Driver design for PWM control with MOSFET

[OzFugah]

Dear Friends,

Along with greeting, I am designing and manufacturing a small PWM actuator based on an MCT2E optocoupler and IRF640 MOSFET. Initially the design was like this (generic image):

The load is a 100 VDC motor and R2 originally was 330 ohm. The system controlled the speed consuming only 500mA max, but the MOSFET was heated intensely, burning in a minute. After several research (https://www.electro-tech-online.com/threads/driving-a-mosfet-irf640-correctly.109914/) I discovered that the MOS (irf640) was being "saturated" by a very low current in its Gate, which caused that on each rise or fall, the system operated in a linear zone, heating up to die due to the circulation of those 0.5A in the load (the MOS allows more than 10A).
The solution that I found after reading a bit, is to use a pre-amplifier stage (Totem pole), because the MOSFET to reach a fast saturation, requires aprox. 400mA in its Gate. In this way I have tried to mount a couple of improved versions, taking this design as a reference (https://electronicsforu.com/electronics-projects/solid-state-relay)

The problem is now R2. Using 10K on pin 5 of the opto does not switch to 12V (I dont know why), so the base of T1 is never saturated. Now if I decrease R2 to 1K for example, effectively the switching of T1 does occurs. If transistor is short the voltage in R4 is 12v, but when it saturates, the voltage in R4 is not zero, it is rather 4V. That means that after going through the Totem Pole, the signal in the MOS Gate is mounted on a constant of 4 or 5 V. Obviously that causes the MOS not to commute accordingly.
Someone here will have some experience with this type of configuration?
Any type of contribution is appreciated.
Best regards and thank you very much.
OZ

 

[ChrisP58]

While the 2nd circuit is faster than the first, it's still designed as a fast static switch. Meaning that it's fast enough for occasional turn on and turn off, but may not fast enough for repetitive operation, such as a PWM controller. That will depend on your PWM frequency.

I would suggest just using a dedicated mosfet driver. Yes, you can do your own, but why?

https://www.digikey.com/product-det...orage/TLP155E-TPLE/TLP155E-TPLE-CT-ND/3056622

 

[OzFugah]

Dear, thanks for the quick response. The switching frequency is 10KHz. The integrated that you mentioned is not available in my city (small town in Chile), however TC4452 and MC33153 are available. I can notice that they do not have an opto-isolated stage, but do you think that they can do the job in the same way? Best regards.

 

[rjenkinsgb]

You are trying to short out the first transistor base with the opto-isolator.
That's not a good design as it means the switching point is around 0.6V

Normally, the opto transistor should be between the base and +12V, with a series resistor to set the "on" current to the base, plus a base-emitter resistor on the transistor to ensure turn-off.

Unfortunately the MCT2E opto is a very "slow" device, according to the datasheet it can take up to 50uS to turn off. That's an entire half cycle at 10KHz.
You need an opto isolator that takes eg. no more than 1% of your cycle time, for optimum efficiency.
See page 4 of the datasheet here: **broken link removed**

Adding a resistor between the base and emitter connections of the opto transistor will speed it up a bit but also decrease it's sensitivity.


With bipolar devices you can speed up the turn-off time somewhat by adding a schottky diode between collector and base, with anode to base & cathode to collector.
That prevents full saturation as the diode takes the base drive away when the collector gets down to 0.2 - 0.3V and can make the turn-off time quite a lot faster in switching applications.
You are making a "Schottky transistor" with discrete components:
http://www.mikrocontroller.net/attachment/14537/pullr_6.gif

It can help with optos that have the base available on a pin, as well as general transistors like T1 in your schematic.
Both that and the base resistor on the opto may each give some improvement, though a much faster opto is really needed.

 

[ChrisP58]

Dear, thanks for the quick response. The switching frequency is 10KHz. The integrated that you mentioned is not available in my city (small town in Chile), however TC4452 and MC33153 are available. I can notice that they do not have an opto-isolated stage, but do you think that they can do the job in the same way? Best regards.

Either of those would be fine as a mosfet gate driver for a PWM running at 10 KHz. The MC33153 has extra features that you don't need, and it's input to output is inverted, but it would would still be and OK part for the job.

Yes, neither of them are isolated. I suggested the part I did assuming that you did need isolation. Meaning the ground of the control circuit and the ground of the power circuit are not connected together. This is frequently done with motor drive circuits, but that does not mean that it always has to be done. If you don't need isolation, then you can tie the output of your PWM controller directly to the input of whatever mosfet driver circuit you choose.

Reasons why you DO need isolation:
1) The motor power is derived directly from the AC line power and not transformer isolated. Even though it's only 100 Volts, it is still directly connected to a *lethal* voltage potential.
2) There is enough electrical motor noise that it corrupts and/or destabilizes the PWM controller.
3) other.

At this point, only you can make the decision for isolation or not. But if you're not sure, feel free to ask more questions or give us more info on your application.

If you do need isolation, you'll need to find a optocoupler that is faster than the MCT2E.

 

[OlPhart]

Can't say on parts availability, but anything other than low frequency / low amp load Needs a MOSFET driver.
Recent project: 2.5A @ 12V load, 6A @ 30V FET, ~125Hz PWM.
Simple NPN w/PU resistor yielded FETs too hot to touch in ~6 sec.
Prior version was simple on/off so NPN worked fine, cust wanted next version PWM by sensor.
Changed to MIC4126, barely perceptible temp rise regardless DC% (design is 75-99% based on sensor input).

