pwm fan controller - mosfet sizing help needed

[heychris]

I've been playing with low cost options of pwm control of a fan based on temperature.  I've tried the MCU based solutions, but wanted to try out the maxim 6643.  The maxim 6643 is purpose build to control 2 and 3 wire fans.  The only additional components are a couple resistors and a mosfet.  It has a built in sensor as well as capability of measuring external sensors.

http://www.maxim-ic.com/datasheet/index.mvp/id/4317

I wanted to try this out with all though hole components.  So I tried to get as close as possible duplicating the eval circuit diagram.  When I breadboarded it up, I get full speed for the startup sequence then it drops to zero then stops.

I'm thinking I didn't translate the mosfet correctly and was hoping to get some help.  The eval kit used a Fairchild NDS351AN.  The one I tried was a onsemi NTD4960N.

http://www.fairchildsemi.com/ds/ND/NDS351AN.pdf
http://www.onsemi.com/pub_link/Collateral/NTD4960N.PDF

My method was to use digikey to search.  I then filtered as many things as I could. 
1. Through hole
2. 30V
3. MOSFET N-Channel, Metal Oxide

Is there something else important that I missed?

My other question is about the controller supply voltage.  I used a bench power meter, but would like to figure out how to supply 3.0V to 5.5V volts at  500µA.  Should I just use a resistor for a voltage divider or some other way?

[TerminalJack505]

My other question is about the controller supply voltage.  I used a bench power meter, but would like to figure out how to supply 3.0V to 5.5V volts at  500µA.  Should I just use a resistor for a voltage divider or some other way?

This question makes me wonder if you are limiting the current to the controller.  The datasheet says the part will operate from 3.0V to 5.5V and typically consume 500uA.  You don't want to limit the current.  Let it use what it needs.  Just power your circuit at anywhere between 3 and 5.5 volts and don't use the constant current feature of your power supply.

So far as the MOSFETs go, they look similar enough that the ones you substituted should work.  The only parameter that I see that could be marginal is the worse-case Vgs(th) for your substitute.  It is 3V and the other part's is 2.5V.  This might be a problem if you're powering the controller at the low end of the supply range.  Since you're just testing, use a 5V supply to avoid any issues.

[heychris]

The current limit is at it's maximum current which is 5 amps.  I've tried to use a few capacitors in case it could have to do with noise.  That doesn't seem to help either.

[TerminalJack505]

Are your fans 12 VDC?  If so then you need to power the fans from 12V and not 5V.  Most 12V fans wont even spin at 5V or less.  In other words, you need two supplies: 5V for the controller and 12V for the fans.

If you can post a schematic it will help a lot.

[heychris]

Thanks..  I much appreciate the help.  This has been stumping me. 

My two power supplies are 12V and 5V.   I have an old laptop power brick that connects to an ebay DC-DC step down with volt and amp meter for 12V.  The 5V is from a DC-DC step down that only controls Voltage, no current limit.

My first go around, I followed the eval kit's circuit diagram.  I completely rebuild everything for a second try with the same result.

The controller has a full speed setting.  When I trigger that, it still acts the same way.

[TerminalJack505]

Hmm...  I don't know.  I don't see any problems with your circuit.  Maybe the minimum duty cycle isn't enough to keep the fan spinning.  Have you looked at the PWM_OUT pin with a scope to see if there's a pwm signal present?  Have you tried heating the chip up to see if the fan will kick in?

I'm not sure why the FULLSPEED# pin wouldn't work.  From what I can tell, if you ground that pin then the duty cycle should be set to 100%.

[electroguy]

I wanted to try this out with all though hole components.  So I tried to get as close as possible duplicating the eval circuit diagram.  When I breadboarded it up, I get full speed for the startup sequence then it drops to zero then stops.

I'm thinking I didn't translate the mosfet correctly and was hoping to get some help.  The eval kit used a Fairchild NDS351AN.  The one I tried was a onsemi NTD4960N.

http://www.fairchildsemi.com/ds/ND/NDS351AN.pdf
http://www.onsemi.com/pub_link/Collateral/NTD4960N.PDF

My method was to use digikey to search.  I then filtered as many things as I could. 
1. Through hole
2. 30V
3. MOSFET N-Channel, Metal Oxide

Is there something else important that I missed?

both fet's should work, the NTD4960N is obviously much more powerful. Are you sure you connected them correctly? or is that what you are asking? according to the schematic you posted, D to the fan, S to GND, and G to the IC.

