Weird results from air-core transformer

[Chancy99]

I'm seeing some strange results from an air-core transformer - basically two coils of 30ga wire. Perhaps someone can shed some light ?

Primary coil is 170 turns of 30ga wire that was wound on a 2" dowel. The secondary is 100 turns the same way. The two coils are sitting on the desk, the secondary on top of the primary with about a 1/4" gap.

The primary is driven by an IRF540 - the gate goes via 200ohm resistor to a micro generating a 15kHz 50% duty cycle PWM signal. The Source goes to GND. The coil is across the Drain and VCC. The secondary goes through a bridge rectifier.

I'm seeing about 23Vdc across the secondary. I was expecting around 8Vdc or so. 100/170 * 14 right ? Hrmm. The 23Vdc is what works out from 170/100 * 14, so it's as if the primary and secondary are backwards.

But if I reverse them, put the 100 turns as primary and the 170 turns as secondary, it puts out around 90Vdc.

The other oddity is that I can increase the 14V up to around 30V, and the output only rises from 23V to about 27V or so.

Dean.

 

[Chancy99]

Madcow -

Excellent - thanks very much for the link. I do wish it was easier to embed images here - I would like to show the schematic as I have it.

Futzed a bit with it tonight. Let me see if I can describe the schematic.

NPN 2n3904 to act as a gate-drive switch for the mosfet. Base has 15kHz PWM square wave to it via 1k resistor. This is fed from the microcontroller as a 0-5V signal. Collector goes to VCC/14V. Emitter has 1k resistor to LED then to GND. This is to show it's on/switching and also to bias the mosfet gate to GND.

Emitter also goes via 200ohm to gate of IRF540. Source goes to GND. Drain goes to primary coil, then to VCC/14V. So power goes from VCC through coil, through switch/drain to GND/source.

The snubber consists of a diode (anode) going from the drain/coil junction to (cathode) a parallel resistor/capacitor, which then go to VCC/14V. This seems to be the classic snubber circuit for a flyback from my reading.

Heh, not too many had examples of *values* though, but for one. 1uF cap, 1k resistor, which I tried. It helped some, but not as much as I thought it would - it did remove the spikes, but the peak-peak was still pretty high.

22nF = 62Vp-p ugly spiky waveform
1uF = 41Vp-p clean square wave, rounded on the bottom
10uF = 33Vp-p clean square wave

1K resistor got hot. I mean it got HOT. I put a ~60ohm load across the secondary so it would pull some current.

I'll read that snubber design link tomorrow. Hrm, would be nice to know the inductance of the coils. They're from 30ga magnet wire wound on a 2" form. The primary is 170 turns, the secondary is 100 turns. I know there's a formula out there to calculate it, but I'm too tired right now to search it out

More tomorrow.

Dean.

 

[Chancy99]

Hrm - a thought. Instead of feeding 14Vdc through the mosfet to the coil, what about regular old AC ? Use a 12-20Vac wall-wart. Use the PWM to lop off the tops of the waveform to control the total power feed to the secondary.

Given that this is air-core, and with some distance (5-10mm) between the coils, how badly would the change from 15kHz to 60Hz affect the power transfer ?

Dean.

 

[Chancy99]

OK - take a look at this :

http://seawall.areyes.com/PWM-mosfet.pdf

I have this breadboarded up right now. Works fine, except that there are two "situations" ...

If the 0.1uF bypass cap is in place, then the mosfet gets blisteringly hot. This is regardless of the presence of the snubber components. This setup produces the nicest sine-type waveform, so the snubber isn't really needed - no nasty spikes.

If the 0.1uF cap is removed, then the 1k resistor in the snubber gets really hot. But the mosfet stays nice and cool. The snubber is required, or we see those 90V spikes on it.

Any reccomendations ?

Dean.

 

[melone]

The reason your IRF540 gets so hot during normal operation is because you have essentially shorted power to ground, thru your coil. Try putting some series impedance before the transistor and see if that cools it down (it should limit the current).

Move D1, R3, and C1 to be connected across the relay coil. The current location of these devices won't clamp the flyback voltage since Q2 essentially isolated them. All you have to do is disconnect the anode of D1 and connect it to the Drain of Q2 and everything should run nice and cool.

