Circuit for MosFets in parallel for extra current capacity.(Solved)

[spec]

I will attach a picture of a smaller rating Mosfet switch I have made to fit to another type of aircraft. This plane will be internal combustion engine powered, but will of course have radio control. The radio receivers and servos require 6 volts in this application and I have chosen to use an electronic switch instead of a common mechanical one for reliability reasons.
I will also fit this type of switch to my unpowered slope soaring gliders. Some of the higher performance slopers use high power demand servos, so a good reliable supply is essential.
This design is by Vollradth and was posted in the link above.

Interesting. That circuit agrees with my calculations which show that 25A per MOSFET is a practical limit. This means that your 100A version would need four parallel MOSFETs to get a good solid reliable design. But if you used some really beefy MOSFETs, the heatsinking requirements would be small. Time allowing, I hope to post a High Side Switch PMOSFET design using four £3UK PMOSFETs, which will hardly need any heatsinking.

[Jim-0000]

Hi Spec,

Thanks for providing the three draft circuits.
However, judging by the way the circuitry is arranged, I fear that I have still no described my application and where about in the model power system wiring, this planned switch will go.
The motor is 3 phase. The ESC is an electronic commutator. There is no direct connection from battery + or - to the motor.
The 3 connections from the ESC go to the motor.
The two wires from the battery go to the ESC.
The attached diagram might clarify this. The switch will go between the battery and the ESC. It will not go between the ESC and the motor.

Edit: The current from the battery to the ESC is DC. While the current from the ESC to the brushless motor is 3 phase AC! .

[Jim-0000]

This is an oscilloscope display of the waveform to a typical 3 phase brushless motor.
The first is at partial throttle.
The second at full throttle.
I hope this clears up some confusion.

Jim.

[spec]

Thanks Jim-0000 for the information in replies #26 and #27.

Yes, that does explain the situation well and means that the only option is high side switching, either with NMOSFETs or PMOSFETs. I would recommend a high side PMOSFET design with beefy PMOSFETs which would need minimum heat sinking: a number of suitable PMOSFETs are available and at a reasonable price too. What do you think?

The circuit that I have in mind would be configurable, where every PMOSFET handles 25A, so you just add PMOSFETs to suit the application: 1 x PMOSFET = 25A switching capability, 2x=50A, 3x=75A, 4x=100A, and so on.

If you are interested in the theory, it is the RDSS of a MOSFET, when it is turned on hard, that ultimately dictates the current handling capability of a MOSFET. The formula for the power generated in a MOSFET is ID² * RDSS. Where ID is the drain/source current and RDSS is the resistance between the drain and source (RDSS is often just labeled RDS on data sheets).

A beefy PMOSFET can have an RDSS as low as 0.01 Ohms (10 mili Ohms), so at 25A ID the power dissipation would be 25² *  0.01 = 6.25 Watts per PMOSFET. Multiply by four PMOSFETs gives a total power dissipation of 25W, which I suggest would be manageable in your model aircraft.

If you are wondering why four beefy PMOSFETs are required to provide a 100A switching capability, rather than just one beefy PMOSFET, here is the answer:

100² * 0.01 = 100 Watts.

[Jim-0000]

Thanks Jim-0000 for the information in replies #26 and #27.

Yes, that does explain the situation well and means that the only option is high side switching, either with NMOSFETs or PMOSFETs. I would recommend a high side PMOSFET design with beefy PMOSFETs which would need minimum heat sinking: a number of suitable PMOSFETs are available and at a reasonable price too. What do you think?...........

Spec,
I will have to do some reading up on the definitions of the terms your use above.

The circuit that I have in mind would be configurable, where every PMOSFET handles 25A, so you just add PMOSFETs to suit the application: 1 x PMOSFET = 25A switching capability, 2x=50A, 3x=75A, 4x=100A, and so on.

If you are interested in the theory, it is the RDSS of a MOSFET, when it is turned on hard, that ultimately dictates the current handling capability of a MOSFET. The formula for the power generated in a MOSFET is ID² * RDSS. Where ID is the drain/source current and RDSS is the resistance between the drain and source (RDSS is often just labeled RDS on data sheets).

A beefy PMOSFET can have an RDSS as low as 0.01 Ohms (10 mili Ohms), so at 25A ID the power dissipation would be 25² *  0.01 = 6.25 Watts per PMOSFET. Multiply by four PMOSFETs gives a total power dissipation of 25W, which I suggest would be manageable in your model aircraft.

