I need a power router to divert excess photovoltaic power into a water heater.

Context: Solar panels are connected to a grid-tied inverter. A smartmeter measures how much power the house draws, and the inverter uses this information to adjust its output power to compensate, resulting in zero average power drawn from the utility if there's enough sun. So this question is about AC not DC, I'm not looking for a solution to feed DC power to the water heater.

The goal is to use excess power without drawing from the grid. This is usually implemented with a dimmer. All required data is available over Modbus from the solar inverter and smartmeter, so this question is only about the style of dimmer to use for this application, not how to control it.

A popular choice is a triac dimmer, which draws distorted current from mains and makes a lot of noise.

Another option would be to use two back-to-back MOSFETs to make an AC switch, and PWM it at high frequency. This makes the AC current more sinusoid-like, but the tradeoff is more HF noise.

The best option, which would be pulse density modulation using whole cycles, is unfortunately not available as the utility meter counts energy on a period by period basis and does not do net metering. So if available power is not enough to cover the full power of the heating element, it would result in importing energy from the grid, which would defeat the purpose.

This is a hobby project, so there is no plan to pass EMC certification.


What kind of filter can I use on both options (triac or PWM) to make this circuit "civilized" in regards to conducted EMI?

Here's a candidate filter for the triac version. It keeps di/dt drawn from mains at turn-on under 0.4A/µs. The 1mH inductor with 10A saturation current is a problem, but I've found a suitable core.

enter image description here

I'm not sure where to put the load either (after the filter or before like in this variant):

enter image description here

  • \$\begingroup\$ Well, you could use a super large choke (and a small bead in series to eat HF energy) and even achieve a sort of sinusoidal current with the triac. So maybe add a small info on budget. I think PWM with an intermediate inductor is a good middle ground, and this is also what active PFC uses in kW supplies. If you only care about HF noise but not about Power factor, then a common mode choke with caps to PE before and after should be fine \$\endgroup\$
    – tobalt
    Jan 12, 2023 at 20:23
  • \$\begingroup\$ I suppose TRIAC is out as well, because it would be intermittently drawing/delivering power to the grid anyway, just within a cycle versus over multiple [whole] cycles? \$\endgroup\$ Jan 12, 2023 at 21:40
  • \$\begingroup\$ @TimWilliams The meter only measures active power, and on a cycle it can't tell the difference between a triac dimmer and the usual reactive load which also draws/delivers power in turn. So a triac dimmer does fit the bill. \$\endgroup\$
    – bobflux
    Jan 12, 2023 at 22:08
  • \$\begingroup\$ Ah, it looks reactive (or harmonic, to a similar end), alright. \$\endgroup\$ Jan 12, 2023 at 22:19
  • \$\begingroup\$ @tobalt Budget should be reasonable, after all this is about lowering the electricity bill so it should not cost more than it saves, say a hundred euro for the whole thing. I'd like to use easily available parts (no big iron) because other people are interested in it. Pretty much everyone is using aliexpress SSR's and no filtering at all. \$\endgroup\$
    – bobflux
    Jan 12, 2023 at 22:19

1 Answer 1


PWM at AC mains works same as any other, but you need bidirectional switches (e.g. anti-series MOSFETs), and can't use a catch diode, it has to be synchronous. You also can't use, like, obnoxiously large electrolytics for filtering the input or output; it has to be film caps, and with a cutoff frequency somewhere between Fmains and Fsw, suitable to get adequate attenuation of switching noise, without impairing line regulation through the converter. Most likely, gate drive will need to be isolated; anti-series pairs can be wired source-to-source, so two such drivers will be required. Control can be open loop (fixed percentage or voltage-mode PWM), but current limiting should still be employed to protect the switches.

It also needs to handle the full range of mains transients, so, consider 1200 V Si or SiC MOSFETs, and a MOV at the inlet. Or if adequate protection is available in the grid-tie inverter, and an auxiliary outlet is provided after such protection, that might save a little bit of work. Do not modify approved equipment, or do so entirely at your own bodily, property and financial risk.

If a common DC (solar input?) rail is available (again, in a safe manner, preferably without modifying equipment), a load there would be much simpler.

  • \$\begingroup\$ Hmm, yes, good point about transients... Triacs can self-protect by putting a TVS on the gate that turns the triac on if there is a voltage spike, so the triac absorbs the spike by turning on for a half cycle... but that wouldn't work with FETs. Although, I could use that TVS+Triac as a protection circuit! \$\endgroup\$
    – bobflux
    Jan 12, 2023 at 22:16
  • \$\begingroup\$ DC would indeed be easier on the MOSFET side, but it has several drawbacks: the inverter's MPPT algorithm would probably object to random current being drawn from its input ; I have 4 PV strings so they would have to be combined ; last but not least, the heater's electromechanical thermostat will definitely stay as a safety measure and it is a bimetallic switch, not rated for DC. In fact, from the safety point of view I think DC would bring even more problems... \$\endgroup\$
    – bobflux
    Jan 12, 2023 at 22:20
  • \$\begingroup\$ @bobflux Fair points. \$\endgroup\$ Jan 12, 2023 at 22:25
  • \$\begingroup\$ This looks like it's going to end up with a huge inductor, and I'm having a lot of trouble finding something in the 400-1000µH range with 10A saturation current! \$\endgroup\$
    – bobflux
    Jan 13, 2023 at 9:42
  • \$\begingroup\$ I don't know what motivates that range. Perhaps you should open a new question discussing a particular design in more detail? \$\endgroup\$ Jan 13, 2023 at 10:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.