MOSFETS in parallel in switchmode improves effciency?

MOSFETs,

Rdson.

drivers in

use the driver

by mounting

allowed,

current?

Yes. N MOSFETs in parallel = 1/N times the resistance.

Except, I would expect the actual reduction in losses to be slightly better than 1/N. If you reduce the power dissipation in each MOSFET, the temperature should drop, resulting in a further decrease in resistance.

I've seen discussions of high-power BLDC motor drivers (aka ESCs, electronic speed controls) for large-scale RC aircraft using 36 MOSFETs (3-phase H-bridge with 6 MOSFETs for each "switch").

Dunno about that; my knowlege of inductors begins and ends with v=L.di/dt ;)

Are you using diodes or synchronous rectification? At low voltages, the diode losses can be significant.

urrent.

Z44A MOSFETs,

/ Rdson.

t.

2 drivers in

can use the driver

heat by mounting

llowed,

same current?

I once played around doing this in a small 1W supply. Did the paralleling of MOSFETs with the same assumptions you had, but the loss was not reduced. Also did a simulation, showed more loss with the dual solution, but never had time to chase the answer. Sorry, not much help, but an interesting discussion

Intuitively the RDS on losses should be lower, but you cannot be sure they turn on a the same time, so swithing losses are probably the same or worse.

Regards

Klaus

On a sunny day (Tue, 24 Feb 2009 21:54:33 +0000) it happened Nobody wrote in :

MOSFETs,

Rdson.

drivers in

use the driver

by mounting

allowed,

same current?

Nice.

Wow.

I am using a S30SC4N dual 30A schottky. The idea of synchronous rectification is attractive. Maybe I will play with that later. Just hope the body diodes of the MOSFETs do not get in the way. Will have to look up some of that stuff.

allowed,

same current?

One thing that springs to mind is whether you had sufficient gate current. A second MOSFET on the same driver means that you need to move twice the gate charge; if current is limited, that would increase the switching time.

e current.

IRFZ44A MOSFETs,

U^2 / Rdson.

rent.

ave 2 drivers in

you can use the driver

at heat by mounting

t allowed,

the same current?

/dt ;)

I don't know if its a buck you are playing with, but you could try to arrange the inductance according to the input voltage range/output voltage/etc to let it function only in discontinous mode, since reverse recovery losses of the freewheeling diode is minimized at the expense of more ripple current

/Klaus

On a sunny day (Tue, 24 Feb 2009 14:13:02 -0800 (PST)) it happened Klaus Kragelund wrote in :

It is a SEPIC. I found a nice article via a link on the IRF site, that describes some issues with synchronous rectification using MOSFETS:

I think I can do the secondary controller for the MOSFET with a PIC 16F690, using its hardware comparator, and doing the 'hold' delays in the interrupt :-) LOL

MOSFETs,

Rdson.

Incredible. The laws of Ohm and Joule at work.

drivers in

use the driver

by mounting

allowed,

current?

Sometimes we have to parallel up to four MOSFETs (current ~ 100A).

Keep in mind that the capacitive switching losses are increasing almost proportionally to the number of FETs. So, at some point, the Rds losses are getting equal to the other losses.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

using its hardware comparator, and doing

Fie, shameful PIC. BTW, I control the 1.5KW SMPS with AVR. Synchronous rectification, PFC bridge, protection and all stuff.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

dt ;)

Same here, that is all you need right? There are a few issues though, with coils. First there is the resistance in Ohms, a high current causes I^2 x R losses. Then there is the fact that the windings are close together spiralled, causing concentrated heat, like the filament in a light bulb is wound to increase heat.. This heats up the core material too. And then there is the skin-effect, already at frequencies of 16 kHz can it be of significance. To reduce this, you sometimes see litze wire used. Due to the skin effect the current actually flows more through the outside then the centre of the conductor, so increasing diameter of wire may not have the effect you expect, With even higher frequencies (several hundred kHz) this should be taken into account. Already in my simple setup, the losses in the coils equal or exceed those in the MOSFET switch, and the diode: The coils get very hot. It is interesting, as I can change software to get any frequency from say 19 kHz too 150 kHz, to experiment I found a nice program that will give you the wire length, inductance, resonance frequency for a given capacitance, and number of turns for almost all regular ferrite ring cores:

It runs in MS windows, and in Linux in the wine windows emulator. From that it is easy to get the wire length, and then for a given diameter, calculate the resistance. These low resistances are very difficult to measure, sometimes only a few turns are used. So les turns spaced further apart also reduces heat loss, but means less inductance, needs a higher switching frequency, and there the switching losses increase again. Some balancing act is required to get the most efficient setup.

allowed,

same current?

That is a major problem when using Mosfets. For switching power applications you should choose the Mosfet with the smallest gate capacitance (i.e. the one that just fits the bill). Oversizing is overkill -literally-.

--
Failure does not prove something is impossible, failure simply
indicates you are not using the right tools...
                     "If it doesn\'t fit, use a bigger hammer!"
--------------------------------------------------------------