If you are wondering why four beefy PMOSFETs are required to provide a 100A switching capability, rather than just one beefy PMOSFET, here is the answer:

100² * 0.01 = 100 Watts.

Yes, I understand that. Thanks. Leave it with me for a few days to do some catch up reading.

[David Hess]

I agree with all that has been said here. One thing I am slightly confused by is the need for a gate resistor, it has been said that a design seen in the wild did not use gate resistors and got away with it, but why would this be bad? If you just connect all gates in parallel and drive them with one hefty gate driver chip, I would have thought that all the gate capacitances would appear in parallel and be charged in parallel, leading to all MOSFETs switching on at the same time (subject to slight differences in their Vgs as discussed).

The gate resistor suppresses parasitic oscillation during switching and prevents gate drive overshoot.  With a low capacitive impedance at the source and drain, the input impedance is negative at high frequencies like a Colpitts or Clapp oscillator.  Slightly larger values of gate resistance may be used to match the impedance of the line between the gate and driver and suppress any series inductance lowering overshoot.

Gate stopper resistors for tubes and especially for tetrode tubes served the same purpose.  The same problem comes up with bipolar transistors driving a capacitive load when the base impedance is too low.

The whole subject was discussed here where I posted an old article with details and math.

This is one of those problems which might only be described as weird until you put a fast enough oscilloscope on it.

[spec]

Hi nick_d

I missed your question of post #15, which DH fielded so informatively.

Gate stoppers are often a source for discussion because, at first inspection, they appear to have no function, so below are a few points.

As well as helping to prevent parasitic oscillations, gate stoppers can have other functions:

• Along with the MOSFET's parasitic capacitances, suppress unwanted induced signals from transmitters: radio, TV, RADAR, WiFi, switches, etc
• Along with other components, shape the gate waveform to optimize high speed switching (not applicable in the OP's application)
• Protect the gate from ElectroMagnetic Pulse (EMP) (caused by lightning mainly)
• Where you have MOSFETs in parallel, individual gate stoppers can help prevent odd interactions between the MOSFETs
• In push pull applications, gate stoppers can be used, with other components, to delay turn on and thus provide some dead time so that both MOSFETs are not on at the same time (a fatal condition for the MOSFETs and possibly any transformer involved)

One thing to bear in mind is that MOSFETs are phenomenally fast but have huge parasitic capacitances in the nF region, especially power MOSFETs.  This is a formula for parasitic oscillations.

The other thing to bear in mind is that the nice schematic that you see on paper is nothing like the real life physical circuit: even a piece of wire has resistance, capacitance, and inductance, and acts as an antenna, that can both transmit and receive electromagnetic signals. And when you consider, transistors, transformers, inductors, capacitors, and resistors the situation is even more complex.

Finally, the physical layout of this high current switch is critical and, hopefully, will be described in later posts, but just to say that gate stopper resistors will only be effective if they are connected directly to the MOSFET gate terminals using leads as short as possible.

You mentioned that there are MOSFET circuits without gate stoppers that seem to work OK. This can be the case, but will the circuits work in all situations- probably not. The other point is that a circuit may be oscillating but the user may be quite unaware of it.

[nick_d]

Thanks guys, I did not know that. I will use gate stopper resistors in future.

The theory warrants a closer look than I can give right now. I understand negative resistance more or less, as I was at one stage planning to build an audio power amp that used op-amp circuits to generate say -3 ohms to drive a 4 ohm speaker. Apparently this gives a form of motion feedback to partially compensate box and driver response.

I don't understand Colpitts oscillators well or why they have negative resistance. I will revisit it some time.

cheers, Nick

[spec]

Thanks guys, I did not know that. I will use gate stopper resistors in future.

The theory warrants a closer look than I can give right now. I understand negative resistance more or less, as I was at one stage planning to build an audio power amp that used op-amp circuits to generate say -3 ohms to drive a 4 ohm speaker. Apparently this gives a form of motion feedback to partially compensate box and driver response.

I don't understand Colpitts oscillators well or why they have negative resistance. I will revisit it some time.

cheers, Nick

No probs from me

You need to adjust the gate stopper resistor value according to the MOSFET and the circuit function. With switch-mode power supplies, typically switching at a frequency of 50kHz to 4MHz, the gate stopper is used mainly for shaping the gate waveform, rather than preventing parasitic oscillations.

