Help with Water Pump

[alec_t]

I used a SPDT to switch over if I get flick mode

Good idea.

Oddly, when the pump that toggled and ran the way it's supposed to, the run time (toggle time) decreased from about 14 sec. to about 11 sec.

Maybe the pump current is causing a voltage drop in the ground line which is affecting the tidal timer module. I'll play with a sim to see if I can replicate the effect. Enjoy the camping trip, Joe.

 

[ronv]

Alec,

1-Seems like he said the input was 0 to 12 at 15 and 5 seconds. That tells me he isn't bothered with flick.
2- LED on motor off led me to the low voltage across the pump. Confirmed at 3.1 volts.
This makes me think the motor resistance is lower than we think or there is some other load or the .22 isn't right.
3- I think there is a hole in the logic where depending on the FET threshold voltage the motor can be in current limit without tripping the first logic gate. Depending on what FET you have in the model try making the diode from the curent sense to the gate a 5717. In my sim with the thresholds set at 6v +/- 1.2 volts the first gate would not toggle for some levels of over current. You can also step temperature to watch it.

 

[alec_t]

2- LED on motor off led me to the low voltage across the pump. Confirmed at 3.1 volts.
This makes me think the motor resistance is lower than we think or there is some other load or the .22 isn't right.

I must confess I haven't got round to analyzing why/how the pump voltages vary. I'm sure we can take the motor resistance as 2.14Ω, since KISS has actually measured this on a gutted pump. The .22Ω is less certain, as the solder and wiring resistances at the Rsense terminations, in series with the actual sense resistor, could be a significant fraction of that and would result in the trip happening at lower current than designed.
Stiction would obviously vary from pump to pump and could account for why 2.4A (or less, depending on that nominal 0.22Ω) allows one pump to start but not another.

3- I think there is a hole in the logic where depending on the FET threshold voltage the motor can be in current limit without tripping the first logic gate

A distinct possibility. Sounds like you've investigated that further than I did. Any suggestions? Someone mentioned many posts ago that we might need to raise the current threshold level if pumps were reluctant to start.

Edit: Threshold level is most easily increased by putting a shunt resistor from Q1 base to ground.

 

[ronv]

Yep, I think there is still some stuff wrong. Your right probably finding out why the motor voltage is so low is the first one. I suspect your right, another little bit of resistance in the FET ground path would do it.
To get it to trip the overcurrent latch I think another diode (or 2) in series with D5 would make the input to the latch go lower by a few tenths before the FET was current limited. But this will just get Salty some noise to go with the stalled pump. Not sure which one to try and fix first since right now it gives him 15 seconds to read the voltage across the pump and after the alarm works only about a second.
The other thing I noticed was the second start pulse triggers the alarm sooner because the cap hasn't had a chance to fully discharge before it gets hit again with the second start pulse. This could explain why it would run once. Could just make the time out a little shorter than the fastest cycle time???

I wonder if there is a time when both coils are energized. That would also fit with the low voltage.

 

[alec_t]

I think there is a hole in the logic where depending on the FET threshold voltage the motor can be in current limit without tripping the first logic gate

I think you've hit the nail on the head there, Ron. I've now looked into this a bit more. Re-visiting the datasheet for the IRF3205 which Joe is using, the graphs show that for a 'typical' 3205 with Vgs = 4.5 the Vds is ~ 0.2V for a drain current of 3A. In other words the FET isn't fully turned on at Vgs = 4.5V. Even so, it would cause Q1 to pull the U1 input down just below the Schmitt threshold (using your example of the thresholds being 6V +- 1.5V) and trip the latch. So operation is marginal.
However, Joe's FET examples won't be 'typical' . Looks like they need a slightly higher Vgs to turn them on enough to give ~3A and cause the motor to run. In which case they won't necessarily get U1 input below the trip point and will leave the motor in a twilight zone, as is now happening.
Three options occur to me to overcome this and make tripping more dependable:
1) replace the FETs with logic-level types,
2) use an op-amp instead of Q1 to set the trip,
3) add another transistor stage.

