[AKADAP]

I bypassed the fuses, and the open sprinkler pi appears to be working now. It is drawing about 300 mA when a valve is open.

Summary:
I believe the primary cause of the problem was a bad power supply.
I replaced the solenoids, but I don’t think this made any difference.
I shortened the 100′ cable by 48′. I think this extra length might have helped trigger the bad behavior of the original power supply.
I replaced the AC power supply.
I removed the power supply from the USB C connector. I believe that having this power supply defeated an accidental safety behavior. If something shorted one of the solenoids, it will kill the AC voltage. If the USB C power is not there, this will cause the PI to reset, and shut off the shorted solenoid. If the USB C power is is place the Pi remains powered, and the solenoid will remain on with the likely result of the triac burning up.

[AKADAP]

I got a replacement open sprinkler pi, and replacement solenoids. I also bought some self resetting fuses rated for 500 mA (supposed to trigger at 1A, and handle 60V. I modified the open sprinkler pi to have a fuse in series with each valve connection.
I replaced the solenoids, and measured the resistance at the end of the cable, I got reasonable numbers.
I measured the solenoids I removed from the sprinkler valves, and they all measured 25.4 ohms.
Anyway, at this point I reconnected the sprinkler controller and tried to test it. The sprinklers would turn on then less than a second later turn off.

I removed the resetable fuses an tried again, this time with a current probe on the output of the power supply. I was seeing 0.9A when a sprinkler was running with occasional spikes to 2A. I was running each sprinkler in sequence, one minute each, when it got to the sixth and last sprinkler, I saw a spike of more than 4A (I was on the 4A range on the meter), and a triac and its associated gate resistor smoked.

I replaced the triac with a new one, and I checked the cabling. There were no cuts in the cable, on only the very end near the sprinkler valves has ever
gotten wet. I did notice that there was a bunch of extra cable coiled up near the sprinkler valves in the crawl space under my house.

I tried connecting up the sprinkler controller again. This time I only used the AC to power it, I did not connect the USB C power.
When I tried to test it, it turned on the 1st sprinkler, the current spiked, the pi reset, and the sprinkler turned off.

I removed the sprinkler controller, and connected the transformer directly to one of the sprinkler valves. The current was hovering around 0.5A with occasional spikes to 2A.

I tested each of the solenoids I had previously removed from the valves (not the new ones) and found that they only drew between 300 and 330 mA.
At this point I suspect that the power supply is behaving badly with the long cable. The power supply I was using has an LED, which makes me believe that it internally has something more than just a transformer. If there is some form of protection circuit that is partly triggering such that it only shuts off the output on the positive, or only the negative half of the waveform, it could be driving a large DC component into the solenoid. The DC signal would not see the inductance of the solenoid, and therefore could drive much more current. This is a rather sketchy theory, but it is the best I can come up with.

I removed the extra cable from the sprinkler valve end of the cable, this measured 42′, so originally I had 100′ of cable, and now I was down to 58′.
I have another power supply that I am pretty sure is just a transformer. It is only rated for 0.8A.
I tested the solenoids installed in the sprinkler valves by directly connecting them to the transformer. I found that they draw between 250 and 300mA.

I re-install the resettable fuses. These fuses have about 1 ohm of resistance. I re-connect the controller and test. I find that while the solenoids buzz, the sprinklers either don’t turn on, or barely turn on.

Tomorrow, I will bypass the fuses and try again.

[AKADAP]

My first suggestion would be to replace the triac with a solid state relay. The triac is a bipolar device, and has a significant voltage drop. This means that power is dissipated in the device causing it to heat up. Solid state relays have very low on resistance so they dissipate very little power. You can also get them with high enough current ratings that the transformer for the AC could not provide enough current to fry them. Also in most cases, the input is an LED, so with a current limiting resistor for the LED, there is no chance of frying the drive circuit. An example of a part like this is this: https://www.digikey.com/product-detail/en/toshiba-semiconductor-and-storage/TLP241AF-TP4-F/TLP241AF-TP4FTR-ND/7652708
Yes, this would increase your BOM cost by about $10, but that is better than burned boards.
These parts are also completely isolated, so one side of the AC would not need to be hard wired to GND, and you could use a full wave rectifier instead of the half wave rectifier you are currently using for the 5V power supply. It would be nice to have a big enough power supply to power a Raspberry Pi 4 with a few peripherals.

Resettable fuses are so cheap you could use one one every channel: https://www.digikey.com/products/en/circuit-protection/ptc-resettable-fuses/150?FV=69%7C409393%2C686%7C150723%2C686%7C150724%2C686%7C153583%2C686%7C157293%2C686%7C160817%2C686%7C213650%2C686%7C257642%2C686%7C257644%2C686%7C257645%2C686%7C287637%2C686%7C287639%2C-8%7C150%2Cmu1.1A%7C683%2Cmu1.2A%7C683%2Cmu1.5A%7C683%2Cmu1A%7C683%2Cmu2.2A%7C683&quantity=0&ColumnSort=1000011&page=1&stock=1&rohs=1&nstock=1&k=fuse&pageSize=25&pkeyword=fuse

[AKADAP]

20 ohms through 50′ of cable is not within the acceptable range, 20 ohms at the solenoid is at the very low end of the acceptable range.
Lightning is something we get about once every other year here, and then usually at least 5 miles away, so no it is not lightning. At this location I seldom have power fluctuations. It is only the triacs and gate resistors that have fried. The only way for current to go through them is through the solenoid. The only way for enough current to go through the solenoid to fry it is if the solenoid is bad. I have six identical solenoids, and the resistance is significantly different between them.
It is the only plausible explanation that I have been able to come up with.