For low frequency switching, as in this application, switching times are relatively unimportant but, all the same, you don't want the MOSFETs to turn on/off too slowly or they may exceed their safe operating area (SOA). Besides which, it is just a waste of power. So a good value of gate stopper in these cases is 50R. 50R crops up a lot in electronics and 50R to 100R, with most resistors, is the magic value range where the reactances of the resistor's inductive and capacitive components tend to offset one another, so you get good frequency characteristics. When you do high speed designs with ECL or PECL you use 50R PCB lines with a ground plane and, of course, you get 50R coaxial cable for video and RF.

All oscillators rely on positive feedback to sustain oscillation. The Colpitts oscillator is just another method of providing positive feedback, as are the Phase Shift, Wayne Bridge, Hartley, Clapp, and so on. The Colpitts oscillator just happens to have good high frequency characteristics.

Negative resistance is no big deal.  With a normal resistance as you reduce the voltage the current goes down according to Ohm's law. With a negative resistance when the voltage is decreased the current goes up. And that is all there is to it.

And one final bit of cracker-barrel advice is to pay particular attention to decoupling. This vitally important area is quite often missed on the vast majority of circuits that you see on the net and in books, probably because decoupling components tend to clutter the schematic and make it harder to follow. 

[spec]

If all this stuff about gate stoppers, decoupling, routing etc sounds a bit fussy and over the top perhaps the following two stories will illustrate the consequences of not doing it right.

The first story concerns a friend's high-end MOSFET audio power amplifier. He said that it sounded very good but there was something about the sound that was not right somehow, and he couldn't quite put his finger on it. I had a listen and on his outfit it sounded superb... at first. But, after a while, I too detected an odd characteristic to the sound, so I took his amp home to investigate.

To short a long story, I found that the output power MOSFETs were bursting into oscillation at random points of the output voltage waveform. The amplitude of the oscillations varied and, for a while, they would be absent. There were no gate stoppers and relatively long traces between the MOSFETs and the driver circuit. So, obviously, I fitted gate stoppers. This greatly reduced the tendency for the MOSFETs to burst into oscillation and lowered the frequency too. But it was not a complete cure.

The MOSFETs were decoupled, but only with high-value aluminum electrolytic capacitors, so I added polypropylene decoupling capacitors. The result was that there was no sign of the parasitic oscillations and the owner was over the moon that his amp had regained it's original pure sound.

The other story concerns a maritime system built into a tall 19 inch cabinet. Right from the start this system was troublesome and gave spurious results. And after many fancy investigations with no conclusion, I was asked to arrange an investigation. As it happened, we had just taken on a new graduate, and he got the job. Once again, to short a long story, he found that far from being some esoteric technical issue that was causing the problem, it was the basics that were wrong.

The first thing he found was that the 74 series TTL chips on the cards that made up the equipment only had around 4V between their actual 0V pin and 5V supply pins (min allowable is 4V75), although the central 5V high-current power supply was pushing out 5V2. He also found that the analog and digital circuits had inadequate decoupling. Armed with these findings, he deigned a modification scheme to correct the issues, and when that was implemented- guess what?  The system was transformed and never caused a problem again. And even better, we never heard another squeak from the 'experts' who had been deriding our elementary approach.

[Niklas]

I used to have the same hobby, flying 10 cell (NiMH) F5F electric gliders, but have not been active for at least 10 years. Brushless motors were new back then and the LiPos were just introduced.
There might be an easier way to implement the safety feature, but it depends on the speed controller's safety. Usually there is a microcontroller that does the control signal processing and commutation calculations. Upon power on, the software waits for the control signal to be "motor off" before enabling the speed controller function. It is also common that it will use the motor windings to beep when this is done.
There is also another software function that performs a watchdog timeout function. The control signals from the receiver is sent out at approx 20-50 times per second, depending on the manufacturer. The signal is a 5V pulse, positive logic, with pulse width varying between 1.2 and 1.8 milliseconds, depending on the demanded speed or servo angle. If that signal is not received, or the pulsewidth is out of range, then the software should disable the speed controller.
A small switch, inserted on the control signal between the receiver and the speed controller, and a pulldown resistor (10k, signal-GND on the ESC side of the switch) could be an alternative. The control signal wire is either yellow, orange or white and the GND wire is usually black or brown. If you use an extension cable, then you don't need to modify the speed controller itself.

