# High frequency p-channel mosfet pwm dimmer problem

The circuit below is the final one, after all the advice from Mishyoshi and especially Andy aka.
It includes Zener diode protection against over-voltage between gate and source on the Mos-fets and also a back-emf diode to deal with below-zero pulses caused by cable inductance to the lamp.
It seems to work well with the 20Khz PWM dimmer I needed to use, at least on the workbench. In the future I would like to add a circuit to cut the supply if it falls below 10 volts to safe guard the Mosfets. Hopefully this will help anyone faced with the same problem as I couldn't find a practical circuit when I looked.

simulate this circuit – Schematic created using CircuitLab

Hi everybody! I am a newbie and although I have read a lot of stuff on this forum this is my first post.

I am an electrician working on a feature film and have been given the task of making several American trucks headlights dimmable. Nowadays every light the camera sees needs to be dimmable by remote control from the lighting desk. The industry standard wireless DMX dimmers (in the UK at least) are Lumenradio LV4's. These run on 12-24volts dc and give a Pulse width modulated output. They have four channels, at 4 amps each.

Unfortunately as well as being under-rated current-wise for the lamps,they are also designed for LEDs so they are common positive. The headlamps are hard-wired as common negative (the lamp fittings have one terminal going to the chassis of the truck which is connected to the negative of the battery).

This means I had to make a high-side p-channel mosfet switcher to drive the lamps. The lamps are 190 watt halogens (there are two filaments for high and low beam, one at 100w, the other 90w).

I did a lot of internet searching trying to find a circuit that would do the job but found it quite exasperating. There was a lot of 'be careful you don't make this value too high or this bad thing might happen, but if it's too low this bad thing might happen' but precious little in the way of a circuit I could just copy. I will admit I am no good at designing circuits to component level. For most electronics needs there are standard circuits which can easily be adapted to specifics, but on this subject It seems like there is an amount of witchcraft involved.

Anyway, I built the above circuit and it works but not with the preferred LV4 dimmers, but it does work with these cheap chinese versions;

I presume this is because they work at only 400HZ, whereas the LV4's work at 20KHZ. We settled for this as a solution but now they are going to be using high speed cameras, around 2000 frames per second so they are worried about flicker. I have seen high speed film of incandescent lamps and they do pulsate.

So now I have to make a circuit which will work at 20 KHZ. I do not understand what it is about the circuit which makes it frequency dependent. I wonder if it is because I used strip-board or if it is a more fundamental design/component choice problem.

P.S. I fitted these filters;

to the truck battery side to deal with transients from the 'dirty' automotive side as I read that MOSFETS are prone to voltage spikes.

P.S. I used an opto-isolator because I didn't know what practically the power set up would be like and it seemed prudent as it keeps the dimmer safe from dodgy auto-motive power.

When I tried to power the dimmer and the power mosfet circuit from the same supply just to see if it was an option it didn't work, the lamp was on full constantly. Why? I am sure it obvious to most of you but I don't see what the problem is.

Thanks again.

This is the output at the drain of the p-channel mos fet . Dimming level is about 40%. The PWM frequency is 20Khz.

• Whoa, what a wall of text, you should make it more readable by splitting it in paragraphs. – Bimpelrekkie Feb 7 '18 at 18:18
• Ok. Will do in future. – gazmono Feb 8 '18 at 9:58
• What prevents you from pressing "edit" and doing it now? – Bimpelrekkie Feb 8 '18 at 11:05
• I tried it but when I hit 'save edits' it didn't. – gazmono Feb 8 '18 at 11:27
• Not gonna read The Great Wall of Text. – Olin Lathrop Feb 8 '18 at 12:09

Halogen bulb are not expected to be dimmable. They are supposed to work at some designed-for temperature. If you go lower, the halogen bulb will more or less burn (reduce lifetime significantly). Also, reducing the voltage will reduce the temperature of the filament. If you go too low, the filament won't emit light.

Apparently (as pointed out by winny) I was wrong and it is possible to dim halogen bulbs without damaging the bulb.

Now, that being said: your circuit kind of works with a slower PWM source. Therefore, you have switching speed issues. We can't debug a circuit from the schematic, especially with missing information (opto-isolator). Try to check which part of your circuit is not switching. If you have access to an oscilloscope, try to see what point is not changing value at 20kHz: (opto-isolator emitter,R2); (collector of M1) or (emitter of M2). If everything is switching, then you have a temperature issue.

If your opto-isolator is like the one pictured here, it would benefit from being driven low-side (put R2 to upper rail instead).

Switching speed of your FETs is given by RC constant formed by say R2 to M1 gate capacitance or R3 to M2 gate capacitance. Gate capacitance (Input capacitance) in datasheet must be fully charge/decharged. To get a good switching, you should have your signal period smaller than 5RC or 10RC (for much cleaner switching).

Therefore, you should have:

FreqRatio = 1/(10*f*R2*Ciss_m1)


Where FreqRatio should be greater than 1 to avoid problems.

FreqRatio_m1 = 1/(10*20kHz*1kOhm*1760pF) = 2.84
FreqRatio_m2  = 1/(10*20kHz*10kOhm*3400pF) = 0.147
FreqRatio_m2slow = 1/(10*400Hz*10kOhm*3400pF) = 7.35


First MOSFET circuit looks fast enough for 20kHz, but we can see that the issue is probably your high-side FET being too slow. It could only work up to 0.147 of the speed you want to go (~2.9kHz). A possible solution would be to reduce your R3 resistor by that amount (being 10kOhm*0.147 = 1.47kOhm; value => 1kOhm would do it with some headroom).

The chinese stuff at 400 Hz works, because the circuit is fast enough. If you have other issues, check your opto-isolator. It might also be slow to respond. There are opto-isolators fast enough, but most cheap/common may not be.

• Thank you so much for your fantastic reply. I will work on it today, I do have an oscilloscope. I forgot to label the opto-isolator, it is an HCPL4504 which I chose because it claimed to be good at high frequencies. – gazmono Feb 8 '18 at 10:07
• That "halogen bulbs are not dimmable" is highly exaggerated. After the first halogen bulbs in the 70's or 80's, the pressure and gas mixture has been modified in serious manufacturers’ bulbs to prevent blackening. Also, if it does go black, you are able to recover the deposited tungsten from the wall by running it at full blast for a few seconds. Please see the excellent site www.lamptech.co.uk for more information and also high speed videos of the halogen process. – winny Feb 8 '18 at 11:53
• @winny All right. I wasn't aware of that. Thank you for the information. I will update my answer accordingly. – Mishyoshi Feb 8 '18 at 15:26
• I have changed the p-channel mosfet to IRF5210PBF which as well as being 100v has an input capacitance of 2780pF which following the formula you gave gives a value of 2.5 so I have high hopes. Thanks for the clear explanation. – gazmono Feb 8 '18 at 19:28

That filter cannot be guaranteed to be suitable - they say (in the comments) that it can't protect a 12 volt television so it doesn't sound like it will protect a MOSFET from too much gate voltage. Gate to source breakdown may occur as low as 20 volts and your circuit doesn't protect the gate in case the 12 volt rises too high. No I don't think the so-called protection filter (with no real guarantees) AND your driver circuit will work for very long.

If the unit has failed then I suspect this is what has happened. Ideally you would protect gates with a series resistor and a zener diode of maybe 15 volts. Under normal operation the zener won't have to do anything other than sit there.

• Comments are not for extended discussion; this conversation has been moved to chat. – Dave Tweed Feb 9 '18 at 20:55