Help with Water Pump

[KeepItSimpleStupid]

There are 3 leads to those devices. In your pic, the left lead to the tab (solder blob) in both directions is the most useful.

I suspect, you don't have a diode mode on your DVM? That's the mode you should use. Can you get a hold of a 9V batter, clip and ~1K (Value not important) resistor? If you can, put the 9V battery in series with the resistor. Put the ends acoss the various 2 port combinations and measure the voltage across the combination. Reverse the directions and repeat.

Usually we would think of the orientation with the leads down, but we can use leads up. Just as long as we know.

 

[()blivion]

!WARNING! Wall of text mode has been engaged!

They are both soldered super tight to what seems like a heatsink, so I cut the legs and tested them with my DMM on the 2000K setting.

By just this, I don't know which legs you tested exactly, I'm assuming your doing the two you cut off the board?

One of them registered 1235, but with the probes reversed registered 1 (as though the probes were touching nothing). That does not make sense to me, so I did it a number of times with the same result. The other registered 1, reversing the probes also registered 1.

You can call off the exorcists and Steven hawking, nothing is possessed and no laws of physics were broken, this is just the diode effect. Diodes only allow electricity to go one direction, like an electronic reed valve. A transistor is an evolved form of a diode, so these results are hardly surprising. These numbers could be a good BJT, or a bad FET. Or possibly a good FET *IF* you did your test differently than I am currently assuming.

In any case, we are most likely going to have to start over again, since I'm not sure exactly which legs you were testing (The crystal ball is getting more smoke put in it right now)

Just as KeepItSimpleStupid has said. There are three(3) leads to those black things, they are most likely FET's... just as what you have been using for your pump circuit. Same gift, just a little different wrapping paper is all. The important thing to get is that the metal tab that is soldered to the heat sink is ALSO the middle pin. I have never seen one that was not this way, but I'm sure they could do it. (If they wanted to be really mean to us that are used to whats normal)

First test:
I want you to test the middle leg to the right leg and swap the probes back and forth if you would. A good part should MORE OR LESS conduct electricity one way (probes healed together), and not conduct the other way (probes held apart) with this test. The way that conducts will likely not show up FULLY conducting with a DMM. It will prolly read a few hundred ohms or so instead. If not, use KISS's test instead.

Second test:
Test the left leg (right leg in your first picture) to BOTH the other legs, and also swap the probes back and forth on that test as well. A good part will *NOT* conduct electricity (probes held apart) at all with this test. Any thing less that ~100,000 ohms is bad. (most of the ones I just tested were greater than my meter could read)

You will be testing BOTH parts, with EACH test. So four(4) tests in all.

Also note, you *MAY* need to use something different than your DMM as some meters are finicky for FET testing. However, this is all I ever use and it has never failed me. Then again, I also know what exactly it is I'm looking for and I have a meter with diode testing function. If you get odd results, build the circuit that KISS suggested. It works better in some ways than a DMM.

Also, it looks to me like in your pictures that the heat sinking is actually unetched copper the manufacturer left on the circuit board just for heat sinking reasons (very common). Or is it that it's an actual full blown metal heat sink under there some how? If it is a full heat sink, like a large piece of actually separated metal then the black things *MAY* be BJT's and not actually FET's. I find this unlikely but I exclude nothing. If you could read us the numbers off them, that would be even better. After you clean them off, if they are hard to read, a little acetone makes them REALLY stand out clearly. Try not to use acetone in an enclosed space, unless you have something really cleaver you want put on your head stone. But you prolly already knew that

Here is another YouTube video for you. This one is on testing MOSFET's. It kinda sucks, and I may make one of my own and upload it to YouTube myself later, (if I'm not still really lazy instead). Suck or not though, it's correct and should help you sort things out a little. *REMEMBER* THE METAL TAB IS THE MIDDLE PIN ALSO*.

 

[salty joe]

DMM at 2000K.
With probes on middle leg and right leg and then with probes reversed, I got values of 1 and 1097 on one FET and values of 1 and 1103 for the other FET.

With probes on left leg and middle leg and then probes reversed I got values of 1 and 1. On the same FET, I got a value of 1 with the probes on the right and left legs. I also got a value of 1 with the probes reversed.

On the other FET, with probes on left leg and middle leg and then probes reversed I got values of 0 and 0. (On that FET the left leg is soldered to the heatsink. The heatsink is a mass of solder.) On that FET, I got a value of 1100 with the probes on the right and left legs. I got a value of 1 with the probes reversed.

