Low-Power Control Systems

The GHurd dump controller has hysteresis built in but can still switch very fast if the following two conditions are being met:

1. The power source is providing a lot of power (high winds against a wind turbine, for example)
2. The dump load draws heavily on the battery

Fast switching is not good for powering motors (and, I'm told, inverters) so some way of keeping the switching rate under control is necessary.

I've experimented with doing that by tacking a one-off timer circuit on to the GHurd dump controller. The timer is initialised when the GHurd dump controller switches to dump mode but stays on for as long as the timer is set to stay on - even if the GHurd dump controller switches off again before the 'on-time' period ends.

This allows you to have the GHurd dump controller switch on, say, an electric pump (IE a motor) for 15 minutes each time the GHurd dump controller switches into dump mode. Regardless of how soon the GHurd dump controller detects that the battery voltage has fallen below the 'dump' voltage threshold again.

The 'off-time' will be determined by how quickly the power source is able to bring the battery voltage back up to 'dump' voltage threshold and the hysteresis level of the GHurd dump controller.

Obviously, this arrangement means that dump load will be discharging your battery a little below your desired voltage, but this should not matter too much. For best battery life, this fact suggests that you should use this one-off timer configuration for the higher threshold voltage dump controller in a two-dump controller configuration.

That allows you to set the one-off timer arrangement to trigger on when the voltage reaches a charge or float voltage and to set another dump controller to trigger on when the voltage reaches a point between the battery's (lower) 'long-term trickle-charge' voltage and it's higher 'short-term float charge' voltage.

I built the one-off timer using a Maplin Electronics (UK) monostable multivibrator kit (Maplin item N32FL and Scroogle 'N32FL.pdf' for the circuit diagram). That £6.99 kit was so cheap that I could barely put the parts together with a decent PCB myself for the same price.

Here's how it looked while I was experimenting with extracting a trigger signal and how to take off 12v power for the monostable from a GHurd kit that was sensing over-voltage on a 24v battery bank.

The red wire coming from the GHurd dump controller provides the trigger signal to the monostable. Note which switch (SW1) and switch hole it is soldered into on the monostable's PCB - the 'available' SW1 hole to the left of the red wire is actually grounded and so not usable for this application.

The monostable uses a relay output, but it is possible to wire power N-fet in its place. I wired the gate of the GHurd kit's IRFZ44N power mosfet into the appropriate position on the Maplin monostable kit's output and, sure enough, it worked fine. You can see where the Maplin monostable supplies the trigger voltage for the mosfet on the photo at the bottom of the page. It's the little silver terminal pin sitting by the bottom edge of the 'RLY1' box on the Maplin monostable board.

1. Power supply to the timer circuit
2. Power supply to the timer circuit if the GHurd dump controller is running at 24v
3. How to get appropriate high and low trigger signals from the GHurd dump controller
4. Rules for testing.

1. Where to tap off the power supply for the 12V monostable? In short, it can be powered from the same supply the GHurd dump controller is monitoring. You should not, however, power it through the GHurd dump controller's leads.

1. Power supply to the timer circuit if the GHurd dump controller is running at 24v. You need a 12v supply to the monostable and the GHurd dump controller is going to be running at 24v or higher so you will need a separate 12v supply whose negative rail is common with the negative rail of the GHurd dump controller. Another 12v battery? For testing purposes, I did run my monostable off the right hand side of the GHurd dump controller. IE from the 12v rail created by the zener/capacitor combination that you place in the 'empty holes' when modifying a GHurd dump controller for 24v use.

1. How to get appropriate high and low trigger signals from the GHurd dump controller? I initially thought I would have to invert the high signal that the GHurd dump controller produces at the drain of its P-fet when it switches on. Similarly, I thought I would need to invert its low signal when switched off? Why? Because the Maplin monostable I used for the one-off timer is triggered on by sending its trigger input low, not high.

So I carefully built a little pulse inverter around one gate of a spare CMOS 4001 I had in my junk box. However, when I tried it, I realised that I didn't need it. I realised that I could take a low trigger signal directly from the junction of R3, P-fet gate, and N-fet drain. That worked fine and I dumped the CMOS inverter as unnecessary and a potential additional point of failure. KISS, as they say.

1. Rules for testing? Testing this in prototype form bent my mind a little. In fact, I thought the design was not working until I suddenly clicked. Here's the deal:

I initially tested by winding the input voltage to the GHurd dump controller up and down, watching the monostable timer turn on and turn off. I used the variable voltage supply to provide that easily-variable trigger voltage - and saw the benefit of having that useful and cheap piece of test equipment to hand.

Then I realised that the timer was turning on and off as I varied the input voltage to the GHurd dump controller. I was not seeing any sign of a one-off timer.

Then I got it. For testing purposes the one-off timer needed to be set so that I could see a really appreciable lag before it turned off. *And* the test was to turn up the voltage until the GHurd dump controller triggered and immediately turn the voltage back down below dump voltage threshold again. This let me see that the timer was actually staying on even though the dump voltage had dropped back. IE, it more accurately simulated a discharging battery.

Below is how it looked as I began to mount it in a box. The red wire linking the two circuit boards shows where the trigger signal is taken from the junction of R3, P-fet gate, and N-fet drain on the GHurd controller and where it enters the monostable circuit.

You can just see where the Maplin monostable supplies the trigger voltage for the power mosfet. It's the little silver terminal pin sitting by the bottom edge of the 'RLY1' box on the Maplin monostable board.

Once I had got the above working, much of the remaining work went into reconfiguring the signal LED at the output of the monostable. Somewhat bizarrely, the supplied Maplin kit keeps the LED on when the monostable output is low or not 'on'. Similarly, it turns the LED off when the monostable's output goes high. To me, this is counter-intuitive so I experimented with various ways of inverting the LED's 'on-offness'.

Failed methods included switching the monostable's LED and resistor combination from pin 3 signal to ground. If I recall correctly, the prevented the FET from firing.

I then tried a variant replacing the Ghurd's output stage p-FET with an n-FET but this caused the monostable output to stay on all the time.

I also tried a couple of simple methods using an NPN transistor as an LED driver (and therefore insulating the LED current from the power MOSFET) I found that although I could get the LED to come on when the monostable went high but not the MOSFET.

What worked was to grab an NPN transistor from the junk box, drive the LED through GHurd's 2.2k resistor at the NPN transistor's collector, and pop a spare 51k resistor into the base line of the transistor. Why did this work? Because - I later discovered - the base resistor allows enough voltage to to build up over the resistor that the gate of the power MOSFET still sees enough voltage to drive its gate into 'on' mode. What I hadn't understood was that transistors I was using earlier had passed so much drain current that the voltage on the MOSFET gate stayed lowe enough to keep the MOSFET off.

I had wondered if I should put a resistor in the gate of the MOSFET too but apparently a MOSFET's input impedance is so high that no effective current flows through them. Therefore no resistor is required.