[spec]

I used to have the same hobby, flying 10 cell (NiMH) F5F electric gliders, but have not been active for at least 10 years. Brushless motors were new back then and the LiPos were just introduced.
There might be an easier way to implement the safety feature, but it depends on the speed controller's safety. Usually there is a microcontroller that does the control signal processing and commutation calculations. Upon power on, the software waits for the control signal to be "motor off" before enabling the speed controller function. It is also common that it will use the motor windings to beep when this is done.
There is also another software function that performs a watchdog timeout function. The control signals from the receiver is sent out at approx 20-50 times per second, depending on the manufacturer. The signal is a 5V pulse, positive logic, with pulse width varying between 1.2 and 1.8 milliseconds, depending on the demanded speed or servo angle. If that signal is not received, or the pulsewidth is out of range, then the software should disable the speed controller.
A small switch, inserted on the control signal between the receiver and the speed controller, and a pulldown resistor (10k, signal-GND on the ESC side of the switch) could be an alternative. The control signal wire is either yellow, orange or white and the GND wire is usually black or brown. If you use an extension cable, then you don't need to modify the speed controller itself.

Brilliant

[David Hess]

The first story concerns a friend's high-end MOSFET audio power amplifier. He said that it sounded very good but there was something about the sound that was not right somehow, and he couldn't quite put his finger on it. I had a listen and on his outfit it sounded superb... at first. But, after a while, I too detected an odd characteristic to the sound, so I took his amp home to investigate.

To short a long story, I found that the output power MOSFETs were bursting into oscillation at random points of the output voltage waveform. The amplitude of the oscillations varied and, for a while, they would be absent. There were no gate stoppers and relatively long traces between the MOSFETs and the driver circuit. So, obviously, I fitted gate stoppers. This greatly reduced the tendency for the MOSFETs to burst into oscillation and lowered the frequency too. But it was not a complete cure.

The MOSFETs were decoupled, but only with high-value aluminum electrolytic capacitors, so I added polypropylene decoupling capacitors. The result was that there was no sign of the parasitic oscillations and the owner was over the moon that his amp had regained it's original pure sound.

My stories involving MOSFET linear output stages without gate resistors usually involve the MOSFETs breaking into a roughly 10 to 100MHz howl and self destructing and the better the low loss high frequency decoupling is, the more quickly they self destruct.  Testing shows no problems until the output cable is attached.

For switching stages time spent in the linear region is short so oscillation is very rarely a problem but larger values may be used to control ringing in the gate drive signal.  But an ESC is usually driving a relatively long lead (transmission line) to a motor which can lead to problems.

The safe rule of thumb is to include a lossy element on at least one lead of any transistor.  Strong decoupling and a capacitive load are especially problematical.

[spec]

Ha Ha I love these stories. It is a consolation to know that others have problems too.

Before, I got the hang of decoupling, stopping, etc, I went though a phase where everything I touched oscillated, unless I tried to make an oscillator that is.

We had a name for intermittent circuits- 'Thursday Circuits/Equipments'. Thursdays were often the day for customer demonstrations/handover and, guess what, that would be the day that the gremlins would show themselves.

By the way, I started a new thread on this general topic: https://www.eevblog.com/forum/chat/thursday-circuits-gremlins-repairs/

[spec]

The safe rule of thumb is to include a lossy element on at least one lead of any transistor.  Strong decoupling and a capacitive load are especially problematical.

Good idea, presumably a lossy feritte bead would be one way.

Quite right about strong decoupling being critical, but you often need strong decoupling to get a good frequency response, supply line rejection, etc. Also poor decoupling can cause parasitic oscillations/ringing too.

[David Hess]

Before, I got the hang of decoupling, stopping, etc, I went though a phase where everything I touched oscillated, unless I tried to make an oscillator that is.

The first time I ran into this problem was when I was like 13 and had hooked up a 7805 regulator to my 723/2N3055 based bench supply.  How does a 5 volt regulator running on 8 volts put out 12 volts?  One 0.1 microfarad ceramic disc capacitor at the regulator's input solved it.

The safe rule of thumb is to include a lossy element on at least one lead of any transistor.  Strong decoupling and a capacitive load are especially problematical.

Good idea, presumably a lossy feritte bead would be one way.

