Inductive load switching with FET

Go To Last Post
45 posts / 0 new
Author
Message
#1
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I have a PWM circuit that controls the RMS voltage to an incandescent bulb (actually 3 in parallel for testing) using PWM.
I'm testing with the following parameters. Vin=38V, Vout(rms)=14V, RMS current out = 29A. The load is switched by a pair of 75V D2Pak FETs on the low side. Switching frequency is ~244Hz.
My problem is the "inductive kick" when the FET shuts off. Voltage at the Drain shoots up to ~77V above the source pin for about 5.7uS then drops off to the Vin of 38V. I've tried adding a 2A Schottky diode across the load, but the voltage at the Drain still shoots up to ~77V for 4.2uS under the same input/output parameters.

Is there some sort of "special" diode I can look for? Or a diode with a certain spec? I've been looking for diodes with a fast Forward and Reverse recovery times.
Am I missing something?

Jim M., Rank amateur AVR guy.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quote:
My problem is the "inductive kick"
Where are you getting the "inductive kick" from? Do you have some coils or transformers anywhere being switched?

John Samperi

Ampertronics Pty. Ltd.

www.ampertronics.com.au

* Electronic Design * Custom Products * Contract Assembly

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Fast recovery Schottkys have always worked for me.

If I did the math right you are looking at about 5uH of inductance. Assuming no coils, inductors, relays, solenoids, transformers, etc it must be elsewhere. I have never measured them but several sources say incandescent bulb inductances are surely <1uH and maybe near 0uH (in any case not enough to explain your problem).

How long are the wires between the FETs and the load? Copper wire is about 0.5uH/foot so 5 feet back and forth would be 5uH. Are you sure it is not your scope probes? You might check them against the reference on the scope. (BTW - analog or digital scope?)

The snubbers should be across the inductive load. If the wires are the inductors (and not the bulb) putting the diode across the bulbs is across the resistive but not the inductive load. Try moving them across the FETs.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

If you are pulse modulating a load at the end of some wire or cable, you likely would be getting some "kick". The bulb itself should not be a problem because its inductance is shunted by its resistance, hence a low Q. I suggest placing a nice fat Transorb across your fet.

Tom Pappano
Tulsa, Oklahoma

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

The "Inductive kick" I believed was from the 3 bulb filaments. The bulbs are 100W halogens. Their filaments are quite large. There are no other coils,chokes, etc in the circuit.
But, aero1's suggestion that it may be the wire looks very interesting. There IS a lot if wire. I'm going to re-wire the test setup to see what effect it has.

Jim M., Rank amateur AVR guy.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quote:
I have a PWM circuit that controls the RMS voltage to an incandescent bulb using PWM.

I read that you are controlling AC? What is the AC frequency & what is the PWM period. Can you just describe that in a bit mote detail perhaps the circuit of your switch, expected waveform and actual waveform?

Charles Darwin, Lord Kelvin & Murphy are always lurking about!
Lee -.-
Riddle me this...How did the serpent move around before the fall?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

The supply is DC. Provided by a Mastech 5020 supply. On the supply output I have a 75V/350,000uF cap to help the supply deal with the big current spikes drawn when the FET turns on. I've been testing with up to 50V so far. The PWM frequency is 244 Hz.

Jim M., Rank amateur AVR guy.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quote:
The supply is DC.

you had me confused writing about RMS voltages & currents, which imply AC!
Quote:
Voltage at the Drain shoots up to ~77V above the source

So are you saying that the load voltage goes from +38V when fet is ON, up to +ve-ve77V at turnoff and then back up to +38V.

The words you use seem to imply that the volatge goes up to approx +ve 77 V (or +ve 115V???) and then comes down to +38V again??????
I would have though a back emf would be -ve at that time????
Are you sure it is not just overshoot on the power supply when the load current goes to zero?
An oscilloscope shot would be helpful!

Charles Darwin, Lord Kelvin & Murphy are always lurking about!
Lee -.-
Riddle me this...How did the serpent move around before the fall?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

The supply is DC, but the FETs switch it on and off resulting is a PWM wave at the bulb. That wave's duty cycle is varied to result in an RMS voltage that is less than the supply. Vrms=SQRT(Vin^2*D)
The power supply does show a tiny bit of overshoot, but no where near 77V.

I'll get some scope captures at Vin and Vdrain later today.

The load voltage doesn't hit 77V. The voltage at the drain hits 77V when the FET turns off.

