Skip navigation
Digi-Key TechXchange Communities > Projects and Designs > Discussions
52351 Views 10 Replies Latest reply: Apr 20, 2016 1:07 AM by bnebeker RSS
Currently Being Moderated

Nov 3, 2013 11:04 AM

How to balance parallel MOSFET's

I want to place several mosfets in parallel to handle more current, but am confused about the Gate resistors and something refered to as a "balance" resistor.  From what I've read, this balance resistor should be between 1.0 and 0.1 OHM and intended to protect the MOSFET's in the array, or "balance" the array for differences in the RDSON values that might cause one (or more) of the FET's to consume more of the load than the others leading to thermal catastrophic destruction of the entire array.


In addition, I'm not sure how to calculate the correct GATE RESISTANCE.  I'm driving each gate with op-amps, so I am not clear why I even need a gate ressitor.  I'm also not sure how to calculate the BALANCE resistor nor even where to connect this BALANCE resistor parallelly to each fet in the array.


If the RDSON value is, say, 0.036 ohms, then a 10,000 ohm resistor in parallel would maintain 0.036 ohms.   However, how the heck do you attach a 10,000ohm resistor in parallel with an RDSON condition?   In addition, this BALANCE resistor has always been mentioned in values of 1.0 to 0.1 ohms.


As always - THANK YOU EVERYONE fro your support.

  • donmiller Novice 3 posts since
    Oct 11, 2013
    Currently Being Moderated
    Nov 4, 2013 8:34 AM (in response to JMAR)
    How to balance parallel MOSFET's

    The "balence" or "BALLAST" resistor goes in series with the source.  If you place, say 50 milliohms or 100 milliohms in series with each FET in this way, variations (part to part) in the 36 milliohm Rds(on) will be a smaller percentage of the total series resistance per branch, and current hogging will be less of an issue.


    The gates of power Mosfets tend to have a LOT of capacitance, so you might add gate series resistance to keep your Op-Amp from going unstable due to the capacitive load.  Keep the series gate resistor as small as you can get away with, however, because the gate capacitance will form a lowpass filter with this resistor.

  • SteveSinn Novice 1 posts since
    Nov 14, 2013
    Currently Being Moderated
    Nov 14, 2013 2:40 PM (in response to JMAR)
    How to balance parallel MOSFET's

    You can also count on the fact that the FET "on" resistance (Rds) will increase as it's temperature rises so parallel FET will "self balance".  So as an example, if you have three FETs in parallel, the one that conducts the most current will also be the one that runs the hottest.  As a result, it's Rds(on) resistance will rise and the other two FETs will conduct more current and thus balance the current load.

  • jbuske Novice 3 posts since
    Nov 14, 2013
    Currently Being Moderated
    Nov 14, 2013 3:35 PM (in response to JMAR)
    How to balance parallel MOSFET's

    Hi Don,

    Mos fets have a positve temperature coeff so they tend to share current.  However gate threashold varies from device to device.  If you sort parts by threashold V with perhaps 25mv bins that will help a great deal.


    • laserdude Novice 2 posts since
      Nov 14, 2013
      Currently Being Moderated
      Nov 14, 2013 5:31 PM (in response to jbuske)
      How to balance parallel MOSFET's

      ...or just make sure that they're fully on if they're just switching. (check the datasheet for Vgs ON)

      • jbuske Novice 3 posts since
        Nov 14, 2013
        Currently Being Moderated
        Nov 14, 2013 8:02 PM (in response to laserdude)
        How to balance parallel MOSFET's

        Hi Laser Dude,


        Depends on your application.  Doing a RF push pull driver into a magnetic? Uneaven switching (treashold voltage/Rdson) will put a DC bias in the core and worse case the core saturates and fets go up in smoke. 


        Data sheets are fine part specs vary device to device.   Doing Rdson measuremts of sub 5 ohm parts should use 4 wire ohm meter  and may need to pulse current 1-20 amps to get a good reading. 


        Threashold voltage measurement use a DAC to drive the gate to a test current few ma or where ever you like.   Then can match up parts with similar swiching performance. 


