On a sunny day (Mon, 23 Jan 2023 11:38:09 +0000) it happened James <news@oxdrove.co.uk> wrote in <tqlrj2$3jo73$1@dont-email.me>:>[As I understand] There is no gate current when just on/off.Right.>The capacitance causes a current flow when going between on and off. The >charge is small but the period is short which leads to the values as >high as 10A being suggested. Gate drives are deigned to supply this >brief transient current. [There is a power requirement if cycling >rapidly at high frequency - I am not.] > > >> The drive circuit impedance then determines how fast his capacitor is loaded, so >> how fast the rise-time of the gate voltage and the controlled current as result of that. > >So a 1 Amp gate driver will not fuse if the capacitance asks for more >than 1A? It will just fill/dump more slowly and switch on/off more >slowly? ie, gate drivers are current limiting devices.There are no infinities in practice, so all drivers have a current limit.... The famous equation Q = C.U = i.t goes. So the charge time t to voltage U by a constant current i is t = (C.U) / i If the drive circuit is resistive and not a constant current than the normal exponential RC charge curve time goes. Note that a very fast drive pulse rise - or fall time may still exceed gate capabilities as then the gate current peaks to very high values. That is more a theoretical case, normally that would not happen. If the rise time of the gate voltage is slow then the MOSFET will dissipate more during the linear part of the drain current Ids versus gate voltage Vgs curve. Not what you want in a switching application. My idea, get some MOSFET and play with it,,,,,
P-channel MOSFET gate driver
Started by ●January 21, 2023
Reply by ●January 23, 20232023-01-23
Reply by ●January 23, 20232023-01-23
On Monday, January 23, 2023 at 3:38:17 AM UTC-8, James wrote:> On 23/01/2023 09:51, Jan Panteltje wrote: > > The gate electrode is separated from the main current channel by a thin insulating layer > > So normally no 'gate current' is needed, much like a grid in a electron tube. > > There is however, due to that construction, a CAPACITANCE, the gate and its > > insulating layer and main channel form a capacitor. > > This maybe be a very small value capacitor for for example small signal RF MOSFETs (a few pF) > > or a rather big one for power MOSFETs (several nF). > [As I understand] There is no gate current when just on/off. The > capacitance causes a current flow when going between on and off. The > charge is small but the period is short which leads to the values as > high as 10A being suggested. Gate drives are deigned to supply this > brief transient current. [There is a power requirement if cycling > rapidly at high frequency - I am not.] > > The drive circuit impedance then determines how fast his capacitor is loaded, so...As a practical matter, the driver either has intrinsic current limiting, like a transistor with emitter resistor, or one can add a simple series resistor to the gate.
Reply by ●January 23, 20232023-01-23
On Monday, January 23, 2023 at 4:07:42 PM UTC-5, whit3rd wrote:> On Monday, January 23, 2023 at 3:38:17 AM UTC-8, James wrote: > > On 23/01/2023 09:51, Jan Panteltje wrote: > > > The gate electrode is separated from the main current channel by a thin insulating layer > > > So normally no 'gate current' is needed, much like a grid in a electron tube. > > > There is however, due to that construction, a CAPACITANCE, the gate and its > > > insulating layer and main channel form a capacitor. > > > This maybe be a very small value capacitor for for example small signal RF MOSFETs (a few pF) > > > or a rather big one for power MOSFETs (several nF). > > [As I understand] There is no gate current when just on/off. The > > capacitance causes a current flow when going between on and off. The > > charge is small but the period is short which leads to the values as > > high as 10A being suggested. Gate drives are deigned to supply this > > brief transient current. [There is a power requirement if cycling > > rapidly at high frequency - I am not.] > > > The drive circuit impedance then determines how fast his capacitor is loaded, so... > > As a practical matter, the driver either has intrinsic current limiting, like a > transistor with emitter resistor, or one can add a simple series resistor to the gate.He doesn't have to "switch" anything. The FETs have a VDS on the order of a diode drop across them. All he has to do is enhance the channel with a 10mA grade gate drive, no need to shake the house down with a multi-ampere job.
