Reverse Phase Dimmer and Mosfet Destruction

[Bskitter]

Hello Everybody, I have built a series of Edison light bulbs (20 of them 1200 Watts) a few years ago. Obvisouly I did not want to run them on 100 % but around 10 % of their power gives the perfect orange light.

I originally dimmed the Bulbs with a Basic Triac Circuit and a beefy hestsink, the buzzing noise was also apparent from the cutting of the phase. (All commonly known things).

Now I am building a new control circuit where I can turn the lights on and off via a web page. All of this has been completed along with temperature and current sensing. While testing I was using a Triac controlled via a MCU and some large inductors to reduce the EMI, but they buzz like crazy.

So before finalizing the project and creating a PCB I am investigating Reverse Phase dimming which should completely eliminate the EMI.

So About the circuit I have used. Most of the circuit has been used from this website, the author has documented the project incredibly well.

Below is my circuit.



When testing it with one 60 Watt bulb the circuit acts fine. But once I connect the 1200 Watt load I instantly destroy both Mosfets and I do not know why, Im guessing it is the inrush current of a Voltage Spike although the voltage spike would be suppressed by the Varistor parallel to them.

Any advice would be much appreciated.

Byron 

[Andy Watson]

My first thoughts are that there is nothing significant to turn the MOSFETS off. R7, 10k is not going to make much impression on the Miller capacitance, so the MOSFETs will be spending a large amount of time in their linear region - dissipating a lot of heat. The ESP design uses a 555 which will have an active pull-down.

[Bskitter]

Could reducing the resistance to about 5k be sufficient or would it be better to implement some logic that will tie the gate directly to ground when there is no trigger ?

But could this really be the issue ? Or any other thoughts, I have already destroyed 4 pairs of these FETs and they are not cheap

[MasterT]

1200 W bulb is a hell of a load, having cold resistance 10x lower than nominal, it's about 12.000 W, or 60A for 200V.
If it's me I'd continue to use a triac , but instead of phase regulation that was good in no-MCU era, better to regulate power by skipping half-circles, so switching would happened at zero voltage/ current and inductors could be omitted.

[Bskitter]

better to regulate power by skipping half-circles, so switching would happened at zero voltage/ current and inductors could be omitted.

You are referring to pulse phase control ? I was thinking about this as well but read in many places that this causes observable flickering and the dimming resolution is reduced dramatically.

1200 W bulb is a hell of a load, having cold resistance 10x lower than nominal, it's about 12.000 W, or 60A for 200V.

Good point, the avalanche current for the IRFP460 is 20 Amps, .. what about putting maybe two or 3 FETs in Parallel ? so 6 in total ?

[MasterT]

Agree, flickering may be an issue when power is low.
I'm thinking aloud, something like smart start-up algorithm needed. For beginning, you need a current sensor - uCPU should always monitor current and do fast MOSFET shutdown if current above 20A. Than, phase monitoring, opto-couplers etc.
uCPU - knowing that voltage goes down, lets say 10-20V , issues a start-up command to MOSFET. Current -5A or so, and than uCPU is slowly driving MOSFET earlier each circles, when voltage is higher, bringing current up to 20 A. Than, bulb start to warm up and current, of course, would drops, so uCPU would continue to decrease a fire-up angle till required power is reached.

[james_s]

How about just running the lamps from a low voltage source? Depending on the voltage of the lamps and the power they draw when dimmed, a regular iron power transformer might be feasible.

[JS]

How about just running the lamps from a low voltage source? Depending on the voltage of the lamps and the power they draw when dimmed, a regular iron power transformer might be feasible.

For a fixed level might be a good idea, in not that low auto transformer is the way to go, as they go down quite fast as they go down in voltage, but you said 10%...

(20 of them 1200 Watts)... Obvisouly I did not want to run them on 100 % but around 10 % of their power

That's 2400W piece of iron, expect quite a few kg but perfect clean power delivery and consumption.

Power mosfet need to be turned off hard, not just wait for it. You could also consider IGBT for this, but again, turned off hard, don't wait, drive it hard low. This could be done with IC drivers or with transformers, there isn't anything obscure about it.

JS

[Giaime]

1200 W bulb is a hell of a load, having cold resistance 10x lower than nominal, it's about 12.000 W, or 60A for 200V.

This. You need properly sized switches (MOS or IGBT), current measurement, uC that handles soft-start...
I also would not use the semiconductors "as-is", either I would use a low frequency inductor in series with the load or an high-frequency choke and run the switches in PWM (search for "AC chopper").

[Siwastaja]

You are lacking both gate driver and MOSFET current sensing. While you can design those using discretes, integrated gate drivers with desaturation detection exist.