        Would take extra hardware but, you could measure average drain current and change each gate drive pulse width to servo the currents to be the abouth the same?  New CPUs have extra pins perhaps set one PWM/Timer for 50% the other PWM to 45% to better share the current for different Rdson or PCB layout resistance etc...?  Also threashold voltage changes with temperature I don't know how well the it tracks part to part.

  • bnebeker Novice 3 posts since
    Nov 17, 2013
    Currently Being Moderated
    Nov 17, 2013 3:37 PM (in response to JMAR)
    How to balance parallel MOSFET's

    Controlling large currents with MOSFETs is something I took a great interest in many years ago when power MOSFETs  first became available. They seemed from the data sheet to be so  powerful but as I observed in using them, they are very easy to kill.  This prompted me to learn as much about them as possible. The body of  knowledge that is available has increased tremendously over the years  and understand how to make your MOSFETs perform reliably takes some time  to understand. So I am presenting some of the important issues you need  to consider in properly driving a MOSFET.


    The great thing about MOSFETS is that the due share the load very well without the need to any balast  resistor and if you need to parallel multiple devices you certainly  don't what to add resistance into the path. This however must be  qualified with using them as switches ie fully turned on or fully turned  off. Driving the gate of the devices however does present some issues  that can best be resolved by using a gate driver circuit. One device may  turn on as 4V and another of the same part number may turn on at 4.5V  however this does not mean that either of them is fully turned on.  MOSFETs  are often used are switches but they also have a linear region. At the  minimum turn-on voltage the transistor starts to conduct as a low level  and the level increases at the gate voltage increases, at some point the  device will reach nearly complete turn-on. At this point the resistance  will be very close to it's minimum value and may be low enough for your  purpose if you are not switching large currents. By raising the gate  voltage a bit more you get to the point where the resistance is at it's  absolute minimum also know as it's saturation point. Also note that the  required gate voltage will vary with the temperature of the device and  the current flowing through it. For this reason when switching large  currents you should drive the gate beyond where it seems to fully turn  on. Due to the very low resistance of these devices they are capable of  handling large currents, low current testing won't reveal much  information about how they will perform under large loads. Applying 12V  to a gate of most devices will ensure complete turn on, even with variances in the devices.


    There are a several good reasons for using a gate resistor when dealing with large currents.

    1. It is often nessacary to control the turn-on and turn-off rate  to prevent large current spikes in the system. While this results in  more time in the linear region thus generating more heat it can however  save components from destruction due to large voltage spike that happen  when you try to change current going though an inductive device too  quickly.

    2. Due to the large capacitance of high current devices between  the gate and the drain current can flow from the drain through the gate  capacitance and into your driver circuit, and kill low power driving  circuits. This can be resolved with slowing down the rate of change,  slower rise and fall times, and slowing down how often you switch. Put a  gate resistor in place will slow down the rise and fall times and will  also limit the current that can flow through the gate capacitance.

    3. When dealing with large currents you in a system the fact that  conductors have some resistance can cause some problems that you may  not have accounted for. On of the biggest issues is that the source of  the transistor may be at a different potential during portions of the  switch cycle than you would expect from the circuit diagram. This can  result in the transistor turning on at high resistance when it should be  turned off. To prevent this you need to ensure that the driver circuit  pulls the gate voltage down to the source potential strongly enough to  keep it from turning back on. The best way to ensure this is using a  drive circuit than is capable of 1 or more amperes sources or sinking  current. The down side is the current that can flow into the drive may  be exceeded due to the gate or drain capacitance. However with a minimal  value gate resistor you can ensure that the gate can still be switch  rapidly enough for your needs but also keeping the maximum current low  enough to prevent destroying the driver circuit.

    4. When a device that is handing large power levels fails, the  reaction can be very violent. It is very possible that they device could  short between the drain and the gate and cause a large current to flow  into your driver circuit. Limiting the current via the gate resistor  could keep the damage to fewer components than if the current is not  controlled at all. This is called fault tolerant design and is very  important when dealing with high voltages or high currents.