My other question is about the controller supply voltage.  I used a bench power meter, but would like to figure out how to supply 3.0V to 5.5V volts at  500µA.  Should I just use a resistor for a voltage divider or some other way?

If price isn't a problem, you could use a 7805 (or 78L05 or any other 5V regulator) and a 100uf 16v capacitor, that way at least you get a regulated 5v output to your maxim IC. Using resistors is a really bad way because if you ever draw more or less than 500ua, your voltage will change (you could use a resistor and 5.1v zener, but it's still not the best way).

[heychris]

The only thing different from the example is the mosfet.  The onsemi NTD4960N mentions it's recommended for high side.  When the pwm duty should be 100%.  I checked the voltage between the 5V and the pwm output.  It is 12V.  Does that sound right?


In general  is there recommendations on picking a Mosfet?

I don't have a scope, maybe I should start looking for one.

I used a heat gun to raise the temperature.  It didn't change anything.

[heychris]

both fet's should work, the NTD4960N is obviously much more powerful. Are you sure you connected them correctly? or is that what you are asking? according to the schematic you posted, D to the fan, S to GND, and G to the IC.

It is connected properly.  I'm only concerned that the ntd4906n is too powerful.  The datasheet for the NTD4906N mentions high side is recommended.

If price isn't a problem, you could use a 7805 (or 78L05 or any other 5V regulator) and a 100uf 16v capacitor, that way at least you get a regulated 5v output to your maxim IC. Using resistors is a really bad way because if you ever draw more or less than 500ua, your voltage will change (you could use a resistor and 5.1v zener, but it's still not the best way).

I was thinking the same thing.  It's $.40 or so, but I tend to not cut corners.

[electroguy]

if you are using a normal computer type fan (12v, approx 200ma or there abouts...) then the NTD4906N is overkill. It can handle 55A.
Either way, it will still work, so your fault is elsewhere, not with your choice of mosfet, the mosfet is fine. Even your other mosfet you mentioned, NDS351AN will work fine.

If you draw up exactly every component you have and how it is hooked up maybe we can help you further... Don't rule out the possibility of the IC being faulty/fried, have you got a spare?

[electroguy]

I checked the voltage between the 5V and the pwm output.  It is 12V.  Does that sound right?

no, something is wrong. Check these things:

• the mosfet connected wrong
• the mosfet is damaged
• the chip is damaged
• or you don't have common ground

Are you sure the ground from the 5V is connected to the ground from your 12V with a wire? I know it is going through your 12v-5v dc-dc, but that doesn't always mean it is connected depending on the converter...

[heychris]

I've got several extra components.  I've tested and replaced them all at least once.  The fans are standard 12V.  I've tried several confirmed good.

Good point on the ground.  I had measured the resistance to verify the connection, but it's better directly connected.

I tried to stay as close as possible to the reference circuit.  I really appreciate the help.  I hand drew the diagram with my changes in the hopes that it sheds some light on the problem.

[TerminalJack505]

So the fan does come on at power-up?  If it does then that tells me it's likely wired-up correctly.

Does the fan come on if you just pull the MOSFET's gate up to 5V through that 4.7K resistor?  (Disconnect it from the IC if you try this.)  It should, obviously.  If doesn't then that would tell you something.

I do notice that their example schematic doesn't include a flyback diode on the fan.  It might not be a bad idea to include one since the fan's an inductive load.  I doubt it is the cause of the problem, though. 

Just put a diode on the breadboard between the 12V and GND pins of the fan.  Put the diode's anode on the GND pin and the cathode on the 12V pin.  Use a fast recovery diode.  (Pretty much anything except a rectifying diode.)  Something like a 1N914 or 1N4148.  An ultra fast recovery diode would be even better.

[heychris]

The fan does come on.  It goes through the startup which includes fan at 100%.   
 
I just realized i could disable failed fan detection which comes from the fan tach.  I also have a usb scope i could use to test pwm output.  Its not ideal, but better than nothing.

Good call on the diode.  I'll try that too.

