Confused - electrolytic vented? (and LC filters)

[Things]

So I have a MCU controlling a few blower fans using PWM, and I installed a low pass filter to try and quieten them down a bit. This seemed to work fine for a while, but I went to fire it up today and after around a minute at full power, I heard the fans start slowing down and heard a hissing noise. Immediately I knew it was the cap

This is the circuit here:



The circle is the blower fans (drawing about 5A at max power), the cap is rated at either 16 or 25V (haven't pulled it apart to re-check yet), and it's running off 12V.

PWM frequency is only 490hz.

Anyone have any idea why it could have failed? Perhaps the inrush on each PWM pulse was too much for it? Was only a cheap cap so maybe I just got a bad one - but it hasn't been running for more than maybe 6 hours total.

[German_EE]

Two guesses:

1) Electrolytic capacitors might not like being fed a 490 Hz signal at power levels.

2) If the fan is just a motor without an internal controller there may be large voltage spikes across the terminals. Try fitting a reverse diode just as you would with a relay.

[Things]

They are DC brushless blowers, so they do have internal controllers. I have the same thing driving a normal brushed DC fan as well, but that cap seems to be fine.

Thought 490hz would have been pretty low and safe for an electrolytic .. Guess I'll try swap the cap out and see what happens, could have just been a bad one too.

[wraper]

They are DC brushless blowers, so they do have internal controllers. I have the same thing driving a normal brushed DC fan as well, but that cap seems to be fine.

Thought 490hz would have been pretty low and safe for an electrolytic .. if anything I thought higher would be even less stress? Guess I'll try swap the cap out and see what happens, could have just been a bad one too.

Actually maximum ripple current is lower at lower frequencies. What you are doing here is intentionally killing capacitor in this schematic. I bet power supply is not happy at all about it too. You basically make almost a short circuit when mosfet opens. PWM cannot be done in a such a way. You should either do not use a capacitor at all if the fan can handle raw PWM or make a controlled DC/DC converter (needs coil).

[tggzzz]

Here's a naive analysis that may or may not offer some pointers...

Assume the capacitor has zero ESR, and the motor is "sufficiently" inductive for the effects hypothesised below to have some validity.

Start with the mosfet non-conducting, then turn it on. Initially there will be zero current through the inductor (motor), and the current through the capacitor will be limited by the 250m? resistor to 12/0.25 = 48A.

So, does the capacitor have sufficient ESR to keep the inrush current within its stated max current rating?

After that, the motor's inductance will ring with the capacitance, producing voltages that might exceed the capacitor's rating (in either direction).

To reduce those effects I'd insert a resistor in series with the capacitor and a resistor+reverse biassed diode in parallel with the motor. (The resistor reduces the duration of a current spike through the diode at the expence of increased reverse voltage spike).

I'm also curious why you chose a 4700uF capacitor - it seems rather high and may well have too much internal inductance to effectively suppress RF. 4700uF across the power supply seems more reasonable.

[Things]

Well, I have a feeling the inductive kickback from the motors won't be great, as they have their own internal controllers that should be copping most of it. Just the current spikes on the PWM pulses I feel could have been too much for it - it's a cheap cap, nothing fancy. Maybe if I use a bunch of smaller caps in parallel I could share the stress across more of them?

Without the filter the fans are just far too noisy, as in it'd probably drown out most laptop speakers on full volume even, and DC/DC converters aren't really an option cost-wise. I'll have to stick to something like this, just somehow reduce the stress on the capacitor(s).

Would using more, smaller caps in parallel help you think?

[trevwhite]

Maybe just put a small resistor in series with the capacitor and maybe a resistor across the fan? See what it looks like on a scope?

[tggzzz]

Well, I have a feeling

An ounce of measurement is worth a pound of speculation!

the inductive kickback from the motors won't be great, as they have their own internal controllers

What do these "controllers" do? Is your PWM compromising their ability to control?

that should be copping most of it. Just the current spikes on the PWM pulses I feel could have been too much for it - it's a cheap cap, nothing fancy. Maybe if I use a bunch of smaller caps in parallel I could share the stress across more of them?

Capacitor tolerance => unequal sharing.

Without the filter the fans are just far too noisy, as in it'd probably drown out most laptop speakers on full volume even,

Drown out acoustically or electrically? If electrically, what is the conduction path between the high currents and sensitive electronics?

and DC/DC converters aren't really an option cost-wise. I'll have to stick to something like this, just somehow reduce the stress on the capacitor(s).

Would using more, smaller caps in parallel help you think?

Possibly, but I think you need to do some research into noise suppression. In particular, look at ferrites, common mode inductors, appropriate wiring to keep high currents away from sensitive electronics. Googling will find many many application notes and guides; I don't have a favourite.

[Things]

What do these "controllers" do? Is your PWM compromising their ability to control?

