Great, thank you, Sgt.! I can now confirm the dead phase. Without power, I noticed that when I turn the motor, it has exactly 12 spots where the magnets allow the rotor to stop. I simply turned the motor to each spot and applied power to see if it would start. You will not be surprised to know that 8 of the 12 spots started every time and 4 of them did not. Interestingly, these 4 spots are two adjacent pairs directly across from each other. Makes sense given the three phases.

Now to test the mosfets...

 

Testing the T-lead voltages was a bit more complicated. I tried probing the mosfets, but didn't get far - to hard to reachespecially with the motor running. I followed the T-leads and soldered on test leads to the back and checked for equal resistance to assure my soldering was okay.

I then ran a ton of tests. I won't bother reporting all the results, unless you want any,

Did you test with the leads disconnected ? This is the test that will tell you if what I suspect & what wookie said is correct: dead phase.
Remember, if you have a dead phase and you test with the leads connected, you will still read voltage on all the phases; that doesn't mean that all the phases are working.

Sounds like one of the phases is dead, which is why it only fails to start occasionally. If the rotor happens to stop on the dead phase, it won't have any torque to start it moving.

You might start by checking each MOSFET, and trying to see if you get output from each Hall-effect sensor.

 

Great, thank you, Sgt.! I can now confirm the dead phase. Without power, I noticed that when I turn the motor, it has exactly 12 spots where the magnets allow the rotor to stop. I simply turned the motor to each spot and applied power to see if it would start. You will not be surprised to know that 8 of the 12 spots started every time and 4 of them did not. Interestingly, these 4 spots are two adjacent pairs directly across from each other. Makes sense given the three phases.

Now to test the mosfets...

ah you beat me to the reply.

 

Yeah, for the first time, we have definitely found a bad component on this board! I simply tested the diode function of each mosfet with the fluke and found one that leaks. It is a bad mosfet. I assume that explains the bad phase. I'm going to replace that one to see what happens. Will report back how it went.

 

After testing the diodes function of the mosfets on the board, I found one that leaked as I previously reported. I ordered new ones that just came in. I retested carefully and noticed that the ones that didn't "fail" had lots of potting gook on their pins. After scraping and retesting, they all "leak". I've tried to make sure there is no charge getting to any pin, but something seems amiss. I removed what I thought was the bad mosfet (a n-type) and tested it separately and it seems to work find and just the same as the new one. I put it back on the board and it failed again.

My hypothesis is that one can't test the mosfets diode function in a circuit like this because testing one activates others or charges caps that end up doing the same thing. I can play with this more to see if there is any pattern, but so far it's as if they all leak when I test in circuit.

The way I'm testing the n-type ones is with diode test on the fluke with the negative probe on the source and the positive on the drain. I touch all the pins to short it out and when I do this test, it shows OL like it should. However, when I then move the positive lead off the drain to the gate, it charges up the gate so to turn on the mosfet. I then put the positive lead back to the drain and it shows much lower resistance, i.e. the mosfet shorted between source and drain. Again, this is what you want and when I test a new one it work this way.

Now the p-type ones are different. I put the negative lead on the drain and the positive on the source after shorting it all out and meter shows OL and I then touch the negative lead to the gate and back and then I get the closed circuit like I want.

It took me half an hour to get the first mosfet out without braking anything, so I hate to remove the rest. They all fail equally when I test in circuit. I suppose pulling them all and testing separateley would help prove things, but I'm guessing the mosfets are all okay.

What do you think? Should I go to the effort to remove test and replace them al, or is here another place to focus my attention?

 

Have you tested for continuity in the individual stator coils? While these motors are a three phase they are three phases of DC, not the same as three phase AC. The three phase AC is joined together at a single point, three phase DC is only from plus to ground in three places, not joined together.

http://ww1.microchip.com/downloads/en/AppNotes/00885a.pdf

 

Thank you, Shortbus. I understand that the motor is 3 phases but using DC. However, I don't understand is how the windings work. I'm revieweing the terrific doc from microchip to learn more, but perhaps you can answer this. There are 12 coils and that makes sense given that I found 12 poles where the rotor stops. However, there are only three wire leads in total. No ground connection. Is the magnetic field formed just by energizing the coils, but without allowing current to flow or am I just not seeing a required common ground somewhere like perhaps in the standoffs?