[electroguy]

Does the fan come on if you just pull the MOSFET's gate up to 5V through that 4.7K resistor?  (Disconnect it from the IC if you try this.)  It should, obviously.  If doesn't then that would tell you something.

Great idea, try this. What do you get? you should be able to turn the fan on and off, by connecting the gate to 5v (fan on) and gnd (fan off). Let us know if this works.

[heychris]

I disabled the fan fail detection.  It's interesting the effect.

It goes through the initial startup.  Then drops down to what my fluke 73 reports as 4.78 volts across the fan + and -.    With fan fail enabled, it dropped to .6V or so.  This fan works at 4.78 V, so the fan spins slowly.  Before I could add heat to the mix to watch the fan speed up, I noticed it getting faster on it's own.  After about 5 minutes it's up to 11 Volts and blowing pretty hard.  My guess is something is feeding back into the pwm circuit.

I did try disconnecting the IC and did verify that 5v turns on the fan and gnd turns off the fan.  Good idea.  I should have thought of that.

Do you think the diode would solve this?  Or maybe I should add more capacitors to the power supply.

Thoughts?

[electroguy]

Do you think the diode would solve this?  Or maybe I should add more capacitors to the power supply.

try both ideas.
The 100nf you have drawn in your diagram, should be as close as possible to pin 16 on the IC. Also try and add 10uF as close as possible to pin 16 of the IC (but keep the 100nF still there, use both).

Also, measuring the voltage on the fan won't help you, as it is PWM and you really need to check the frequency and duty cycle too (ie, 12V at 50% duty cycle *could possibly* read as 6v on your DMM, depending on frequency, etc...). And it depends on the frequency response of your DMM too. Use a CRO instead if you have one?

[TerminalJack505]

I was typing this up as electroguy made his post so some of it is the same...

I would definitely add the diode.  A big electrolytic cap (10uF to 100uF, >= 16V)on the +12V rail close to the fan's positive supply pin wouldn't hurt. 

A standard 100nF bypass cap for the IC never hurts.  (I just noticed you have one in your schematic.  Make sure it is close to the IC's +ve supply pin.)

So far as the fan fail detections goes, this is an output, right?  Are you saying you had it wired-up like I've shown in the attached schematic and disconnected it from the FULLSPD# pin?

Edit: The datasheet is really confusing so far as the FULLSPD pin goes.  The example schematics seem to indicate that this is an active-low input but the table on page 4 says it's an active-high input.

[heychris]

I've seen several inconsistencies in the datasheets. 

In regards to your markup of the diagram.  Yes.  I removed the fanfail to disable the detection of a failed fan.  I grounded the fullspd in order to deselect 100% output.

I'm curious about the caps.  I've seen a fairly standard setup with microcontrollers which is typically one electrolytic capacitor and one ceramic capacitor.   What is the difference between each?

I added a 100uF electrolytic and 10nF to each rail.

Is a fast recovery diode the same as a Schottky diode?  I've got a bunch of 1n5819 Schottky diodes.
http://www.fairchildsemi.com/ds/1N/1N5817.pdf

I also have some Zener diodes.

I've been looking at the max6643 eval diagram.  http://datasheets.maxim-ic.com/en/ds/MAX6643LBEVKIT.pdf
They've got several additional pieces.  I'm mostly wondering what r4/short and r5/open are supposed to be for.

 

[electroguy]

Is a fast recovery diode the same as a Schottky diode?  I've got a bunch of 1n5819 Schottky diodes.
http://www.fairchildsemi.com/ds/1N/1N5817.pdf

I also have some Zener diodes.

Don't use zeners for this application.
Either of the 1n58xx diodes you listed should be fine (or just use a 1n4004 or anything else you have)

[heychris]

I added a 1N5819 with no affect.  I've got a contact at Maxim.  Maybe I'll ask for pointers

[SeanB]

Just as an aside the most common tacho output on fans is an uncommitted open collector output from the fan driver, normally integrated onto the chip inside the fan. This will not work in this circuit, as when the mosfet is off it will feed whatever the fan voltage is ( less a volt or so across the parasitic diodes in the little IC in the fan) direct into the Maxim controller. If the fan is run from 12V then the tacho input will get around 10V applied to it, at whatever current the fan allows to flow through the diode, probably near half of the rated current ( the fan will basically be running on the difference between the rails, less the drop across the fan diode in reverse across the tacho output (part of the chip process and never supposed to be used other than as an ESD clamp) and the forward voltage drop of the EDS clamp diode in the Maxim part) which will rapidly kill the controller, and at best cause it to malfunction.

