HobbyKing.com New Products Flash Sale
Reply
Thread Tools
Old Sep 26, 2009, 10:18 PM
Look at my little red dot
On the Snap on truck
Joined May 2004
806 Posts
Discussion
Does stacking fets, really increase power handling?

I was wondering if stacking fets really does increase power handling capability? some say if "stack the fets" you double,triple etc etc power handling, some say no ?
It's for a rc tank with known issues with burning up fets
http://www.rctankwarfare.com/Forums/...p?topic=2397.0
I could not just repost the picture ,top picture.
TIA
Joat is offline Find More Posts by Joat
Reply With Quote
Sign up now
to remove ads between posts
Old Sep 26, 2009, 11:32 PM
Registered User
PMB-NZ's Avatar
Wellington, New Zealand
Joined Aug 2009
186 Posts
It does increase current capacity but the layout/circuit-board or wiring (depending on how it's arranged has to ensure the load is shared. If the layout favours one fet (even by the smallest margin) it will blow then the next and so on. The theory looks simple but it's tricky to make reliable. Small brushless ESCs are pretty good but they group the fets very closely and use short fat circuit board tracks.

Paul
PMB-NZ is offline Find More Posts by PMB-NZ
Reply With Quote
Old Sep 26, 2009, 11:44 PM
Dimension Engineering
Akron, Ohio
Joined Jan 2002
1,705 Posts
Best answer is "maybe, but probably not"

Doubling the fets up will halve the continuous conduction losses and double the switching losses, roughly speaking. If the design of the drive is such that the conduction losses are dominant, then you'll probably see an improvement. If the switching losses are dominant, then it will get worse. Parasitics get worse (ringing, radiation, etc increase)

Switching to a better FET (lower switching losses, lower conduction losses, or both) will usually help some. Of course, then you have to start considering the copper losses, and whether the increased amperage is causing the FETs to avalanche, etc.

Motor drive design is tricky. FWIW, I have a new dual brushed driver in the pipeline, and the development budget for it is approximately $100,000. That's for ONE design, and that's considering that we already make five similar products.

Really, though, for a fan cooled speed control, the EASIEST thing you can do to increase the current rating a little bit is put a bigger fan on it. That will almost always yield some improvement.
Comatose is offline Find More Posts by Comatose
Reply With Quote
Old Sep 27, 2009, 12:59 AM
Registered User
vintage1's Avatar
East Anglia, UK
Joined Sep 2002
29,703 Posts
Whilst everything Comatose says is perfectly true, I think he is a little pessimistic. But then Dimension are probably the best designers in our little world of RC, and don't take risks..

the reality is that most ESC designs are not so well specced, and that they are probably cost reduced ripoffs of someone else's.

IF the drive circuitry is up to it, and the thing doesn't go unstable, paralleling FETS is a quick and simple way to increase power handling. The current sharing problems are not as severe as suggested above: FETS have more resistance than the tracks feeding them and positive temperature coefficients, so they current share well without any special steps needing to be taken.

And many lower power designs are often high power designs with the FETS reduced in size or removed..

I will take issue on the fan cooling bit though. MOST of the gains of fan cooling come at prettty low airspeeds. Adding more fans doesn't often do a great deal more. Not without some special heatsinking like a CPU sink.
vintage1 is offline Find More Posts by vintage1
Reply With Quote
Old Sep 27, 2009, 02:26 AM
Look at my little red dot
On the Snap on truck
Joined May 2004
806 Posts
Quote:
Originally Posted by Comatose
Best answer is "maybe, but probably not"

Doubling the fets up will halve the continuous conduction losses and double the switching losses, roughly speaking. If the design of the drive is such that the conduction losses are dominant, then you'll probably see an improvement. If the switching losses are dominant, then it will get worse. Parasitics get worse (ringing, radiation, etc increase)

Switching to a better FET (lower switching losses, lower conduction losses, or both) will usually help some. Of course, then you have to start considering the copper losses, and whether the increased amperage is causing the FETs to avalanche, etc.

Motor drive design is tricky. FWIW, I have a new dual brushed driver in the pipeline, and the development budget for it is approximately $100,000. That's for ONE design, and that's considering that we already make five similar products.