Reply by ●January 25, 20232023-01-25
On 23/01/2023 12:26, Jan Panteltje wrote:>>> The drive circuit impedance then determines how fast his capacitor is loaded, so >>> how fast the rise-time of the gate voltage and the controlled current as result of that. >> >> So a 1 Amp gate driver will not fuse if the capacitance asks for more >> than 1A? It will just fill/dump more slowly and switch on/off more >> slowly? ie, gate drivers are current limiting devices. > > There are no infinities in practice, so all drivers have a current limit....How does the current limit manifest itself? If I plug a cooker into a 1A supply it is current limited, the fuse blows / trips and if it didn't fail it wouldn't bake a cake at 50 degC in 10 hours. The workings of the driver are opaque to me; I assume there is some cleverness to justify the their existence else we would just use a single transistor. Do I need a gate driver with enough "zap" to match the MOSFET gate charge? [Does the cleverness mean at high frequency the charge coming out is stored and fed back in on the next cycle?] Will the driver fuse? Will it run out of power halfway though switch on? Do I need to add a current limiting resister between the driver and the gate? [As shown on some application schematics but no actual values are given, how to calculate?]> If the rise time of the gate voltage is slow then the MOSFET will dissipate more during the linear > part of the drain current Ids versus gate voltage Vgs curve. > Not what you want in a switching application.This is probably only an issue at a high frequency. I think mine will be acceptable to switch in 60ns rather than 30ns.> My idea, get some MOSFET and play with it,,,,,Buy 10 at a time. Prepare for whiffs of smoke... I'm still not clear if the eg Microchip TC44xx series work with N or P channel. Thank you.
Reply by ●January 25, 20232023-01-25
On a sunny day (Wed, 25 Jan 2023 09:52:20 +0000) it happened James <news@oxdrove.co.uk> wrote in <tqqu4m$k564$1@dont-email.me>:>On 23/01/2023 12:26, Jan Panteltje wrote: > >>>> The drive circuit impedance then determines how fast his capacitor is loaded, so >>>> how fast the rise-time of the gate voltage and the controlled current as result of that. >>> >>> So a 1 Amp gate driver will not fuse if the capacitance asks for more >>> than 1A? It will just fill/dump more slowly and switch on/off more >>> slowly? ie, gate drivers are current limiting devices. >> >> There are no infinities in practice, so all drivers have a current limit.... > >How does the current limit manifest itself? If I plug a cooker into a >1A supply it is current limited, the fuse blows / trips and if it didn't >fail it wouldn't bake a cake at 50 degC in 10 hours.Yep, always something will give way, no infinities. mamaticians came up with infinities doing their divide by zero etc ..>The workings of the driver are opaque to me; I assume there is some >cleverness to justify the their existence else we would just use a >single transistor.No, you likely need a totempole: delivering power for voltage going high and voltage going low, to be able to charge / discharge the MOSFET gate capacitance fast enough.>Do I need a gate driver with enough "zap" to match the MOSFET gate >charge?Yes>[Does the cleverness mean at high frequency the charge coming >out is stored and fed back in on the next cycle?] >Will the driver fuse?Well even in a totempole its drive will limit current.>Will it run out of power halfway though switch on??>Do I need to add a current limiting resister between the driver and the >gate? [As shown on some application schematics but no actual values are >given, how to calculate?]Likely not, more often gate resistors are added to prevent oscillation in a linear MOSFET circuit at some very high frequency, but in a switcher these are likely not needed.>> If the rise time of the gate voltage is slow then the MOSFET will dissipate more during the linear >> part of the drain current Ids versus gate voltage Vgs curve. >> Not what you want in a switching application. > >This is probably only an issue at a high frequency. I think mine will >be acceptable to switch in 60ns rather than 30ns.That seems pretty fast to me... You can start with the switching time required, find the gate voltage swing needed for full conductance to no conductance look up the gate capacitance, then you can calculate how much current the driver needs to supply.>> My idea, get some MOSFET and play with it,,,,, > >Buy 10 at a time. Prepare for whiffs of smoke...Na, just measure one on the bench, Vgs versus Id at some Vdd voltage... etc>I'm still not clear if the eg Microchip TC44xx series work with N or P >channel.Never used that particular chip.... seems every semiconductor company makes driver chips... Maybe these days using just transistors is safer as no teling if a particular chip is still available in a few years. Same for transistors, but there is more choice there.. Strange no chips no cars, old diesel would just work and get you everywhere. Can people still find their way without GPS these days??? I only got lost once at night, a $1 compass helped me out... Too much tronix is no good ...
