Confused - electrolytic vented? (and LC filters)

[wraper]

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

How did you calculate that? If using Ti tool http://www.ti.com/tool/buck-convcalc I get 2x higher ripple with such capacitance. I think there is very good reason why film capacitors are not used for buck converters, size and price. Why to use big very expensive capacitors if you will get worse performance. I would at least use ceramic or tantalum capacitors if completely do not want to use electrolytic ones.

[wraper]

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?

That inductor has iron powder core with very low permeability. Well, you can use it but need to wind much more turns compared to the ferrite one.

[Things]

The main thing stopping me from trying it is I don't have any diodes really beyond some basic rectifiers, nothing even close to fast enough to up the PWM frequency.

But then if I'm gonna buy some diodes, I may as well just buy the proper inductors and caps while I'm at it. I just can't seem to find any info on what values I should be looking at

Basically, I could use either 62Khz or 32Khz PWM, 12V, would need to handle ~6A constant, and as for ripple, probably even 1V would be more than quiet enough, but if it's as simple as changing the capacitor or inductor value slightly for better ripple, then that's great I just don't know where to start. 10uH? 100uH? 1000uH? 1H?? 10uF? 100uF? 1000uF?

[wraper]

Use 62KHz, can use values given with it. Industance can be reduced to about 47 uH and even lower if you are not going to use it with a load lower than 1A. As of capacitor, 1000uF 25V low esr would do well, or use 2x of 1000uf for general purpose to be safe. If you have computer psu, then you can get schottky diode from it. You can use inductor like this: http://lv.farnell.com/bourns/2309-v-rc/inductor-toroid-v-47uh-15-10-7a/dp/1929759

[Things]

Looks pretty similar to this one I got out of the computer power supply, do you think it'd work too?

[wraper]

Looks pretty similar to this one I got out of the computer power supply, do you think it'd work too?

This one is too small. Only if rewinding biggest inductor.

[Things]

OK, will see what I can do, thanks for your help

[wraper]

Actually it might work, but then you would need to know how to measure if it is doing well.

[Things]

I do have a crappy pocket scope, so I can probably at least do comparisons using various inductor + cap combinations. I'll have to mess around with it in the morning.

[SeanB]

It would work, but may run warm. Try it, worst case it heats the core up till it changes colour.

[Things]

Just tried it at 3.9Khz with that inductor and a 1000uF cap - didn't work all that well. Will try a higher freq tomorrow. I also noticed the FET heated up a lot faster too 

[DanielS]

Just tried it at 3.9Khz with that inductor and a 1000uF cap - didn't work all that well. Will try a higher freq tomorrow. I also noticed the FET heated up a lot faster too 

There are three main reasons for it to heat up more than it did before bumping the frequency up:
1- switching too slowly (dissipates more power during rise/fall since there are more transitions)
2- not switching correctly (ex.: too high impedance gate drive causes the gate voltage to oscillate or do other stuff due to C_DG and Miller effect)
3- slow diode causes voltage across the inductor to rise and trigger avalanche operation (the FET starts conducting even with V_GS near 0V)

[mikerj]

Just tried it at 3.9Khz with that inductor and a 1000uF cap - didn't work all that well. Will try a higher freq tomorrow. I also noticed the FET heated up a lot faster too 

Just to check, have you definitely fitted the required flyback diode?  If not then you can expect bad things to happen to the FET.

[Things]

Forum is back finally!

So I've been experimenting and confusing myself with this a lot .. and I have a working solution now (almost)

Just using the random inductor I posted a pic of earlier, a 1000uF cap, and a FR104 diode out of the same PSU, running at 32Khz, everything is dead silent, which is great .. best of all it works, and nothing is heating up excessively.

Measuring directly across the legs of the inductor, I seem to get about 70Vpp, however measuring across the FET between drain and source, I get about 20Vpp, which seems acceptable. The FET isn't heating up either, so it must be OK with it.

However, now I seem to have an EMI issue with the microcontroller This may have been an issue even with the RC filter - but I assume it's because I upped the PWM frequency to 32Khz, so it's just spewing out RF everywhere. The inductor probably isn't helping either.

What would be some ways I could try cut down on this EMI a bit? Ferrite bead on the wires running between the FET and filter perhaps?

