Help with Water Pump

[KeepItSimpleStupid]

The ground leg of the 12V regulator is probably the best point for linking the heavier conductor PowerGnd wire to the lighter conductor SignalGnd wire.

OR you can take it all the way back to the power supply (-) terminal.

 

[alec_t]

Yup, that will do nicely.

 

[ronv]

I'm back to the 317 idea.

Here is my thinking:

Limit the output voltage to 20 volts just in case that is the maximum gate to source voltage on the FET.

The LM317 has safe operating area protection that will limit the maximum current to 3 amps similar to the factory controller. See page 14 of this app note.

https://www.electro-tech-online.com/custompdfs/2012/06/f4.pdf

Using this should eliminate the need for soft start as the current limit will limit the inrush to 3 amps from 9 or 10 amps.

Q3 can be used to controll on off and to shut it down for an overcurrent of 2.5 amps for a second or 2, whatever is decided to get started.

A small heatsink would be needed for normal operation - 4 watts at say 6 or 8 C per watt.

Seems given the absence of specs on the pump parts this would be about as safe as normal controller. With the addition of a pot it could also be made adjustable like the regular controller.

 

[alec_t]

Alas, I can't get that sim to put out more than 15V and a current of 800mA, and only then if the pump resistance is modelled as 18Ω If we set L1 = 3.4mH and series resistance = 2.14Ω (the actual measured values for the pump) then the output drops to a 0-4V oscillation at 2.5kHz

Edit:
According to the datasheet the 'typical' current limit for the LM317 is quoted as 2.2A and the minimum as 1.5A. Tempting as it is to use one, if it internally limits at either of those values then, as has been questioned many posts ago, would there be enough oomph to get a pump started in water? The OEM controller can actually put out 6A peak, according to its spec.

 

[salty joe]

. I was assuming you'd salvage bits from your present build; or will you be starting from scratch? .

Everything is soldered really tight to the board-the legs are bent and soldered-so I'll be starting from scratch.

It only makes sense if the pumps are still under warranty. They'd be better fitted with heavier duty plugs and corresponding sockets for reliability. Just for testing you can probably get away with the present connectors. Just watch out for smoke signals .

Makes sense. Heavy duty jacks and plugs it is.

Salty:

Do yo know about www.mouser.com (TX), www.jameco.com (CA), www.digikey.com (Minnesota) and www.newark.com (IL)?
Allied electronics is more of an industrial supplier of electromechanical parts: https://www.alliedelec.com

I did not know of all of them. Thanks.

OR you can take it all the way back to the power supply (-) terminal.

So far, 22 gauge single strand wire has been used for everything. I have 19 gauge single strand wire, should I solder that to my heavy duty jacks?

Most of the parts are already in my basement. I'll get the rest ordered tomorrow.

Thanks everybody for all your help. Thanks for hanging in there and making this thing happen.

 

[()blivion]

Alec_t. Nice work on the schematic as usual.

Just a thought. Can we put a NTC thermistor in the PCM board in thermal contact with M1 and wired into the alarm circuit? Or maybe kill two birds with one stone and mount the PTC in thermal contact with M1 since PTC's also trip by heat? Just in case god decides that 1.1A X 24V would be better dissipated in our non-heatsinked FET than do actual useful work. I know it's unlikely to be needed with this layout, but... I like thermal protection and it should be a trivial addition. :shrugs:

Also, can't the PCM circuit us just a normal diode forward biased instead of a zener for the trip circuit? Or am I missing something?

@Salty

Everything is soldered really tight to the board-the legs are bent and soldered.

I, like many others here have a desoldering setup that would make short work of that problem, if it's worth the shipping? Or you could go to RadioShack (the cell phone store) and get a solder sucker yourself for ~10$ (That is if they haven't discontinued that as well.)

 

[ronv]

Alas, I can't get that sim to put out more than 15V and a current of 800mA, and only then if the pump resistance is modelled as 18Ω If we set L1 = 3.4mH and series resistance = 2.14Ω (the actual measured values for the pump) then the output drops to a 0-4V oscillation at 2.5kHz

I think we may both have the same problem with the 500 ma 317 model. But having said that take a look at the LT1084 instead. I think it is 5.5 amps minimum. If not enough go to the 1083. The inductor model is not really to good as the the pump circuit probably won't turn on until it has 4 or 5 volts across the pump and the inductance with the rotor installed is probably closer to 10 or 12 mh.
The motor FETs are probably okay at 10 amps for a couple seconds, but who knows.

