Help with Water Pump

[salty joe]

I need to order some parts in a couple days, 8 amp fuses will be part of the order. Thank you very much.

 

[4pyros]

I need to order some parts in a couple days, 8 amp fuses will be part of the order. Thank you very much.

Thats good, never fuse higher than the smallest wire can handle.
Because you are running motors you may want a slow blow fuse type to deal with the start up current rush.

 

[salty joe]

Well, I ended up with a problem with the circuit from post 236. The bench test with 40w light bulbs went well, so today I put four pumps in a bucket of water and connected them to the circuit. At first everything went as planned-one pair of pumps toggling and the other pair flicking every minute or so. After 4-5 hours, one FET got so hot the metal seperated from the plastic and was smoking. Alec, if you could ponder this and share your thoughts, I would appreciate it a lot.

This is the FET. https://www.electro-tech-online.com/custompdfs/2012/05/A-104.pdf

The power supply was pretty warm and was shut down. I have not had a chance to check it out yet.
The entire circuit including the tin box enclosure got quite warm which makes me apprehensive about using a heat sink. Just seems like too much heat. Just wondering, is there a way to provide a dedicated FET for each pump and still have the pumps work as a pair? Would that even be helpful?

 

[alec_t]

Oh . The FET spec looks ideal. In theory when a pump pair is running they draw ~ 3A and their FET gate voltage is ~ 12V. The specified Rds(on) for the FET when the gate voltage is >=10V is only 8mΩ max, so the FET should be dissipating ~ 3 x 3 x 8 = 72mW max continuous. That would normally cause slight warming but not require a heat sink.

After 4-5 hours, one FET got so hot the metal seperated from the plastic and was smoking

No, that wasn't part of the game plan, and that FET will need to be replaced! Clearly the dissipation was higher than it should be. It could have been a rogue FET, or perhaps stray inductance/capacitance around it causing it to oscillate. If both FETs got hot (rather than warm) then it might indeed be necessary for a circuit rethink.

Edit: Just remembered the pair switch-over is only after 6hrs, so at 4-5hrs only 1 FET will have been on continuously. Safest to assume the other FET might have over-heated too. That being the case then, as you suggest, a separate FET for each pump would certainly reduce the chances of over-heating (the pumps could still operate as a pair). Heatsinking of the FETs would help too, but would probably mean extending the wires to the FET terminals; which increases the risk of oscillation. A series resistor (10-100 Ohm) very close to the gate terminal damps oscillation, and a ferrite bead on the gate wire might be advisable (so I read, I've no experience of using beads). I'll redraw the schematic to show the added FETs.

Further edit: Did you by chance check the current drawn from the power supply when the pumps were running? Anything over ~ 3A would indicate an overload somewhere, which could also have accounted for the over-heating.

 

[salty joe]

Thanks for looking at this, Alec.
It did not occur to me to check how much power the circuit was pulling. I'll make that standard procedure from now on.
A redraw is much appreciated. Any chance you could include the earlier revisions showing a pot at R3 (flick pot) and the connection between Ua2 and pin 10 of U1? Another revision to get the flick working was adding a 10uf cap paralell to C10 (0.47uf).

If C10 were 4.7uf, would that allow for finer tuning of the flick? As it is with a value of 10.47uf, (10uf + .47uf) the difference between a good "flick spin" (almost too much) and nothing is a blonde hair turn of the flick pot.

Also, I installed a soft start that is not shown on the schematic for each pair of pumps. BTW, I used a 390k for R6.

I'll get some more FETs and a few other parts coming my way, so in a week or two I'll get after this again.

Thanks Alec!

 

[alec_t]

I'll do a redraw incorporating those updates.
It occurs to me that the power supply getting hot suggests an overload somewhere. I don't see how that can occur unless there is some extra current flow in parallel with the pumps. That's a mystery as the circuit (apart from the path via the 12V regulator) is essentially in series with the pumps! Have you tried powering the pumps directly from the supply (minus the control circuit)?

