switching dc motors with mosfets

[lumley32]

i am looking at switching a 12 dc motor (130A) with MOSFETs, as well as using them for general power switching (40A). just need to check a few things first.

1) is it ok to parallel them up to increase current handling?

2) am i correct in thinking that i dont need to worry to much about heat as they will only ever be on or off?

3) do i need any "driver" circuit or can i just put 12v into the gate?

i have simulated all of this but would just like to 100%!

Thanks Rich

 

[(*steve*)]

1) In general, mosfets can be paralleled much more easily that BJTs.

The reason is that as they heat up, their channel resistance rises, so they tend to shed current to the other mosfets (BJT's do the opposite, so the one taking the highest current tends to want to take more until it goes *poof*)

However, if you're doing this, you need to realise that the sharing isn't perfect.

2) There is some truth in the "only on or off, so no heat" argument, but it's not that simple. If the mosfets are not being switched often, then heat is still generated whilst they're turned on because the channel has a resistance. That resistance (Rds(on)) will cause power to be dissipated according to I^2.R.

If the mosfets are being switched on and off repetitively, due to the finite time it takes to do this switching, there may be significant rates of power lost during the very brief period the mosfet transitions between on and off or off and on. This adds up and can be a significant amount (actually dwarfing the power generated in the constant "on" case on many occasions).

Whether or not you need to consider this depends on how, and how often you switch the mosfets.

It is also important to consider the actual gate voltage (Vgs) that you will be using. He channel resistance of the mosfet will be dependant on this.

3) maybe... A lot depends on (2) above. How often are we switching? Mosfets require a certain amount of charge to be transferred to the gate to switch (it makes the gate look like a capacitor with some slightly odd characteristics). If your driver is capable of only a very little current, the switching speed will be slow. If it is capable of high current, it will be fast. Special gate drivers allow high gate currents to be used.

So the answer to this depends on (a) the load, (b) the switching frequency, and (c) the current available.

 

[lumley32]

ok that's interesting, i wasn't going to run them with no heat sink, just bolted to an ally casing ~ 4mm thick. i think ill do some testing on that as weight is a problem so dont want to be carrying heat-sinks i don't need!

as for the switching side of it, this is to replace relays no frequency as such. the high current load will be on for ~ 5 seconds then off for how ever long, could be day could be mins! the low current side will be either on or off.

i am thinking of doing something with the gate side as it will need to be reversed in some cases to work with existing controllers.

the other thing, all the circuits i see show the fet on the - side, ie switching the current return, i will be switching the + side, is that a problem?

 

[(*steve*)]

Bolting them to a 4mm thick casing *may* be sufficient.

Let's say you have an Rds(on) of 0.1 ohms and you parallel 4 of these devices. Whilst the current is flowing, the mosfets will dissipate 490 watts. That's an enormous amount of energy to dissipate.

However, if they have an Rds(on) or 0.02 ohms, and you parallel 10 if them, the heat you need to dissipate falls to under 40W. Depending on the ON time (and the duty cycle) the case itself may have enough thermal inertia to keep the devices cool, even if it can't easily radiate all that heat. As long as the off time is long enough, the case will cool between events.

You're probably going to want to pick a mosfet that can handle a fairly huge maximum current too. There is certainly the possibility that all the mosfets may not switch at exactly the same time (although that would be your target).

Switching the +ve side is not a problem. Typically P channel MOSFETS are used for that. P channel devices have somewhat lower specs than N channel devices, but this may not be a huge issue.

Here is a device that *might* work. You may also be able to select fewer mosfets with a lower channel resistance, but that's a design issue for you. This one is rated at 55V 74A and 0.02 ohm channel resistance (but watch the gate voltage needed for this).

You could also look at something like this, which has a channel resistance about half of the above, and can handle twice the current.

If you employ a suitable driver, you can use an N channel device, and maybe something a little exotic like this. However, you do need to concern yourself with details like.. Are the legs of the device going to melt with 140A flowing through them?

 

[lumley32]

well i was thinking about running 6 of these for the high current

http://docs-europe.electrocomponents.com/webdocs/0791/0900766b807910ff.pdf

well over what's needed, but many hands make light work! i think i will go for some sort of heat sink as it will all be in a in-closed space

interesting that they make fets that would switch 500A with legs that look like they wouldn't take 50!

 

[(*steve*)]

That has a 44milliohm Rds at Vgs = 20V. If we assume 60milliohms at 12V, then the power lost at 140A would be around 200W (that's about 35 W per device, but it could be up to (say) 70W for some if the devices if you allow for some spread in the Vgs).

A lower Rds(on) would benefit you (in single digits would be nice)

 

[lumley32]

yea my bad, that's the wrong link! too many pages open! the one im looking at now is 1.5millioms

is there likely to be that much difference in Vgs? i would have expected it would split fairly evenly.

thanks for your help!

 

[(*steve*)]

What you mean to ask is "Is there likely to be any significant variation of Rds between devices for a particular Vgs?"

No, they are likely to be quite similar, but it would be prudent to allow for some difference.

If you're making a single unit, you could measure them at an appropriate current and see what the voltage drop is.

If you're planning to produce more than one, the tedium of measuring (and possibly matching) mosfets can be overcome by assuming (say) a max +/- 10% difference. This also makes life easier if you ever have to replace one.

The same effect can be achieved by ensuring that none of the devices is actually operating too close to an edge of the safe operating area.

 

[lumley32]

well i am going to be running very much over the top, for the main circuit 4 200A fets, just playing with layouts and how to connect them all up!

i am thinking of useing copper bus bars to make life easer for myself, something like this

as you can tell im more of a design engineer than an electrical!