Good luck and keep us posted!

 

[Chancy99]

Dang it ! The breadboard is actually correct, the bloody schematic is wrong. It's late ...

The anode end of the diode is actually connected to the drain of the mosfet Q2. I just redid the pdf - so it now reflects the breadboard appropriately. Note - heating effects are still as noted above.

I added in a series resistor in the schematic between the mosfet source and GND. Haven't done it yet - tomorrow. Or should it go between the coil and VCC ? Does it make a difference ?

Dean.

 

[melone]

It should be between VCC and the coil. Otherwise, you will raise the Souce voltage on Q2, which will reduce Vgs, i.e. not turn on the transistor as hard.

 

[melone]

Also, eliminate R3. This will clip the flyback voltage to Vd.

 

[melone]

Get rid of C1 as well.

 

[melone]

Sorry, last post. Why do you have C2 across the Drain to Source? Why don't you put in on the Gate leg, and simply slew the turn on / off time, and generate your shape that way?

One more thing, you might want to place a 10K-100K resistor from Gate to Ground. This will help evacuate the charge from the Gate of the transistor.

 

[Chancy99]

Cool - more comments. No such thing as last post. Here's the link again.

http://seawall.areyes.com/PWM-mosfet.pdf

Once I thought about it, putting the resistor between the VCC/coil and the mosfet makes sense.

The C1/R3 combo is the snubber circuit - that clips the spikes down from their 100V high to something more manageable. At least that's what I read.

C2 is across the drain to source to round off the waveform going into the coil. The coil power transfer deals better with a more sine-wave form than a square wave, and it also helps to reduce EMI ("fewer" harmonics). If it's across the gate leg, then the mosfet won't just be switching, it will also be operating in the linear region, so heat up, won't it ?

There is the 1k resistor and LED from the gate to GND - shouldn't that help evacuate the gate charge ?

With the snubber in place, R3 gets hot, dumping the back-emf, and I measure about 150mA through the primary coil. The waveform is a nice clipped sine wave. Without the snubber, the mosfet gets cooking hot, and about 100mA through the coil. The waveform is more ramp-shaped, with much higher peaks. Those are DC current measurements on the meter. AC-rms are somewhat lower. The coil measures 6.7ohms across it.

I used a capacitor decade box to arrive at the 0.1uF cap across the mosfet. That value was the best compromise between waveshape rounding and power transfer. Smaller values still had large spikes, while larger ones damped it down too much.

I need to get about 400mA at about 5V across to the secondary.

Dean.

 

[melone]

- If the diode is placed correctly, you will clamp the flyback to ~0.7V.

- By moving C2, you will definitely cause Q2 to operate in the linear region. This will translate to heat.

- The led needs about ~2V to turn on. Therefore, once Vgs drops below ~2V, you will no longer have a path to evacuate the charge (other than the leakage thru the LED).

 

[Chancy99]

OK - take a look at the link again. I've updated the schematic to what I will test with tonight. You may have to refresh your browser. Changes are :

1) Increase 1k resistor in snubber circuit, to limit the current it's carrying. This will reduce the amount of clamping it will do, but that should be offset by the rounding effect of C2. May even be able to remove the snubber components (D1, C1, R3) completely depending on the size of the spikes.

2) Add in R7 at about 10 ohms to start with. This will limit the current through the primary coil, and hence the mosfet heating.

3) Move C2 low leg to below R7. Hrm - should it be right at GND, or right across the mosfet ?

I'll also add in the 3.3V LDO linear regulator to see how stable the output voltage is.

Dean.

 

[melone]

- Don't put R7 in the Source leg, it must be in the Drain leg of the circuit (for the reasons mentioned above).

- Increase R8 to 10K-100K. If R8 is too small, you will create too much of a divider, and therefore not bias the gate hard enough.

- Get rid of R3 and C1. They won't do you much good.

 

[Chancy99]

Get rid of just R3 and C1, or D1 as well (the whole snubber) ?

Dean.