With the diode setting slected, I got values of 535 on one FET and 549 on the other FET with probes between the right and middle legs.

I tried real hard to find numbers on these FETs-no luck.

For orientation, I placed the FET's legs facing me and called the leg on the right the right leg. (opposite of the picture)

If it would still be helpful, I can scare up a 9V battery and run your test KISS.

Thanks!

 

[KeepItSimpleStupid]

The "0 0" on the "other FET" means it bit the dust.

Your diode mode is wierd. If it read 0.535, it might make sense, but 535 doesn't at all unless your diode mode is wierd. e.g. resistance with enough voltage to turn on a junction.

It might be "instructive" for you to try the battery test. Radio shack has all of the parts.

If you want to, you can send me the motor and I can run some other tests such as:
1. winding resistance
2. Winding inductance
3. Parameters of the good FET

If the windings check out, I could even try to replace the FET. No payment, just postage. It would be for our curiosities sake. I'm on the east coast in the US.

 

[salty joe]

The "0 0" on the "other FET" means it bit the dust.

Your diode mode is wierd. If it read 0.535, it might make sense, but 535 doesn't at all unless your diode mode is wierd. e.g. resistance with enough voltage to turn on a junction.

It might be "instructive" for you to try the battery test. Radio shack has all of the parts.

If you want to, you can send me the motor and I can run some other tests such as:
1. winding resistance
2. Winding inductance
3. Parameters of the good FET

If the windings check out, I could even try to replace the FET. No payment, just postage. It would be for our curiosities sake. I'm on the east coast in the US.

On the 0-0 FET, one leg was melted into the solder heatsink. Could it have gotten THAT hot when the pump fried? There was only 1/8" or so of epoxy coating at that point. It showed no sign of damage. Until I cooked it with a torch, that is.

I double checked the diode reading and there is no decimal point. It is a cheap DMM.

If there's a possibility that you taking a look at the dead pump might help find a way to get the circuit to reliably drive these pumps, then heck yeah-I'd be glad to send it to you. If it helps, I could also send a new pump, along with return postage so you could send it back. The dead pump is pretty thouroughly butchered, it's a lost cause except maybey figuring out what went wrong and why. I don't understand what went wrong.

 

[()blivion]

On your meter "1" must be open circuit, IE... the same as probes NOT touching each other eh? I also suspect that when you say "there is no decimal point" What your really saying is "the LCD screen simply doesn't have one" If so, then as KISS said, that's an odd (cheep) little DMM you got /)

If it would still be helpful, I can scare up a 9V battery and run your test KISS.

Yeah. I guess if you still really want to know which part is blown and feel like doing it your self, then best to do the KISS test now and report in. And as KISS said, it should be educational. Or you could take him up on the repair/testing offer.

I got values of 535 on one FET and 549 on the other FET with probes between the right and middle legs.

KISS beat me to it, but yeah. That should be 0.535 and 0.549 volts, as 0.6 volts is about the voltage drop of a single diode, which is what the diode mode is meant to test. These values would be expected from that test with a good part. If you have a spare diode, give it a test, if I'm right it should read around 600 one way on the same mode.

I tried real hard to find numbers on these FETs-no luck.

Rats... but that's fine, it happens. Some times the manufacture cuts costs by not having numbers printed on the part. Since that's a whole new machine they would need to have operating. Well.... more like a whole different wing of a building, but same thing. They could have also been printed on, and not actually stamped or LASER etched in. In which case, they may have just rubbed right off on you. To bad, but "shikata ga nai".

On the other FET, with probes on left leg and middle leg and then probes reversed I got values of 0 and 0. (On that FET the left leg is soldered to the heatsink. The heatsink is a mass of solder.) On that FET, I got a value of 1100 with the probes on the right and left legs. I got a value of 1 with the probes reversed.

Forgive my ignorance, but I'm a little confused. Just to be clear, Is the left leg actually cut and lifted from the board? Didn't you say you cut them all off? But the highlighted comment kinda makes it seem like you left that one on. If you didn't cut it, and left it on the board still, then you may be finding another path through the circuit board that's reading as 0 ohms.

If you *DID* cut them all off, then as KISS said you found your bad part with 0-0. The left leg should NOT show 0 ohms to ANY other pin *EVER*. That would mean the one FET's gate is shorted to at least the drain. That's never good for a FET. The gate pin is a very sensitive thing, they can die for almost no reason it seems. It was more likely a short high voltage spike that did the deed. This can happen for a number of reasons. I'd say voltage spikes on the power line, or back EMF from the coils to the internal FET's. Most FET's limit is 10~30 volts, and your pump runs at 24V, so we are already riding on the edge.