Really, though, for a fan cooled speed control, the EASIEST thing you can do to increase the current rating a little bit is put a bigger fan on it. That will almost always yield some improvement.
Ok, a bit way beyond me, I was planning on doing a fan mod ,as from the factory there are none, I may try to put a heatsink.
as in the link,some posted that on the kyosho miniz the factory double them up? was this more of a failsafe then a power handling issue?-your best guess is still better then mine
Joat is offline Find More Posts by Joat
Reply With Quote
Old Sep 27, 2009, 02:29 AM
Look at my little red dot
On the Snap on truck
Joined May 2004
806 Posts
Quote:
Originally Posted by vintage1
Whilst everything Comatose says is perfectly true, I think he is a little pessimistic. But then Dimension are probably the best designers in our little world of RC, and don't take risks..

the reality is that most ESC designs are not so well specced, and that they are probably cost reduced ripoffs of someone else's.

IF the drive circuitry is up to it, and the thing doesn't go unstable, paralleling FETS is a quick and simple way to increase power handling. The current sharing problems are not as severe as suggested above: FETS have more resistance than the tracks feeding them and positive temperature coefficients, so they current share well without any special steps needing to be taken.

And many lower power designs are often high power designs with the FETS reduced in size or removed..

I will take issue on the fan cooling bit though. MOST of the gains of fan cooling come at prettty low airspeeds. Adding more fans doesn't often do a great deal more. Not without some special heatsinking like a CPU sink.
So you think stacking might actually make a difference? I was looking at heatsink and fan cooling anyway
Joat is offline Find More Posts by Joat
Reply With Quote
Old Sep 27, 2009, 02:45 AM
Registered User
Joined Apr 2009
188 Posts
Quote:
Doubling the fets up will halve the continuous conduction losses and double the switching losses
Why exactly do the switching losses double? During the switching wouldn't there also be just half of the current flowing through each FET?

For IMHO for these RC-Brushless ESC's the switching losses are dominant.
OiOi is offline Find More Posts by OiOi
Reply With Quote
Old Sep 27, 2009, 04:48 AM
7000mw of raw power!
rich smith's Avatar
New Hampshire (not the old one)
Joined Dec 2006
5,987 Posts
Baloney not withstanding, I've had excellent results stacking FETs on small Turnigy controllers. Bottom line: it let me use bigger props without blowing up.
rich smith is offline Find More Posts by rich smith
Reply With Quote
Old Sep 27, 2009, 04:57 AM
Registered User
vintage1's Avatar
East Anglia, UK
Joined Sep 2002
29,703 Posts
Quote:
Originally Posted by OiOi
Why exactly do the switching losses double? During the switching wouldn't there also be just half of the current flowing through each FET?

For IMHO for these RC-Brushless ESC's the switching losses are dominant.
switching losses are about charging up capacitors - parasitics. So they do double - twice the stray capacitance. If that starts to make the output FETs slow down so they are half on for an appreciable period, then heat can rocket.

Comatose is right in every respect except that he doesn't have an accurate figure on how many controllers would be worse rather than better with stacked FETS. I am inclined to think more benefit than are made worse.

But its not as sure fire as Rich seems to indicate.

Proceed with caution.
vintage1 is offline Find More Posts by vintage1
Reply With Quote
Old Sep 27, 2009, 05:10 AM
Registered User
Joined Apr 2009
188 Posts
Quote:
switching losses are about charging up capacitors - parasitics. So they do double - twice the stray capacitance. If that starts to make the output FETs slow down so they are half on for an appreciable period, then heat can rocket.
Yes that's clear, but his only happens if the driving isn't able to deliver enough current for the second FET. But regardless of the gate-cap to charge the FET needs some time for switching, and this shouldn't increase. Or am I wrong. For the before mentioned doubling of the switching losses doesn't make sense only in relation to the gate cap charging as this can usually easily be fixed.

Surely, the parasitic effects due to increased gate capacitance will increase, no doubt about that!
OiOi is offline Find More Posts by OiOi
Reply With Quote
Old Sep 27, 2009, 05:22 AM
Registered User
vintage1's Avatar
East Anglia, UK
Joined Sep 2002
29,703 Posts
Quote:
Originally Posted by OiOi
Yes that's clear, but his only happens if the driving isn't able to deliver enough current for the second FET. But regardless of the gate-cap to charge the FET needs some time for switching, and this shouldn't increase. Or am I wrong. For the before mentioned doubling of the switching losses doesn't make sense only in relation to the gate cap charging as this can usually easily be fixed.

Surely, the parasitic effects due to increased gate capacitance will increase, no doubt about that!
MMm. First of all 'can easily be fixed' != 'simply stack the FETS.'

Also, its not clear as to what actually does determine the switching times. IF it is no more than charging up parasitics, then it will increase anyway.


My memory of FET theory is rusty, but ISTR that is more or less what it is.

I.e. switching times is more ore less the stray capacitance inherent in the chip, times the miller effect, times the stray resistances inherent in the chip (or necessarily placed in series with the gate for stability).