If you want to use the tacho, you will need to use an optoisolator to provide isolation of the current path.

[heychris]

SeanB,

You are right.  The tach lead in this case is only used to detect a failed fan.  With open collector fans, they recommend a current sensing circuit.

Many fans have open-drain tachometer outputs that produce periodic pulses (usually two pulses per revolution) as the fan spins. These tachometer pulses are monitored by the FAN_IN_ inputs to detect fan failures. If a 2-wire fan with no tachometer output is used, the fan’s speed can be monitored by using an external sense resistor at the source of the driving FET (see Figure 3). In this manner, the variation in the current flowing through the fan develops a periodic voltage waveform across  the sense  resistor .  This periodic waveform is then highpass filtered and AC-coupled to the FAN_IN_ input. Any variations in the waveform that have an amplitude of more than ±150mV are converted to digital pulses. The frequency of these digital pulses is directly related to the speed of the rotation of the fan and can be used to detect fan failure

[heychris]

I've finally figured out the source of my problems.

The documentation states that Pin 8 connects to an external temperature sensor.  In the pin description of the datasheet for pin 8 it mentions “Connect to GND  if no remote diode is used”.  I intended to use the internal sensor.  I grounded pin 8.  When I did this, the behavior was such that it continued increasing the fan speed until 100%.  I tried disconnecting it.  When I did, it started working properly.

The documentation states a grounded pin is read as 0 degrees.  If this were the case, it should have used the higher of two temperatures and still been accurate. per documentation.

I removed all the extra capacitors and it's working like a champ.

My intention was to make a few dozen little boards with a surface mount pad for the IC and through hole components for the rest.  I figure that their may be a few other people that have fans they'd like to quiet.  Can anyone recommend a PCB service that would allow others to buy if they so please?

Before I build the board I have a couple questions.

1. Should I choose a different mosfet?
2. I've notice a couple other similiar controllers that use the resistor/diode method of feeding the microcontroller it's reduced voltage.  Does anyone have an experience where this is a problem?
3. I'd like to optimize the capacitors.  The documentation calls for specific ones, but the eval kit has quite a few more.  Which is a better target?

[heychris]

All,

Thanks again for all the help getting this straighten out.

I'm looking at a couple voltage regulators for the microcontroller.

I like the  to-92 case.  I did a digikey search to filter to-92 and 5V.  I tried to get the other specs close to the LM7805.  I came up with a 100ma 5v regulator.  What do you think?
http://search.digikey.com/us/en/products/L78L05ACZ/497-2952-5-ND/634716

Maybe this mosfet is a better fit?  http://www.fairchildsemi.com/ds/FD%2FFDD3N40.pdf

In the past I've put a 220uF electrolytic capacitor on ATTiny Vcc.  I also add a small ceramic.  I don't really know the logic here, but seems to be common.

How does that sound?

[TerminalJack505]

Glad to see you got it figured out.

The voltage regulator looks fine. 

I liked the other MOSFET better than this new one.  The problem with this new one is it has a poorer worse-case V_gs(th).  It is 5V and the other one is 2.5V or 3V if I remember correctly.  Since you will be driving it at only 5V it may not be as efficient.  When using a MOSFET as a switch you typically want to drive the gate at least a couple of volts over the worse-case V_gs(th).

[heychris]

Interesting..  I'm not quite sure what the impact is of the Vgs.  The one I tried  is rated identically to the one in the maxim eval kit.
MAXIMUM RATINGS
Parameter Symbol Value Unit
Drain?to?Source Voltage VDSS 30 V
Gate?to?Source Voltage VGS ±20 V

Looking this up did cause me to realize that the one I tried was not a logic level gate.  That seems like it could be important.  right?

Maybe something like this?

http://www.onsemi.com/pub_link/Collateral/NTD5867NL-D.PDF

[TerminalJack505]

That one looks good.  Look at firgure 1 in the datasheet.  This will give you an idea of how Vgs affects the maximum current.  A 5V Vgs is more than adequate to drive this particular load.