Help with Water Pump

[()blivion]

()B, Kiss is right, it is almost done (I think).

Yeah, what was I thinking, it was a passing thought we can forget I said anything. Would have been better idea if we were to go after it from the start. Though if we were looking to engineer the best product possible that does the most things in the least space, then it would be feasible to restart. But I think it's best for all if we just get the job done.

So yeah... best to just commit to what's down now.

 

[KeepItSimpleStupid]

so Joe would need to add a resistor from the FB pin to ground?

Example: Wrong numbers, but an example nonetheless:

Suppose the ratio is 10,000/1000 that gives 10, no big deal.
Now suppose we put a 1000 K in parallel with that 1K using a transistor. Now we have a number like 500 in the denominator, so 10,000/500 is 20.

So adding a resistor in parallel with the existing 1K INCREASES the output.

There are 2.2K 1% MF resistors connected to the bases of these transistors.
With T2 turned on, you will have the HIGHEST voltage.
With T1 turned on, you will have the MIDDLE voltage.
With T1 and T2 OFF, you will have the lowest voltage.

Hence, pull the end of the resistor attached to the transistor and ground preventing T1 and T2 from turning on and you get the lowest voltage.

You can lift both of the 2.2K resistors (the lead away from the transistor as well) and get the same effect.

There is a fixed divider and the denominators get smaller when one of the transistors is on. A total of 4 resistors are used in setting the gain. Two fixed and 2 that are insertable as parallel, circuit wise, to the denominator resistor.

The formula is something like 1.25(1+R2/(R1||Rn) where Rn = infinity, ~2K and ~6K (selected by combinations of T1 and T2. R2 and R1 are fixed.
T1 T2
0 0 Lowest output 10.5
0 1 Highest (22.5 V)
1 0 Middle
1 1 Not used, but would be > 22.5 V

 

[ronv]

Okay, I got it now. I think. () B did this to me with his old quote.

I think.

 

[KeepItSimpleStupid]

ob():
We have to give it our best shot, but it can fail due to no fault of anybody here. I think we are at the t- "proof of concept" stage. t- read as approaching very close. I'd very much like to see the system work with 18V or less be applied to the pumps. That, however, may be function of the coral, not the water.

 

[leonardason]

It is very doable You have 4 pumps that run on 24 VDC and draw about 1.3 amps each and yes that would be about 31.2 watts but what a pump really draws would be a function of how much work it is doing. You want to run two pumps per controller using two independent controllers. Pumps on about 30 seconds and pumps off about 5 to 10 seconds. Timing could be done using a 555 timer chip configured to drive the relays used to switch the pump(s) on and off.

 

[alec_t]

The power in the fet is around 50 watts at this point so a fair sized heat sink would be needed even if it's only on for a few seconds.

Trip time would be < 1 sec.

If we went MCU with this, we would be throwing away good parts of Alec_t's fine engineering and schematic work, and I wouldn't think he would like that much. Alec_t, would you be able to work with me schematic and simulation wise if we went full MCU core with this project? Would it bother you at all? Most of the timing and control work you have done so far would be all for not, as we would be much better off to do it all in the micro.

It doesn't bother me if Joe were to decide on the MCU route, and I'd certainly be willing to help with interfacing the MCU to the hardware. I don't think the timing side of things is the main problem, though; it's getting the volts and amps right for the pump that's the real headache.

so Joe would need to add a resistor from the FB pin to ground?

It's already on the OEM board and is transistor-switched. The add-on board simply supplies a base signal to the OEM transistor.

 

[alec_t]

@leonardson

Timing could be done using a 555 timer chip configured to drive the relays used to switch the pump(s) on and off.

If only life were that simple...... However, the OP wants to switch between pump pairs at 6 hour intervals and a 555 is unsuitable for that long a period. Also, solid state switching is used instead of relays.

 

[ronv]

Trip time would be < 1 sec.

Don't forget the start time of .4 seconds was at 24 volts. It will be slower at lower voltage.
My guess is it will take about 10 volts for the pump to start (40% of rating) or around 5 amps.
That leaves 14 volts at 5 amps across the FET or around 70 watts. Based on this thread https://www.electro-tech-online.com/threads/thermal-response.128051/ a heatsink will be required. Probably one like is on the controller would give a long enough time constant to protect it.

