The Delon bridge is the key to harvesting as much energy from the CT as possible:

The capacitors can’t really be changed, they are the reservoirs of whatever energy the two diodes can pump into them. Harvesting takes place whenever the input (which is a non-sinusoidial AC voltage) exceeds the current voltage level in the capacitor.

Note that these capacitors do have to withstand a certain amount of stress: the constant charge / discharge cycles leads to ripple currents. Good quality electrolyte or polymer caps will improve the MTBF and reduce loss of capacitance due to ageing.

Diodes are a different story. Every real diode has a small “forward voltage drop”. This introduces some losses in the harvesting circuit, since it reduces the voltage that actully reaches the capacitors. MOSFETs can be used to avoid this, but it requires extra circuitry.

The diode drop of a 1N4148 is about 0.65..0.70V.

Schottky diodes have a considerably lower diode drop, somewhere between 0.2V and 0.3V usually. We could replace those two 1N4148’s by 1N5819’s. Some quick measurements:

• with 1N4148’s at a certain input current and load, harvester output is 4.21V
• with 1N5819’s at the same input current and load, harvester output is 4.81V

The extra voltage in itself is not of much use in the current MPS, because the rest of the circuit simply throws away anything over 3.6V (a switching regulator could do better). Nevertheless, the extra margin will allow the MPS will continue working at lower input levels, corresponding to smaller AC mains loads.

Another optimisation we can explore, is to what extent that fancy TLV431 is a better zener than a red LED. This can be done with a Component Tester, which applies a varying voltage and measures the current at each point:

On the left a red LED, on the right a TLV431 with a 1 MΩ / 470 kΩ voltage divider.

The scale of these two screen shots is -2 .. +6 V horizontally and -2 .. +6 mA vertically. As you can see, they both start conducting at around +1.8V, although the TLV has a somewhat sharper bend and a more vertical line, so it’s more “decisive” in it’s zener-like action.

The TLV also has a reverse diode protection at -0.6V, but that’s irrelevant for our circuit.

These images suggest that a red LED might be just as effective after all. This is something to keep in mind once the MPS is completely done. Maybe we can save a few components.

But at this point it makes little sense to embark on such micro-optimisations – we’re better off first getting the complete MPS off the ground, including powering up the radio module and sending out packets with it. Then we can tweak – and make sure everything still works.

[Back to article index]