LDEVRIES wrote:
Quote:
The supply is DC.

you had me confused writing about RMS voltages & currents, which imply AC!
Quote:
Voltage at the Drain shoots up to ~77V above the source

So are you saying that the load voltage goes from +38V when fet is ON, up to +ve-ve77V at turnoff and then back up to +38V.

The words you use seem to imply that the volatge goes up to approx +ve 77 V (or +ve 115V???) and then comes down to +38V again??????
I would have though a back emf would be -ve at that time????
Are you sure it is not just overshoot on the power supply when the load current goes to zero?
An oscilloscope shot would be helpful!

Jim M., Rank amateur AVR guy.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Try adding a fast recovery diode, anode to drain, cathode to +38V supply. The diode conducts for voltage > (38 + Vf), where Vf is about a volt. The energy will be absorbed by the large capacitor. This might solve the problem.

It all starts with a mental vision.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

The BIG capacitor will probably not absorb
the spike because its own parasitic inductance.
I usually use a fast Schottky diode near the
MOSFET with an additional 0.1 uF capacitor
accross the DC supply very near to the MOSFET.
Show us your setup with a photo.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

tpapano wrote

because its inductance is shunted by its resistance, hence a low Q

I would have thought the inductance of the filament is in series with resistance of the filament as the currents through them are in phase, in which case the Q will be higher. It will still be relatively low because XL/R is quite low.
A CRO shot & a photo would be nice to have.

Charles Darwin, Lord Kelvin & Murphy are always lurking about!
Lee -.-
Riddle me this...How did the serpent move around before the fall?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Thanks for all the feedback. I'll try to get a couple of pics and scope caps soon. I couldn't make it into the shop last night.

For some of the replies above.
I have tried a fast recovery diode from Drain to V+. No significant change.
This is a test setup. The big cap and power supply will not be used in practice. The normal supply will be a battery.
A 0.1uF cap will do nothing. The RMS current is ~ 29 AMPS. Peak current is ~92A. The load is ~0.4 Ohms

Jim M., Rank amateur AVR guy.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quote:
I would have thought the inductance of the filament is in series with resistance of the filament as the currents through them are in phase, in which case the Q will be higher. It will still be relatively low because XL/R is quite low.

You are correct, the inductance as series is the proper equivalent circuit, since an ideal inductor would have zero resistance. The overall Q should still be low, whether an inductor has resistance in parallel or in series, I would think. I will stick my L-meter on a light bulb and see what it says.

I still think OP needs to put his suppression circuit *at* the fet rather than at the bulb. I'm betting the inductance of the leads is contributing. I suppose since it is such a low resistance lamp, the Q is not being reduced that much, anyway.

Tom Pappano
Tulsa, Oklahoma

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quote:
I'm betting the inductance of the leads is contributing.

I agree. It seems to me that if the supply leads are the same length and the -Ldi/dt is the same in each the induced voltages should cancel each other out. It is important that any measurement be done using the FET's source as a reference point especially if the inductance of the leads is much greater than the inductance of the load. The OP does not mention his reference point but if it is the -ve of the power supply it will lead to a huge error.

Charles Darwin, Lord Kelvin & Murphy are always lurking about!
Lee -.-
Riddle me this...How did the serpent move around before the fall?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

The scope ground reference is at the Source of the FET.

Jim M., Rank amateur AVR guy.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Make sure that the -ve of the power supply is not grounded (as in mains power ground), because -ve supply impedance will be quite different to the +ve supply impedance and will give you a big error.

Incidentally, have you observed the power supply output(at the power supply) during a complete PWM cycle? How stiff is it?

Charles Darwin, Lord Kelvin & Murphy are always lurking about!
Lee -.-
Riddle me this...How did the serpent move around before the fall?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Well, I've been in the shops now for a bit. Long enough to blowup the 3 bulbs I was using for testing. I'm using a single bulb now.
The first capture is the Vdrain at Vin=30V. 20V/div & 1uS

The second is the Vdrain (top) and Vin at the regulator (bottom) both 20V/div & 1uS.

I've been looking at snubbers, RC and DRC types to put around the bulb. But now looking at the Vin wave form, it looks like the 350,000 uF cap I'm using isn't stiffening it up enough and I'm also getting overshoot on Vin when the FET shuts off. I may have to just get 3 car batteries to do testing.