Option 1 could be expensive; option 2 would mean quite a change to the circuit layout; option 3 seems simplest with least change to the layout.
Here's option 3....PDM_Mk11 :
View attachment 67938
In this circuit the component changes are within the dotted-line. Q3 and R11/12 have been added and U1d is now redundant (but it's inputs need to be connected to +12 or ground, whichever is convenient).
I've run a sim on this using various FET models, plus forcing various voltage levels on Q1 collector. The circuit should now trip when Q1 collector is anywhere within the 2-10V range, so should cope with almost any N-FET.

Edit:

Could just make the time out a little shorter than the fastest cycle time?

A possibility; but a shorter time would raise the average power dissipation in the FET, which is already pretty high when the trip kicks in.

 

[ronv]

Hi Alec,

I couldn't get that one to work right. For some reason the trip level on U1A was around 3 volts instead of more like 6.2 volts. That's probably the difference in our sims. I used inverters and set the hysterisis based on this data sheet:

https://www.electro-tech-online.com/custompdfs/2012/10/CD4093BC.pdf

I also made R9 a little smaller-- maybe 1.5k.
Along the way I discovered that the diode can't have very high leakage in the reverse direction or it effects the time out. So I just used a 1N914.. Don't know what Salty has for diodes.
I also added a little diode resistor network to speed up the discharge of the cap when the input to the driver turns off to kind of reset it faster.
Attached is a sim with the 2 diodes in the current limit path. These should also be silicon so they have the larger voltage drop.
Take a look and see what you think.

Ahh I think the 4093 model keeps the 5 volt thresholds even when running at 12 volts.

 

[ronv]

One other thing just struck me. When we tested the pump at 6 volts there was no current limit, so if per chance the pump "wakes up" with both coils on we would see just 3 volts across the motor. If that's the case we may need to add one of Joe's 47Ufd caps from base to ground of the current sense transistor to give it a "kick" to get it started.

 

[alec_t]

For some reason the trip level on U1A was around 3 volts instead of more like 6.2 volts. That's probably the difference in our sims. I used inverters and set the hysterisis based on this data sheet

I think the 4093 model keeps the 5 volt thresholds even when running at 12 volts

On my sim the Schmitt levels for the 4093 are 5V and 7V on a 12V supply. As you say, there are differences between our two sims.

Don't know what Salty has for diodes.

1N4148, as far as I know.

Take a look and see what you think.

Ran your sim. That R/D addition certainly brings the cap voltage down quicker, but I didn't find it necessary in my sim. If the trip circuit misses the occasional reset it hardly matters in practice, as it only delays by ~30 sec the next attempts at re-starting a stuck-rotor pump.

if per chance the pump "wakes up" with both coils on

I don't see how it can. The Hall chip in the pump will only allow one coil at a time to be energised.

we may need to add one of Joe's 47Ufd caps from base to ground of the current sense transistor to give it a "kick" to get it started.

I've already simulated that, thinking to delay the current limiting slightly. The problem then is that the FET oscillates at a few kHz, the current limit varying +- 0.2A about the mean 2.9A, and there would be a lot of RFI.

 

[ronv]

On my sim the Schmitt levels for the 4093 are 5V and 7V on a 12V supply. As you say, there are differences between our two sims.

Hmm. interesting. Where did you get your CD4000 collection? I think mine came from Eric, but yours are closer to correct. As close as I could guess from the data sheet they would be 4.8 and 7.2.

1N4148, as far as I know.

Thats cool.

Ran your sim. That R/D addition certainly brings the cap voltage down quicker, but I didn't find it necessary in my sim. If the trip circuit misses the occasional reset it hardly matters in practice, as it only delays by ~30 sec the next attempts at re-starting a stuck-rotor pump.

I wasn't thinking about the reset as much as the 1 second delay gets shorter when the pump starts the second time.

