Help with Water Pump

[salty joe]

Yeah, the 0.22 is like a sausage. I'll try adding a 1.8K parallel to the 4.7K. That would be 2.9K, right?

Yes, the pump did start with 6 amps way back when.

 

[KeepItSimpleStupid]

Actually 1.3K. A parallel combination of any number of resistors is ALWAYS LESS than the lowest valued resistor.

R=1/(1/1.8+1/4.7) gives answer in K ohms.

Two of the same resistors, half the value. e.g. Two 8 ohm speakers in parallel means it looks like a 4 ohm load.

 

[salty joe]

Yeah, the 0.22 is like a sausage. I'll try adding a 1.8K parallel to the 4.7K. That would be 2.9K, right?

Yes, the pump did start with 6 amps way back when.

Thanks KISS_makes sense.
The pump started with 6 volts way back when.
Could you guys give an opinion on which of the three proposed fixes (D9-R13, snubber, resistor across Q1 base-emitter) Would be the least likely to cook the pumps?

 

[alec_t]

I'll try adding a 1.8K parallel to the 4.7K.

No. Use the 1.8k (or 2.2k) instead of the 4.7k.

Could you guys give an opinion on which of the three proposed fixes (D9-R13, snubber, resistor across Q1 base-emitter) Would be the least likely to cook the pumps?

Adding that resistor across Q1 base-emitter is the riskiest of the fixes. It's the only one which affects the current limit. But certainly add D9/R13 anyway.
I've simmed a mod which instead delays the onset of limiting by ~ 150mS, giving the motor a ~7.5A start current then retaining the 3A limit thereafter. The FET doesn't oscillate with that mod (though its current is chopped by the coil commutation) so the snubber probably isn't needed. That mod would, I think, be safer for the pump (in the case of a stall) than a steady 6A limit. I'll post it as PDM_Mk13 shortly.
Whichever mod is used the turn-on current will exceed the Schottky diode D2's steady-current max (3A) for at least 150mS. Not sure if it will survive. Have to check the datasheet.
BTW I wouldn't put all your eggs in one basket for further testing. Stick with one pump that didn't start, and one PDM for now.

 

[alec_t]

Ta-da.... The PDM_mk13
View attachment 68219
And this is the theoretical motor coil current waveform for a 2 sec run
View attachment 68220
I've also attached the .asc file for anyone who wants to play.

Differences from the Mk12 version
are highlighted by dotted rectangles and are:-
1) R4 is now 4.7k,
2) R6/D6 have been removed since the D9/R13 mod makes them redundant,
3) C4 is added,
4) R5 has been re-purposed and with C4 serves to delay current-limiting,
5) D5 cathode now connects to C4 instead of Q1 collector.

Simulation
The model of the motor is drastically changed and now accounts for coil commutation by the Hall IC in the motor. The motor is modelled as the series connection of a coil, a back-emf voltage source and a switch.
The model is based on these assumptions:-
1) Rotor torque α [is proportional to] coil current
2) Rotor acceleration α (torque - counterforce)
3) Counterforce [= windage + friction + viscosity] α rotor speed
4) Rotor speed = integral of acceleration
5) Back-EMF α rotor speed
6) Commutation rate α rotor speed
7) Hall switch duty cycle ~ 90%

Constants of proportionality (k1, K2, k3) used in the sim were chosen by trial and error to give a motor which gets up to speed in ~ 1 sec, with an rpm ~2000-3000 and a running current ~ 1A.

 

[KeepItSimpleStupid]

Joe:

The snubber thing is benign. Here is a tiny bit of reading material: https://en.wikipedia.org/wiki/Snubber Manufacturers' might omit to save cost.

The reverse biased diode and the TVS are both snubbers. In terms of transient protection, you already have it covered. An RC snubber does something else that transient protection can't. It limits the rate of rise of current (dv/dt). Ceramic capacitors by themselves across a motor, close to the motor, are used suppress RF (Radio Frequency).

 

[salty joe]

I am working with a PDM with D9-R13 installed. Trimmer set at 25.something. D9 cathode is connected to 12V. It does not have a resistor at Q1. Pumps A, C and E don’t toggle reliably-pumps B, D and F toggle reliably, same as PDM with 4.7 resistor across Q1 emitter-base.
I get the same values with both PDMs. Pump stalled- pump 21.3, Q1 gate 10.8, Q1 drain 0.3, Q1 source 0.1. I switched the DMM from 200VDC to 20VDC to get better numbers for the FET. Q1 gate 10.75, Q1 drain 0.22, Q1 source 0.03.
I plan to stick with this PDM and pump A.
Thanks for all of your help everybody!

Two posts snuck right in there, thanks.

 

[salty joe]

Does this look like a good fit for D2?

**broken link removed**

 

[alec_t]

Seems very expensive. Ok apart from price. What diode do you currently have for D2? Hopefully it will cope with a brief overload. (The MBRS340 used in the sim can take 80A for ~10ms non-repetitively, so should handle a couple of 7A 50ms pulses)

EDit:

D9 cathode is connected to 12V

If you're toggling the PDM control voltage (+Vt) you also need to toggle D9 cathode at the same time for the D9/R13 mod to have any effect. It's not going to make a reluctant pump start, but is to allow repeated toggling of a willing pump.

 

[salty joe]

This is my current D2 https://www.ebay.com/itm/50-x-1N582...922?pt=LH_DefaultDomain_0&hash=item3a77dceae2

I'll toggle D9.

The price of the 7.5 A schottkey includes shipping, so maybe not so bad?
Thanks for a solution-the only bummer is I don't have room for a heatsink on D2 and it looks like the 7.5A schottky needs one. I guess I could mount the 7.5A schottky remotely with a heatsink to test and then re-build. Or maybe another small board with all six 7.5A schottkeys and heat sinks.

