parallel FETs

V

V8TR4

Guest
Hello,

I want to make a controller for a golf cart motor and am considering using
paralleled MOSfets or HEXfets. I know that there can be problems with the
individual xsistors not sharing the load equally. Would it help if I had a
seperate line driving each one rather than hooking up to some common rail to
activate them? Any info would be much appreciated.

I am looking to control a PWM with a current up to 300ampere surge and
80ampere running.

Thanks
 
"V8TR4" <v8tr4@REMOVE2EMAILMEadelphia.net> wrote in message
news:3f-dnWROg77O-gfcRVn-pQ@adelphia.com...
Hello,

I want to make a controller for a golf cart motor and am considering using
paralleled MOSfets or HEXfets. I know that there can be problems with the
individual xsistors not sharing the load equally. Would it help if I had a
seperate line driving each one rather than hooking up to some common rail
to
activate them? Any info would be much appreciated.

I am looking to control a PWM with a current up to 300ampere surge and
80ampere running.

Thanks
Click to expand...

Here are some thoughts:

* Take a look at the Fairchild ISL9N302AP3. Rds less than .002 Ohms (25C,
10V). One of these can handle all your current in the silicon, but the
TO-220 package is only rated for 75 Amps. Use several in parallel or look
for this sort of device in a better package (or both).

* I would not add any resistance. Yes, I would use individual wires to the
battery (motor, whatever) for each device. You need to consider how you're
going to sense current so you can protect the devices. Something I've seen
in some PWM controllers is that they watch Vds. You could possibly do that,
maybe where the drain wires get connected together. There will be noise on
this point.

* Give some thought to how you're going to handle the ringing you're going
to have on the drains. Even a couple of feet of wire can cause a nasty spike
if you shut off the current fast enough. You may want to consider tailoring
the gate drive to help this problem.

* I would probably not depend on the integral body diodes in the FETs.

* Drive each gate through its own resistor.

* Layout and wire dressing will be important.

* Be sure you know what you mean by 300 Amps surge and pay attention to the
peak current vs duration curves for the device(s) you choose.

Good luck.

John
 
V8TR4 wrote:

Hello,

I want to make a controller for a golf cart motor and am considering using
paralleled MOSfets or HEXfets. I know that there can be problems with the
individual xsistors not sharing the load equally. Would it help if I had a
seperate line driving each one rather than hooking up to some common rail to
activate them? Any info would be much appreciated.

I am looking to control a PWM with a current up to 300ampere surge and
80ampere running.

Thanks
Click to expand...
The key to successful paralleling of FETs is tight thermal coupling. The
+ve tempco of Rdson has a self-sharing kind of effect: the FET carrying
more current gets hotter, thereby increasing Rdson, which in turn
decreases the current thru that particular FET. If all the FETs share
the same thermal environment (eg closely spaced symmetric layout) then
direct paralleling can and will work. Normally one uses individual gate
resistors. The same approach works for IGBTs - I have direct-paralleled
6 x 600A IGBT.

Or you could buy a semikron SK300MB075 300A 75V dual FET in a SemiTop
package. Great thermal coupling, good thermal interface etc.

The real killer is switching - at which point stray inductance tends to
dominate. if dI = 300A and dt = 100ns say, then dI/dt = 3kA/us. 10nH
will drop 30V! Symmetry is a saviour here - if the layout is symmetric
then all the strays will be the same. If it is not, they wont.

Cheers
Terry
 
Boris Mohar wrote...
V8TR4 wrote:

I want to make a controller for a golf cart motor and am considering
using paralleled MOSfets or HEXfets. I know that there can be problems
with the individual xsistors not sharing the load equally. Would it
help if I had a seperate line driving each one rather than hooking up to
some common rail to activate them? Any info would be much appreciated.

I am looking to control a PWM with a current up to 300 ampere surge
and 80 ampere running.

IEEE Transactions on Electron Devices Vol. ED-31, No.7, July 1984
An Analysis and Experimental Verification of Parasitic Oscillations
in Paralleled Power MOSFETS.
Also appears in Siliconix MOSPOWER APPLICATIONS book ISBN 0-930519-00-0
http://www.smpstech.com/books/booklist.htm
http://www.smpstech.com/severns/papers.htm
Click to expand...
V8TR4 is on target. Most parallel switching FET oscillations are
easily solved using John Smith's advice, "Drive each gate through
its own resistor." And it would be wise to heed the rest of his
advice as well. The FETs in the article mentioned above (no author)
had their gates directly tied in parallel. That's a real no-no.


