Help with Water Pump

[alec_t]

There you go, Joe. The concensus is that 10uF (electrolytic, >30V rated) and 0.1uF (ceramic) caps are advisable at the input of both the 20V and 12V rails, per PDM. You've got the PDMs working now, but these may prevent nuisance effects where more than 2 pumps are involved in the future.

 

[KeepItSimpleStupid]

@alec

I was just wondering if the Emitter of Q1 should be included in the power ground.

Based on the patents of the brushless motor controller IC, I think we made a pretty good guess. At one point I looked that up.

@Joe
I didn't include the possibility of yet another ground. Although this is for nothing but a TEST POINT.
Isense could be labeled Isense+ and a wire attached to the ground end of that resistor and call it Isense-.

These connections or test points make it possible to read the motor current accurately although the 0.22 resistor should be easy enough, but maybe not necessarily convenient.

I'm actually envisioning a 6 to 8 position @ pole selector switch attached to these points with a set of binding posts or even a built-in meter.

The idea would be to measure these voltages when the motor is first put into service. Over time due to wear and gunk, the currents should increase. This could be an indication when to clean the pumps.

At the very least, when you initially start to do any long term testing, record the voltage across the motor connector and the voltage across the 0.22 resistor when the motor is operating.

 

[alec_t]

Post deleted.

 

[alec_t]

I was just wondering if the Emitter of Q1 should be included in the power ground.

I think that would be best, since Vbe is key to the current-limiting.

Based on the patents of the brushless motor controller IC, I think we made a pretty good guess. At one point I looked that up.

IIRC that was one 4-legged IC picked at random (?) and there are alternatives out there with higher voltage ratings. For example Melexis do an 18V IC with a 6A rating; also a 30V IC with a 2.5A or 3.5A rating. 18V was a calculated long-term average supply voltage to the pump based on the OEM controller typical 'on' times and pump speeds.
Perhaps the dummy load high-power resistors, or similar, could be used to drop the 20V supply down to 18V to be on the safe side?

 

[KeepItSimpleStupid]

Nope, not necessarily at random.

See: **broken link removed**

On that page, there is this statement:

These features are combined with the Melexis patented no-VDD design to fit the IC in a small 4-pin VK package.

You can ignore the 4-pin, because the 4-pin is the tach or the other alarm output for this similar IC. The IC in the motor is 3-pin.

That and the circuit configuration is what makes Melexis a VERY GOOD MATCH and there doesn't appear that there a 3-pin 24 VDC part available.

The US91 (4-pin) mentions the patent. The US78 (3-pin) doesn't. Both are a no Vdd design.

 

[salty joe]

There you go, Joe. The concensus is that 10uF (electrolytic, >30V rated) and 0.1uF (ceramic) caps are advisable at the input of both the 20V and 12V rails, per PDM. You've got the PDMs working now, but these may prevent nuisance effects where more than 2 pumps are involved in the future.

Sounds like each PDM needs separate inputs.
Each PDM has three 12V inputs. (at R6, R11 and U1) Should these 12V inputs be tied together at the terminal block and have the caps installed at that point? Am I correct to assume the negative leg of the caps goes to ground? Does it matter what order the caps are installed? For the 20V, could the caps be installed at the D1-D2 junction?

With C4 gone, will there still be a safeguard to deal with power glitches and outage issues?

For the next solder session, I'll separate the prolong hooks, install input caps, lose C4 & install 4.7K on Q1 and add a dedicated 19g ground to Rsense. Not tonight though-I'm tired from working. I've made way too many soldering mistakes when I was tired. Thanks for all the help you guys!
BTW Pryros, C1 is connected directly to IC pins 7 & 14.

 

[KeepItSimpleStupid]

Sounds like each PDM needs separate inputs.
(Not sure what this means when combined with the statement below)

Each PDM has three 12V inputs. (at R6, R11 and U1) Should these 12V inputs be tied together at the terminal block and have the caps installed at that point?

Bingo!, it only needs one +12 input.

+12 can be tied together anywhere, but only one external +12 terminal is needed.

One 10 uf and one 0.1 ceramic cap in parallel at the terminal block. The Notation of +12 is just a way of avoiding clutter in a schematic. Grounds are done the same way. Usually triagular shapes with letters in them. Chassis ground (looks like a rake) and Earth (Horizontal equal spaced lines in the shape of a triangle) have different symbols. The inverted triangle with symbols in them denote various "common points". i.e. All the ones with a D in it arer assumed to be connected together. All the ones with an A in it are assumed to connected together. All the ones with a 1in it are assumed to be connected together.

So, U1(7) and Trip @ ground and the (- end of C4) goto to one terminal called (Digital Ground).

You should also have Supply+ to a terminal. And the modified power ground, which has been modified again as the ground side of Rsense AND the (Emitter of Q1). So, this terminal contains the ground side of the 0.22 resistor and the emitter of Q1 and call this POWER GND or HI CURRENT GROUND.