 

Thank you, Shortbus. I understand that the motor is 3 phases but using DC. However, I don't understand is how the windings work. I'm revieweing the terrific doc from microchip to learn more, but perhaps you can answer this. There are 12 coils and that makes sense given that I found 12 poles where the rotor stops. However, there are only three wire leads in total. No ground connection. Is the magnetic field formed just by energizing the coils, but without allowing current to flow or am I just not seeing a required common ground somewhere like perhaps in the standoffs?

I don't understand that either. Shortbus's document answered a lot of my questions, but created more questions for me in the process. I started my own thread about it. From what I see in the document, the 3 windings are connected in a star configuration exactly like a 3 phase AC motor, no ground connection. I do not know. hopefully someone will shed some light on it either here in this post or in mine.

 

The microchip doc explains the commutation sequence on page 12. It is a star or wye and one winding is positive, one negative, and one left open. The chip and FETs basically use PWM signals in a series of 6 different combinations of + - and open to drive the motor. The hall sensors feed back the position to keep the timing of these PWM pulses in balance. The 6 sequences must coincide with the n and p type FETs, but this is only starting to become clear. I suppose a sequence is defined by one p type and one n type being activated to set the voltage of two of the windings with the last one open.

Since it is a wye type circuit, each of the three windings leads are connected so should have continuity. That's easy to test and, no. I haven't done that yet. will report back in the AM.

 

After all the reading I have just done on these BLDC motors I think you got it right in post #6. Hall sensors. As I understand it, (assuming the hall sensors are giving input to the mosfets) in any position of the motor, two windings will either have a positive or a negative voltage on them, so you should have some torque. considering there are only 3 hall sensors and 6 MOSFETS, one hall controls a pair of opposing MOSFETS. I doubt that coincidentally both mosfets failed. more likely that the hall that controls these two mosfets failed. but still with no schematic I am about 50% certain of this.

 

http://www.motioncontrolonline.org/files/public/BrushlessOperation.pdf
check out the 3rd page. theres a hall/mosfet logic diagram that might help you narrow it down.

 

The BLDC motor can also be thought of as being similar to a stepper motor. With a difference.

Stepper motors have rotor magnets with ribs/teeth. And a stator with coils and ribs/teeth.

A BLDC has a single magnet rotor and a Stator with separate coils and poles that cover a segment of the inside diameter of the stator.

The control logic switches the coils on and off to make the magnetic fields 'rotate'. they don't really move, it's just the switching on and off that simulates that.

The hall effect sensor tells the logic that one field has energized and the next one should turn on and the first one should shut off.

I'm not good at explaining things in text, as you can tell. You can't see my hands move to illustrate what I mean.

The usual thing to go bad in a motor is a winding, not the electronics. concentrate on the continuity of the windings. Could even be the connections to the board, that is my thinking since it is an intermittent problem. If a winding isn't energizing the hall effect sensor has nothing to make it work.

 

What portions of the motor is reachable while it is running? If you have a frequency counter on your fluke, try to measure the outputs on the hall sensors after you've used your human motor starter circuit. Compare them to each other, or check for irregularities. Also, with your usb scope check the gates and sources/drains of the mosfets while in operation (maybe check voltage levels to be sure they're not switching 120v which may too high for usb scopes?)

 

I agree, it's probably not the windings or the mosfets. Continuity of the windings checked out as expected. This may explain the difficulty I had trying to test the diode function of the mosfets. The diode test produces a charge so when you bring a gate up, the mosfet closes which effectively powers all the windings which are connected to all the other mosfets making it impossible to isolate them in circuit.

I need to test the hall sensors which will be tricky. The pins are under the rotor so I'll need to solder test leads to each pin and lead them to the logic analyzer.

I can't locate the data sheet for these sensors, but I think they were made by Allegro Microsystems because the logo "A" matches and allegro makes hall sensors. The sensors are each labeled: A89E 949). The attached data sheet is the closest I can find and assumes this is for a A3189.

Before I start on that, I need some help with something that might be the big problem. As a reminder, this thread is about solving the start problem for this motor. I've learned that there is no specific start circuitry in a brush-less DC motor like this. This then suggests that the start sequence to get from still to motion is critical. How then is a hall sensor able to determine motor position prior to the rotor spinning?

There are 12 positions where the rotor will stop. To me, this implies 12 north or south poles with 12 opposing poles. The hall sensors are each positioned on an equivalent pole, but appear to be slightly out of order so they would fire in a rapid but not simultaneous sequence. But where to start?