I'm glad the LC filter works, and works well

[TMM]

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

How did you calculate that? If using Ti tool http://www.ti.com/tool/buck-convcalc I get 2x higher ripple with such capacitance. I think there is very good reason why film capacitors are not used for buck converters, size and price. Why to use big very expensive capacitors if you will get worse performance. I would at least use ceramic or tantalum capacitors if completely do not want to use electrolytic ones.

LTspice, ideal voltage source to generate the PWM and ideal inductor/capacitor. I see that TI's spreadsheet accounts for parasitics, that should explain the worse ripple.

I suggested a plastic film because I figure the average hobbiest would most likely already have through hole electros and polyester caps on hand to prototype, or at least easily available to them. If it is being optimised for mass production, absolutely consider using MLCCs and tants instead.

[mikerj]

Measuring directly across the legs of the inductor, I seem to get about 70Vpp, however measuring across the FET between drain and source, I get about 20Vpp, which seems acceptable. The FET isn't heating up either, so it must be OK with it.

The peak-peak inductor voltage should not be greater than the supply voltage for a buck configuration.  Could you post a schematic to show how your components are connected, as it sounds like something is not wired correctly?

[Things]

I also tried it like this, but didn't seem to change much (if anything)



Waveform across source-drain on the FET:



It could also possibly be the scope, but I'm not sure.

[mikerj]

That looks ok, but the high voltages suggest your diode may not be doing anything (maybe open circuit?).

I'd personally go for the second configuration as it means you won't have high frequency switching voltages on one of your fan wires which could cause EMI problems.

[Things]

Is the FR104 diode even suitable? I just randomly ripped it out of the computer power supply from near the inductors. They're fast recovery and rated at 400V 1A, but I'm not sure exactly what specs I need. The diode definitely makes a difference though - stops the power supply sounding angry and the FET runs a lot cooler

[Things]

So I've ripped 2 similar inductors from computer power supplies, and wired them up with 2 of the FR104 diodes in parallel as they seemed to be getting quite hot at higher powers.

One of the inductors seems to be getting pretty warm though, even at "low" power. Not entirely sure how much current it's handling, but the wire should be more than sufficient for 6A.

Is there anything besides excessive current that can cause an inductor to heat? Or do I simply need a beefier inductor?

Edit: Just measured using my DMM (so it's probably only getting an average), but it's heating even at only 1A .. seems odd.

[mikerj]

Is the FR104 diode even suitable? I just randomly ripped it out of the computer power supply from near the inductors. They're fast recovery and rated at 400V 1A, but I'm not sure exactly what specs I need. The diode definitely makes a difference though - stops the power supply sounding angry and the FET runs a lot cooler

If you are trying to supply up to 5 amps then a 1 Amp diode isn't really going to be up to the job  

Check all the TO220 packages in your PSU, you might find some of them are high current fast rectifiers even if they have three legs.

[Things]

I found a nice beefy dual diode, though it's maximum repetitive reverse voltage is only 40V, not sure that'd be high enough considering the ~80Vpp I measured across the inductor ... Though it is good for up to 250A surge

[wraper]

I found a nice beefy dual diode, though it's maximum repetitive reverse voltage is only 40V, not sure that'd be high enough considering the ~80Vpp I measured across the inductor ... Though it is good for up to 250A surge

Should work. That voltage you are seeing with your oscilloscope likely is not true at all. Usually you will see very high and short voltage spikes which actually are not there.  Anyway voltage on the mosfet drain actually cannot go below zero as there is another diode in the mosfet itself. As of fighting the noise, add another capacitor between +12V and GND as close as possible to the schematic.

[wraper]

This is not good at all although will work. Reason is that your fan power rails will be floating with PWM frequency. Therefore you are making radio transmitter  .

[DanielS]

Is the FR104 diode even suitable?

1.3V forward loss and only 1A rating, not really. For a low-voltage switching supply, you want to use a Schottky diode which has no reverse recovery time and a much lower 0.3-0.6V forward drop. You also want it to be rated for at least half your load current since it will be seeing full load current for part of the time where the FET is off.

This will make your circuit even more efficient and clean up some of the ringing and spiking you might be seeing with fast diodes.