If all you want to just get the most work done in the least amount of time, I recommend to do as KISS said before and put on the extra protection parts or something. Then you can relax knowing things are safe now. But it doesn't hurt to be thorough. Your choice.

I'll be around.

 

[salty joe]

On your meter "1" must be open circuit, IE... the same as probes NOT touching each other eh? I also suspect that when you say "there is no decimal point" What your really saying is "the LCD screen simply doesn't have one" If so, then as KISS said, that's an odd (cheep) little DMM you got /)


Yes, my DMM reads 1 when probes are not touching. And yes it is a cheap one.

Forgive my ignorance, but I'm a little confused. Just to be clear, Is the left leg actually cut and lifted from the board? Didn't you say you cut them all off? But the highlighted comment kinda makes it seem like you left that one on. If you didn't cut it, and left it on the board still, then you may be finding another path through the circuit board that's reading as 0 ohms.

I did lift the legs from where they were connected, but there is a blob of solder right at the body of the FET connecting that leg to the heatsink. After closer inspection, I noticed a circular depression in the solder around that FET. I now believe the FET got so hot it melted the surounding solder and deposited the blob.

 

[salty joe]

A MOV and reverse biased diode close to the motor and a fuse sized to protect the motor would be your best insurance for the motor itself. A self protecting switch rather than a simple FET would be the best change for the circuit reliability.

I can't mount anything close to the pump because it will be submerged.

The existing circuit has a single soft start and a single FET for each pair of pumps. Do you think I'd have a fighting chance if I used a dedicated soft start and a dedicated FET for each pump, as Alec kindly drew in post 367? Alec drew this before we knew what was inside the pump.

 

[KeepItSimpleStupid]

It's possible that your 535 is in ohms with a higher current. What's the model of the DMM? There might be a datasheet somewhere on the web. I have a $4.00 harbor Freight meter that I got free and it measures the diode voltage correctly as well as lots of other meters.

Here is an interesting IC: https://www.allegromicro.com/en/Pro...~/media/Files/Datasheets/A4930-Datasheet.ashx

It kinda gives you an idea how things are supposed to be designed. That particular IC used a motor with a single coil and apparently two hall sensors. I didn't figure out exactly how it's supposed to work.

In any event, a locked rotor condition could potentially destroy a motor. I'm grasping at straws, but what if all of the motors tried to start at the same time AND the power supply could not supply the short-circuit current of all of the motors simultaneously. Suppose this occurred at start-up or after a power glitch. The basic result is the motor would have been supplied with a reduced voltage that would not cause the motor to turn.

That circuit is too simple and there looks like there is NO PROTECTION, so potentially it could easily fail.

Wire lengths, wire size, short circuit current might need to be looked at.

Short circuit current is basically the current that is defined by the DC resistance of the motor and wiring.

This is up for discussion.

 

[KeepItSimpleStupid]

I need to sleep on it. Post 367: https://www.electro-tech-online.com/threads/help-with-water-pump.124250/#post1060587

 

[()blivion]

but there is a blob of solder right at the body of the FET connecting that leg to the heatsink. After closer inspection, I noticed a circular depression in the solder around that FET. I now believe the FET got so hot it melted the surounding solder and deposited the blob.

Bingo! that might be all that was wrong. If so I throw away my previous assumptions in favor of getting new data. Try cutting, bending, smashing, slicing, or otherwise removing the solder blob then retest the part. I think that blob was just shorting things out. If the reading 0 is basically the same as the probes touching, then it's a no brainier that if there was solder connecting the two legs, 0 is the reading you would get. When the blob is removed it should act like the other FET does if I'm right.

Still... need to know why it got hot in the first place if it even did at all. It's also possible that it didn't actually get hot enough to melt the solder, but that the blob was always there. It was just not actually touching until we started playing with it. We won't know until we do more tests. Cut the blob and retest. Then we will know.

That circuit is too simple and there looks like there is NO PROTECTION, so potentially it could easily fail.

It's ironic that KISS would say this, but it's not a flawed statement. I mostly agree, there needs to be more safety. It *IS* electronics in salt water after all. In alec_t's defense though, I think he was going for a preliminary design to work out the bugs in this circuit as of right now. So problems are somewhat likely I would say. But now I think it might be time to start adding the armor.