 

[Chancy99]

OK. Another iteration. The snubber is gone, and it looks like this now :

http://seawall.areyes.com/PWM-mosfet.pdf

Guess what ? The mosfet *still* gets hot. Even with a heatsink on it, after a few minutes it's very hot. The bench power supply is only giving it around 80mA at 14V, and the meter between the coil and 14V is showing about 70mA consumption through the coil.

So why is the bloody thing heating up so much ? It should be simple : NPN acting as a switch to 14V to act as a mosfet gate driver. Mosfet to act as a switch for a coil. On off on off on off. 0.1uF cap to smooth out the square wave. Ba da bing ba da boom

Dean.

 

[melone]

The problem might the transistor that you are using to turn on the FET. The maximum voltage that you will be able to see should be ~ (5V-0.8V)*1000/(1200) = 3.5V (assuming you use a 220ohm series gate resistor). By the way, does Q1 get hot?

Is there any way to increase the voltage on your PWM generator, to say, 14V? I have a sneeky suspicion that Q1 might be the source of your problem. I think that Q2 isn't properly turning on because Q1 isn't fully active.

Some quick and dirty simulation indicates that you are not sufficiently turning on your FET since you can only hope to achieve 3.5V, and the gate needs time to fully charge / discharge. Try using the following values for your components:

R8 = 810
R2 = 47
ADD a series resistor of 2K between Q1 collector and 14V (ABSOLUTELY ESSENTIAL!!!)

This should hopefully solve your problem if you can't get the PWM voltage increased. Also, try decreasing your PWM rate to give Q2 a chance to cool down.

Finally, keep D1 in your circuit. The extra capacitor and resistor probably won't do you a lot of good, but the diode can REALLY help you.

Good luck and keep us posted!

 

[Chancy99]

Hrm. It's certainly acting like the mosfet is still operating in the linear region. Right now I have a 33ohm gate resistor as you had mentioned earlier.

The AVR is running at 5.1V, and the PWM signal entering the 1k base resistor shows a 5.6Vp-p on the scope.
Q1 gets only very slightly warm. The scope shows a p-p voltage at Q1-emitter of about 5.6V as well - this is the first time I've actually checked that, I thought it would be higher.

What is the 2k in the Q1 collector for ?

Right now the whole snubber circuit is out (D1, C1, R3). The 0.1uF cap across the mosfet rounds off the wave nicely - no spikes. So it looks like the cap isn't needed. I'll try a smaller cap (more spikes), and add back in D1 to shunt the spikes.

R7 - the coil current limiter, is 27ohms. It gets pretty warm. I would expect that dissipation to reduce when there's a real load on the secondary, as more power will be getting transferred over rather than being dissipated in R7.

Getting there. Supposed to be simple, never is, but getting there, and I'm learning, which is key

Dean.

 

[melone]

The 2K is required since you are running from 2 supplies. Your PWM can only get as high as 5.6V, but you want to drive the gate with 14V. The problem is that with the NPN, the voltages around the device must sum to 0, otherwise, things don't work. That 2K drops enough voltage so Q1 can operate properly.

 

[Chancy99]

Hokay - I think we got it. Woo hoo. Check out the links again :

http://seawall.areyes.com/Air-core-xformer.pdf

I finally broke down and ran it through spice, playing with values. This was SwitcherCAD from Linear Tech. I used close-fit parts for the mosfet and the diode.

http://seawall.areyes.com/Air-core-xformer.asc

In practice, it's really close. The gate voltage is actually around 11.5V rather than the 9.5V predicted. Mosfet stays nice and cool, as does the 2n3904. The drain shows a nice 16V square wave - it would be nice to round that off a little. What gets hot now is the coils !

To be expected I suppose - 170 turns of 30ga on a 2" form. Pushing 400mA through gets it nice and toasty. I may have to redo it with 28ga instead.

The secondary coil is 100 turns, the same. It only shows about 2.5Vp-p on the scope. The coil ends are going through a pot set to about 20ohms. One leg has a schottky diode on it instead of the bridge.

Which means the efficiency is lousy. Might swap the coils ...

Dean.

 

[melone]

Change R3 -> 1K and R2 -> 100K. This should help limit the current through Q1 (1K ~14mA, while 470 ~28mA), and turn on M1 harder ([14V-0.1V]*100K/[100K+1K]). Otherwise, congratulation on getting it working.