I don't see how it can. The Hall chip in the pump will only allow one coil at a time to be energised.
I've already simulated that, thinking to delay the current limiting slightly. The problem then is that the FET oscillates at a few kHz, the current limit varying +- 0.2A about the mean 2.9A, and there would be a lot of RFI.

I don't either. It just seems like it fits with the 3 volts he is seeing across the motor. Guess some more measurements are in order.
I didn't see the oscillation. What FET are you using?

PS Would you mind sharing your CD4000 files?

PSS what do you think about the 2 diode fix?

 

[alec_t]

@ronv

Where did you get your CD4000 collection?

Direct from the LTSpice user group on Yahoo (as did Eric, I think; so you should already have the CD4000 collection that I do). Your 4093 model seems the same as mine, but I see you were using the generic LT inverter model rather than the CD40106 model.

I wasn't thinking about the reset as much as the 1 second delay gets shorter when the pump starts the second time

It does. Which is why C3 is best biased just below the lower Schmitt threshold by R5/R6. The first time is arguably the most important.

That R/D addition certainly brings the cap voltage down quicker, but I didn't find it necessary in my sim.

I realise now that I was running the sim with a 'stuck-rotor' pump model. I've since made a (albeit crude) free-running pump model and found that C3 voltage then does indeed fail to drop low enough between successive pump turn-on attempts (if their period is less than the trip reset period, without your R/D addition. That seems to be the cause of 'it only toggles once'. Good call.

It just seems like it fits with the 3 volts he is seeing across the motor

It's possible. I'm guessing the 3V (and the other odd voltages Joe found) is more likely due to the FET not switching on fully, so not providing enough current either to operate the trip or get the pump up to full speed. At half speed, for example, the back-emf could still be low enough to allow a current of, say, 1.5A. The twilight zone.

I didn't see the oscillation. What FET are you using?

IRFH5207 (don't have a model of Joe's IRF3205). BTW the oscillation frequency with 47u on Q1 base is ~300Hz, not a few kHz as I mis-remembered.

what do you think about the 2 diode fix?

Assuming the lower Schmitt threshold is 5V then by my reckoning any FET which needs Vgs > 5V to give Ids > 2.9A (the current-limiting threshold) won't work reliably in the Mk10 circuit. The 2-diode fix (i.e. another diode in series with D5) would, I think, enable the use of FETs needing Vgs up to 5.6V which might well solve the problem. Adding Q3, on the other hand, would allow for FETs needing Vgs up to 10V or so. It's Joe's call.

Here's (yet another!) revision .....PDM_Mk12 (Joe, forget the Mk11 ). The differences from Mk10 are within two dotted-line fences. R13 is 33k because Joe already has that value elsewhere, but could be anything from ~10k-100k or more. I think this version, with ronv's suggestions, now incorporates what's needed to overcome (a) the twilight zone with the FET not fully on when it should be and (b) the 'it only toggles once' problem.View attachment 67979

 

[ronv]

Hmmm. Do I need to change anything but Vdd to get the thresholds right?
I used the inverter so I could set the thresholds when the 4093 didn't work right for me.

Yes, probably should understand the low pump voltage before anything else.

The transistor solution will allow a wide range, I was looking for a simple rework with the diode.

 

[alec_t]

Do I need to change anything but Vdd to get the thresholds right?

I just set Vdd to 12V and the model automatically gives 5V and 12V thresholds. I've never tried tinkering with the threshold parameters (Vt, Vh) directly. For this project I just monitored the Q1 collector voltage to assess whether a hypothetical 4.5V or 7.5V threshold would have been reached. Eric's a whizz with sims. A PM to him perhaps if you want to check?

 

[salty joe]

Enjoy the camping trip, Joe.

Thanks Alec. It's amazing that a 3 hr. drive can put you in place where you can see the milky way. No cell phone reception, kids playing in the stream, hiking Cook's Forest, it was nice.