Reply by ●January 25, 20232023-01-25
On Wednesday, January 25, 2023 at 1:52:29 AM UTC-8, James wrote:> On 23/01/2023 12:26, Jan Panteltje wrote: > > >>> The drive circuit impedance then determines how fast his capacitor is loaded, so > >>> how fast the rise-time of the gate voltage and the controlled current as result of that. > >> > >> So a 1 Amp gate driver will not fuse if the capacitance asks for more > >> than 1A? It will just fill/dump more slowly and switch on/off more > >> slowly? ie, gate drivers are current limiting devices. > > > > There are no infinities in practice, so all drivers have a current limit...> How does the current limit manifest itself?Maybe by adding a resistor in series with the gate, maybe with an inductor to tame the surge (might be just a ferrite bead).> The workings of the driver are opaque to me; I assume there is some > cleverness to justify the their existence else we would just use a > single transistor.Don't assume that! Make the "driver" manufacturer show you a justification.> I'm still not clear if the eg Microchip TC44xx series work with N or P > channel.The marketers of that series reuse the same prefix for a variety of different units, some for single NMOS drive, some for PMOS, some for bridge drive... so the best answer is 'yes, no, and maybe'.
Reply by ●January 25, 20232023-01-25
On 25/01/2023 9:52 am, James wrote:> On 23/01/2023 12:26, Jan Panteltje wrote: > >>>> The drive circuit impedance then determines how fast his capacitor >>>> is loaded, so >>>> how fast the rise-time of the gate voltage and the controlled >>>> current as result of that. >>> >>> So a 1 Amp gate driver will not fuse if the capacitance asks for more >>> than 1A? It will just fill/dump more slowly and switch on/off more >>> slowly? ie, gate drivers are current limiting devices. >> >> There are no infinities in practice, so all drivers have a current >> limit.... > > How does the current limit manifest itself? If I plug a cooker into a > 1A supply it is current limited, the fuse blows / trips and if it didn't > fail it wouldn't bake a cake at 50 degC in 10 hours. > > The workings of the driver are opaque to me; I assume there is some > cleverness to justify the their existence else we would just use a > single transistor. > Do I need a gate driver with enough "zap" to match the MOSFET gate > charge? [Does the cleverness mean at high frequency the charge coming > out is stored and fed back in on the next cycle?] > Will the driver fuse? > Will it run out of power halfway though switch on? > Do I need to add a current limiting resister between the driver and the > gate? [As shown on some application schematics but no actual values are > given, how to calculate?] > >> If the rise time of the gate voltage is slow then the MOSFET will >> dissipate more during the linear >> part of the drain current Ids versus gate voltage Vgs curve. >> Not what you want in a switching application. > > This is probably only an issue at a high frequency. I think mine will > be acceptable to switch in 60ns rather than 30ns. > > >> My idea, get some MOSFET and play with it,,,,, > > Buy 10 at a time. Prepare for whiffs of smoke... > > I'm still not clear if the eg Microchip TC44xx series work with N or P > channel. > > Thank you. >The TC44xx series will work fine with N channel devices if the mosfet source is at or close to GND potential and will work fine with P channel devices if the mosfet source is at or close to VDD potential. But you said you were not switching at many kHZ - this is just a DC transfer switch? So assuming your load has some bypass capacitance to ride out a microsecond glitch then you really shouldn't need any kind of rapid gate driver - a simple DIY descrete circuit will be fine. You might be able to reduce your 4 mosfets to just 3 and still retain the isolation and swicthing you want. piglet
Reply by ●January 26, 20232023-01-26