Attachment(s): 

Jim M., Rank amateur AVR guy.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Put a resistor between the gate and the uC. Lower your switching freq to the lowest that is practical. Increase the gate resistor until you are either happy with the voltage transient on Drain or FET is cooked. If Fet gets cooked add large heatsink. If Fet still gets cooked, add fan. If Fet still gets cooked, conclude that God hates you.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

toalan,
The FETs are driven by an FET driver. The driver supply is 11.7V. I have a 10 ohm resistor between the driver and the gates. slowing down the FETs until I have massive fan cooled heatsinks is not my answer. I have to more accurately simulate the condition under which it will operate in normal use. That means using a battery supply that can't overshoot. Once I've acquired some appropriate batteries (3 or 4 car batteries), I'll see if the ringing is still there. If so, I'll look at some sort of snubber. Anyway, I don't have room for massive heat sinks and fans. These FETs fail under 90C case temperature.
But I appreciate all the help and advice everyone has given me.

Jim M., Rank amateur AVR guy.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I might be heretical,but i think all those spikes in on/off pulses are just normal,especially in higher currents.Waveforms are captured and analyzed by digital osciloscopes,showing the rise and fall of a spike of 3us.
But i think it is normal.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

The capacitor issue, as mentioned by Ossi, is important. To effectively deal with high frequency currents, it is common practice to use two or more small caps in parallel rather than one big one, to better present a low impedance to high frequency current.

The other biggie is the circuit *layout*, including exactly how your instruments are connected. It is time for that picture!

Tom Pappano
Tulsa, Oklahoma

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

The circuit traces on the board are simple short and wide. 2oz copper on both sides. The wiring to connect it is another thing. I trimmed off all the extra wire I could but still have the 12ga superflex going to the cap then power supply. I'm going to get rid of the supply and cap entirely for future testing. I haven't purchased batteries yet, but when I do I'mm be connecting them with the shortest 8ga wires I can. Until then, a picture of the wiring set up I have now is pointless. Unless you want a pic of the board.

Jim M., Rank amateur AVR guy.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Quote:
Unless you want a pic of the board.

That would help

Tom Pappano
Tulsa, Oklahoma

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Capacitors may be required with the battery supply.

A snubber circuit using a resistor and capacitor in series source to drain may help. There are a few white papers, search for "mosfet snubber". Snubbers may add to switching losses.

It all starts with a mental vision.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I sometimes use the following technique for
identification of the source of ringing. I
add additional "parasitics" into the circuit
by paralleling capacitors to switches or increasing
length of wiring.

If the frequency of the ringing changes, the value of
the parasitic you add has some influence and therefore
is probably a component involved in ringing.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

OK Below is a pic of the board. As you can see, the power section is on the top. I can't make those traces any bigger. Also, before everyone jumps on the the routing of the gate control, this thing only runs at 244Hz. The top power traces are mirrored on the bottom with the exception of the Drain trace's cutouts for the gate pins.

Attachment(s): 

Jim M., Rank amateur AVR guy.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I would like to see a photo of the complete setup
with supply, capacitors, wiring, load etc.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Between the two fet pads is a perfect spot for a fat little Transorb. Assuming there is any length of wire at all going from the 'bulb' terminal out to the load, you will have the flyback pulse you see on the scope due to the wiring/lamp inductance. The Transorb will clip that pulse to keep it below your fet's maximum rated voltage.

Tom Pappano
Tulsa, Oklahoma

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

ossi wrote:
I would like to see a photo of the complete setup
with supply, capacitors, wiring, load etc.

I'll see what I can do.

I'll also lookup "transorb".

Jim M., Rank amateur AVR guy.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

transorbs are just a particular brand of TVS diodes. You can get them capable to handling 1000W or more for short durations. All that kickback energy will come out as heat from the TVS and you might be trading noisy +ve supply for a noisy ground (more noisy ground).

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

If the load is incandescent bulbs, just put a holdon R across the FETs. Problem solved.

Imagecraft compiler user

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I can buy/sample a few TVS diodes and see how things look noise-wise. The electronics ground plane is separted from the power ground plane and only tied in one spot. The electronis V+ supply is fed from the same supply that feeds the bulb. It has a 10 ohm series resistor then a 160V/10uF aluminum cap before the primary regulator and >10uf of ceramic caps after the primary (12V) before it goes into the 5V regulator. That has 4.7uF after it as well as 0.1uF at the AVR.
I think my next steps will be as follows:
1. Get batteries instead of power supply.
2. Get a few TVS diodes
All along the way I'll be scoping the FET drain, Electronics V+, 12V and 5V busses for noise.
There's no sense blowing up more FETs with a whacky power supply though.