I have ordered a replacement Open Sprinkler Pi. I will put a fuse in series with the AC. I have also ordered replacement solenoids. We will see if the new unit survives more than a month.

I mostly do digital electronics, and some low voltage analog, so I am not too familiar with power electronics, but I’ll take a look, and see what solutions are available.

[AKADAP]

I was looking at the specifications for the OpenSprinkler pi (reproduced here):
Specifications:
Input​ ​Voltage:22V​ ​AC​ ​to​ ​30V​ ​AC.
DC​ ​Output​ ​Current:500mA​ ​@​ ​5V​ ​(to​ ​power​ ​RPi).
Number​ ​of​ ​Zones:8​ ​on​ ​the​ ​OSPi,​ ​expandable​ ​by​ ​linking​ ​zone​ ​expansion​ ​boards.
AC​ ​Output​ ​Current:800mA​ ​continues​ ​@​ ​24V​ ​AC​ ​per​ ​zone​ ​/​ ​station,​ ​8A​ ​impulse​ ​/​ ​inrush.
Over-voltage​ ​Protection:48V​ ​bi-directional​ ​TVS​ ​on​ ​each​ ​zone,​ ​AC​ ​input,​ ​and​ ​rain​ ​sensor​ ​terminal.
Over-current​ ​Protection:2A​ ​on​ ​AC​ ​input;​ ​1A​ ​on​ ​5V​ ​DC.
Size:135mm​ ​x​ ​100mm​ ​x​ ​32mm​ ​(5.3”​ ​x​ ​4”​ ​x​ ​1.26”)
Weight:150g​ ​(5.3oz)​ ​w/o​ ​RPi

I don’t see where the over-current protection comes from.
For the AC, I see one connector directly connected to GND, the other connected directly to the COM connector. The triac connects the pin of the solenoid directly to GND. I see no circuitry to limit the current.
For the 5V, I see a switching power supply configured for 2A with a 12V input This part has a current limit, but it is significantly higher than 3A. I don’t see any other current limiters on the 5V.
I do see the TVS diodes for voltage limiting, I just don’t see any circuitry for current limiting.
Can you tell me how you came up with the over current protection numbers?

[AKADAP]

Once I disassembled the setup it was obvious what had failed.

Three of the triacs are obviously fried. One so badly that its pads have come loose from the circuit board.
All of the 220 ohm gate resistors are damaged. They show discoloration in the center, and none measure 220 ohms. The lowest resistance I got was 300 ohms, some were in the 100s of k.

I am running a Raspberry pi 4. This Raspberry pi can draw more than the 1 amp that the OpenSprinkler pi can provide, especially if one wants to connect anything to its USB ports which I had not done, but was thinking about adding to the system.
I followed the recommendation to power the PI through the USB type C port.

The following is my best theory of what happened:
In measuring the solenoids through the 50′ of wire, two of them measured 20 ohms. This is lower than it should be.
There appears to be no current limiting other than the transformer itself on the AC input.
There appears to be some accidental protection against shorted solenoids since usually the raspberry pi is powered by the AC, and if there is a short, the AC voltage will drop to near zero, killing the power to the Raspberry pi, resetting everything, and shutting off the Triacs.
By adding the additional power supply on the USB C port, this shutdown did not happen, the Triacs stayed on, drawing more current than they were designed for, overheated, and eventually burned out. I’m not entirely sure why the gate resistors also burned, but this happened on every single gate transistor, including two that never had a solenoid connected to the associated triac.
For safety, I think there should be a fuse between the AC connector, and the Common connector.

If OpenSprinkler decides to spin the board to add a fuse, I would suggest that they also spread out the traces between the two 40 pin headers. Packing those traces so close together on a two layer board is begging for crosstalk. You get away with this now since you are not using those signals, but anyone trying to use that second connector is going to have problems. The alternative is a four layer board with two ground layers, but that is more expensive.

The package you are using for the triacs is designed to transfer heat to the circuit board. Unfortunately, this design does not have any place for that heat to go. Ideally you would have about a square inch of copper connected to the large pad to act as a heat sink.

I’d also like to see some way for the Raspberry pi to detect when a solenoid is malfunctioning. These solenoids are out in the elements and will eventually fail. I’d rather not have to replace my sprinkler controller every time a solenoid fails, or run for months thinking I’m watering when I am not.

Attachments:

1. 

   FriedTransistors.jpg
   

[AKADAP]

Perhaps interfacing with WeeWX http://www.weewx.com/docs/hardware.htm#vantage_notes would alleviate the problem of too many APIs. They seem to support a lot of different weather stations, and it can be run on a raspberry pi.
I have not tried using it yet, so I’m not sure how hard it is to connect to my Davis Vantage Pro2 weather station.

[AKADAP]

There was another thread a few years ago about weather underground changing their API key policy and not allowing access.
As far as the API expiring, I found that my old API key was gone, and created a new one on weather underground, and found that it had an expiration date six months from when I created it.

Yes I am referring to using the PWS as the weather data source. The advantage is that I would not have to worry about the api key expiring, or the web site being purchased by someone else and being screwed up.

As far as attempting to modify code, I once attempted to modify the mouse driver in linux to support a different mouse. I wasted a bunch of time trying to figure out even which files the part I wanted to change was buried in. I never found it. These projects usually have a steep learning curve that one must climb before one can do anything useful in them. I am a hardware guy, and have not done serious programming in 20 years. It is fine to point me at the source code, and for you it might be a trivial change, but for anyone else the learning curve to be able to do anything useful is a huge barrier to even making a trivial change.

[Author]

Posts