Also, in H bridge type situations, there is always the possibility that and upper and lower pair can become switched on together, Certainly CMOS chips used to do just that, drawing BIG spikes at switchover. If that happens doubling the FETs will also double the losses.
vintage1 is offline Find More Posts by vintage1
Reply With Quote
Old Sep 27, 2009, 05:32 AM
Registered User
Joined Apr 2009
188 Posts
Quote:
MMm. First of all 'can easily be fixed' != 'simply stack the FETS.'
OK, I agree on that.

If the driver has enough punch you would just make the gate resistor smaller or add a second gate resistor for the other FET.
Wouldn't that work?

Quote:
I.e. switching times is more ore less the stray capacitance inherent in the chip, times the miller effect, times the stray resistances inherent in the chip
As i still remember it's dependant on many factors, but these factors are fixed so by the design of the FET, that it needs such and such time to switch, no matter how much current you would be able to drive into the gate, as you also only have a limited voltage to drive. And in most designs the current driver is good enough to be able to drive a second one.
Surely maybe the time would increase a bit. But i would doubt that this would double the switching losses.

Quote:
Also, in H bridge type situations, there is always the possibility that and upper and lower pair can become switched on together, Certainly CMOS chips used to do just that, drawing BIG spikes at switchover. If that happens doubling the FETs will also double the losses.
Lol, I doubt any ESC switches High and Low Side together. But surely due to the maybe increased switching time it could happen, that they get into a short situation (if the timing was close before), then the losses would drastically increase.
OiOi is offline Find More Posts by OiOi
Reply With Quote
Old Sep 27, 2009, 06:26 AM
Registered User
vintage1's Avatar
East Anglia, UK
Joined Sep 2002
29,703 Posts
Don't be bemused by specs of 'switching times'

What those mean are basically that the internal resistance between the gate pin and the actual gate layer itself.. times the actual and virtual capacitance on the gate due to Miller effect, becomes the RC time constant. Given that you can only drive the gate so hard voltage wise, that sets the upper limit on switching speed.

During the transit time, the FET may or may not be passing more or less current. That's a hard one to predict with a complex load.

As far as two fets on together..its something you normally would do, to get best switching speeds, or simply because you are lazy and drive the on signal synchronously with the off, and the delays in both lead to a 'both half on' scenario.

By the way approximately ALL the heat in a typical MOS CPU chip is 'switching losses' - charging and discharging stuff through the internal resistances as its clocked..
vintage1 is offline Find More Posts by vintage1
Reply With Quote
Old Sep 27, 2009, 07:41 AM
Registered User
Joined Apr 2009
188 Posts
What I actually wanted to say is, that the said "switching losses would double" is IMHO not correct. Am I right here? Sure it's quite some time now since I studied...Wouldn't bet my hand on that.

Quote:
By the way approximately ALL the heat in a typical MOS CPU chip is 'switching losses' - charging and discharging stuff through the internal resistances as its clocked..
For older CPUs certainly true. New ones in very small processes tend to have more and more leakage current. Therefore complex driving patterns are necessary which disable the power to whole circuits to save some power, when not needed. Also process changes were made with high-k dielectric, SOI, etc.
And what should a CPU drive anyway, LOL, just needs to drive the next gate caps in the logic. Sure there are mainly switching losses.
OiOi is offline Find More Posts by OiOi
Last edited by OiOi; Sep 27, 2009 at 08:11 AM.
Reply With Quote
Old Sep 27, 2009, 08:06 AM
homo ludens modellisticus
Ron van Sommeren's Avatar
The Netherlands, GE, Nijmegen
Joined Feb 2001
10,676 Posts
Thoughts about and methods for better controller cooling:
www.mgm-compro.com
-> tech. info
-> controller cooling (Acrobat pdf file)

Vriendelijke groeten Ron
Ron van Sommeren is offline Find More Posts by Ron van Sommeren
Reply With Quote
Reply


Thread Tools

Similar Threads
Category Thread Thread Starter Forum Replies Last Post
Question Does a prop saver rob power or increase load? rpavich Power Systems 4 Jul 15, 2008 11:20 AM
Parallel NIMH packs to increase current handling/runtime? BrianRickman Batteries and Chargers 12 Jul 18, 2003 09:47 PM
Does the RappE really need an 11 oz motor? HankF Electric Heli Talk 13 Apr 25, 2003 09:42 PM
Does it always take more power to hover at higher headspead? HankF Electric Heli Talk 15 Mar 25, 2003 04:57 PM
Does flying in winter increase flight times? Mike Wood Beginner Training Area (Aircraft-Electric) 37 Nov 25, 2001 07:37 AM