--
Thanks,
- Win
 
On Wed, 17 Nov 2004 15:04:06 +1300, Terry Given wrote:

IEEE Transactions on Electron Devices Vol. ED-31, No.7, July 1984

An Analysis and Experimental Verification of Parasitic Oscillations in
Paralleled Power MOSFETS.

Also appears in Siliconix MOSPOWER APPLICATIONS book
ISBN 0-930519-00-0

Bwahahahahahaaaa! I have a hard-cover of that :)

The first time I EVER saw a Routh-Hurwitz stability analysis performed
on a circuit. A lovely bit of maths.
Click to expand...
Argh. That paper isn't in my old data book. How long is it? I've got
some papers on that analysis technique, but haven't seen it used
aside from that.
--
Best Regards,
Mike
 
In article <Kvvmd.2994$9A.112752@news.xtra.co.nz>,
Terry Given <my_name@ieee.org> wrote:

The key to successful paralleling of FETs is tight thermal
coupling. The +ve tempco of Rdson has a self-sharing kind of
effect: the FET carrying more current gets hotter, thereby
increasing Rdson, which in turn decreases the current thru that
particular FET. If all the FETs share the same thermal
environment (eg closely spaced symmetric layout) then direct
paralleling can and will work.
Click to expand...
I can't make sense of that Terry.

If all FETs are forced to run at the same temperature
then the current (or power) ratio will be constant.

Shouldn't there be some thermal elasticity between FETs,
to allow the device with the lowest RdsON to run slightly
warmer?

--
Tony Williams.
 
V8TR4 wrote:

Hello,

I want to make a controller for a golf cart motor and am considering using
paralleled MOSfets or HEXfets. I know that there can be problems with the
individual xsistors not sharing the load equally. Would it help if I had a
seperate line driving each one rather than hooking up to some common rail to
activate them? Any info would be much appreciated.

I am looking to control a PWM with a current up to 300ampere surge and
80ampere running.

Thanks
Click to expand...
I see all the comments about paralleling *FETs. (All well and good, if you
insist on paralleling.) Am I missing something by suggesting that the given
peak and continuous ratings are well within the capabilities of a single
device?? Probably less cost, too.

Thanks, Steve
 
Hi Winfield,

V8TR4 is on target. Most parallel switching FET oscillations are
easily solved using John Smith's advice, "Drive each gate through
its own resistor." And it would be wise to heed the rest of his
advice as well. The FETs in the article mentioned above (no author)
had their gates directly tied in parallel. That's a real no-no.
Click to expand...
It depends on the size of it all. With very small FETs that are right
next to each other you can reach stability without. When the FETs are
large, on a heat sink and far apart that is, of course, a different
matter and resistors are a must. If have seen PWM converters where the
turn-on-off loss increases due to these series resistors were
responsible for more than a 1% penalty in overall efficiency.

Regards, Joerg

http://www.analogconsultants.com
 
Tony Williams <tonyw@ledelec.demon.co.uk> wrote in message news:<4d0f3af58ftonyw@ledelec.demon.co.uk>...
In article <Kvvmd.2994$9A.112752@news.xtra.co.nz>,
Terry Given <my_name@ieee.org> wrote:

The key to successful paralleling of FETs is tight thermal
coupling. The +ve tempco of Rdson has a self-sharing kind of
effect: the FET carrying more current gets hotter, thereby
increasing Rdson, which in turn decreases the current thru that
particular FET. If all the FETs share the same thermal
environment (eg closely spaced symmetric layout) then direct
paralleling can and will work.

I can't make sense of that Terry.

If all FETs are forced to run at the same temperature
then the current (or power) ratio will be constant.

Shouldn't there be some thermal elasticity between FETs,
to allow the device with the lowest RdsON to run slightly
warmer?
Click to expand...
Even if they are mounted side-by-side on the same heatsink their can,
and will be, temperature differences between them. Heatsinks, too,
have thermal resistance.
 