It doesn't bother me if Joe were to decide on the MCU route, and I'd certainly be willing to help with interfacing the MCU to the hardware. I don't think the timing side of things is the main problem, though; it's getting the volts and amps right for the pump that's the real headache.

I think everone agrees this is almost done.

It's already on the OEM board and is transistor-switched. The add-on board simply supplies a base signal to the OEM transistor.

With your latest design the OEM controller is no longer required. That was just to test where the pump would start and how long it would take to come out of current limit.

Keep up the good work it's almost done!

 

[alec_t]

With your latest design the OEM controller is no longer required.

That's what I think too. But I would say that, wouldn't I? Others may differ . It's Joe's call. I agree a heatsink is needed regardless of the route we go. If the concensus is that 3A is too low a limit it's simple enough just to reduce the sense resistor to, say, 0.15Ω for a ~5A limit.
My PDM-Mk7 doesn't itself have any speed selection feature (which the OEM board does), so if that feature is essential then a switcher to drop from ~20V to ~12V or whatever would be an additional module for a later date. Perhaps that could be a hacked OEM controller? The 2576 is rated 3A (6A peak), but I doubt it could be used to drive 2 pumps at the same time, which is a requirement of Joe's tidal flow system.

 

[ronv]

I forgot 2 pumps at a time, but there is a switch for each pump, right?

I checked the IRF 3205 and 10 volts drain to source at 5 amps is outside its safe operating area for a single 10 ms pulse so a better FET is needed.
This one would barely make it. Let me see if I can find a better one.

https://www.electro-tech-online.com/custompdfs/2012/06/auirf3504.pdf

Wrong dataa sheet here is the right one:

https://www.futurlec.com/Transistors/IRF3205.shtml

 

[ronv]

Here is one that will have margin.

https://www.electro-tech-online.com/custompdfs/2012/06/CD00003356.pdf

 

[alec_t]

That looks better. Extrapolating the SOA data (Fig 6 of the datasheet) for the IRF3205 suggests that it should handle 5A for ~1 sec with Vds of 10 (i.e. 50W dissipation) and a case temperature of 25°C.....just. So it would be marginal. Yet the same datasheet states "The TO-220 package is universally preferred for all commercial-industrial applications at power dissipation levels to approximately 50 watts", from which I infer that it should handle 50W indefinitely

 

[salty joe]

I sent you one of each of these headers:

https://www.sparkfun.com/products/115
https://www.sparkfun.com/products/116

Thanks KISS, it arrived. I have not had a chance to figure out what everything is yet. Quite a collection of goodies-Thank you! I am looking forward to using the braided desoldering wire. I read somewher that putting a bit of flux on the wire is helpful.

Will Feed really be when the pumps are operating? Is feeding time within a pump cycle? I was assuming that the pumps would have to start at the lower voltage as they do with the OEM controller.

The tank could be fed with the pumps running wide open. It would be nice, but not critical, to be able to slow the pumps for 5 minutes or so at a time with the push of a button.

Joe, Any chance you have a 12 volt power supply laying around that could be used to see if the pump will start at 12 volts?

@Joe
You could always take it out to the car battery. Put a voltmeter across the battery first. It will likely be little higher than 12.

Would a car battery charger do the trick? Mine can be set for 6V or 12V.

That's what I think too. But I would say that, wouldn't I? Others may differ . It's Joe's call. I agree a heatsink is needed regardless of the route we go. If the concensus is that 3A is too low a limit it's simple enough just to reduce the sense .

No, no it's your call. I have heatsinks.
My computer decided not to alert me to all the activity on this thread. I was going to thank KISS for the goody bag and wow! you guys have been going after it. Thanks again KISS and thanks all for all the king sized brainpower help.

 

[ronv]

That looks better. Extrapolating the SOA data (Fig 6 of the datasheet) for the IRF3205 suggests that it should handle 5A for ~1 sec with Vds of 10 (i.e. 50W dissipation) and a case temperature of 25°C.....just. So it would be marginal.

My bad. 3205, 3502 -- all looks the same to me.

Yet the same datasheet states "The TO-220 package is universally preferred for all commercial-industrial applications at power dissipation levels to approximately 50 watts", from which I infer that it should handle 50W indefinitely

Yea, the fine print is the case temperature of 25C. If you bolt it to the frame of the car it's ok.