On 25/01/2023 22:07, piglet wrote:> The TC44xx series will work fine with N channel devices if the mosfet > source is at or close to GND potential and will work fine with P channel > devices if the mosfet source is at or close to VDD potential.Perfect answer, thank you. Sorry for not understanding this but the specifications does not make it explicit although I could see no reason from the functional diagram why not (but I am not an electrical engineer, hence my need to ask the experts). Also eg www.mouser.co.uk filters them as "low side".> But you said you were not switching at many kHZ - this is just a DC > transfer switch? So assuming your load has some bypass capacitance to > ride out a microsecond glitch then you really shouldn't need any kind of > rapid gate driver - a simple DIY descrete circuit will be fine.Correct and understood. I need some isolation between the 3.3v logic to the power side and some current amplification device - connecting a ESP32 through a resistor (40mA max) or using 4N25 works for low powers but maybe not at when switching 10 Amps with high capacitance gates. I thought a gate driver was as easy as anything and at under £2 for dual device affordable.> You might be able to reduce your 4 mosfets to just 3 and still retain > the isolation and swicthing you want.I want to run either input on with the other off. The feeds in are off-the-shelf parts so I can't know if back EMF is a problem, caution says reverse protect both. Thank you. James.
Reply by ●January 26, 20232023-01-26
On 26/01/2023 11:33 am, James wrote:> On 25/01/2023 22:07, piglet wrote: > >> The TC44xx series will work fine with N channel devices if the mosfet >> source is at or close to GND potential and will work fine with P channel >> devices if the mosfet source is at or close to VDD potential. > > Perfect answer, thank you. > > Sorry for not understanding this but the specifications does not make it > explicit although I could see no reason from the functional diagram why > not (but I am not an electrical engineer, hence my need to ask the > experts). Also eg www.mouser.co.uk filters them as "low side". > > >> But you said you were not switching at many kHZ - this is just a DC >> transfer switch? So assuming your load has some bypass capacitance to >> ride out a microsecond glitch then you really shouldn't need any kind of >> rapid gate driver - a simple DIY descrete circuit will be fine. > > Correct and understood. I need some isolation between the 3.3v logic to > the power side and some current amplification device - connecting a > ESP32 through a resistor (40mA max) or using 4N25 works for low powers > but maybe not at when switching 10 Amps with high capacitance gates. I > thought a gate driver was as easy as anything and at under £2 for dual > device affordable. > > >> You might be able to reduce your 4 mosfets to just 3 and still retain >> the isolation and swicthing you want. > > I want to run either input on with the other off. The feeds in are > off-the-shelf parts so I can't know if back EMF is a problem, caution > says reverse protect both. > > > Thank you. > > > James.Is your 3.3V logic that controls the switching itself powered by the switch. Just curious if a startup issue could arise? piglet
Reply by ●January 26, 20232023-01-26
On 26/01/2023 14:14, piglet wrote:> Is your 3.3V logic that controls the switching itself powered by the > switch. Just curious if a startup issue could arise?One side is battery backed (and solar powered) and can be assumed always available. Even depleted it will be enough to power the controller. I do need to consider the default power path and maybe add some "passive" components to achieve this (not wait for a micro controller to boot/react). Now you have made me wonder if each MOSFET pair and driver needs to be powered from its own input side. The voltages on each side are within 0.1v of each other so gate should be close enough to source to turn off from either.