Near the Supply + terminal and POWER GROUND, there will be another set of caps in parallel: The 0.1 ceramic and the 10 uf cap.

Am I correct to assume the negative leg of the caps goes to ground? Does it matter what order the caps are installed? For the 20V, could the caps be installed at the D1-D2 junction?

That's the effective place that the + of the cap should be installed. Ceramic caps have no polarity.
The other end of this parallel combination effectively goes to the ground side of the 0.22 resistor and the emitter of Q1.

NOTE: The break and the move and the diagram that I drew isn't quite right now, MOVE THE BREAK to the right of the emitter of Q1

With C4 gone, will there still be a safeguard to deal with power glitches and outage issues?

Alec? The 10 uf caps will help ride out short power glitches.

For the next solder session, I'll separate the prolong hooks, install input caps, lose C4 & install 4.7K on Q1 and add a dedicated 19g ground to Rsense AND NOW INCLUDE THE Emitter of Q1.

Not tonight though-I'm tired from working. I've made way too many soldering mistakes when I was tired.

Now we're talking 20/20 foresight rather than hindsight.

Thanks for all the help you guys!

You bet!

 

[4pyros]

C1 is connected directly to IC pins 7 & 14.

Vary good and good night.

 

[KeepItSimpleStupid]

Melixis IC

Bad me. When I went searching my email I found a response from Melexis many months ago:

Are you working on a fan design? I would like to hear about it.

The voltage at the pins of the US79 should be less than 35V. With a conventional fan design, the Voltage at the US79 output pins can be double the fan voltage. This is due to the back emf from the fan. This is why we specify the fan supply voltage as 18V maximum.

The 24V schematic uses the 600 ohm resistors to drop the voltage, and uses the internal Zener diodes to clamp the voltage. With this special schematic and the external transistors, you can make a fan with 24V supply.

Melexis has a similar part that works at 24V without extra components. This is the US91.

The resistors that were in the pump were 520 ohms. That is a difference of about 15% from 600. If you take 15% of 24, you get 24V. SO, based on that we should not go over 20 Volts or maybe 22V.

24*24/600 = 0.96 W and 22*22/520 = 0.93 W

Poor cooling would make erring on the low side of the voltage a better choice.

 

[()blivion]

That's just for the hall sensor correct? Would this in any way explain why Joe blew up that pump way back when?

Good work BTW KISS.

 

[KeepItSimpleStupid]

If you take a look here: https://www.engineeringtoolbox.com/thermal-conductivity-d_429.html

and note the thermal conductivity of Epoxy, Aluminum and Copper, so we are probably fighting against those odds as well.

OB():

Well, I think we knew all along that a block of Epoxy has very bad thermal conductivity and we knew that the OEM controller operated the pump for a very short time at close to 24V. I forget that number, so power dissipation is a big issue for the hall IC and the drivers and 24/7/365 is a big stress on them.

There is no way to measure the temperature of the drivers or the hall IC, but the post-mortem, I think suggested a shorted driver because one of the coil forms was melted and their were shorted windings. But what happened first? The driver IC failed or a FET failed? Classic Chicken and egg problem. I was unable to determine if the hall IC was good or not. Physical damage and it could have been Joe's hammer.

Unlike the OEM controller which allowed a short 6A surge and a 3A normal current, Joe's supply allowed like 15 Amps to flow and the circuit had no TVS etc.

So, i think the whole point of this exercise is to do all what we can without having temperature monitoring at the FET drivers. We need to protect against transients, over voltage and over currents. The hope, I think, is that the PTC fuse would integrate out the thermal issues.

And as the pump does it's thing, the bearings will wear and it will pick up crud increasing friction and the pump may fail to start as well.

I am thinking that power dissipation was the killer, but whether the hall IC died first or a driver, I have no idea. The solder melting around the driver, suggests that the driver got really hot.

Joe had a few fuses for the pumps, but they never blew suggesting that something was limiting the current.

All we can do is hope for the best.

 

[salty joe]

I'll use a dedicated 19g power ground for Rsense and Q1 emitter per PDM. Is it OK to combine the digital ground from two PDMs, giving me three grounds per two PDMs? (two PDMs per board)

My power supply has three connections for ground. Would it be OK to solder three wires together for each ground connection at the power supply? Assuming I need a total of nine ground wires.

The existing pair of PDMs has all six 12V inputs on the same terminal. I'll seperate them to add caps. Two sets of 10uf & 0.1uf ceramic caps per PDM installed on 20V and 12V inputs, is that right?

I need to order the caps, but the rest of the soldering will get done this weekend.

 

[4pyros]

The existing pair of PDMs has all six 12V inputs on the same terminal. I'll seperate them to add caps. Two sets of 10uf & 0.1uf ceramic caps per PDM installed on 20V and 12V inputs, is that right?

Yes thats right. Can you put them where the power is connected to the board?