 

[alec_t]

In alec_t's defense though, I think he was going for a preliminary design to work out the bugs in this circuit as of right now.

Indeed. As my signature says, all my designs should be regarded as experimental.
The OP is a newcomer to electronics so the original circuit was designed to be relatively straightforward to build. Of necessity there are compromises to be made in any design. Bells and whistles can be added but have to be justified as they add complexity and cost.

But now I think it might be time to start adding the armor.

Okay. Any suggestions? Is the armour to protect the switching circuit, the pumps or both? Is protecting the circuit cost-effective? Would adding current-limiting to the circuit be effective, or would it cause complications by interacting unpredictably with whatever electronics is inside the pumps? Is the soft-start feature (here a slow FET turn-on, rather than a PWM type) helpful or detrimental for brushless motors? Would PWM drive of a brushless motor with unknown internal electronics be risky?

 

[salty joe]

It's possible that your 535 is in ohms with a higher current. What's the model of the DMM? There might be a datasheet somewhere on the web. I have a $4.00 harbor Freight meter that I got free and it measures the diode voltage correctly as well as lots of other meters.

.

I have the el cheapo CEN-TECH model 92020 fron Harbor Freight. The directions say when in the diode setting, the readout is in mV. So 535 is reasonable?

I decided not to cut out the blob. I am afraid of messing it up and losing available info. Plus, I want you to see what this looks like KISS, before it gets cut up. I'll get it in the mail this weekend-thank you KISS for offering to take a look at it.

 

[KeepItSimpleStupid]

535 mV is indeed close to a diode drop. Most people use 0.6, some use 0.7. Schottky diodes https://en.wikipedia.org/wiki/Schottky_diode s, a bit less.

The idea is that you either want to see diode drops or opens and not the same value in each direction. For a bipolar transistor there is an inherent E-B diode and E-C diode.

FETS are a bit harder to judge.

 

[()blivion]

Salty, you can ignor all of this. It's mostly tech talk with alec_t about how to modify this circuit to be more safe for your pumps.

Okay. Any suggestions?

I think some simple circuitry for sensed current-limiting is a great idea (As well as slow blow fuses). We can do it on a per motor basis I hope. And we need HV transient protection of some kind. It also occurred to me that we might want to make a *SOFT OFF*, since an abrupt off could cause back emf in the motor and fry the FET's. I'm wondering if the motors internal FET's are using the natural body diode, or if they have one added for safety. KISS is going to have to figure this one out when he receives the blown pump.

Would PWM drive of a brushless motor with unknown internal electronics be risky?

Some what risky... yes. But it all depends on how you do it. It already has an internal speed controller that switches at a natural rate, which is likely what your worried about right? Adding another frequency that's anywhere close to the pumps internal speed controller frequency would most likely cause bad bad bad interference in the pump's electronics. But if you make the PWM frequency really low, then it could run with out problems, though it will likely be in the range of human hearing this way. Or you could do PWM, then integrate it to an analog voltage on our board before it gets to the pump. Then it should not be a problem since it's no longer a frequency of any kind.

The brushless motors that are in PC "box fans" are about the same circuit idea as these pumps are, just on a smaller scale. They can be powered with PWM without dieing. However they have internal capacitors that integrate the PWM to analog to a degree. I don't think Salty's pumps do. Also, PC fans are some what more noisy when done this way. All my PC's fans use NTC thermistors as part of a voltage divider to bias an FET in the linear for thermal speed control. (I dislike noisy fans.)

Is the soft-start feature (here a slow FET turn-on, rather than a PWM type) helpful or detrimental for brushless motors?

Not sure. No matter what, the FET's are in series with the field coil. And since coils have a natural inductance, they would oppose any massive current shoot through that might destroy the FET's. How much inductance is up to KISS to figure out. If it's to low, then the current of a "hard start" might climb up enough to go over the max of the FET before the rotor turns enough to activate the other FET. I think this would be why the manufacturers would recommend a soft start.

On the flip side of things, *IF* the soft start is making the motor operate the FET's in a liner region with a medium to high current. Then the FET's might make excessive heat and self destruct. However, FET's have a small natural negative feed back because Rs-on increases with temperature. Also, in accordance with ohms law our boards FET's would have to see the same current. So this being a real problem is less likely.

So... going off the odds alone, I'd say keep the soft start. But we will know more when KISS tells us how the pumps circuit works exactly.

