MOSFET Gate driver problem

Thread Starter

Schniz2

Joined Apr 15, 2010
22
I have set up an IR2011 High and low side gate driver to drive a single n-channel MOSFET on the high side.
When i apply power to the circuit, the gate stays at ~15V (the voltage i am supplying the gate driver with) and so the MOSFET never turns off. The maximum forward voltage of the diode is 1.27V, everything else is marked on the circuit diagram.
The PWM i am supplying is at about 5V and a duty cycle of 0.125.

Not sure what i'm doing wrong.
 

Attachments

ashokcp

Joined Mar 8, 2007
50
What is the load? Is there an LC filter at the output?
There is normally a LC filter with a freewheeling diode at the output of the buck regulator, which is not there or shown in the circuit.
At power on, the bootstrap cap is not normally charged, and is expected to be charged thro' the LC filter - ouput cap is expected to be atleast 10times that of Cbootstrap, so Cboost will get most of the voltage. There are other conditions too, when the bootstrap cap cannot be charged. Please refer to AN-978 from IR.

Are you sure that the MOSFET is always On, the voltage on the gate is w.r.t. source and not w.r.t. the ground? My feeling is the MOSFET is not ON at all!
 

Thread Starter

Schniz2

Joined Apr 15, 2010
22
Here is the full circuit... i think the gate driver is working ok now (i found a track that wasn't cut properly on my prototyping board ugh!!)

Is there a way to calculate what current the gate driver will be pulling from the supply? i overheated my voltage regulator even with a (smallish) heatsink on it... i measured the gate driver to be pulling 200mA at around 8V (its supposed to be 15V but i think the regulator is heat damaged). The switching frequency is 20kHz and the output current of the gate driver is 1A for <10ns.

I thought 200mA was a bit high personally... is there a way to roughly calculate what it should be pulling by using the gate charge of the MOSFET and the switching frequency?
 

Attachments

ashokcp

Joined Mar 8, 2007
50
Nice to hear it works. Leaving complexities of gate charge at different voltages for simpler calculation :
The gate charge has to be replenished every cycle. So multiplying this by switching freq. should give you total coulombs reqd per sec (52nC * 20kHz), which is the current. Add any quiscent current reqmt should give the current required. Regulator output divider and quiscent currents - Well, everything together also cant be more than 10 to 20mA. Hope this helps.
 

Thread Starter

Schniz2

Joined Apr 15, 2010
22
i was wrong, its not working properly... but a little better...

I have hooked up the gate driver to a separate DC supply because ive overheated two 300mA regulators somehow.

The gate driver supply is at about 17V and i only measured up to about 2mA steady state current which seems more realistic.

I recorded the following for a duty cycle of D=0.76 and a buck converter load of 60Ohms... all power paths and devices are rated at about 5A continuous or greater.

Buck converter: Vin = 2.3V; Vout = 2V; (Vout/Vin)=0.87;
MOSFET gate switching between 0V and 16.6V

Buck converter: Vin = 10.94V; Vout = 8.7; (Vout/Vin)=0.79;
MOSFET gate switching between 0V and 16.6V

Buck converter: Vin = 13.45V; Vout = 8.8; (Vout/Vin)=0.65;
MOSFET gate switching between 0V and 16.6V

**Note: all these voltage are w.r.t. ground.
**Note2: When i increase Vin beyond 13V, the power supply 'clicks' however at that stage it is only supplying 0.87A at 13V, which is a lot less than it's ratings (its a Topward 6303A)

... as i understand, shouldn't the gate driver be at a voltage of (Vcc + Vs) for the "on" period of the MOSFET? Where Vcc is the supply to the gate driver and Vs is the source voltage?
not sure where i am going wrong :S
 
Last edited:

ashokcp

Joined Mar 8, 2007
50
the gate driver be at a voltage of (Vcc + Vs) - YES.

When the MOSFET is turned off, Vs goes down to one diode drop below Gnd! That is the freewheeling diode's drop as it is conducting. So, the zero or slightly less if you can see closely.
But when the MOSFET is on, almost Vin appears on Vs, after an Rdson drop, and, the gate should be Vcc+Vs ~ Vcc+Vin, but it appears to be steady at 16.6. When Vin goes above 13, Vgs is too low to turn it on.

I am guessing the gate driver could be faulty.
 

Thread Starter

Schniz2

Joined Apr 15, 2010
22
I have replaced the gate driver now, and its working better but still not how i would expect.
Attached are oscilloscope traces for a Duty Cycle of 0.75, Load Resistance of 50 Ohms. All measurements are wrt ground
The Blue trace is the output voltage at the load in all images.
Vd is the input of the buck converter
Vgd is the input of the gate driver.

000- Vd=15, Vgd=15 - yellow trace is gate voltage
001- Vd=15, Vgd=15 - yellow trace is input PWM to gate driver
002- Vd=15, Vgd=15 - Yellow trace is source voltage
003- Vd=15, Vgd=18 - Yellow source voltage
004- Vd=18, Vgd=15 - yellow source voltage
005- Vd=30, Vgd=15 - yellow source voltage

Firstly, the traces are horrible, i was expecting some noise but i dont think this much is normal (everything is connected as close as physically possible on a prototyping strip board).