"V8TR4" <v8tr4@REMOVE2EMAILMEadelphia.net> wrote in message news:<3f-dnWROg77O-gfcRVn-pQ@adelphia.com>...
Hello,

I want to make a controller for a golf cart motor and am considering using
paralleled MOSfets or HEXfets. I know that there can be problems with the
individual xsistors not sharing the load equally. Would it help if I had a
seperate line driving each one rather than hooking up to some common rail to
activate them? Any info would be much appreciated.

I am looking to control a PWM with a current up to 300ampere surge and
80ampere running.

Thanks
Click to expand...
This may be of interest:

SCHÖNKNECHT AND DE DONCKER: PARALLEL CONNECTION OF SOFT-SWITCHING
HIGH-POWER HIGH-FREQUENCY INVERTERS
IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 39, NO. 2,
MARCH/APRIL 2003

Abstract—Inductive heating applications like pipe welding or
steel strip annealing require electrical power ratings of several
megawatts at frequencies up to 100 kHz and higher. The large
power-frequency product represents a significant challenge for
today's semiconductor technology. As the absolute maximum
rating of a single-stage inverter is often far below rated power,
several inverters or several devices have to be connected in parallel.
This paper presents a novel topology, consisting of parallelconnected
soft-switching high-frequency inverters. Distinctive
features include flexible configurations, negligible shunt currents
between inverters, and equally shared power among inverters.
Furthermore, compared to a single-inverter system, no additional
reactive components are necessary for connecting inverters in
parallel and there is no need of a high-frequency transformer to
adapt the impedance of the load to the inverters.

A novel topology for connecting soft-switching inverters in
parallel is proposed in this paper. The following demands, which
often cause difficulties, have to be taken into consideration:
1) no or negligible circulating currents, even if inverters
switch nonsynchronous;
2) equal power sharing among parallel-connected inverters;
3) flexible variation of the number of parallel inverters, enabling
a wide power range and, thus, the possibility to
minimize the number of inverters for the given power;
4) no additional devices for connecting inverters in parallel;
5) minimization of losses in semiconductors by ZVS softswitching
operation at near zero-current.
 
In article <mq2lp018rl2dj73kt0q7l476msojmgsh52@4ax.com>,
Spehro Pefhany <speffSNIP@interlogDOTyou.knowwhat> wrote:
[...]
When used as switches MOSFETs tend to share rather than hog the current.

Note that they share better if given individual heat sinks rather than
one big heatsink, so that the temperatures can vary a bit.
Click to expand...
Yes the currents are more equal but, unless I am missing something it
really doesn't matter that much whether you use one big or individual
sinks since the die temperatures end up the same in the two cases and it
is the die temperature that really matters to the survival of the devices.

BTW: at low currents in linear operation MOSFETs do hog current. As
temperature goes up Vth and gm decrease and Rds(on) increase.

--
--
kensmith@rahul.net forging knowledge
 
Ken Smith wrote...
Spehro Pefhany wrote:
[...]
When used as switches MOSFETs tend to share rather than hog the current.

Note that they share better if given individual heat sinks rather than
one big heatsink, so that the temperatures can vary a bit.

Yes the currents are more equal but, unless I am missing something it
really doesn't matter that much whether you use one big or individual
sinks since the die temperatures end up the same in the two cases and it
is the die temperature that really matters to the survival of the devices.

BTW: at low currents in linear operation MOSFETs do hog current.
As temperature goes up Vth and gm decrease and Rds(on) increase.
Click to expand...
Unless one is fortunate enough to be using the elegant lateral power
MOSFETs from Hitachi, etc.


--
Thanks,
- Win
 
In article <cnh5rv$rh5$1@blue.rahul.net>,
Ken Smith <kensmith@green.rahul.net> wrote:

Yes the currents are more equal but, unless I am missing
something it really doesn't matter that much whether you use one
big or individual sinks since the die temperatures end up the
same in the two cases and it is the die temperature that really
matters to the survival of the devices.
Click to expand...
There are slight current or power sharing differences
between one sink or individual sinks.

I can work out the current/power sharing sums for two
devices on individual heat sinks, and for one heatsink
where Theta(j-sink) is zero, but have so far not managed
to work out a generalised sum for one heatsink where
Theta(j-sink) is non-zero. I suspect that what happens in
that last case could depend on the ratio of Theta(j-sink)
against Theta(sink-air).

--
Tony Williams.
 
