Using IGBT or Mosfet

[sabu31]

Hi all,

I need to make a simple boost converter for boosting from 48 V to 192V at 1.2kW.
I have need a suggestion regarding which would be apt device for the active switching device (IRF300P226 Mosfet (2 in parallel) ) or (FGH50T65UPD IGBT (2 in parallel)). Which would give lower overall loss. Switching Frequency is around 50kHz.

Also presently I am using Diode SL80F040 for the boost diode, should i replace the diode with mosfet and go for synchronous boost.

 

[D.A.(Tony)Stewart]

Ron = 0 is never ideal, unless your reactance parts are so lossy that they provide the necessary damping factor or Q at minimum load.

Thus, you need to define your max/min load ratio or discrete values with time and determine your loop stability margin in each state for a step load in either direction to avoid a massive overshoot in one of the conditions. Then you will realize that efficiency and stability or step load error or load regulation error are all tradeoffs. Unless you can predict disturbances you must rely on voltage AND current feedback with whatever phase or gain margin you have at a given Q. This means you need all these specs and more to decide which is best. such as OVP margin, OTP margin, thermal runaway margin. Certainly the switch be it diode, FET or IGBT with the lowest RC=Tau value and the right value of Ron for OVP stability AND temp rise is best. These are the SiC and similar types which have a much lower Ron*C(0v)=Tau than Si types. These are also indicators of fast reverse recovery time which affects dynamic power loss directly with time during transition..

FETs naturally have a much higher Cout than IGBT's with a BJT output. Depending on your low side PWM switching f and resulting slew rate can make dynamic losses greater than static losses. Your primary average impedance R = 48V^2/ 1200W= 1.92 Ohms. Naturally you might estimate for 1% loss use a switch< 1% of 2 ohms.

IGBT's have a quasi linear curve of Vout= 0.7+ Imax * Rce. You can also estimate Rce for any transistor or diode or IGBT by consider the heatsink, Pd required to dissipate all the heat with < 60'C junction temperature rise for adequate reliability. The case will be cooler. My rule of thumb is Rce< k/ Pd for k = 0.25 to 1. If > 1 don't use it. Lower is better quality.

Thus for any transistor , diode or IGBT, Rpn=Rce= (Vmax-0.7) / Imax which is then approx = k/ Pd for k = 1 max for worst case.

Not to dwell on this but there is high correlation between thermal resistance and electrical ON resistance and 0V capacitance.

Read more what the experts say about switches and also read about how to design the optimal RC-D values of the high-side voltage booster.

https://www.mitsubishielectric.com/semiconductors/powerdevices/application_notes/sic-mosfet_super_mini_dipipm_e.pdf

It seems everyone is computing power incorrectly and not including dynamic power loss.
For Miller Capacitance I recall the dynamic peak will escalate rapidly above static loss at if the date drive resistance is not low enough <= Rg input. Then the dynamic FET loss Pavg increases with f but does not change with d.f. since there is high boost voltage gain on Vds/Vgs feedback on boost regulators. There is an optimum balance between static + dynamic loss with gate exponential current, RdsOn and Ciss. Although Mitsubishi must have assumed this in their choice of components.

I wonder why has no one commented on Mitsubishi's equations for power loss in FETs.
Power is always the average of 1 or N cycles but may be the product of Pavg=Vrms*Irms for 1 or Ncycles.

 

[Easy peasy]

dynamic losses are heavily dependent on the gate drive - this usually rests with the designer and his/her expertise ( or otherwise ) hence it appears fruitless to comment on dynamic ( switching ) losses until the layout and gate drive are known.

 

[KlausST]

Power is always the average of 1 or N cycles but may be the product of Pavg=Vrms*Irms for 1 or Ncycles.

for power dissipation of a transistor this almost never is true.
(indeed I can only think of two extreme cases: 0% duty cycle and 100% duty cycle)

Example: I have an 10mOhms MOSFET switching a 10 Ohms resistive load with 50% duty cycle:
* then we have a square wave of about 0V/10V with 50% duty cycle, making it 7.07V RMS across the MOSFET (this is what a true RMS meter shows across DS)
* and the current is a square wave of about 0A/1A with 50% duty cycle, making 0.707A RMS through the MOSFET (this is what a true RMS meter shows as I_D)

According the formula P_MOSFET = V_RMS x I_RMS = 7.07V x 0.707A = 5W (the true conduction power dissipation is about 0.005W)
But indeed the P_tot = integral [from 0 to t_period] (V(t) x I(t)) /t_period

The formula is true for:
constant ohmic loads .... which a switching transistor never is.

Where am I wrong?

Klaus

 

[danadakk]

@danadakk The fact that the senseless term "RMS power" is used in an ADI application note doesn't give it more sense. See https://en.m.wikipedia.org/wiki/Audio_power#Continuous_power_and_.22RMS_power.22

--- Updated Apr 26, 2024 ---

Thanks to Easy peasy for the detailed calculation. I would usually answer the design question with a short consideration.
1. IGBT isn't well suited for 48V input due to high Vce,sat
2.IGBT operation at 50 kHz is no good idea

Seems to me ADI is pointing out its senseless, which does give it standing (sense), its senseless :

You do not want to calculate the rms value of the ac power waveform. This produces a result that is not physically meaningful.