 

[alec_t]

Yer tiz.
I've circled a few areas in the schematic where minor changes have been made, but with the major change of adding FETs (and their associated diodes etc) and the soft-start components I've now renumbered most of the parts in the right half of the schematic. Note R3 added in series with the flick pot to 'stretch' its useful adjustment range. I haven't shown the power supply part, as that remains unchanged. I think I've incorporated all the mods to date, but if I've missed something let me know.

 

[KeepItSimpleStupid]

A reminder that the junction temperature of the FET is 175 deg. C., so the FET internally operates at a very high temperature.

Just about any power transistor will self-destruct without a heat sink. Heat sinks can be sized using thermal resistances. I'm a little rusty doing it. The ideas is to use the "junction temp, ambient temp, junction to case thermal resistance, and the thermal resistance of the insulator to find the minimum thermal resistance of the heatsink that is required.

One of the leads of the FET is usually at one of the power supply potentials and therefore it is usually electrically insulated from the heatsink with a thermal washer. Their are two different techniques used. One uses thermal grease and the other does not. You can buy a "mounting kit". The mounting kit generally consists of a shoulder pad, screw, nut and a domed lockwasher. The domed lockwasher is better at applying constant pressure.

1. A sil-pad doesn't require thermal grease. It's texture is not smooth.
2. the other method traditionally used a mica washer. This has been replaced by other types of materials. It is important that only a "very thin" layer of thermal grease be used on each side.

When you drill the mounting holes make sure that their are no burrs.

 

[Boncuk]

A reminder that the junction temperature of the FET is 175 deg. C., so the FET internally operates at a very high temperature.

Just about any power transistor will self-destruct without a heat sink. Heat sinks can be sized using thermal resistances. I'm a little rusty doing it. The ideas is to use the "junction temp, ambient temp, junction to case thermal resistance, and the thermal resistance of the insulator to find the minimum thermal resistance of the heatsink that is required.

One of the leads of the FET is usually at one of the power supply potentials and therefore it is usually electrically insulated from the heatsink with a thermal washer. Their are two different techniques used. One uses thermal grease and the other does not. You can buy a "mounting kit". The mounting kit generally consists of a shoulder pad, screw, nut and a domed lockwasher. The domed lockwasher is better at applying constant pressure.

1. A sil-pad doesn't require thermal grease. It's texture is not smooth.
2. the other method traditionally used a mica washer. This has been replaced by other types of materials. It is important that only a "very thin" layer of thermal grease be used on each side.

When you drill the mounting holes make sure that their are no burrs.

All that tells me to eyeball the required heatsink as I have done for 50 years. The cooler the better!

Nowadays you have so many selections that it makes it impossible not to cool sufficiently. Use a copper core CPU cooler with fan to gain a thermal resistance of 0.4K.

Boncuk

 

[salty joe]

Thank you Alec for that redraw. And thank you for allowing a finer tuning of the flick! I have connected the pumps directly to the PS, and they run. I'll do so again and let you know what they draw.

I could thermal epoxy this heatsink https://www.ebay.com/itm/2-x-Heatsi...258?pt=LH_DefaultDomain_0&hash=item3cc0ef55a2 to the FETs if I don't crowd the FETs with other components. The FETs could still be connected by bending the legs and soldering directly to ground and other components-no wires needed. A little cooling capacity with no further risk of oscillation. Does that sound right?

Way back when, the very first circuit ran four pumps independantly, each pump with a dedicated FET (the same 3205) and a dedicated soft start. I ran four pumps all day in a bucket of water and the entire circuit stayed cool. Fingers crossed.

Thanks KISS and Bonuk for your input.

 

[alec_t]

Does that sound right?

Yes, but bear in mind that the heatsinks will be 'live', i.e. at the FET drain voltage (which can be 24V via the pump winding); so keep them clear of other wires/box. As you say, fingers crossed.
As for the flick, I think that 150k resistor in series with the pot will help. If not, you could experiment with other values (say 33k-470k).

 

[ronv]

Sorry to come in in the middle - or should I say the end.