Jim M., Rank amateur AVR guy.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

OK. I'm back after buying 4 lawn tractor batteries.
With 38V input (3 batteries) I'm still seeing almost that that same spike at FET turn off. I've tested a schottky diode across the bulb, with zero effect. However, I think it may have a lot to do with the lead length. I can't really make the leads shorter than they are now. A total of about 4.5 feet including the jumpers between batteries. However, I connected an 82,000 uF Aluminum Electrolytic cap at the regulator. The below pic shows the scope at FET turn off. Green is with the cap, Light Blue is without. So, aside from slowing the FET down so it makes a ton of heat, would the best solution be a 75V transorb/TVS across the FETs?

Attachment(s): 

Jim M., Rank amateur AVR guy.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Yes, perhaps with a clamp voltage less than the fet's maximum rating rating, but more than the maximum expected supply voltage.

Tom Pappano
Tulsa, Oklahoma

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Thanks. I'm reading up on TVS Diodes now. Standoff voltage versus Breakdown, etc. I have a nice little place on the boards I've already had made where I can just drop one in.

Jim M., Rank amateur AVR guy.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

No one thinks the keepon R will work? Just try it? We used em for years. Select R to keep the filament just at the threshold of incandescense. Instant on, no warmup surge.

Imagecraft compiler user

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

One could compare similar plots using one and few, or, alternatively, different bulbs. Btw is serial bulb connection an option?
Looks like supply topology is possibly causing this phenomena. Halogens are nasty loads - measure cold filament resistance. So second for preheat resistors.
PWM 244 Hz chopped 34V is no way the same as 14,4V DC. You already proved this by burning first set of lamps, imho.
OP mentioned regulator in circuitry and significant lengths of wires. A diagram with all these details would help (imho proving wire ringing):
regulator placement
capacitor placement
bulb wiring.
Shall this be happening from ringing, even winding/straightening wires would have immediate effect on scope plots.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

The keep-on Rs may work very well. But my application battery powered. Thus drawing any significant current while off can't really be done.

Halogens are nasty loads. Their start-up current can be as high 10 times their running current.

PWM is not the same as DC. Correct. However, to the thermal mass of the bulb filament, it's close enough.
34Vin @ 18% duty = ~14.4VRMS. The FETs failing in a mostly ON state is what blew up the bulbs. I've been using 244Hz PWM to run lots of bulbs with no problems. It's just that this new design is supposed to handle almost double the voltage and 3 times the current as the previous design. What was a tiny spike in the previous design, is now a monster spike. That's what I'm trying to quantify. Series bulbs will not be used. Typically only a single bulb will be used. Possibly 2 in parallel.

I'm going to try rearranging the wiring to run the supply and return wires next to each other.

Jim M., Rank amateur AVR guy.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I've made another measurement. This time ty-wrapping the supply and return wires together as close as I could from the batteries to the regulator/bulb. The spike has diminished.
In the above scope capture, the blue was the "without cap". Now the pink (just less than the green)is the "without cap". The Orange is the with cap now.
Just more indication that the spike is mostly the fault of the inductance of the wiring. In practice, the wiring will only be several inches, not a few feet. I'm still going to buy some TVS diodes, just to be sure.

Attachment(s): 

Jim M., Rank amateur AVR guy.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

The best solution believably would be turn-off snubber occupying free place between MOSFET's.
As an alternative would suggest using of TVS on source lines, automotive specific.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Can you just use a higher voltage rated FET?

(including a snubber would still be recommended)

oddbudman

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I've ordered some 70V (1500W) TVS diodes. The regulator will be used with several different kinds of batteries and bulbs. Designing an effective snubber would be difficult since the inductance of the load/wires would be nearly impossible to calculate.
Of course just adding a "good enough" RC snubber across the FET is do-able.
I'd really rather not have to return the FETs I've purchased.

Jim M., Rank amateur AVR guy.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Well, it looks like the TVS diode is doing the trick. The spike is still there but is chopped off.
The FETs had been running at ~90C at 38Vin at 14V/29A Out RMS. Now they run at 56C at 50Vin 15V/30A out RMS.

Jim M., Rank amateur AVR guy.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Excellent!

Tom Pappano
Tulsa, Oklahoma