PWM drive for a 2KW universal motor/router

[LZ1TWB]

Hello all,

I am designing a PWM drive with a closed-loop speed control but have some major problems with the power stage. The schematic is pretty simple: TL494 pwm generator, mosfet driver and a single mosfet/igbt at the end. A fast diode with a soft recovery across the motor. The problem is at the switching between states of the power transistor. I have very high peaks of current, measured across a source to ground resistor of 0.22Ohms. I read something like 25-30Amps for a period of 0.3usec. I have tried to increase the gate drive resistor and got some results but the switching speed degraded significantly and switching loss was getting high. Is there a way that I bring these peaks low enough so I don't blow my transistors? Btw, power supply is kept low - around 80V DC. Target is a rectified 220V AC line.

I have tried a resistive load in forms of a light bulb and got much lower values of the peak current. This is not the case with the motor connected however.

I have attached a photo of the voltage across the sense resistor, this is with a 1KOhm gate resistor /pretty much I know/. 20usec/div and 0.5V/div measurement.

Thanks,
Todor

[Al_The_Man]

I believe traditionally universal motors have been triac controlled, there is a Motorola IC out there designed for such an application using a simple inductive pick up tach.
TDA1085. There was a treadmill maker that used this that had the triac fire into a full wave bridge and the motor connected to the bridge DC side.
Al.

[Larken]

Sounds like you have some capacitance in the load. A 1k is much too big. it should be less than 50 ohms for the gate resistor.

Can you post the output part of the circuit ?

Larry K

[LZ1TWB]

Thanks for the replies. I am very close to the moment that I abandon this project and really do it with the TDA1085. Anyway here is a part of the schematic I drew recently. I managed to keep the positive peak of the current low with the BC547 transistor but the negative caused by the freewheeling diode remained. I still cannot understand why the mosfet shorts out and what is the cause. I've got the gate protected, power supply voltage if far less that the device rating, I've got no ringing at the voltage across the FET, no excessive heat dissipation. It is some transient process that causes the damage. Sometimes it is done at startup, sometimes when I change the programmed speed. It is always connected to some major change of the input. The 494 is wired to have a current limit, and also the BC547 acts just like one, so I don't think I overloaded the thing. By the way the sense resistor is 0.1 Ohm now.

I will be glad to see any light for my problem guys. Thanks.

Todor

EDIT: I removed the choke and the filter capacitor inside the motor at the very beginning. It should be his internal capacitance only now. Maybe it is something to the with the motor being series wound.

[Larken]

I would replace that DS12 diode with a mosfet with the gate and source connected acting as a diode.
I couldn't find specs on the diode, but it may have too much capacitance.

I would think the BC547 can cause the mosfet to be under driven and make it fail. you would need that transistor to be fully on or fully off .
BTW,the mosfet you are using is pretty wimpy.

Driveing an inductive load may be causing the diode to saturate on the reverse kick and then its too slow to recover when the mosfet turns on.
Is it possible that your turning the mosfet back on as the coil is colapsing ? maybe reduce the pwm frequency.

[LZ1TWB]

Hi Larken,

Recently I did some more tests and measurement. It looked like I didn't take into account that the sense resistor has some inductance and the voltage that I see across it is not quite the current through the transistor. I obtained a 120Mhz current probe for the scope and the output seemed more friendly than I thought, you will see it on the attachment. I removed the BC547 that was on the gate.

I reassembled the current limit circuit, removed it from the input op-amp section and put it on the dead-time control input of the TL494 as manufacturer suggests. This should do a pulse-current limiting that works better. It looks like my previous limiting circuit was slow enough so that the current at turn-on was rising faster than it can sense and that eventually blew the switching FET.

Tuning the PID settings is a bit tricky also and the interesting thing is that I couldn't find a suitable info on the net about driving a DC motor with a single OP-AMP PID controller. There is of course a lot of theory but almost no one put some real values of the resistors and capacitors on the feedback circuit. I doubt mostly on the impedance of the circuit. Say I can make the same time constant with a 1mF cap and 1MOhm resistor and with 10mF and 100KOhm. I'm not sure though which one will do better and how it would influence on the control process. I managed to do some good results as minimum slow-down when torque is applied and less ringing at a change of speed but I realize it should do a lot better if I find the proper scale of the values of elements there.

Anyway, datasheet of diode DSEP 12-12A says it is designed especially for "free wheeling diode in converters and motor control circuits". It is rated 15A continuous at 1200V max, 40ns recovery time. The IRF634 is weak and I used it at lower voltage for testing purposes only. The final version should be with an IGBT. I will change it with IRG4PC40W and make some more tests at higher voltages, possible reaching a rectified 220V AC, which goes up to 300V.

Attached is the voltage across the FET at 50V/div and the current through it at 2A/div.
Todor

[Mariss Freimanis]

The 300ns turn-on spike is a sum of the diode reverse recovery current and motor inter-winding and winding to motor case capacitive currents. Your diode is fast enough (50ns trr) so I'm guessing it's capacitive current for the most part. The capacitance can be substantial, someplace between 1nF and 10nF. Place a 10uH to 100uH ferrite core choke in series with your MOSFET drain to motor connection near the MOSFET end. This should limit the di/dt to a reasonable value.