Just plain ol brushless motor controllers like you'd find in a computer fan - unfortunately I don't think they have a PWM input on them directly (No wire at least, but I didn't see any unconnected pads either).

I've run them PWM before and the controllers seemed OK with it - but using the low pass filters they're basically silent once they've actually started spinning.

Drown out acoustically or electrically? If electrically, what is the conduction path between the high currents and sensitive electronics?

Acoustically. They're mounted in a large wooden box so the entire thing just acts as a speaker - and at 490hz, it's more than audible!

I have tried above-audible PWM frequencies, but neither of the fans seemed to work at all that high.


I've stuck a 35V 2200uF cap in it just for now - even at half the capacitance they're still nice and quiet, so you were correct in regards to such a high value being unnecessary. Hopefully maybe that'll help prolong the life a bit too. They're mounted above a large fan so they have ample cooling, just found it odd this one decided to fail so early.

[bobcat]

You don't necessarily have to change the PWM frequency to above audible frequencies to quiet a motor. Try varying the PWM until you find the sweet spot that reduces the noise.

[mikerj]

I have tried above-audible PWM frequencies, but neither of the fans seemed to work at all that high.

Raise the frequency, replace the resistor with a suitable inductor and add a flyback diode from the MOSFET Drain to the supply rail.  This will effectively give you a buck converter.

[wraper]

Without the filter the fans are just far too noisy, as in it'd probably drown out most laptop speakers on full volume even, and DC/DC converters aren't really an option cost-wise. I'll have to stick to something like this, just somehow reduce the stress on the capacitor(s).

Would using more, smaller caps in parallel help you think?

Nonsense. Total BOM will cost maybe slightly above your current part cost. You cannot just add capacitors in parallel. You basically shorting your PSU. Increasing resistor value also is a no way as your PWM will become a linear regulator (Edit: in terms of heat dissipation) with all power dissipated on a resistor. Well, you can add more capacitors, but then your PSU will die with blown capacitors anyway.
Edit:
Your circuit is able to work somehow only because you are sagging your PSU voltage (and killing it slowly at the same time) when mosfet is opening. If your PSU would be able to supply infinite current at given voltage, your circuit would blow up almost instantly. You cannot PWM from DC to DC (on capacitor and fan) without a coil. Well you actually can if using a high value resistor, if the load current consumption is very low. In such a case it would be a low pass RC filter. But heat dissipation on the resistor will be the same as if using linear regulator. Therefore you only can do PWM without an inductor only if: 1. load is completely resistive, like a bulb or some kind of heater. 2. load is inductive by itself like a brushed motor, stepper motor.

[DanielS]

My guess on the capacitor would be excessive AC current ripple.

If you buy random "low-ESR" caps, you often end up with something rated at 40mOhms and ~1A RMS. If you shop for low-ESR caps like you really mean it, you can find caps rated at 8-15mOhms and 3-5A RMS. Even those 5A-rated caps might have a hard time with hard ripple across them in a high-power RC filter.

As wraper said, you want to use a DLC filter - the same thing as a buck converter - so the inductor and capacitor can split the current feed between them. At 500Hz though, you would need a huge inductor so you might want to consider options for raising that to 10kHz or more.

[Hamster]

What is the goal here? Control fan speed? or Reduce the Current draw when you turn it on? You should not feed a Brushless controller with a PWM Signal unless its designed to have one, i would start off at a lower voltage then ramp up to 12v to reduce inrush current.

[Things]

The goal is to quietly control fan speed.

Thing is though, I thought provided I chose a good capacitor value, the capacitor shouldn't fully discharge between PWM off times - otherwise it'd be pretty useless as a filter. Therefore the load on the FET's/PSU is only a fraction of what it'd be if it had to completely re-charge the cap each time. Looking at the voltage across the cap with my scope, the ripple is only around 1V ?

I suppose I could use an inductor LC filter if it's only one addition, but I have no idea how to calculate it

[SirNick]

Looks like you have a circuit sim right there...

[wraper]

The goal is to quietly control fan speed.

Thing is though, I thought provided I chose a good capacitor value, the capacitor shouldn't fully discharge between PWM off times - otherwise it'd be pretty useless as a filter. Therefore the load on the FET's/PSU is only a fraction of what it'd be if it had to completely re-charge the cap each time. Looking at the voltage across the cap with my scope, the ripple is only around 1V ?

Problem is not about discharging at all. It is all about charging current which is virtually unlimited. And how big is a ripple voltage on the PSU?

[wraper]

Simplest you can do. Component values are from a thin air but something thereabouts should work at 20-60 kHz. As inductor must be pretty beefy for 5A, it might be cheaper to get a core and wind it yourself. Also as PWM frequency is pretty high, you must be sure that MOSFET gate is charged fast enough if driven by MCU output directly. It may be not enough and additional driving circuit must be added in a such case. Diode must be fast and preferably Schottky to reduce heat dissipation in it.