    From the information above it would seem that simply using a  single resistor between the gates and the driver should work, and in  some cases it does however there is a good reason for separate  resistors. Using the shortest possible path between the gate resistor  and the gate of the device will result in less stray capacitance from  surrounding circuits. With a separate resistor for each gate the individual resistance can be higher thus providing more protection for each transistor from issues mention above. You may also consider using a zener diode from the gate to the source of the transistor. This will protect the transistor from the easiest way of killing a MOSFET  which is to the exceed the gate breakdown voltage. When switching large  currents ground bounce can cause the potential at the source of the  transistor to not match the ground voltage of the driver circuit. This  could result in a much larger voltage from gate to source than intended.  The zener diode protects the gate from breakdown and the gate resistor protects the zener from conducting excessive current.


    As jbuske mention driving a inductive circuit using a push pull configuration could have some addition problems due to vaiants in turn on voltage. One way of deal with this is to purposely add a capacitor to the gate that is much higher than the internal capacitance  and using a small value gate resistor. This ensures the different in  gate capacitance between the devices is smaller but also makes it more  difficult to switch the devices fast enough. Another method of dealing  with this is to insert a resistor into the source circuit which has two  effects. One effect is it forces some balance in current sharing but  more importantly if reduces the voltage between the source and gate of  the device that turns on earlier thus lowering the drive which will  balance the gate drive.


    As a last note don't trust the first page of most data sheets as they tend to show the good and not the bad numbers. Many MOSFETS  show that they are capable on very large currents in 100's of amperes.  However the pins on the devices are typically limited to 75 amperes. I  usually limit continuous current through a single devices to 20-30  amperes as a rule of thumb. One of the reasons for this is heat. The  great resistance value quoted is usually as room temperature but as the device heats up the resistance increases. One of the reasons I love power MOSFETs is that they can be so efficient and can be packed densely in a circuit if you can keep the heat down. By paralleling transistors you can control large currents with little heating of each transistor and using the nature heat dissipation of each package to avoid the need to heatsinks  in many cases. So using two transistors to control 20 amperes may not  make much sense when each transistor can easily handle 50 amperes but it  could keep power losses low enough to simply use the PC board as a heat sink.  Note that the faster you turn the transistors on and off the more time  is spend in the linear region and harder it is turn fully drive each  transistor to full on. To make things worse when you drive a transistor to it's saturation point it takes more time to turn if off.


    As an example, I have a circuit that I built to  prevent excess current flow in case of a short circuit for trailer  lights. The normal current was less than 10 amperes but the allowed  current was 20 amperes, which allows for the surge currrent that light  bulbs experience. I designed the circuit to be used with higher current  levels up 50 amperes. It also was designed to handle current flowing  into and out of a battery. They difference between the circuit when  using it at lower current levels was just how many transistors I  populated the board with. When using it at a higher current level I used  8 mounted back to back 4 across while only using 2 for lower current  levels. The driver circuit for the transistors was a simple 1K resistor  pulled down by a open collector comparator. Because this circuit is  either on all the time or turns off for several seconds when the current  is exceeded a very simple drive circuit was possible. I did use a  single 100 ohm gate resistor in this case to ensure that a short in the  transistors would not fry the comparator. In my testing using 8  transistors and no heat sinking I was able to handle 30 amperes of current flow with barely enough temperature rise to notice using my finger.

    • Duricam Novice 1 posts since
      Apr 18, 2016
      Currently Being Moderated
      Apr 19, 2016 12:40 PM (in response to bnebeker)
      How to balance parallel MOSFET's

      I have got a set of solar panels with Voc arround 200V. I plan to use the generated electricity to heat water in the boiler.

      The boiler heating spiral resistance is 24 ohm (cold).


      As the boiler is designed to get 240V AC and there is a bimetal Thermal switch - thermostat which would soon die operating on DC, I wanted to substitute it's function using MOSFET. the closest MOSFET I found that would fit the parameters was IRFP264, with Vds max 250V. The idea was to use the boiler's thermostat just with a small current to control MOSFETs.


