PC Fans - single or double phase

[jeremynd01]

For those of you who are in the brushless-motor-know: would a PC cooling fan be considered a single phase or a two phase motor?

In particular, I'm referring to the four pole (arrayed at 90 degrees) variety. I've been testing on a few, and they all operate similarly: each pair of the poles (pairs being the poles separated by 180 degrees) takes turns being on. Basically that means one pole is driven by a square wave and the other pole by its inverse.

I ask because I read in a book that any poly-phase motor can be made self starting (unlike single phase). This makes perfect sense to me for three-phase motors and motors driven by quadrature waves (a sine and cosine), but a sine and negative-sine doesn't seem to fit the bill.

Comments?
-Jeremy

[LBMiller5]

You are correct. The motors in standard PC Case fan is a 2-phase unit with a 4 pole stator. They are driven by 2 square waves that are 180 degrees out of phase with one another. If you take one apart, you will notice a small black device that looks like a transistor with 4 legs. Most use a device known as an APX9140. This device is a combination of a hall effect sensor, a voltage comparitor, and a pair of power transistors. The 2 outside legs are for +12 and Ground, and the 2 middle legs are the outputs. The outputs are configured so that when one is on, the other is off, and vice-versa.

Here is a link to a PDF of the Data Sheet for this part:

http://innov8tivedesigns.com/rcgroups/APX9140A5.pdf

The ring magnet in the fan has 4 poles, which are arranged NSNS. There are 2 phases of wire coils on the stator, and if you started from one pole the winding pattern is ABAB. The 2 A poles are across from each other, and the 2 B poles are also across from each other, between the A poles. The beginning of each wire coil is connected to +12, and the the end of each coil is connected to the outputs of the IC device described above. If the sensor detects a North magnetic pole by it, it turns on output A, which pulls the end of phase A to ground, causing current to flow, and energizes phase A. The fan then rotates 90 degrees trying to go towards the energized pole. Just about the time it gets there, a South pole of the magnet ring passes the sensor, turns off phase A, and turns on phase B. Then the motor rotates another 90 degrees to go towards the new pole. This process repeats over and over, and the motor spins faster and faster until the air drag on the fan equals the power being produced by the motor, at which point the speed stabilizes.

You can control the speed of the fan by controling the amount of voltage to the fan. You need a minimum of at least 3 volts to turn on the sensor and get the fan started, and the IC can handle up to 20 volts. The maximum continuous current on the APX9140 is 400mA, and if you look at the sticker on your case fan, most of them draw around 150-180 mA at 12 Volts DC.

Hopefully that answers your question.

Let me know if you need any more info.

Lucien

[jeremynd01]

Thanks for the feedback, Lucian! I had a feeling you might respond, what with your other thread containing pictures of dozens of fans for your ducted fan project!

A you might have guessed, I'm trying to turn a few of these into simple brushless motors, for two reasons: 1) They're even cheaper than CD rom motors and the ESC is much simpler (and self-commutating!), and 2) everybody else says it can't be done, and I love a challenge!

My first order of business in my prototype was to re-wind the stator with about 15 turns of 30AWG wire. I breadboarded the standard PC fan electronic commutator circuit (schematic shown below) using power BJTs to handle higher current (with the exception that I'm using a square wave frin a function generator to simulate the rotor spinning). The results were less than promising. The problem seems to to be the inability of the transistors to pull their collectors to ground (they usually muster about 80% Vcc, at best). My thought is that the inductance+series resistance of the new windings is too small, and that the stator windings are acting basically like wires. Instead of seeing a nice, 50% duty cycle square wave, I see closer to a 90% duty cycle wave.

Any thoughts on that?

[jeremynd01]

whoops... forgot the schematic

[LBMiller5]

You could use this type of fan. Since they are "Free-running" you would need to control the speed by varying the input voltage. It would take about 3-4 volts to start the fan turning, and would run up to about 15-16 volts at the top end.