 

[OzFugah]

Thank you very much for your advice (I would dare to say that I have taken all your advices in part). First, get away from the MCT2E because of its bandwidth limitations. Second, consider the incorporation of an opto-coupled stage, in order to protect both the micro and the loads (motors). Finally, I have experimented with different topologies with BJT, and after failed attemp (always overheating of the FET). Today I was able to get the VO3120 in the local market. Although it is not identical to the TLP155 that was recommended to me, but I understand that it should work in a similar way. I will be posting the results with that IC.
Best regards.

 

[OzFugah]

Hello friends, it's me again. Finally I was able to get the VO3120 that can deliver up to 2.5 A at its output. I have connected the "free" motor (motor without load, approx. 300mA), as shown. Also change the power for the driver, now 19V (since the driver would not work with 12V). The voltage at the FET gate (irf640) is as shown in the figure. Everything looks great, except that at 30 seconds the mosfet overheats to burn. I have added a small heatsink, but also heats up excessively. Some people recommended to check the bootstrap topology, but I think that in this case it does not apply. Any suggestion is welcome. Greetings and thanks.

 

[rjenkinsgb]

The LED drive current must be a minimum of 7mA to guarantee proper operation, according to the data sheet.

Are you getting that with the 220 Ohm series resistor and only 3.3V possible drive voltage?

 

[alec_t]

..... also,
a) 19V is perilously close to the 20V Absolute Maximum rating for Vgs of the IRF640,
b) The on resistance of that FET is 0.18 Ohms, the starting current of the motor will be a lot more than 300mA, and so you can expect some heating of the FET.
c) Unless the gate resistor is right at the gate and there is only a very short wire between the driver and the FET it is possible the FET is oscillating.

 

[rjenkinsgb]

Another thought - is the diode across the motor a fast-recovery or schottky type?

Normal rectifiers line 1N 4000 or 1N5400 series are not fast enough to work with PWM systems, the turn on & turn off times are rather slow.

Also depending on the characteristics of the motor, a small inductor in series near the FET drain may improve switching.

 

[OzFugah]

The LED drive current must be a minimum of 7mA to guarantee proper operation, according to the data sheet.
Are you getting that with the 220 Ohm series resistor and only 3.3V possible drive voltage?

The voltage at R1 is approximately 2.1 V during the high pulse. With the R = 220 ohm, the current should be about 9.5 mA (7 - 16 mA according to data).

..... also,
a) 19V is perilously close to the 20V Absolute Maximum rating for Vgs of the IRF640,
b) The on resistance of that FET is 0.18 Ohms, the starting current of the motor will be a lot more than 300mA, and so you can expect some heating of the FET.
c) Unless the gate resistor is right at the gate and there is only a very short wire between the driver and the FET it is possible the FET is oscillating.

a) You are right about the maximum limit, Im working very close of it, however Im feeding with a laptop source, with a fairly stable output of 19.3 V. It has never exceeded 20 V. Also, the FET overheated when Vgs was near of 12V.
b) Like you said, the 0.18 ohm and the 0.3 A max RMS that Im measuring with the clamp generates 16.2 mW. Im sure my FET is dissipating more power than that, because it heats up to melt.
c) The resistor of the Gate is actually right next to the MOSFET pin, but I must mention that I have mounted the circuit on a breadboard, and I must be honest that the phenomenon of the oscillation was unknown. I will investigate on this point, thank you very much for mentioning it.

Another thought - is the diode across the motor a fast-recovery or schottky type?
Normal rectifiers line 1N 4000 or 1N5400 series are not fast enough to work with PWM systems, the turn on & turn off times are rather slow.
Also depending on the characteristics of the motor, a small inductor in series near the FET drain may improve switching.

The diode I am using is a RU1A, according to data it is a fast recovery diode, 5kHz tests are suggested according to its datasheet.


Thanks to all my colleagues!!!!

 

[rjenkinsgb]

The voltage at R1 is approximately 2.1 V during the high pulse

That's what I was getting at; the LED will have a voltage drop of 1.8V (or possibly rather more), leaving just 0.3V across the resistor.

That's about 1.4mA current at most.

You need at least 1.6V across the resistor itself to pass the 7mA threshold; that's just a bit difficult with a 3.3V signal less a LED voltage drop of around 2V.

You may need to drive the base or gate of a switching transistor and connect the opto and its resistor to power instead, switching the low side.

 

[OzFugah]

That's what I was getting at; the LED will have a voltage drop of 1.8V (or possibly rather more), leaving just 0.3V across the resistor.

That's about 1.4mA current at most.

You need at least 1.6V across the resistor itself to pass the 7mA threshold; that's just a bit difficult with a 3.3V signal less a LED voltage drop of around 2V.

You may need to drive the base or gate of a switching transistor and connect the opto and its resistor to power instead, switching the low side.

I think there is a misunderstanding, The voltage across R1 is 2.1 V during the high pulse, and across the diode the voltage is 1.2 V approximately also in the high pulse.
By adding both voltages, the 3.3 V is obtained during the high pulse, all this measured with the oscilloscope. Excuse me if my English is not very good. Best regards, colleagues.

 

[rjenkinsgb]

Yes, I misunderstood, I thought you meant the voltage into the resistor was 2.1V.