Shouldn't the source voltage (002) be roughly a scaled version of the input PWM (001)?.. instead it switches between 15 and 4V during the 'on' time when i would expect it to be constant 15V.

As i increase the gate driver supply by 3V, the source voltage amplitude increases by about 3V and the output voltage increases.... does this mean that because of all the switching during the "on" time, the gate voltage is not high enough for the switch to be fully on?

As i increase Vd in 004 and 005, it seems the "on" time switching gets worse, in that it spikes down even closer to zero.

If anyone has ideas on what might be wrong or suggested things to test that would be great.
 

Attachments

ashokcp

Joined Mar 8, 2007
50
Is there noise in LM317 Ouptut (Vcc supply)?

Try adding 0.1uF ceramic caps at the input & output of LM317, and, one more at the power supply to IR2011 (close to IR2011), if it is far.

A small LC filter of 100uH-1mH and 10uF cap at the input.

A snubber will also be in order. An RC of 22 Ohm and 4700pF in series across the MOSFET as a starting point or as per
http://www.national.com/en/power/snubber_circuit_design.html.

Good Luck.
 

Thread Starter

Schniz2

Joined Apr 15, 2010
22
Ahh, it is much better with the snubber on it.. i also discovered my inductor core was being saturated, so i got a different one which seems to be working fine.

The voltage ripple is a fair bit higher than i expected according to my design equations and simulations which include the ESR of the capacitor... i also added a low ESR plastic capacitor to reduce the ESR of the electrolytic but the ripple didnt improve. i was expecting some increase due to parasitic elements but i am getting 250mV output ripple for an output of 36.5V @ 0.73A... maybe it would improve if it was on a PCB.

Blue trace is output voltage, yellow trace is voltage across snubber resistor.

There is also still quite a large spike at turn-on, would increasing the gate resistance improve this by turning the gate on slower? i also read you can include a bootstrap resistor... is this equivalent to having a gate resistor or does it have some other benefit?
 

Attachments

ashokcp

Joined Mar 8, 2007
50
How nice to hear!
Inductor core saturation can be checked with the following formula :
B = LI / NA,
where I is the peak inductor current,
say Iomax + 1/2(dI), dI is the pk-pk inductor ripple current and usually 1.1 to 1.2 * Iomax.

A is the Ae of the core material.

At this condition, B should be below the Bmax of the core.

Ripple voltage is normally due to Inductor ripple current * ESR of capacitor. Some tips of low esr caps - higher voltage rated caps have lower esr (of the same family), paralleling caps to reduce the esr. Proper PCB layout can reduce parasitics as you have rightly pointed out.

The spikes! I havent got good results with increasing gate resistor as the MOSFET turn on is steeper, and, with a higher gate resistance the turn on was getting delayed a bit. It appears to me that adding a resistor in series with bootstrap might delay turn on, but faster turnoff, as against delayed turn on/off with a gate resistor.
The snubber can be tweaked now. Which snubber are you using? Snubber resistance may be reduced to check the spike.
Good Luck.
 

Thread Starter

Schniz2

Joined Apr 15, 2010
22
I tried decreasing the snubber resistor from 22 to 12Ohms but it didn't change the spikes at all.

I measured the ringing frequency with snubber disconnected to be about 19Mhz, and with the snubber connected it is still almost 19Mhz... i am currently using a 6800pF capacitor, and have tried 22000pF and the ringing frequency does not decrease
 
Last edited:

ashokcp

Joined Mar 8, 2007
50
That is intriguing. Just to be sure, the snubber resistor should be non-inductive type, carbon film resistors are better suited, the cap should be of excellent high frequency, and, not to mention that they be very close to the switch and ground.
Also, the ripple current rating of the input cap is preferably atleast 50% of the output current, and output cap atleast double of max inductor ripple current.
 

Thread Starter

Schniz2

Joined Apr 15, 2010
22
**Edit, i "fixed" the problem... i somehow managed to kill another gate driver... not sure if its something i am doing wrong or if i shorted something with a probe :S

I ended up ignoring the spike and am hoping it will go away when i build it on a PCB.

Something else happened today, it was working fine and then i started to get heaps of switching on my gate again (even with the snubber in), which promptly overheated my gate driver regulator.

I haven't changed the circuit at all so im not sure what could have caused that to happen... the only thing i was doing at the time was putting a capacitor across the input of my ADC input on my microcontroller to try and reduce some of the measurement noise... which i have had connected before with no problems.

The first thing that came to mind was that i had ruined a capacitor, because whilst tuning my PI controller i had some periods of really high ripple currents which reached almost the max output current (3.1A). Now the controller is roughy tuned i am only getting about 60mA ripple current.

I pulled out my snubber cap and input cap and their capacitance was unchanged (is this sufficient to test if they have been damaged?). My snubber resistor also tested ok.

Im not sure what else could have failed for the gate ringing to be such a problem again :(... its due in 2 weeks so im starting to freak out a bit :p
 
Last edited:
Top