 

[4pyros]

I have no practical experience of resettable fuses so will start another thread to seek advice.

Sorry I am late I was out doing fireworks today.
Can you put a link to the new thread back in this one?

 

[alec_t]

I have 19 gauge single strand wire, should I solder that to my heavy duty jacks?

Suggest 19AWG from 24V power supply input to TAM and for any connection to PowerGnd; 22AWG for the remainder.

Or maybe kill two birds with one stone and mount the PTC in thermal contact with M1 since PTC's also trip by heat?

I like that. Good idea. (Joe, the PTC is the resettable fuse).

Also, can't the PCM circuit us just a normal diode forward biased instead of a zener for the trip circuit?

I think it's stretching things to call the circuit PCM when the pumps are only switching every 30 secs or so . As for the zener, I was thinking it might give a better-defined trip point less subject to temperature drift than just a diode (and/or the b-e junction of the transistor). I'll play with the sim and reconsider.

@4pyros
Here's the link to the other thread
https://www.electro-tech-online.com/threads/selecting-a-resettable-fuse.128099/
@ronv
Yes, the 1084 would be better than a 317. There would still need to be quite a few components in addition (for sensing that the 1084 was current-limiting) because we need to detect rotor lock or other pump failure).

the inductance with the rotor installed is probably closer to 10 or 12 mh.

Good point. I'll re-run the sims with that value, but it seems to be resistance which dominates anyway.

 

[alec_t]

I've re-run the sims.
Results:
Without the zener diode the setting of the trimpot is critical and the drift in trip point from 20°C-40°C is ~ three times greater than with the diode and is quite noticeable.
Using 10mH instead of 2mH as the coil inductance value had very little effect, except for a slight increase in current rise time at t=0.
Using a 1084 (or presumably any other 3-terminal linear regulator) as a controlled voltage source for the DPM-Mk2 introduces the problem of heat dissipation. With a stalled rotor and with the ADJ input of the 1084 grounded to minimise the 1084 output voltage (and hence the pump current) the 1084 will dissipate 13W unless something else (hey, what about a FET!) is added to switch it off completely. With a short-circuit the 1084 dissipation is 18W.

 

[()blivion]

I think it's stretching things to call the circuit PCM when the pumps are only switching every 30 secs or so .

LOL, I prolly should have used a better acronym eh? "Pump Control Module" is what I was going for.

 

[ronv]

Using 10mH instead of 2mH as the coil inductance value had very little effect, except for a slight increase in current rise time at t=0.

It's only important for inrush current and running current should you ever need it.

Using a 1084 (or presumably any other 3-terminal linear regulator) as a controlled voltage source for the DPM-Mk2 introduces the problem of heat dissipation. With a stalled rotor and with the ADJ input of the 1084 grounded to minimise the 1084 output voltage (and hence the pump current) the 1084 will dissipate 13W unless something else (hey, what about a FET!) is added to switch it off completely. With a short-circuit the 1084 dissipation is 18W.

During locked rotor and the input grounded the output is only 1.4 volts. So the motor FETs won't turn on.
With short circuit the built in short circuit protect takes over.

But the FET is fine. I was just concerned about the motor FETs having a go at 9 amps for a few seconds.

Here is the best P FET I could find at Mouser with a 30 volt gate to source rating in a DPAK. Do you think it can handle 10 watts? If so just put the fuse in, put a low current alarm across it and let it cycle.

https://www.electro-tech-online.com/custompdfs/2012/06/FQB22P10TM_F085.pdf

 

[alec_t]

During locked rotor and the input grounded the output is only 1.4 volts. So the motor FETs won't turn on.

Probably not. I guess it depends on the unknown FET spec. Could they turn on partly?

With short circuit the built in short circuit protect takes over.

True, but there's still the little problem of getting rid of 18W of heat over a lengthy period (6 hours, with a duty cycle of 92%) if there's no additional switch.

Do you think it can handle 10 watts?

Yes, with adequate heat-sinking.

 

[ronv]

True, but there's still the little problem of getting rid of 18W of heat over a lengthy period (6 hours, with a duty cycle of 92%) if there's no additional switch.

No, they won't start into a short.

Yes, with adequate heat-sinking.

Hmmm. I guess I thought that was how we got here.

What are the possibilities?:

The coil just shorted.
The FET just shorted.
The pump stalled and the FET couldn't take it.

 

[alec_t]

No, they won't start into a short.