There are a bunch of different ways to handle it as shown in the article attached below.

A ferrite bead is usually more expensive than a resistor and a resistor also serves as a fuse but I have used either way depending on the application.

[Jim-0000]

A small switch, inserted on the control signal between the receiver and the speed controller, and a pulldown resistor (10k, signal-GND on the ESC side of the switch) could be an alternative. The control signal wire is either yellow, orange or white and the GND wire is usually black or brown........

Excellent suggestion Niklas!  I know that if the good quality ESC's that I use do not receive a control signal, the motor will not operate.
I think this might be the solution...........and a very elegant one at that!

If you use an extension cable, then you don't need to modify the speed controller itself.

I will probably cut the servo lead signal wire and hard wire the switch in series with it including one end of the resistor and the other end to earth. I try to avoid servo lead plugs and sockets wherever possible. We know there are too many poorly crimped servo extension leads around on the market.
I have drawn a little sketch to confirm if I have interpreted your suggestion correctly or not. Please check it for me.
I will be trying this out on the bench before the day is finished. I will cut up an extension lead for test purposes first, before hard wiring the switch and resistor in the ESC receiver lead as described above.

Many thanks,

Jim.

[bitbanger]

Three phase BLDC ESCs can be quite complex (much more-so beyond just this parallel issue) even for the hobby realm. I'm never one to knock anyone's ambition but have you considered that you can buy these readily? However perhaps this is being integrated into an existing PCB? Either way, this doesn't address the parallel question directly but you may find the below reference designs helpful:

https://www.st.com/content/ccc/resource/technical/document/user_manual/group0/06/a3/c1/ae/7d/27/4c/e0/DM00384353/files/DM00384353.pdf/jcr:content/translations/en.DM00384353.pdf

https://www.infineon.com/dgdl/Infineon-Application-Motor_Control-Drone_Electronic_Speed_Controller_ESC-TR-v01_00-EN.pdf?fileId=5546d462580663ef015843a229fe54ea

[Jim-0000]

Three phase BLDC ESCs can be quite complex (much more-so beyond just this parallel issue) even for the hobby realm. I'm never one to knock anyone's ambition but have you considered that you can buy these readily? However perhaps this is being integrated into an existing PCB? Either way, this doesn't address the parallel question directly but you may find the below reference designs helpful:

https://www.st.com/content/ccc/resource/technical/document/user_manual/group0/06/a3/c1/ae/7d/27/4c/e0/DM00384353/files/DM00384353.pdf/jcr:content/translations/en.DM00384353.pdf

https://www.infineon.com/dgdl/Infineon-Application-Motor_Control-Drone_Electronic_Speed_Controller_ESC-TR-v01_00-EN.pdf?fileId=5546d462580663ef015843a229fe54ea

Thanks for the post Bitbanger.
However, we appear to be at crossed purposes. Either that, or you have replied to the wrong thread. Your post has me mystified; I apologise if I may have communicated the wrong message along the line.
Please explain, what is this "parallel issue"?
I already have ESC's (Electronic Speed Controllers plural!). I have no intention of building one any day soon, and most likely, never will. They are readily available, cheap and of good quality: as you suggest.
The aim of this project is to design and build a device that will safely switch the power supply (in this case, a Lithium Polymer 14.4 volt battery), to an already existing ESC.
Not to make one.

Jim.

[Jim-0000]

I...................................
A small switch, inserted on the control signal between the receiver and the speed controller, and a pulldown resistor (10k, signal-GND on the ESC side of the switch) could be an alternative. ........................

Niklas,
Success!  I just did a test on the bench with a cut and modified servo extension lead as per my little circuit diagram above. It works perfectly..........the motor will run with the switch on and not run with it off regardless of the throttle stick positon.
Very good, thank you for the creative suggestion.

Thanks to all contributors.

Jim.

[bitbanger]

Jim - appears my post was the result of insufficient reading/skimming, and the late hour! My apologies for the confusion, please disregard.

Cheers,
James

[Jim-0000]

Jim - appears my post was the result of insufficient reading/skimming, and the late hour! My apologies for the confusion, please disregard.
Cheers,
James

James,
Apology accepted, of course. All good in the end.
Thanks for explaining.

Jim.

[Jim-0000]

This is a short video of my vastly simplified safety switch in operation on the test bench. Circuit thanks to Niklas.