About your BC547 current limit circuit. It puts your MOSFET at risk and is unstable. The collector feedback to MOSFET gate will tend to bias your MOSFET in its linear region as it attempts to limit current. Operation in that region exposes your MOSFET to SOA curve restrictions. MOSFETs are very fragile in that region (simultaneous significant Ids and Vds). It is unstable because it will repeatedly turn your MOSFET on and off at a high frequency (>1MHz).

Instead, your BC547 collector should terminate the PWM cycle for the balance of the switching period. Use an RS flop; a narrow (<1us) "set" pulse to initiate the PWM cycle and a narrow "reset" pulse to terminate it. OR the collector with "reset" signal. This will give you stable and safe cycle-by-cycle current limiting.

Mariss

[LZ1TWB]

First of all, happy new year!

Thanks for the answer mr. Freimanis. The BC547 current limit circuit was part of my first tests, it is removed now. I didn't quite understand your idea of the RS trigger, I guess it is useful when the pulse generator is a microprocessor, which is not exactly my case. I have seen it done with a D trigger that was put before the mosfet driver and powered its shutdown input. This should be similar in terms of operation.

I read an application note on using the TL494 and they give a simple schematic of a pulse current limiter incorporating 3 transistors. Basically it works but is not so stable, I think they missed some base to emitter resistors in it and it should be thought over to make it work as it should. The principle is that when a high enough current is measured, the outputs are disabled until the timing capacitor is discharged. This is exactly my problem that the BC557 does not fully go to the OFF state.

More or less the schematic is working, the last thing to be done is of course the current limiting. I realized that when we are going to high voltages and a stall rotor only the proper operation of this circuit will prevent the output transistor of frying.

Attached is a more finished state of the design, the IR2184 driver is for reference only, final schematic will be with a single driver.

Any advice on improvements are welcome.

Todor

[Mariss Freimanis]

Todor,

Happy New Year to you! I'm not Mr. Freimanis to you; I'm much more comfortable with Mariss.:-)

The idea on pulse by pulse current limiting is it terminates a switching cycle anytime the current limit is reached. A process (the PWM modulator) starts the switching cycle. A single pulse from the current limit detector is sufficient to terminate the cycle prematurely until the next PWM start cycle. The process repeats until current falls below the limit threshold.

Your BC557 will never saturate (switch). As soon as the collector voltage falls below the MOSFET gate to source threshold, current through your sense resistor eases and acts to turn off your NPN transistor. Gate to source voltage rises then, drain current increases and the process repeats itself again and again. Why does it repeat? There is a propagation delay time through the circuit that stabilizes at a -180 degree phase shift. It's not pretty to see what's on the gate of your MOSFET when you look with a scope.

Can you describe your current limit circuit? It looks to be slower Aesop's turtle based on the impedances I see. Keep in mind the limit is not temperature compensated. It will trip at 14A for a 25C NPN junction temp and 10A for a 125C junction temp. There are ways around this you know. The circuit is way too complicated for what it does.

Also loose the bipolar zener and the 10K resistor gate to source. They contribute nothing. Finally, there is no need for the IGBT collector to emitter diode (DSEI12-12A). It is a big expensive part that will never do anything in your circuit; your switching topology doesn't ever have a state condition where it will conduct.

Finally, this looks like a DC motor speed control. Have you ever considered adding IR compensated feedback to get better than 2% speed versus load regulation? All it takes is a few resistors and capacitors ($0.02 cost).

Mariss

[LZ1TWB]

OK Mariss,

The bipolar zener is just for protection, in case something with the gate drive gets corrupted, and the 10K is there because I see many mosfet circuits use one.

I really doubted about the IGBT parallel diode, some designs use it, some don't. I decided to put one just in case. It seems though you are right because it does not even get warm. I'll remove it tomorrow and see if something changes.

I am struggling with the feedback circuit now, it seems that the frequency to voltage conversion that I use as TACH has some ripple at the output and the controller seems to try to correct the speed every revolution or so. This leads to a very fast, jerky movement better seen at low speeds near zero. If I reduce the proportional gain it moves smoother, but the speed lowers significantly under load, which is not good. I have used a DWT wood router for quite long and I hate its feedback loop. It drops under load, it is very slow in reaction, and it overshoots a lot when the load is removed. Seems they didn't make it with the PID settings there, so I intended to design the spindle controller the way I want.

As for the TACH feedback I drew it simple in the schematic, but in the real world it is a UGN3113U hall effect sensor, put near a round magnet with 2 poles, so I have 1 pulse per rev. A LM2917 converter outputs 0-6V which I use in the PID loop.

That is for now, I will give more info after some more testing.

Todor

[Al_The_Man]

For what it is worth, the SCR TDA1085 T.Mill controller I adapted uses a 8 tooth Pick up wheel and hall type pick up as a tach.
I am using it on band saw and it run very stable at low rpm.
Probably much lower rpm than spindle use.
Al.

[LZ1TWB]

The dumb problem that I didn't start the design with the TDA1085 was that I could not find it here in the local shops. I knew the schematic will be simpler and no doubt it will work, I just don't like SCR's from the time we studied their theory in the university. I had a bunch of TL494's here and that's why I did it this way.

Actually this is a revised schematic for higher voltages based on my old design driving a sewing machine PM DC motor with analogue TACH. It also seems that driving a brush motor at DC and at a higher frequency PWM leads to significantly less arcing on the brushes which tends to smooth the collector and brush surfaces, and hopefully produce less wear.

Todor