 

[alec_t]

With C4 gone, will there still be a safeguard to deal with power glitches and outage issues?

Yes. If you install 4k7 across the base/emitter of Q1 the current limit will be ~4.4A, after a brief or prolonged glitch/outage or during normal toggling.
You can re-purpose C4: it can substitute for 2 of the '10uF' for the 12V rail, i.e. one 47uF per board (pair of PDMs). I would hold off building more PDMs 'til we know if this 4.4A is enough to start all pumps reliably.

@KISS
That's useful info re Melexis ICs. I couldn't find a better match either for the IC you found during the autopsy, so chances are it's a Melexis IC in the pump (other manufacturers seem to do 3-phase chips only). Based on that info you've convinced me we should aim for 20V supply max; preferably 18V. A simple, cheap linear Darlington-based dropper with a 2N3055 (or better) would drop from 21V down to 18V while dissipating ~6W when driving two pumps simultaneously at normal running speed, so would suffice for the whole tidal system. Of course the 2576 chips in the OEM controllers could be used instead to get 18V somewhat more efficiently, but I think that would be a more complex solution and not save much power.

 

[KeepItSimpleStupid]

I'll use a dedicated 19g power ground for Rsense and Q1 emitter per PDM. Is it OK to combine the digital ground from two PDMs, giving me three grounds per two PDMs? (two PDMs per board)

My power supply has three connections for ground. Would it be OK to solder three wires together for each ground connection at the power supply? Assuming I need a total of nine ground wires.

The existing pair of PDMs has all six 12V inputs on the same terminal. I'll seperate them to add caps. Two sets of 10uf & 0.1uf ceramic caps per PDM installed on 20V and 12V inputs, is that right?

I need to order the caps, but the rest of the soldering will get done this weekend.

All of that sounds good. But there is one other "option" to consider:

Have the dual PDM have two grounds instead of three. Just make the wire leaving the PDM a heavier guage (16 AWG) and combine the two high current grounds. 16 AWG has 2x the cross-sectional area of the 19 AWG.

Your choice. It's a trade off (suggested reasons) between:
1) # of wires
2) Cost - Whether you have any 16 AWG wire. Cost of a terminal
3) Whether or not the terminals will accept 16 AWG.
4) You have a terminal to spare or need to loose one.
5) Easier to dicern which is power gnd and which is logic ground.

 

[ronv]

When you look at the datasheet the lower voltage is when you are driving the motor directly. When driving a PFET they show it running at 24 volts.

https://www.melexis.com/Assets/US79-DataSheet-4828.aspx

 

[KeepItSimpleStupid]

Yep, they do show that and the guy from Melixis said that an internal Zener diode allows that to happen and the 600 ohm resistor in the Melexis diagram (520 ohm in the motor) limits the current to the unknown Zener diode.

 

[salty joe]

I put the prolong hooks on separate terminal points, made a seperate 12V terminal for each PDM, pulled off C4 from one of the PDMs and added a 4.7K resistor across Q1 emitter-base. Then I gave each Rsense/Q1 emitter a dedicated power ground.

I fed the power grounds and 20V directly into the PDMs.The PDM with C4 removed and 4.7K resistor on Q1 toggled every pump 10+ times. Then I found the alarm did not work-I forgot to connect the digital ground. Now the alarm sounds for every toggle even if there is not a pump hooked up and the trimmer is set to max. I poked at it a little bit, but can't find what I did to make the alarm act up. Maybe if I forget about for awhile, I'll find the problem next time I look.

At least we know the pumps toggle with the resistor on Q1. This sounds like it is safer for the pumps. If the pump is slightly more apt to stall, that's where the alarm comes in and shuts it down-is that right?

 

[ronv]

Yep, they do show that and the guy from Melixis said that an internal Zener diode allows that to happen and the 600 ohm resistor in the Melexis diagram (520 ohm in the motor) limits the current to the unknown Zener diode.

So we agree there is no need to drop the voltage below 20.8 volts.

 

[ronv]

I put the prolong hooks on separate terminal points, made a seperate 12V terminal for each PDM, pulled off C4 from one of the PDMs and added a 4.7K resistor across Q1 emitter-base. Then I gave each Rsense/Q1 emitter a dedicated power ground.

I fed the power grounds and 20V directly into the PDMs.The PDM with C4 removed and 4.7K resistor on Q1 toggled every pump 10+ times. Then I found the alarm did not work-I forgot to connect the digital ground. Now the alarm sounds for every toggle even if there is not a pump hooked up and the trimmer is set to max. I poked at it a little bit, but can't find what I did to make the alarm act up. Maybe if I forget about for awhile, I'll find the problem next time I look.

At least we know the pumps toggle with the resistor on Q1. This sounds like it is safer for the pumps. If the pump is slightly more apt to stall, that's where the alarm comes in and shuts it down-is that right?

Maybe something when you fixed the prolong. Check that string of parts inclucing D7, R8.