      The problem is that with Rdson 0.075 it dissipates about 3.7W at 7 Amperes flowing into the heating spiral. So I decided to use 2x IRFP264 in parallel, as I prefere to heat up water instead of the MOSFET itself.

      As this seemed to be a pretty simple DC circuit I had not considered to add any compensation stuff that would reflect that there are 2 devices in parallel. the circuit seemed to work OK on my test board - in the late winter when there was not much sun during the day. (the thermal switch in the boiler never had to switch off, as the max water temperature had not been reached)

      Suddenly when there was more sunshine during the day I found the plastic box with the circuit half melted.fortunately the MOSFETs seemed to be OK.

      I supposed that when the boiler thermal switch switched off, the Gate voltage went down very slowly due to the Gate capacitance, the MOSFETS got half open for some time and they shared the power with the boiler. so I added a Gate-Source discharging resistor of 1M to prevent it. (before there was only a zener to limit the Gate voltage) - schematic diagram attached.

      Unfortunately I found the box in a bad state again, moreover, one MOSFET of the couple died.

      (the resistance between G an S is about 17 ohms, both polarities).

      Then I did a more proper PCB and used M1 as a G-S resistance.

      Test switch is there to be able to connect/disconnect the circuit without having fireworks. LED glows when the water temperature is lower then pre-set on the thermostat.

      the circuit got crazy immediately after connecting it and one MOSFET died again. (26ohm G-S both ways, about 125 ohm D-S both ways - cca 3 ohms difference for the other polarity)


      As it would be quite difficult to learn "the hard way" I would really appreciate if someone could help.


      I suppose that there must be a kind of situation upon the switch on/off that causes the circuit to oscillate and kill MOSFET by exceeding Vds max or even Vgs max.

      would, for example simple, adding dedicated gate resistors to each MOSFET help to keep the circuit stable?

      many thanks!!!


      • bnebeker Novice 3 posts since
        Nov 17, 2013
        Currently Being Moderated
        Apr 20, 2016 1:07 AM (in response to Duricam)
        How to make your circuit reliable.

        You have four issues to address with this circuit to make it reliable.


        1. Your FET's are driven in their linear region for a considerable amount of time because you have to charged the gate capacitance through the a resistor which takes time. More current less time but more wasted current. Using a gate driver chip would solve this problem. I would recommend a small value 150 OHM resistor to the gate because it will slow the switching down a little and it will protected the gate driver from damage if a FET shorts drain to gate. This does mean having to add a regulator to provide 12V but is not as difficult as it might seem. While the driver may indicate it can source or sink 1 ampere of more it will only be drawing hard for a very short time, using a 100uF capacitor to provide then extra current and you just need a regulator than supply 100mA or so. If you can tap you PV at a lower voltage then your choice of regulator is wide open if you have to using the main supply then a regulator using a discrete transistor and zener diode would be a good choice. Look around on the Internet you should find a example circuit to use.


        2. Thermostat chatter. As the water is heating up it reaches a point where the thermostat turns off followed by the heating element turning off. This can result in the thermostat turning off and back on again. This is not good for the transistors because it cycles them through their linear region over and over again. You need a circuit that would delaying turning off the heating element by several seconds (5 - 10 Seconds). This will allow the water to heat up a little more so the thermostat does not turn back on immediately. A LM393 comparator circuit works well for this type of application.


        3. The heating element by nature can be a fairly inductive load. This could cause a voltage higher than 250v to appear across the transistors which switching off. A snubber circuit which consists of a capacitor and a resistor is your best bet, but for this circuit a capacitor by itself would probably work just as well. I would use a 0.22uF to 0.47uF 500 V foil capacitor placed across the heating element. If will be a little bulky but it should give you long life.