To be or not to be that is, or is not, the question .....

IGBT based welding at 100 Khz : https://www.millerwelds.com/resourc...nables-lighter-more-powerful-welding-machines

Hypertherm, Lincoln more like 50 - 60 Khz.


Knight

 

[sabu31]

19 m-ohm ( worst case for the fet ) x 1.8 for 100deg C in the die, divided by 2 mosfets = 17.1 m-ohm

IGBT = 1.4 V worst case @ 30 A - plus these have poorer turn off losses

48 V to 192 V at 1200W output give s a D == 0.85 to allow for losses

so at steady state, I in = 26.3 amps, I out ave = 6.25A, so the output caps sees 21A pk ripple

Assuming an L that gives +/- 15% current ripple = Ipk on input is 30.3 A, I start each cycle = 22.4 A

rms current in the mosfets is:
View attachment 190372
= 24.43 amps @ D = 0.85

Mosfet cond losses = 10.2 watts ( 5.1 watts each )

IGBT = 34.2 watts losses ( 17.1 watts each )

turn on losses are dictated by Rgate ( on ) and can be similar for both devices ( igbt's can turn on really fast if you allow it - ~ 20nS )

Turn off losses can be near zero for the mosfet due to its non linear Cds, for < 30nS Vgs 10V to zero the losses ( turn off ) are very small indeed

for the igbt however: best case is 0.22 mJ turn off losses @ 30A, 192V, which at 50kHz = 11 watts, 5.5 W ea.

Thanks Easy-Peasy for answer and along with calculation. I needed a decision between IGBT and Mosfet, a mosfet is way to go for this application.

However, there must some cut-off limit say input voltage/ current where IGBT starts making more sense as compared to Mosfet or vice-versa.

--- Updated Apr 26, 2024 ---

Seems to me ADI is pointing out its senseless, which does give it standing (sense), its senseless :



To be or not to be that is, or is not, the question .....

IGBT based welding at 100 Khz : https://www.millerwelds.com/resourc...nables-lighter-more-powerful-welding-machines

Hypertherm, Lincoln more like 50 - 60 Khz.


Knight

Yes. I am seeing IGBT application in commercial half bridge series resonant inverter based induction cooking . They also operate in range of 40k to 60k.

 

[D.A.(Tony)Stewart]

The AD's datasheet plot below indicates for almost same Imax rating in Si MOS vs IGBT, shown for static loads.

IGBT's are the same as FET when using 50% or rated current then IGBT's have lower loss from lower Rce than RdsOn in spite of 0.7V diode offset.
This crossover threshold drops as the temperature rises to max 150'S. Using 50% of I max assumes Tjcn rising is 50% above 25'C with the same system thermal resistance.
T
One can choose I max current rating that is 2x nominal use for reliability using the typ. plot @ 25'C or could use 5x for FETs, but then dynamic power loss must be considered from L/R, RceCce << RdsOnCoss and other time constants. So one must not assume FETs are faster with LC reactive effects and lower Rdson means higher Coss. and longer L/R = T time constant and energy dissipation.

e.g. IGBT @ 25'C Rce (1.35-0.7V)/25A = 26 mohm
vs SJMOS, @ 25'C RdsOn=2.2V/25A = 88 mohm (Silicon Junction Metal Oxide Semiconductor)

Of course, this is subjective, comparing only one pair with ~same I max rating. I marked up below from AD link from @danadakk

 

[Easy peasy]

@sabu31 IGBT's are cheaper - and can be quieter for EMI due to their tail losses - but they generally require more heatsink.

larger die IGBT's can have 10uS short ckt capability - which makes them very rugged compared to mosfets - hence you never see mosfets in motor drive applications.

 

[D.A.(Tony)Stewart]

Correction: #26 link was Tosbiba' not AD https://toshiba.semicon-storage.com/eu/semiconductor/knowledge/faq/mosfet_igbt/igbt-002.html

--- Updated Apr 26, 2024 ---

Hi all,

I need to make a simple boost converter for boosting from 48 V to 192V at 1.2kW.
I have need a suggestion regarding which would be apt device for the active switching device (IRF300P226 Mosfet (2 in parallel) ) or (FGH50T65UPD IGBT (2 in parallel)). Which would give lower overall loss. Switching Frequency is around 50kHz.

Also presently I am using Diode SL80F040 for the boost diode, should i replace the diode with mosfet and go for synchronous boost.

Hi Sabu,

You ought to know that you cannot choose an optimal design without details on dynamic and reactive load impedance ( time and frequency) Please define Z(f) and Z(t), along with voltage error tolerance, but preferably do not restrict yourself to devices or switching frequency (implementation-specific design-spec error) unless you state a good reason.

Cheers from Toronto,
Tony