 

[alec_t]

I see the 1N5822 is rated for a 8ms surge of 80A, so it should be ok as is for ~7A brief pulses. The exact pulse duration in your motor depends on the unknown coil inductance with the rotor in situ in the motor, on the Hall switching characteristics, on the rotor acceleration etc, so is uncertain. If it were my motor I'd just use the 1N5822 and hope for the best, or even omit D2 altogether; but if you're aiming for 24/7/365 reliability you might want to use a higher-rated diode. The reason for having D2 is to prevent reverse drive to the power supply in the unlikely event that there is pump over-run when the power goes off. The Hall IC may already prevent that happening. We just don't know.

 

[KeepItSimpleStupid]

Joe:

What's your aversion to buying things in the US. If you need one you can order one from Digi-Key although there may be a small order charge, not sure, but postage will probably be $10. https://www.digikey.com/product-detail/en/MBR1535CTG/MBR1535CTGOS-ND/1477165

The package has two diodes in it. Heatsinking should not be required because the current is not continuous.

 

[salty joe]

Thank you Alec for Mk13.
I'll leave D9-R13 and toggle D9, then incorporate the Mk13 mod. I have enough parts on hand to modify one PDM, except for the 7.5A schottky.
Should I check V from Q1 emitter - FET and pump V as before?
I do need to have 24/7/365 confidence, so I'll switch out D2 for the 7.5A schottky when it gets here. Is it agreed that a heatsink won’t be necessary? Could it be that the 7.5 schottky runs cool for quite awhile, then something occurs that cause it to cook?
BTW, while I was testing the pumps the alarm did not sound when the motor stalled.

 

[alec_t]

except for the 7.5A schottky.

D2 doesn't need to be a Schottky type, so it may be cheaper to use a standard 8A power diode such as a BYC8D-600:
https://uk.mouser.com/ProductDetail...=sGAEpiMZZMtvcUztdGSumH77jbR4ZZVqdJQVmHvU3uE=

Should I check V from Q1 emitter - FET and pump V as before?

Useful to have those measurements.

Is it agreed that a heatsink won’t be necessary?

Agreed. The ~7-8A current should persist for only tens of millisecs. Thereafter it won't be more than ~3A. Either the pump will run properly, dropping the current down to ~1A after ~ 1sec, or else if the pump stalls the trip should switch it off after ~ 1 sec if the average current is ~3A at that time. So the worst case should be ~ 3A average for ~ 1 sec. Shouldn't get more than warm.

BTW, while I was testing the pumps the alarm did not sound when the motor stalled.

I'm guessing the pump rotor started to turn sluggishly, then the Hall switch commutator shut off the coil current before half a rev, the rotor was pulled back magnetically to its starting point and this sequence was repeated. That could have caused the vibration you detected and kept the average current down to ~ 1A.

 

[ronv]

The only thing I noticed with the new circuit - and I don't know if it happens- is that on initial power on if the timer is on there is no high current pulse because C4 is not charged up. It recovers on the next cycle, but may confuse debug.
Finally understand the oscillation. Once the FET starts to turn off in the current limit cycle one would think the "kick" would go thru the clamp diode, but since the FET is not really off, only turned down, the inductor current continues to flow thru the FET and the sense resistor. I could make it go away by adding a snubber from pump- to ground. Think I used 47ufd and 5 ohms.
The alarm not sounding may be the FET threshold vs the logic threshold thing.
All those seem to be gone in the new circuit.

 

[alec_t]

The only thing I noticed with the new circuit - and I don't know if it happens- is that on initial power on if the timer is on there is no high current pulse because C4 is not charged up. It recovers on the next cycle, but may confuse debug.

Good catch. I don't think it will be a problem in practice if Joe has the timer set for 5 sec on 5 sec off for testing the Mk13. Just means a 15 sec (?) wait from power-up before we can be sure a pump is/isn't going to run. We could probably design round it by modifying the timer circuit to delay +Vt rise after power-up, but I don't think that's really justified at this point. In practice the 24V power supply may power up slowly anyway, so that too could prevent a high current pulse in the first cycle.

 

[KeepItSimpleStupid]

alec: An aside

I once had the just opposite problem. I was using a SBC (Single Board Computer) with a switching power supply and all was well. When I used a linear one, the processor would not Power on reset sometimes. Fortunately, the fix was easy. Replace a chip with the schmidt trigger version and the chips were socketed.

In general, a switching supply powers up faster than a linear one.

 

[salty joe]

Thanks all.
I tried to modify the existing PDM, but was having such an awful time I decided to start fresh. I ordered enough parts for six Mk13s. This time, I'll build just one in case it needs tweaked.

If I understand this correctly, the pumps might not run for the very first toggle with the Mk13. That's is definitely good to know. It sounds like no big deal.
Anyway, as soon as my first Mk13 is built, I'll touch base.

 

[alec_t]

I tried to modify the existing PDM, but was having such an awful time I decided to start fresh.

That's a shame; but at least you should have graduated at soldering academy by now . I expect a matrix board layout is trickier than a stripboard one to modify because you may have several connections to a single point. I'm sorry all the mods have arisen; but each trial seems to reveal some new thing about the pump characteristics which needs a work-around.

This time, I'll build just one in case it needs tweaked.

That's the recommended procedure.

 

[salty joe]

Every time I start to solder I determine to make it perfect, but after a little while realize making it work will have to be good enough. One more shot at a really nice job.....maybe not graduated but getting there.

By Alec "That's the recommended procedure."
Somewhere in this thread that was probably mentioned, IDK, but I'm just gonna consider this a lesson from the school of hard knocks.

This driver has been a tough nut to crack-I very much appreciate all the help.