecessary weight is the definition of scalability. Of course some weight is

added to but it is far from a linear relation. Some parts will remain fixed when adding or

removing output effect while some will be abundant.

From the tables in chapter 4.3. a mathematical approximation can be derived to estimate

the loss of an arbitrary converter with some basic input characteristics. The inputs would

be power in watts, switching frequency in hertz and the phase to phase voltage on the

input. The estimation has a maximum error according to the table of 7% and should be

precise enough to cover at least 20% outside the regions studied.

() ()

()

1515.115.1

59.0

20000084.060000034.0121

400

),,( −+−+

⎟

⎠

⎞

⎜

⎝

⎛

insw

ACin

swACininloss

fVPP

(4.2.1)

The same approximation can be done for the estimated mass using the tables in chapter

4.1. Input parameters are the same and the result is given in grams.

() ()

()

059.1772.0

24.0.0

2000305.0600027.02290

400

),,( −+−+

⎟

⎠

⎞

⎜

⎝

⎛

insw

ACin

swACinin

fVPMass

(4.2.2)

The method used to fit the equations to the values is to assume the relation is exponential

and then start from the lowest value. In this scenario the 200 VAC, 2 kW and 6 KHz

version is used as base. The increase in weight and loss is averaged on the values

considered as fixed when the value at hand is growing, in this case in three steps. Doing

this for all three input values and adding their components makes up (4.2.1) and (4.2.2).

4.3. Efficiency

The estimated amount of loss compared to the rated output power. As can be seen in the

following tables a lot of combinations are given to estimate the power loss for

corresponding combination. For the lower voltage case the lower voltage version of the

transistor and FWD are used as they both have one. For the lower and higher effects

similar components but with different power ratings are taken of-the-shelf usually from

the same series and manufacturer as the components in the example.

230/400VAC / 540VDC 115/200VAC / 270VDC

Rated effect/component

@ 6kHz

2kW 10kW 20kW 2kW 10kW 20kW

Input filter 23 111 220 37 170 460

Rectifier 7 45 85 14 89.6 185

Dc-bus filter w. surge limiter 2.6 27 54 5 42 80

Power supply 1.5 1.5 1.5 1.5 1.5 1.5

Controller and driver 2+4.8 2+4.8 2+4.8 2+4.8 2+4.8 2+4.8

IGBT and FWD 36 109 278 43 172 363

Output filter 4 21 42 7 39 67

Snubber circuits 0.05 0.5 1.476 0.074 0.74 2.21

Sum of loss: 81 322 689 114.4 521 1165

Table 18 : Approximate separated losses @ 6 kHz switching frequency

230/400VAC / 540VDC 115/200VAC / 270VDC

Rated effect/component

@ 10kHz

2kW 10kW 20kW 2kW 10kW 20kW

Input filter 23 111 220 37 170 460

Rectifier 7 45 85 14 89.6 185

Dc-bus filter w. surge limiter 3.2 33 67 6.3 52 100

Power supply 1.75 1.75 1.75 1.75 1.75 1.75

Controller and driver 2+5 2+5 2+5 2+5 2+5 2+5

IGBT and FWD 44.6 145 355 59 188 397

Output filter 5 30 56 9 49 76

Snubber circuits 0.083 0.83 2.46 0.123 1.23 3.68

Sum of loss: 92 373 794 134 558 1230.8

Table 19 : Approximate separated losses @ 10 kHz switching frequency

230/400VAC / 540VDC 115/200VAC / 270VDC

Rated effect/component

@ 14kHz

2kW 10kW 20kW 2kW 10kW 20kW

Input filter 23 111 220 37 170 460

Rectifier 7 45 85 14 89.6 185

Dc-bus filter w. surge limiter 3.8 41 81 8.2 68 120

Power supply 1.85 1.85 1.85 1.85 1.85 1.85

Controller and driver 2+5.3 2+5.3 2+5.3 2+5.3 2+5.3 2+5.3

IGBT and FWD 53 175 433 74 204 431

Output filter 8 47 93 12 67 107

Snubber circuits 0.117 1.167 3.44 0.173 1.73 5.16

Sum of loss : 104 429 924.6 154.52 609.5 1317.3

Table 20 : Approximate separated losses @14 kHz switching frequency

4.4. Reliability

The reliability is dependant on margins and precautions taken to ensure the longevity of