On Tue, 16 Nov 2004 22:31:28 +0000 (UTC), the renowned
kensmith@green.rahul.net (Ken Smith) wrote:

In article <3f-dnWROg77O-gfcRVn-pQ@adelphia.com>,
V8TR4 <v8tr4@REMOVE2EMAILMEadelphia.net> wrote:
Hello,

I want to make a controller for a golf cart motor and am considering using
paralleled MOSfets or HEXfets. I know that there can be problems with the
individual xsistors not sharing the load equally. Would it help if I had a
seperate line driving each one rather than hooking up to some common rail to
activate them? Any info would be much appreciated.

I am looking to control a PWM with a current up to 300ampere surge and
80ampere running.


When used as switches MOSFETs tend to share rather than hog the current.
Click to expand...
Note that they share better if given individual heat sinks rather than
one big heatsink, so that the temperatures can vary a bit.


Best regards,
Spehro Pefhany
--
"it's the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
 
On Wed, 17 Nov 2004 11:51:13 +1300, Terry Given wrote:

V8TR4 wrote:

Hello,

I want to make a controller for a golf cart motor and am considering using
paralleled MOSfets or HEXfets. I know that there can be problems with the
individual xsistors not sharing the load equally. Would it help if I had a
seperate line driving each one rather than hooking up to some common rail to
activate them? Any info would be much appreciated.

I am looking to control a PWM with a current up to 300ampere surge and
80ampere running.

Thanks

The key to successful paralleling of FETs is tight thermal coupling. The
+ve tempco of Rdson has a self-sharing kind of effect: the FET carrying
more current gets hotter, thereby increasing Rdson, which in turn
decreases the current thru that particular FET. If all the FETs share
the same thermal environment (eg closely spaced symmetric layout) then
direct paralleling can and will work.
Click to expand...
Oddly, Spehro Pefhany says exactly the opposite a couple of FUs back.
He says, put them on individual sinks, so each can find its own
thermal equilibrium, or something like that.

I'm not well-versed enough in this area to figure out which is
"right." I have encountered a system with banks of MOSFETs on
common heat sinks, with very carefully calculated individual
source resistors, so each would settle at a _different_ current,
the excuse being some mumbo jumbo about air flow over the heat
sink. And I know even _less_ about thermal transfer stuff. :)

Thanks,
Rich


Normally one uses individual gate
resistors. The same approach works for IGBTs - I have direct-paralleled
6 x 600A IGBT.

Or you could buy a semikron SK300MB075 300A 75V dual FET in a SemiTop
package. Great thermal coupling, good thermal interface etc.

The real killer is switching - at which point stray inductance tends to
dominate. if dI = 300A and dt = 100ns say, then dI/dt = 3kA/us. 10nH
will drop 30V! Symmetry is a saviour here - if the layout is symmetric
then all the strays will be the same. If it is not, they wont.

Cheers
Terry
Click to expand...
 
On Tue, 16 Nov 2004 19:59:16 -0500, Boris Mohar
<borism_-void-_@sympatico.ca> wrote:


IEEE Transactions on Electron Devices Vol. ED-31, No.7, July 1984

An Analysis and Experimental Verification of Parasitic Oscillations in
Paralleled Power MOSFETS.

Also appears in Siliconix MOSPOWER APPLICATIONS book
ISBN 0-930519-00-0

http://www.smpstech.com/books/booklist.htm
http://www.smpstech.com/severns/papers.htm
Click to expand...
At the time the Siliconix article was published, you could still buy
multi-chip fet modules from IR that were miswired, preventing
predictable switching performance.

other refs

AN918 Motorola
http://www.datasheetarchive.com/download.php?pi=29568

AN7513, AB-9, Fairchild
http://www.fairchildsemi.com/an/AN/AN-7513.pdf#page=1
http://www.fairchildsemi.com/an/AB/AB-9.pdf#page=1

AN941 IR
http://www.irf.com/technical-info/appnotes/an-941.pdf

PESC'01 Huang
Characterization of Paralleled Super Junction MOSFET Devices under
Hard- and Soft-Switching Conditions link dead

EPE'99 Jeannin
http://manuales.elo.utfsm.cl/conferences/seminarios/EPFL/pc/papers/614.pdf

RL
 
You must log in or register to reply here.

Welcome to EDABoard.com

Sponsor

Back
Top