MOSFET protection advice, fan driver needs to be ruggedized

[iso9001]

So, I have a 12V automotive fan, it's three connections are +, -, and a very small guage PWM line. I have no specs for the part so I can only go on observation.

I know it's pulse width driven for the speed. I know that a the PWM line is pulled up inside the fan. I know that it's designed to be pulse-grounded by a micro controller. I know I can do this easily enough with a PIC and a mosfet. I know the voltage will range from 10-15V

What I do not know...

1. The current this thing is designed for. It's a small wire, so I'll assume very small amounts and that there is either a micro inside the fan counting and converting the PWM anyhow, or a hardware circuit that's also driving a mosfet or similar. But I should probably be careful anyhow.

2. How to protect the mosfet on my end. The best I can think of is a .1uF cermaic cap, a TVS diode, and an inline diode to prevent backwards current. I can not look at the original controller's circuit, so I'll just have to do my best to guess.

Is there anything else that I can use to protect this circuit from all of the unknowns? I'm concerned that the mosfet could be damaged, so I'm considering using a dual fet and tying them together, and including basic protections. I'm not really up on isolation or similar techniques.

Comments, schematics, suggested alternate parts, recommendations wanted. Thanks!

 

[kubeek]

normally the pwm wire is for sensing the rpm of the fan, not for setting the fan power.

 

[iso9001]

normally the pwm wire is for sensing the rpm of the fan, not for setting the fan power.

Well... Then this isn't "normal" is it? This is automotive. Take what you think about tiny computer fans and toss it out the window

For real though, it's definitely a driver. The positive and negative leads to the fan go directly to the battery (fused). I've watching this line on the scope, I've played around with the timing, even prototyped it. I just want to be as sure as I can that I'm doing enough to protect my circuit.

 

[alec_t]

Is there anything else that I can use to protect this circuit from all of the unknowns?

A snubber across the fan supply terminals? A freewheel diode (anode to fet drain, cathode to V+)? A fet gate resistor (10Ω to 100Ω) to reduce oscillation tendencies?

 

[iso9001]

A snubber across the fan supply terminals? A freewheel diode (anode to fet drain, cathode to V+)? A fet gate resistor (10Ω to 100Ω) to reduce oscillation tendencies?

I'm not touching the wiring of the fan supply. That's factory and staying so. But as it were, not sure what a snubber is.

Freewheel diode, you mean for.... reverse bias protection? I see how you mean to hook that up, just not what it's doing. And of course I'll use a gate resistor, and a small pull down to keep it from accidentally activating.

I was just thinking like, maybe a specific TVS, or a choke or... I don't know... I'm not sure what would be good for "circuit unknowns".

 

[alec_t]

A snubber is typically a 0.1uF cap in series with ~ 100Ω, to provide a bypass path to absorb voltage spikes.
The freewheel diode clamps any inductively-induced back emf to the 12V supply, so the fet won't be killed by voltage spikes.

I'm not sure what would be good for "circuit unknowns".

The more belts and braces the better. Too much protection is better than not enough!

 

[iso9001]

Ok.... so... Here is what I'm thinking. I need to make the PCB before I have all the time to test everything. I need to order PCBs today or tomorrow. I don't have alot of room on my component side of the PCB, but I could put extra parts/pads on the b-side if really needed. I'll probably over-do-it for extras as a just in case.

So I'm looking at the freewheeling diode for sure, possibly a TVS, but those might be doing similar things. Then I'm not exactly sure what kind of smoothing the original driver might have been doing or if that circuitry is inside the fan itself, so I was considering at least adding a spot for two 1206 sized parts, maybe a .1-1uF ceramic and a 2-4uF tantalum capacitor or whatever value ends up being desired.

**broken link removed**

The FET gate pulldown should ensure it never floats regardless of microcontroller state, and the gate to micro resistor further protects the gate right? None of this is exactly my day job, well, it sort of is I guess, I'm just bad at it


edit: One thing I just saw that I'm curious about, isn't that freewheeling / reverse diode going to have a constant forward leakage into ground? This circuit needs to be as power efficient as possible, so I'm just curious about that.

 

[kubeek]

I think that pulldown is too large value, 10k would be more reasonable.
About the diode, one side you say this is a large automotive fan, and on the other side you are worried about some tiny leakage ini a diode. I guess it would be just a few uA so that should be no problem.

 

[iso9001]

I think that pulldown is too large value, 10k would be more reasonable.
About the diode, one side you say this is a large automotive fan, and on the other side you are worried about some tiny leakage ini a diode. I guess it would be just a few uA so that should be no problem.