You did not mention what type of NPN transistor you were using. With 15 turns of 30ga wire, you will be pulling a fair amount of current. You need to look at the gain curves for your output transistor to see what your Hfe is at the current you are pulling. Many transistors have a gain that drops to as little as 2 or 3 at high currents! When this happens your Vce goes way up and the transistor gets real hot real fast! You might want to try a TIP120 type Darlington Power transistor in a TO-220 package. They can handle up to 3 amps of current, and have a Hfe of over 1000. Another alternative would be an N-Channel MOSFET power transistor. They are a voltage controlled device instead of a current controlled device, as in the case of the Bi-Polar Junction Transistors. Many of the MOSFET Transistors have an "on resistance" of about 1/100th of an ohm, so there is virtually no voltage drop accross them. Because of this they run much cooler than the Bi-Polar variety. (That is why all of the speed controllers use FET transistors and not BJT's)

You will also need to gang up several magnets to make it work right. The standard configuration of the PC Fan is a 4-pole NSNS magnet. To get wide enough magnets, you may need to put 12 or 16 together in groups of 3 or 4 per pole. For a 12 magnet rotor it would be NNNSSSNNNSSS with the same pole magnets tight against each other with gaps between the opposite poles. For a 16 Magnet rotor it would be set up NNNNSSSSNNNNSSSS.

Another way to optimize the motor is to be able to vary the timing. this is done 1 of 2 ways: Either by physically moving the location of the sensor with respect to the stator poles, or by having a varible delay timer that changes the degrees of advance as a function of motor speed.

All in all is sounds like a fun project. Good luck with your tests.

Lucien

[LBMiller5]

I just looked back at the schematic and noticed a problem. You are driving the second phase off the output of the first phase. This will not work. Most of the hall effect motor controllers have 4 wires on them: +12, Ground, Output 1, and Output 2 (Which is opposite Output 1). If you have the circuit wired like it is shown, you have to wait for the discharge of the first cap to turn off the second phase. This is why you are seeing the timing problems. Try it again with the 4-legged motor controller with each power transistor controlled by its own output from the sensor, and you should see a much better result. Alternately, you could put a couple inverters in series on the output of the hall effect sensor. Then take the output for phase 2 after the first inverter, and the output for phase 1 after the second inverter (since 2 inversions would give you back your original signal)

If you don't have any of the 4-legged APX9140 type parts, let me know. I have a bunch of them here and I could mail you a few.

Lucien

[jeremynd01]

Thanks again, Lucian. You know, I get a real kick out of reading your posts, (and I read 'em all!) partly for your attention to detail and partly for the excitement you get from every little find (I do the same... much to my wife's chagrin!). In fact, I decided to make one of your goodbyes my signature, just because it hits so close to home!

Anyway, I've got to confess that sometimes I make the dumbest mistakes. Remeber how I mentioned using a function generator for testing? Well, the thing was such a weak POS that it wouldn't drive the low-impedence load presented by the power BJT's (which were 2N39-somethings... I can't remember exactly, but something like 6A sustained collector current). I faked in my own darlington variety system with some signal transistors and it worked much better, but I'm scrapping BJT's altogether, as soon has I can find some salvaged MOSFETs in my drawer-o-junk.

Now about the circuit schematic - I agree, running the second pole of the motor off of the first is pretty flakey (but ok for a $1 PC cooler). I think I'll try a comparator to do the inversion and implement your four-wire idea. I may also try replacing the capacitors with zener diodes, which would keep the transients down without the timing delay. I'm leaning towards driving the Vcc rail with a PWM frequency that is greater than the switching frequency of the stator for speed control.

I'm interested in your idea about ganging up a bunch of magnets. I have been using the stock ring magnet the fans come with (because it's there). Are you thinking that the magnets will reduce the flux leakage or increase the field (or both)? I'd like to get the setup to work with the stock ring, just to make the project a $.50 something to do for those getting in to building their own motors.

One last question on timing (which I understand in concept, but only a little in theory, which bugs me. I'm a physicist by trade, and I love good theory...). I've seen emperical results where guys say "I advanced/retarded the timing x-degrees, and now I get this-much-more torque." Do you have any good references for this that I could browse? It seems to me that, with the hall sensor nestled between two poles (let's define that as 0 degrees) that I would want to delay switching so the rotor can turn an additional 45 degrees. (I can picture this better than describe it... maybe I'll draw something up tonight).

Take care, - Jeremy