[TMM]

The electro popped because it's having to do too much work. Just because it's not fully discharging doesn't mean it's not charging/discharging substantial currents.

Increase the PWM frequency so you can use a smaller plastic film capacitor and inductor.

Circuit as in wraper's post, 50KHz pwm, 100uH inductor, 10uF plastic film cap.

This may be of use in regards to winding your own inductor:
http://www.diyaudio.com/forums/class-d/89018-big-t-6.html#post1393703

[wraper]

Circuit as in wraper's post, 50KHz pwm, 100uH inductor, 10uF plastic film cap.

10uF film capacitor for 5A current? Maybe it will survive, but certainly won't do the job well. And will be more expensive than electrolytic one. 2200uF is not because of the capacitance as much smaller is needed but because of the ripple current it can hold, LOW ESR one is preferable.

[TMM]

The ripple current through the cap is much less than 5A due to the inductor. Assuming a 2ohm load, i'm getting 0.17Arms through the capacitor at 50% duty cycle, 50KHz switching.

Film cap will last for pretty much forever in that application. A 10uF electro will not, even if 0.17Arms ripple current is within it's specification. A 1000uF+ cap will last a lot longer, but then transient response of the circuit is terrible (probably not a huge issue for a couple of fans..)
 
You could get away with as low as 2.2uF with acceptable voltage ripple on the output.

[wraper]

The ripple current through the cap is much less than 5A due to the inductor. Assuming a 2ohm load, i'm getting 0.17Arms through the capacitor at 50% duty cycle, 50KHz switching.

Film cap will last for pretty much forever in that application. A 10uF electro will not, even if 0.17Arms ripple current is within it's specification. A 1000uF+ cap will last a lot longer, but then transient response of the circuit is terrible (probably not a huge issue for a couple of fans..)
 
You could get away with as low as 2.2uF with acceptable voltage ripple on the output.

And what output voltage ripple did you calculate with 10uF cap and with 2.2uF one?

[Things]

Hmm, assuming I'm not able to change the PWM frequency due to other reasons, would an inductor from say, a computer power supply be suitable? Like the large ones they have on the output themselves?

[TMM]

The ripple current through the cap is much less than 5A due to the inductor. Assuming a 2ohm load, i'm getting 0.17Arms through the capacitor at 50% duty cycle, 50KHz switching.

Film cap will last for pretty much forever in that application. A 10uF electro will not, even if 0.17Arms ripple current is within it's specification. A 1000uF+ cap will last a lot longer, but then transient response of the circuit is terrible (probably not a huge issue for a couple of fans..)
 
You could get away with as low as 2.2uF with acceptable voltage ripple on the output.

And what output voltage ripple did you calculate with 10uF cap and with 2.2uF one?

150mVp-p for 10uF
600mVp-p for 2.2uF

Good enough for some fans and better than what the OP was getting with their original circuit.

Hmm, assuming I'm not able to change the PWM frequency due to other reasons, would an inductor from say, a computer power supply be suitable? Like the large ones they have on the output themselves?

The toroidal ones with maybe 20 turns of heavy gauge wiring? not even close for 490Hz.

At 490Hz the inductor needs to be increased to something around 10mH and the capacitor 1000uF.

What might be of use is if you find a powersupply that has passive power factor correction and take it's PFC inductor. You know it's the right inductor when it looks like a transformer about 5x5x5cm but only has two wires going to it. I have one out of an old PC powersupply and it is about 50mH but the DC resistance of the windings are around 1.3ohm which is going to cause a fair bit of loss when your load is around 2ohm. If you found one of those and unwound it to around 10mH the loss caused by the DCR might be small enough to be acceptable.

[Things]

Hmm, damn, was just looking at the exact inductor you mentioned (has about 22 turns). There is another larger one there with a couple of windings it looks like.

OK - The PWM frequency options I have are: 62.5Khz, 31.250Khz, 7.812Khz, 3.910Khz, then a few at even lower frequencies.

When I tried running the fan on the 2 highest frequencies directly, they didn't do anything at all, so I was a bit concerned about using them. However if the filter is going to be smaller/easier/more efficient at higher frequencies (and give me a smoother output), then that's great.

I've tried looking online for low pass LC filter calculators, but there seems to be many different configurations, I can't really find any with just the single inductor, cap and diode. Does anyone have any links perhaps?

I'm just a bit confused as to what I'm actually aiming for here - in terms of ripple attenuation values etc.

I've basically never used inductors in my circuits before, so I was a bit hesistant about the LC filter, but seems like what I'm going to have to go with. Just not sure what component values I'd need to go with assuming those higher frequencies? Are inductors rated purely by Irms? So in my case, I'd aim for something around 6A? Would 62Khz and 32Khz work better with ferrite or powdered iron inductors etc?

Cheers