But the way I read it (The right data sheet) it should be good if it doesn't cycle on and off very fast so it has time to cool. Maybe the retry circuit could be longer??? We could also put the resettable fuse next to the FET (low budget thermal shut down).

 

[ronv]

Would a car battery charger do the trick? Mine can be set for 6V or 12V.

Battery chargers can have a lot of ripple on their output that may mislead us. The "Kiss modification" would be the best.

I have heatsinks.

Are they pretty good sized?

 

[KeepItSimpleStupid]

@Joe

The braided wire contains flux. You can buy braided wire without flux significantly cheaper.

The reason for modding the OEM controller is primarily for the FEED mode, but it did offer the identical OEM protection too. The OEM controller also contains a shutdown pin. An alternative is to have another power supply set for the lower voltage. It's harder to design a high current DC-DC variable voltage converter. The bad part about using the OEM controller is the size of 6 boards. I think individual motor controllers make more sense than one that can handle two motors. The starting current can be significantly larger too and single controllers are better able to protect a single motor. Industry always uses one controller per motor.

Wide open is what MAY have killed the pump. Because the Melexis IC is patented and only good for 18 V, this leaves me to highly suspect the operating voltage >18 V to be a problem. I've been considering calling Melexis. 22.5 V is the OEM voltage, so I would no go past that.

I wouldn't use a car battery charger. The pump controller will work fine by unsoldering two resistors. If there is a need to go variable, we can do that too.

Sometimes your email system can hiccup. Generally if you put the notification address for ETO as one of your contacts, the SPAM filter will always deliver the mail. You can accidentally unsubscribe too. Under "Thread tools", you can add the subscription. Occasionally a thread might be interesting to me, but I won't post, but I will subscribe to the thread.

The long black header thing unplugs into 3 separate pieces. One piece, a jumper, doesn't quite belong. You can use two pins or three pins of the male header and the jumper to "configure options". e.g. Say a center pin can be 0 or 1 for something to be enabled and a low and high voltage are at the ends. You can then place the jumper to enable a feature. Sometimes extra pins are used as storage of the jumper too.

Here is a application of the headers: http://shieldlist.org/ Note the stacked structure of the PC cards. Those are long enough to allow fairly high components to be used on bottom boards.

The wire-wrap pin was just thrown in for no real apparent reason. Wire-wrapping uses square posts and the Wire Wrap wire that was enclosed. Stripping the wire will be difficult without the specialized stripper. A hand or motorized tool is used to WRAP the wire around the square post, so soldering is not necessary. I've seen entire commercial systems use that technique.

I really wanted to send you a push in terminal. Those, I do think you should use for breadboarding because it keeps the components separate from the traces. They pretty much require a $20 insertion tool and are available for 0.042 and 0.062 holes.

Your welcome for the package.

Hey, this isn't real work because there are no imposed deadlines.

 

[alec_t]

If you bolt it to the frame of the car it's ok.

Now there's an idea..... a mobile reef tank!

Maybe the retry circuit could be longer?

Retry period is currently ~16 secs but changing R4 to 2.2 meg in the PDM-Mk7 would double that.

I've been playing with a further sim. Adding a 68μH inductor and 100μF cap to the Mk7 circuit would enable a PWM input to the FET gate to supply a smoothed selectable voltage to the pump just as the OEM controller does, giving us speed selection. Those components could be 'borrowed' from the OEM controller.

 

[KeepItSimpleStupid]

enable a PWM input to the FET gate

So, your filtering (LC filter) the PWM, so it wouldn't be PWM per say which we thought could be detrimental.

 

[alec_t]

Yep. One L, one C, seems to give a reasonably smooth DC. So the pump doesn't 'see' much PWM. Ripple ~ 150mV (IIRC) acceptable?.
The values chosen don't seem too critical (I'm open to suggestions there, as inductors are strange creatures to me), but the OEM values sim ok for a PWM input ~50kHz (close to the 2576's internal 52kHz). The L goes in series between FET drain and Pump-. The C goes in parallel with the Schottky by Pump+.
Come to think of it, with C there is the snubber R/C needed?

 

[ronv]

A lot of headaches trying to build a homemade buck converter. One thing to worry about is the start up current. Spice doesn't model inductor saturation so it is easy for the current to be out of spec (for the inductor) during start up. It might be better to use a chip like the LT one you used before that takes care of all that "stuff'. How about just latching the over current in the one you have so that it must be reset by hand on a failure?