Problem may be the very slow turn on of the FETs. Power dissapation will be very high for a tenth of a second or so. Heatsink territory for sure.

 

[alec_t]

Don't be sorry, Ron . Always good to have fresh thoughts on these things.
Slow FET turn-on arose from a 'soft-start' request and was chosen by the OP, in preference to a PWM approach, for simplicity. I had my doubts on the heating issue but the earlier project, referred to in post #370, also had this slow turn-on yet had no heating problem. The main difference between this second project (up to now) and the first is that currently one FET drives two pumps rather than one. That presumably has tipped the balance, though in theory (a wonderful thing!) there should be no great heating effect. That 1/10 sec or so is only once per 42 sec, worst case.

Edit: The fact that the pumps ran for 4hrs before the over-heating was noticed suggests to me that the problem is a continuous over-current or FET-not-fully-switched-on one rather than a start-up one.

 

[salty joe]

I installed the softstart between the FET gate and 10 ohm resistor. On the redraw, the soft start is on the other side of that resistor. Could that have a bearing on the FET heat issue?

If the soft start from post 173 will be less likely to fire up the FETs, I would not mind using that softstart.

Thanks.

 

[alec_t]

Could that have a bearing on the FET heat issue?

Quite possibly. The 10Ω resistor needs to be physically close to the FET gate terminal to be effective at damping oscillations. Circuit layout can be critical, as lengths of wire have inductance which resonates with gate capacitance.
Before considering the other soft-start I'd stick with the present one (which seems ok with your first project) but ensure that 10Ω resistor is right at the FET gate pin, then run the pump and monitor the FET closely for signs of getting hot.

 

[salty joe]

OK, will do. About the only place I've been using wire is to hook up the ICs. Mostly I've been bending and soldering component leads to each other. I'll be sure to cut the 10 ohm resistor's leg short before soldering it to the FET gate.

Just curious, would using four soft starts have an impact on FET temp?

 

[alec_t]

I'll be sure to cut the 10 ohm resistor's leg short

Good idea.

would using four soft starts have an impact on FET temp?

Depends what caused the heating. If it was FET oscillation then that could still occur with four: but if it was because one FET drove two pumps then going to one FET/soft-start per pump would probably help.

 

[salty joe]

If I understand this, a dedicated soft start for each pump might help, can't hurt. Is that right? If so, I'll throw in a couple more soft starts.

Finally got a chance to go down to the basement for a few minutes to see what the pumps draw. My DMM did not show any initial surge at all. Two of the pumps from the trial run showed amperage draw at 1.13-1.14. One pump showed 1.34 A and one showed 15.14 A. Ouch. The plug from that pump was melted into the jack, I had to cut it loose to to test it. That pump is toast I guess, it does not run. There was a 10A fuse installed on the positive lead directly from the power source that did not blow. This seems completely bogus to me.

The power source must have an automatic shutdown because it was off when i found the smoldering mess a few nights ago. I was certain the PS was a goner, but it still works!

 

[alec_t]

a dedicated soft start for each pump might help, can't hurt. Is that right?

Correct.

My DMM did not show any initial surge at all.

DMMs are fairly sluggish, so unlikely to register initial surge current.

one showed 15.14 A.

As you say, ouch! The 1.13-1.34A draw is as expected. 15A would certainly explain the FET and the power supply getting hot! I wonder what caused the pump to draw that much? Shorted turn in its winding, perhaps? Pity the fuse didn't blow, but unless it was a quick-blow type and the pump fault occurred suddenly that's not too surprising. I've had a failed domestic lamp take out a 30A main fuse yet its local 5A fuse survived! Glad the power supply still works. I'd be tempted to re-try this circuit with soft-starts for just 2 pumps for now and then expand to 4 if all is ok.

 

[salty joe]

hmmmm...If I take that pump apart, could I possibly determine the root of the problem? Besides trying to determine cause of failure, anything I should look for?

It is a quick blow fuse. It does not make sense to me that a fuse will take more current than it is rated for.