 

[alec_t]

Thanks for the input, ()blivion.
I'll get my thinking-cap on re current sensing/limiting.
A varistor or two sprinkled around might help with HV transient protection of the circuit but I'm not clear how we could further protect the motors (which are sealed and remote from the circuit board) from any transients induced in their cabling or produced by the pump windings. The Schottky diodes offer circuit protection from back emfs, so would a soft-off give any advantage? Soft-off is just a simple matter of removing a couple of diodes/resistors from the circuit as it now stands.

Or you could do PWM, then integrate it to an analog voltage on our board before it gets to the pump

Sure, but I think that is equivalent to the present RC integration used for the soft-start.

However I'm still not convinced that the start-up (soft or otherwise) or switch-off has any bearing on the problem, because the pump failed over 3 hours after it had been started and before the end of the timed period.

It's a pity additional protective components can't be added inside the pumps themselves .

We await KISS's findings.

 

[KeepItSimpleStupid]

Question:

How much "head" is this pump pumping against? e.g. (Height of outlet above pump outlet)

Would the system benefit by having a weep hole near the outlet of the pump, so that whatever water is left in the lines empties? This is similar to what is done with a sump pump. A 1/8" hole is drilled very close to the outlet side of the pump. Yea, the pump leaks on purpose. When the pump stops, water drains slowly back into the sump.

 

[()blivion]

Thanks for the input, ()blivion.

No problem, just don't expect a warranty with my ideas either .

I'll get my thinking-cap on re current sensing/limiting.

I'm a big fan of magnetic current sensing myself. I thoroughly dislike shunt based methods as they are wasteful because of insertion loss. They already make purpose built IC's for just this reason. Maybe we could use something like **broken link removed** or this?

I'm not clear how we could further protect the motors (which are sealed and remote from the circuit board) from any transients induced in their cabling or produced by the pump windings.

It's a pity additional protective components can't be added inside the pumps themselves .

Yup, that is a dilly of a pickle. All we can do is try to "think out side the box" (Pump?). Things like soft start/soft stop and MOV's and such should help based on the expected characteristics of the pumps. If we can predict what the pump will do with a given input, we should be able to make it run 100% safe with our circuit. *UNLESS* of course there is a design flaw in the pumps that can't be worked around. Then it would be time for us to team up on the manufacture and complain.

Sure, but I think that is equivalent to the present RC integration used for the soft-start.

Actually... your 100% right. My mistake, disregard my comment. For some reason I was thinking we were doing PWM for full speed control and not the soft on. Which would be a whole different kettle of crawdads. Since we are talking JUST soft start and soft stop, then there really can't be any problem with doing it the way we are now. If there is we will know when our boards FET start to smoke. Which is unlikely with a good pump.

However I'm still not convinced that the start-up (soft or otherwise) or switch-off has any bearing on the problem, because the pump failed over 3 hours after it had been started and before the end of the timed period.

I'm in 100% agreement. IMO it's our circuits FET's that would take the biggest hit, not the motor. So there should be absolutely no problems with it.


EDIT:

How much "head" is this pump pumping against? e.g. (Height of outlet above pump outlet)

The pumps are JUST for agitating water in the aquarium. They are for making waves/turbulence. Coral needs chaotic water flow to stay alive.

 

[KeepItSimpleStupid]

However I'm still not convinced that the start-up (soft or otherwise) or switch-off has any bearing on the problem, because the pump failed over 3 hours after it had been started and before the end of the timed period.

If this is true, then a defective pump due to "infant mortality" or a piece of debris makes sense. Bear in mind, it's not the brightest pump design.

Current limiting has it's use and that would be protecting the driver (alex's circuit), not the pump. e.g. protected low or high side switch.

Current limiting to be really effective on motors has to incorporated with a delay. You let the limit be exceeded briefly for start-ups.

Properly sizing a fuse per pump and using a protected high or low side driver is about as best as we can do.

It's a badly designed pump. A really good pump would have an internal thermal.

 

[KeepItSimpleStupid]

Interesting observation: I looked at 3 pin hall effects sensors at Digi-key that were specified for VCC >24 V and found very few. 24 V is cutting it really close. This https://www.digikey.com/product-detail/en/AH337-PL-B/AH337-PL-BDI-ND/1301094 (diodes Inc AH337-PL-B) was the most interesting and the only one that was Open drain. All of the others were open Collector.

It probably does mean that the motor is susceptible to over voltages and likely the MOSFET body diodes are providing a great deal of protection. There isn't a lot of choices for the hall effect IC.

I suppose that it may be using an N and P channel MOSFET.