        4. By connecting the themostat to the high voltage you have another high voltage wire and multiple connection points to deal with. Your circuit would be much safer if you have only 1 high voltage node (the transistor drains) in your circuit. This node should be clearly marked on the board or with stickers to warn anyone would might come in contact with the board. Make sure then connectors you use are rated for least 500 v. Often they are rated for AC voltages and when used with DC a higher break down voltage is needed due to the fact the DC arcs are continues once started, where AC arcs tend to stop themselves. For this same reason your switch needs to be rated for DC voltage it will be exposed to, which is often difficult to come by when you need high voltages.


        I have a circuit than I use to power a 13A heating element and a 1A fan that is controlled by a CPU. I used a small PC board mounted transformer and optically isolation drivers to handle 120VAC with just very small heatsinks on the FETs. Because I am controlling AC I have to use 2 FETS back to back. In my design I allowed for the use of 4 FETs to share the load for the heating element but it turns out I only needed two. Properly driving your FETs is the key to efficiency and long life.


        You may have not shown it for simplicity, but a fuse is an essential component for this type of circuit I would use a 10 A ceramic fuse with a fuse holder designed for at least 20 amperes. A thermal cut-out between the high voltage and the heating element would be well advised because it could save your boiler if the transistors shorted drain to source. Even the best designed circuit fail so you should take measures to prevent the worst from happening.


        Keep the trace between the gates and the gate resistor as short as possible. If the gates are far apart use two separate resistors. I prefer TO-247 package FETs for high voltage use because they are easier to mount and the pins give a little better air gap.

  • jbuske Novice 3 posts since
    Nov 14, 2013
    Currently Being Moderated
    Nov 17, 2013 10:09 PM (in response to JMAR)
    How to balance parallel MOSFET's

    Laser Dude,


    Don't know if you are driving lasers for range finding i.e. Very high currents for nano seconds. Operating many devices in || reduces each average current and increases switching times as the miller feed back is deduced.


    Parts like the IXYS IXRFD631 are designed for vey high speed switching of 30 amps with rise and fall times of less than 5ns and on time down to 8ns.   Directed energy, EO-Devices and others like Dr. Heller Elektronic KG offer such devices and or drivers.


    Or a avalanche transistor like Zetex ZTX413 can be used generate short high current pulses. 


    With any high speed high current SHORT leads and traces are a must.  


    You want to look for devices with the lowest gate capacitance and charge required to switch.  Super low Rdson devices are generally slower may not be what you want.  As the faster device may run cooler as it speeds less time in the linear region. 


    Low cost devices like 2n6660 can switch in less than 7ns may be an option for your application.  


    Regarding gate resistance 1-50 ohms slows the switching time and can keep the device from oscillating.  PCB layout and transmission lines must be used for the high speeds and currents.


    If you don't blow up a few devices you are not having fun.  Mfg speed a great deal to seal the part to keep the smoke in.


    • davecuthbert Journeyman 65 posts since
      Feb 18, 2014
      Currently Being Moderated
      Feb 28, 2014 12:41 PM (in response to jbuske)
      Re: How to balance parallel MOSFET's

      A "secret" to successful paralleling of MOSFETs in linear mode it to keep things symmetrical, otherwise some of the FETs may oscillate. Since you're driving many MOSFETs from an op amp the turn-on time even in switch mode (FETs on hard) is long enough for one or more to build up a destructive oscillation. So, don't daisy chain the gate drive from FET to FET, run the gate drive traces from the op amp in the center of the MOSFET array and place a 100 ohm series resistor right at each MOSFET gate. Reducing the driver (op amp in this case) output impedance can help. A driver such as the LT1010 can help. I learned all of this the hard way in more than one design.


      In switch mode, series source resistors are often not needed for balancing. In linear mode they can be important as the often cited positive temperature coefficient function may not be strong enough to ensure adequate balancing. In this case the source resistor value should be chosen to produce at least a 1 volt drop at maximum operating current. The required voltage drop depends on the threshold voltage range for the FETs, the transconductance range, the heatsinking thermal impedance, and the maximum die temperature the hottest FET can have. So, there's not a one-size-fits-all answer. Iterative work on paper will yield a good range of resistor values.

More Like This

  • Retrieving data ...

Bookmarked By (0)


  • Correct Answers - 4 points
  • Helpful Answers - 3 points