I'll take your word for that, but LTSpice says otherwise for the LT1084. However, I was forgetting the Hall device in the pump; that needs a few volts before it will turn on the pump FETs. I've re-run a sim with the pump modelled as the coil in series with an NFET, and the Hall device modelled as a 3V3 zener feeding the FET gate. Better model? Grounding the ADJ pin then cuts off the (non-shorted) pump and 1084 current without having to use another FET. With the ADJ pin high the 1084 output switches high and low at ~ 130Hz, so that would enable rotor lock to be recognised and steps taken to ground the ADJ pin. Not sure if the pump would like being pulsed at 130Hz, albeit with current limited to 6A. The concensus in posts above was that pulsing it (PWM) might be detrimental. Both coil and internal FET dissipate many Watts during the pulsing.

What are the possibilities?: The coil just shorted. The FET just shorted. The pump stalled and the FET couldn't take it.

Chicken and egg comes to mind. We know the coil and external FET fried but can't prove whether the pump internal FET was ok or not.

 

[KeepItSimpleStupid]

What are the possibilities?:

The coil just shorted.
The FET just shorted.
The pump stalled and the FET couldn't take it.

Then there is the role of the solder blob, possible poor design of the electronic commutator and the lack of current limiting which exists in the OEM controller and the possibility of rotor lock.

So, externally everything has to be done to reduce the possibility of failure with the three major points being: 1) Prevent damage due to Rotor-lock, 2) slow-start and 3) Spike suppression. (1) and (3) should be in the motor.

So, we are not out of the woods yet.

Coil collapse can still damage the commutating F

 

[alec_t]

I'm having second thoughts about my PDM-Mk2 circuit in post #612, following Ronv's input and further simulation. The real bugbear is rotor lock. Asking the pump to survive 9A for even 0.5 sec may be too much. So don't be in a hurry to build it, Joe! Ronv's suggestion to use a 1084 voltage regulator as the basis for a drive module is looking more promising, with the regulator internally limiting current to 6A. More thinking to do......

 

[KeepItSimpleStupid]

Way, Way back I suggested using the same or similar simple switcher that the OEM controller uses because it offers a few capabilities:

1. ON/OFF control
2. Current limiting that the OEM uses
3. Easy to change the output voltage for feeding mode.
4. Could even rob the parts from the OEM controller or just "cut the section out", PCB and all. Maybe add a standoff here and there.

Combined with the "slow-start" and "current exceeds x for x time" ought to work.

"slow-start" could be done using the output voltage divider similarly as the OEM controller does it.

Add a "Hall effect" current sensor and you don't even need another FET.

 

[salty joe]

So don't be in a hurry to build it, Joe! QUOTE]

Ok, alright! Thanks everyboby.

 

[ronv]

I'll take your word for that, but LTSpice says otherwise for the LT1084.

Now we are getting on the same wave length. It's still a litttle marginal. I'm thinking maybe the heatsink is ~30C per watt so at 6 amps it is still going towards 160 C. But at 9 amps we're looking at 330 C at least. Thats solder blob teritory.

However, I was forgetting the Hall device in the pump; that needs a few volts before it will turn on the pump FETs. I've re-run a sim with the pump modelled as the coil in series with an NFET, and the Hall device modelled as a 3V3 zener feeding the FET gate. Better model?

I think the pump uses a P Fet. I haven't tried that, but I'll give it a shot. I once tried to build a motor model but all I did was pull my hair out.

Grounding the ADJ pin then cuts off the (non-shorted) pump and 1084 current without having to use another FET. With the ADJ pin high the 1084 output switches high and low at ~ 130Hz, so that would enable rotor lock to be recognised and steps taken to ground the ADJ pin. Not sure if the pump would like being pulsed at 130Hz, albeit with current limited to 6A. The concensus in posts above was that pulsing it (PWM) might be detrimental. Both coil and internal FET dissipate many Watts during the pulsing.

I used a lower voltage than 24 so the 1084 would start. That may be the difference. Where I ended up was with a 1/2 ohm in series with the regulator and running it at 20 volt input voltage. The thought being that maybe the FET has a 20 volt gate to source maximum so maybe we should stay under it. It's probably a 30 volt FET so maybe not required.

I haven't been able to duplicate the 130 HZ oscillation with a short. It does oscillate at a low voltage if you make the inductor 0 ohms but a short seems ok. The regulator will get hot but its protection circuity should protect it. ~ 15 watts.

Here is the sim I used.

We would still need something like you had in one post to shut everthing down after a couple of seconds