Solar diverter for water heating Dc/AC

[bolek_bolek]

Hi,
I designed and built this diverter for water heating with two things in mind but I know there are several versions around using the same idea.
1.Use the excess power from PV so it runs only when the battery is full or close to, so it's voltage triggered.
2.Keep using the mechanical thermo-switch that the water heater came with.
It's been in use for months with no issues. I made it simple, cheap and reliable, however it's not like real MPPTs that actively trace the max power point. It's also made with parts from the junkbox except U1 IC and can be built by anyone with soldering skills. PCB is THT except the two ICs.
The big capacitor in parallel to the PV is not on the board and can be any big electrolytic you have available at higher than PV max open voltage. In my case PV is 4s2p old polys so MPP is around 100V dc, diverter is set to 105V and it bounces ~3V around it.
Point 1 is achieved by U2A comparator with hysteresis. U2 is any SOIC8 dual opamp , I used the MC. It keeps the U1 switcher in standby and no current is delivered to the load until PV input is higher than the threshold set by R11 trimpot. At higher voltage the green LED called DIVERT is on and D3 is off so U1 can run.
Point 2 is achieved by the U1 full bridge gate driver and the 4 power transistors. With values show it outputs a ~70Hz square wave with very little deadtime, but it proved enough to let the electrical arc in the heater's thermal contact extinguish naturally without melting the contacts.
U2B is wired as a shunt amplifier to show if the heater is working. When the water is hot enough and the thermal contact is off , the red LED called LOAD is off too.
The linear regulator may look strange , the pass transistor can be a bipolar too, the series ballast resistors make it run cooler. Measured current is ~15mA.

[Seekonk]

Any scope waveforms?  I'm wondering how adiabatic the charging is.  When operating with a charge controller a diode is needed to isolate the capacitor bank and prevent capacitor current from fooling the CC into thinking panel can produce higher current.

[bolek_bolek]

I haven't saved the waveforms unfortunately, and this one is boxed in the attic and working day and night, but they are just squares or bursts of squares. I promise to add some scope images when I assemble the next one.

Indeed a diode is useful to stop the cap from feeding the charge controller. But the MPPT control loop is supposed to be very slow anyway, right?

Also, rumours say that some (fancy?crazy?clever?) CCs also short their PV inputs from time to time to see how far away they are from the real MPP so to avoid local maximums, and they might regret doing so if a milifarads cap is in parralel to the PVs. The Isc is supposed to be not too far on the right from the knee, and in this case it is. Maybe someone having such a charge controller would be kind and try connecting a big cap at the PV input and tell us what happens.

PS: My current setup is all DIY and the charge controller is a constant voltage PV buck manually adjusted for the knee voltage.

BTW not sure what you mean by adiabatic charging?

[Zucca]

Nice, congrats!

Is the heater only connected to the diverter I imagine.

If the above is confirmed why not feed high voltage DC into the heater maybe using PWM? A resistor would heat up with DC as well, don't need the AC.

[bolek_bolek]

Requirement #2, the mechanical thermostat will arch and burn on HV DC. PWM will help only if dead time is long enough.
A crazy idea that I have not tested is connecting a ferrite (auto) transformer at the output to adapt the heater's resistance to the panel string's preffered impedance. Of course Fsw should be much higher and switching losses will rise, but it may help those with lower V higher I pannel strings.

[Zucca]

Make sense, but since you are not feeding a pure AC sine to the heater but more an AC square wave with Vpp 2xVHdc solar panel.... I was wondering if the switch would be okay/surviving with just DC.
Probably not.... so your choice was correct.

Your idea with an auto transformer is sensible, but at point I would have just have installed  a microinverter between the panel and the heater, and turned it on when you wanted it.

[Seekonk]

Sorry for the delay, I was moving.

Congratulations, you are the fist person in ten years that has viewed power point water heating as a diversion problem.

That is a really good first attempt at a water heater control. I see a couple technical issues that make this a work in progress that cannot be recommended in the current state.   In the solar world the standard is better than nothing. On an electronics forum I like to think our standards are higher.  Even working 99.99% of the time isn't good enough when adding a couple components is all it takes.

My big concern is that after 3/4 power begins to be lost.  That is something I can't accept.  This happens when the reset time becomes longer than the recharge time of the capacitor bank.   I've worked a lot with the IR2153 which is about the same architecture.  A soft reset is at 1/6Vcc and a hard reset like you do takes even longer.  My initial design forced a soft reset and that didn't work well enough.  Ended up doing a dual timing resistor arrangement. 

My first 2153 design was with a 300W inverter board about 8 years ago.  Thought this would be the easiest build with solar people who had little talent.  It gave them a good start with a premade board for $7.  Voltage sensing was with a TL431.  Pictured is the SG3525 board, you never knew which board they would send you and clearances were terrible.  Nobody could ever complete the project.

The SG3525 makes a good water heater with deadtime added.  Any OLD non microprocessor MSW inverter has this chip.  Change two pins and you have a great water heater with a nice case. No need to invent anything.

I did a dedicated board with a IR2153 single ended output with arc interrupt and sent a few out. It worked well and that chip can make an interesting two heater with priority.  I abandoned that design because half the 2153 chips out there are fake and don't do the data sheet shutdown as they should. Surprised you were from Romania.  In the US I can get any part.  Even Canada has trouble with part access.  I think your chosen chip is too exotic for international use.  I prefer to use totally jellybean parts. Learned an important lesson, never get involved with solar people.   

My advice is keep it simple for the dummies.  I added simple outdoor temperature sensing requiring only seasonal pot adjustments.  H bridge is nice for small point of use tanks, but single ended is cheaper and easier.  For voltages around 100V, turn the controller into a shunt regulator with a resistor across thermostat. Then maximum switched voltage is less than 30V,  Voc - Vmpp which any switch can take. 

Anyway, water heater controllers can be made out of almost anything.  I have done dozens of designs, each has some shortfall. My favorite is a LM358 design.  Only ten parts, delt V of 3%, no noise and performs flawlessly.  The extra slow slew rare of the 358 filters out noise.

I run three water heaters off the same array in parallel with a charge controller.  Each has a different priority based on voltage.  It is an amazing system! 

You should isolate with a diode. Electrons always know what they are doing.  It is a little cavalier to tell someone to try it without and see what happens.  A large capacitor bank can do a lot of damage to outside components.  I don't do it with one of my controllers but it is isolated being 150 feet away with #12 wire or less. I'll be installing a diode as it does mess with things.

Third picture is the Polish ACTii which is MPPT but never stops hunting and goes all over the place with the poor delta V.