I wonder why the rotor bell has a carefully marked permanent line painted on that I haven't been able to explain. It appears to mark a "special" or reference north or south pole, but why is this needed? Since the rotor magnets allow it to stops in one of 12 positions, why is only one marked? Is there some significance to this one position with regard to the hall sensors and the start sequence?

I suppose I could try a process of elimination, but it is not intuitive that the rotor would need to have a special start position. The mechanical attachment of the rotor is only via friction, so it could slip over time. If it were so critical, they'd have a detent of some sort to avoid slipping, right? Couldn't be it, but then why the heavy permanent mark?

Any suggestions or insights appreciated as usual!

 

What portions of the motor is reachable while it is running? If you have a frequency counter on your fluke, try to measure the outputs on the hall sensors after you've used your human motor starter circuit. Compare them to each other, or check for irregularities. Also, with your usb scope check the gates and sources/drains of the mosfets while in operation (maybe check voltage levels to be sure they're not switching 120v which may too high for usb scopes?)

The rotor covers everything up when spinning. However, after removing the heat sink from the mosfets, I can get to them but not easy. I'll have to solder leads onto the hall sensor pins to test them. They each have three pins, supply, output and ground times three.

 

I hooked it all back up with the leads soldered on for the hall sensors and luckily the motor worked. I'll have to try later to test the logic analyzer. Thanks for your help!

 

The usual thing to go bad in a motor is a winding, not the electronics. concentrate on the continuity of the windings. Could even be the connections to the board, that is my thinking since it is an intermittent problem. If a winding isn't energizing the hall effect sensor has nothing to make it work.

The problem isn't exactly intermitent. He's stated that in 1/3 of the available starting positions, it won't start. This is a repeateble error.

I noticed that when I turn the motor, it has exactly 12 spots where the magnets allow the rotor to stop. I simply turned the motor to each spot and applied power to see if it would start. You will not be surprised to know that 8 of the 12 spots started every time and 4 of them did not. Interestingly, these 4 spots are two adjacent pairs directly across from each other.

If you look at the document I linked to in post#51, on the 3rd page it shows the logic table for the hall sensors.
in any of the 12 available positions, the logic will have inputs from either 1 or 2 hall sensors and determine which phase to fire next. if the logic had 0 inputs, I assume the output would be 0.
for example, looking at that logic table, if hall#2 were not working, the logic table would be completely different and you would never get a binary output of 2, so in positions 4 & 10 the motor would not rotate (i assume). Now that only accounts for 1/6 of the starting positions, and he's said that it won't start in 1/3 of the positions. but...

The hall sensors are each positioned on an equivalent pole, but appear to be slightly out of order so they would fire in a rapid but not simultaneous sequence. !

it appears his hall sensors are positioned in some wacky fashion, so his logic table may be completely different and this could account for a zero output in 1/3 instead of 1/6 of the positions (I assume).

 

There are 12 positions where the rotor will stop. To me, this implies 12 north or south poles with 12 opposing poles. The hall sensors are each positioned on an equivalent pole, but appear to be slightly out of order so they would fire in a rapid but not simultaneous sequence. But where to start?

The time it takes for the hall sensors to latch one after another is the input for the controller IC. That way, it knows when the motor has stopped, and it's internal logic can switch the mosfets in whatever desired speed, gradually increasing to max.

 

Guys, I think I'm close to throwing in the towel. I must have somehow causes a short when soldering the jumpers to the hall sensors. I smelled something and immediately pulled the wires just as a resistor smoked. At least I think it was a resistor. All the diodes are larger. Anyway, I studied the circuit and tried a couple of different resistors in sizes the like others on the board in a similar section of the circuit, but no luck. Maybe I haven't hit on the right resistor, or maybe I smoked something else too. Without the schematic, I'm flying blind.

I still have continuity on the windings, so that's good, but I'm not sure now how to proceed.

 

Guys, I think I'm close to throwing in the towel. I must have somehow causes a short when soldering the jumpers to the hall sensors. I smelled something and immediately pulled the wires just as a resistor smoked. At least I think it was a resistor. All the diodes are larger. Anyway, I studied the circuit and tried a couple of different resistors in sizes the like others on the board in a similar section of the circuit, but no luck. Maybe I haven't hit on the right resistor, or maybe I smoked something else too. Without the schematic, I'm flying blind.

I still have continuity on the windings, so that's good, but I'm not sure now how to proceed.

how about post a pic of your burnt thingy. you have photos from the other day. maybe we can reconstruct what it was from before & after shots.