Yes, it is automotive. I wasn't aware the diode would be only a few uA, I seem to remember someone telling me 1.6V which I know those two things could be true simultaneously I was thinking it would be a larger amount of current. This module must remain under 1mA in standby, and it's already pushing that considering all the other things going on. Real "sleep" isn't possible because of the bus comms so "idle" is as low as I can go for now, so nothing that's going to open up even 100uA of current draw is ideal.

Good point on the pulldown. I'll adjust that to 10k.

 

[kubeek]

When the output of the micor is low, then the fan should be off, right? Then the diode going to the +12V rail will not conduct since the pullup will be at the same potential. The only thing might be the TVS, as fully charged battery is close to 14.4V if i remember correctly, add to this some rise when charging and I would rather up this to 18V.
Another thing is the input to the mosfet, is it 5V or 12V? Some TVS on this couldn´t hurt too.
Also those caps seem too large, no idea what the pullup value is but I think that 0.1uF is about the max you´d want there.

 

[iso9001]

When the output of the micor is low, then the fan should be off, right? Then the diode going to the +12V rail will not conduct since the pullup will be at the same potential. The only thing might be the TVS, as fully charged battery is close to 14.4V if i remember correctly, add to this some rise when charging and I would rather up this to 18V.
Another thing is the input to the mosfet, is it 5V or 12V? Some TVS on this couldn´t hurt too.
Also those caps seem too large, no idea what the pullup value is but I think that 0.1uF is about the max you´d want there.

Yes and No on the first part. When the micro is low the mosfet's gate will be closed/off/unpowered, this will mean the fan is pulling the pwm line high but not spinning. Likewise, when the micro is high, gate is engaged, pwm line is low fan is still not engaged. Only when the line is cycling high/low at the specific frequency does the fan actually run. It's not clear to me how it works, but only a steady duty cycle actually engages the fan. I don't believe there is a micro in the fan, but there could be.

Good point on the TVS being at 18V.

The mosfet will see 5V at the gate, 12V-14V+ at the drain/source. You mean a TVS on the gate? I already have plans for a TVS at the drain.

I'm not clean on the caps either. The frequency isn't that high, 12-60hz or somewhere around there. 12V signal, I think it's probably a 220 to 2.2k pullup as that's the common value in current automotive for pullups. I was just going to leave pads open for two 1206 sized parts to ground. That way if I needed to I could put another TVS or Zener or caps on the line. I'll start with a .1uF and see how the signal looks.

I think I have enough to go on for this, I was just kind of expecting someone to recommend a zener or a choke or something, or reduncancy. On that note.... Is there any merit to instead of using one mosfet, I use two driven off of the same gate? If one goes bad ever the fan still runs on the other? Or waste of time/money?

 

[kubeek]

There is no reason to use two parallel mosfets, they typically fail shorted so that wouldn't help. Better use some beefier device like 100V just to be sure. Also note that the mosfet needs to be logic-level one, so that it can be activated with just 5V on the gate - for example IRFL110. Also I would put some 5.6V zener on between gate and source just to be sure nothing goes back to the controller.

 

[iso9001]

Also note that the mosfet needs to be logic-level one, so that it can be activated with just 5V on the gate - for example IRFL110. Also I would put some 5.6V zener on between gate and source just to be sure nothing goes back to the controller.

I thought practically all mosfets would trigger at 5V because only a very tiny amount of electrons would be required to trigger the gate... But alright. Duly noted. Good idea on the 110V although I'll look around at a cost/size compromise if I must.

On the Zener... I was not aware anything could "backfeed" through a mosfet... Interesting.... Wait, between the gate and the source? Like this? I'm definitely not following what this would do.

**broken link removed**

 

[kubeek]

Yes, like that. The mosfet has some small capacitance between drain and gate in the order of 100pF, but with fast enough voltage burst on the drain it is possible for it to backfeed to the gate and then to the micro.

 

[alec_t]

I'm not clear why you have the caps on the PWM input. If the fan internal electronics are expecting a PWM rectangular-wave signal then the caps will round off the waveform to some extent, so pulse edges will be less easily recognised by the electronics.

 

[iso9001]

I'm not clear why you have the caps on the PWM input. If the fan internal electronics are expecting a PWM rectangular-wave signal then the caps will round off the waveform to some extent, so pulse edges will be less easily recognised by the electronics.

Because I'm uncertain what is inside the fan. If I hook this up on my board with this hardware, if I scope it and it looks peak-y, I might need some smoothing. I don't plan to populate the caps at this time, just adding them to the schematic so I have a place for them if needed later. They're currently optional at best.