U1d is now redundant (but it's inputs need to be connected to +12 or ground, whichever is convenient).

So, I need to disconnect the three pins from U1d and either run them to ground or 12V?

If I have time tomorrow and you give me the green light, I'll try this mod on one driver.

Thanks one and all.

Meanwhile, would it be useful to record the resistance of each pump?

 

[KeepItSimpleStupid]

See what happens when you go away? Lots of activity.

U1d: Only the input (the pin on the left side of the schematic) needs to be tied to ground or Vdd. Your choice, but since it's an inverter, I would tie it to +12 because there would less power dissipation on the output. Very slight to the point it doesn't matter, but since this is going to be running 24/7/365, I'd go with +12.

For the benefit of Salty, CMOS logic chips MUST have the inputs tied to Vss or Ground, otherwise they dissipate more power and oscillate. Sometimes when they oscillate, they destroy themselves because they get too hot.

Joe:
You probably can't measure resistance accurately. I'd guess +-1 ohm. Remember what i suggested?
Take the power supply in series with the 0.22 resistor. Mark it, so you know which one you used.
Turn each pump on.
Measure the voltage across the motor and the 0.22 resistor.

Now you can computer R, to the accuracy of the 0.22 resistor. R=V/I On a relative standpoint, you will be riight on,

 

[ronv]

Welcome back. Sounds like you had fun. I'm not long back from Colorado. My Son took me fishing where we could walk across the Colorado river. Nice to get back to nature once in a while.
If you have a chance while Alec is sleeping maybe you could measure a few points with your trusty volt meter.
With the ground lead close to the emitter of Q1.
The drain, source and gate of the FET and at the pump.
If you can do it with the pump that runs once then stalls it would give us a good idea.

I don't think measuring the pump will help because the driver is in the way, but you could try it.

 

[KeepItSimpleStupid]

@Ronv @Joe:

I want Joe to measure the pump without the driver. Just the power supply, A 0.22 resistor and the pump. Sorry if I wasn't clear.

 

[ronv]

Sorry KISS, I was responding to Joe's post not your edit.
I think he will pretty much get that measurement if he uses the pump that runs once the stalls once, but it sure wouldn't hurt to know the run current.

 

[KeepItSimpleStupid]

I hate when that happens. Two people responding at nearly the same time.

 

[salty joe]

See what happens when you go away? Lots of activity.

Very cool-thanks!

U1d: Only the input (the pin on the left side of the schematic) needs to be tied to ground or Vdd. Your choice, but since it's an inverter, I would tie it to +12 because there would less power dissipation on the output. Very slight to the point it doesn't matter, but since this is going to be running 24/7/365, I'd go with +12.

On the left side of U1d, I connected pins 8 and 9 together and ran it between R4 and D5. Are you saying take it away from R4/D5 and send it to 12V?

Joe:
You probably can't measure resistance accurately. I'd guess +-1 ohm. Remember what i suggested?
Take the power supply in series with the 0.22 resistor. Mark it, so you know which one you used.
Turn each pump on.
Measure the voltage across the motor and the 0.22 resistor.

Now you can computer R, to the accuracy of the 0.22 resistor. R=V/I On a relative standpoint, you will be riight on,

OK, put a .22 resistor on the +side of the power supply. Hook the pump to the resistor and the – side of the power supply. Measure V by putting black DMM lead on ground and red DMM lead between .22 resistor and pump. Is that right?

“With the ground lead close to the emitter of Q1.
The drain, source and gate of the FET and at the pump.
If you can do it with the pump that runs once then stalls it would give us a good idea.”

I’m guessing the emitter is pin one, which in Q1 goes to ground. So you’re saying hook up a pump that only toggles once, touch DMM ground lead to pin 1 of Q1, then check the V at all three FET pins. Is that right?

 

[ronv]

Yes you have it right. These measurement will give a lot of info on what is going on.
- pump current running and stalled
- if there is a ground shift making the current limit to small
- the trigger voltage of the FET.