the device. Functions implemented to protect the device during errors or unwanted

conditions are a necessary factor to prevent permanent damage to the device. Also how

effective the components are cooled will result in a reliability factor as higher

temperature reduces lifetime in most semiconductors and especially electrolyte

capacitors. It may be needed to pinpoint every component that is subject to high power

loss and evaluate which temperature it will reach and how it will affect its lifetime.

Protective measures such as snubber circuits which will reduce voltage transients over

already very warm components can improve lifetime significantly as heat, junction

voltage and cosmic radiation are the main factors of reducing semiconductor lifetime

[21].

4.5. Simulations

Some of the simulations in this master-thesis are simply to illustrate the qualitative

behavior of some circuits while some are just to measure noise and disturbances. A few

Simulations has also been done to verify the calculated values of some components. For

the snubber considerations only Pspice have been used to illustrate the effects a snubber

circuits can inflict on the voltage transients. Any tries to reduce power loss has been

disregarded as it is very time consuming to find components with values that actually can

be realized.

For component value verification Matlab and Simulink have been used, the layout of the

system can be viewed in appendix B. A fully working system using the surge current

limiter can be found in appendix A1 where the graphs shows the growing voltage of the

DC-bus, the switched voltage from an IGBT couple and the same low-pass filtered

voltage. Under the same circumstances the currents running through the diode bridge and

the transistor bridge have been measured and the results are presented in appendix A2.

First is an averaged form of the following diode currents. The average current seen in the

first graph eventually corresponds to the steady-state value as expected. However the

instantaneous current seen in the second graph is of larger interest as in the case of no

surge current limiter used these currents as seen in the top graph in appendix A4 rise to a

dangerous level. The current running through the IGBT bridge can be seen being

regardless of using the surge current limiter or not as seen in the bottom graph in

appendix A3 and A4.

In appendix C1 the frequency spectrum at the input terminal has been measured using the

model set up as in appendix B. The voltage spectrum consists of the fundamental

frequency and its harmonics without any filtering. In appendix C2 the current harmonics

have been measured. The very same voltage measurement has been done in appendix C3

but this time it is filtered. The most disturbing harmonics as well as the total harmonic

distortion has been reduced significantly. In appendix C4 the same voltage measurement

has been carried out but this time a 12-pulse rectifier is used. In addition the current

harmonics have been measured in appendix C5. As can be sent the harmonics are much

higher in frequency but much lower in magnitude making it much easier to filter if

needed, in this scenario it is not.

5. Conclusions

The most obvious conclusion from performing this master-thesis is that the selection of

switching transistors should be given some time as the amount of power loss can be

reduced significantly with a delicate choice of components. However, as seen in this

thesis there are other parts of the converter contributing to a vast amounts of loss and in

particular weight not foreseen in the beginning of the thesis. In the converter studied in

this master-thesis the filters add up to almost 25% of the total weight, although in some

other configurations it becomes more or less prominent. The power lost in the filters is in

the same size as in the switching devices making them worth further more attention.

Altogether it would make good choice to address the filters the same detailed approach as

the switching devices.

The choice of using a SiC FWD ended in slightly less calculated loss in the switching

devise due to the negligible reverse recovery energy; however this should be simulated

extensively under appropriate conditions to be verified. Although uncertain loss it can

easily be said that the SiC technology requires less cooling as it can work under very high

temperatures. However using a FWD not integrated in the package introduces additional

capacitance and inductance possibly slows down the switching process canceling the

reduction gained. In addition the need for snubbers circuits increase when having higher

stray inductance.

As it turns out, using cooling relying on passive convection only in an aircraft a high

altitude would be too inefficient and the built-in cooling system would be the only

choice; regardless of which coolant used. In an aircraft flying on low altitude on the other

hand the high velocity winds could be used to cool the equipment.