The Electronic NiCad

This site was last updated on February 15, 2008.

We like NiCads a lot for their low internal resistance, their convenient natural voltage, and the fact that they can be recharged. However, for our glow driver, we'd like more:

we'd like a battery that holds its charge longer,
we'd like a battery that's less particular about how it's charged,
we'd like the voltage to be adjustable, to accommodate different plugs, and
we'd like to use long wires, for safety.

To meet these requirements, Steve Sacco designed the first Electronic NiCad. Later, Howard Rush copied Steve's circuit and improved it a bit (consulting with Bob Carver). Nowadays, it meets the requirements listed above, plus:

includes an Ammeter and a Voltmeter for feedback, and
shuts off in the event of a short (but resets instantly) to avoid melting the battery.

Basically, it's a constant voltage source. It measures the voltage drop across the plug and adjusts its output to to keep that voltage constant (regardless of the resistance of the plug or the length of the wires). Therefore, when the engine floods, more power is delivered to help burn off the flood.

It's build around a pair of the 2-volt Cyclon cells. These batteries can be recharged using a variety of methods and will hold their charge for weeks, leaking very slowly. They also have plenty of capacity (we use the X-cells which are 5 Amp-Hours), so they'll last all day, even under heavy use.

The basic idea is to use one pair of wires to carry current from the driver to the plug and a second pair of wires to sense the voltage across the plug. Then, if the voltage across the plugs drops (for instance, when the engine floods, the plug gets wet, and the resistance drops), the circuit can notice the change and turn up the current to maintain a constant voltage.

The driver circuit (see the complete schematic) has 4 main sections:

the voltage definition,
the integrator,
the current booster, and
the foldover.

The voltage definition section (show in the schematic above) lets us set the target voltage (across A and B) for the glow plug. By making it adjustable, we can use the driver for a variety of plugs.

The schematic is drawn so that higher voltages are toward the top and lower voltages are towards the bottom. D1 is a voltage reference, LM385BZ-2.5 (you can get all these parts from Allied Electronics), that maintains a drop of 2.5 volts across its leads. Since B will be close to ground, there should be about 1.5 volts across R1 and about 0.8 mA of current, which is enough to keep D1 alive.

Don't substitute a zener here! Use the voltage reference; it works.

The series of R2, R3, and R4 act as a voltage divider, with 0.75 volts across R2, 1.14 volts across R3, and 0.61 volts across R4. So, as we adjust R3, a linear-taper potentiometer, A will vary from 0.61 volts through 1.75 volts, with respect to B. Of course, the exact values will vary, depending on the resistors, but none of it is too critical. The goal is simply to provide a reasonable range of adjustment.

Finally, C1 smooths the jitter as we adjust the pot, giving the controls that silky feel so prized by combat fliers around the world.

The (somewhat abstract) schematic above shows the overall control loop. The glowplug is represented by R9. Notice there are 2 connections at each end of the plug. The heavy lines represent heavy wire that carries the current to/from the plug. The thinner lines represent smaller wires used to sense the voltage across the plug. In this manner, we are able to set the voltage accurately, independent of losses in the high-current leads (this is known as a Kelvin connection).

The heavy leads to the glow plug need to be able to carry about 5 Amps without melting. 18 gauge wire is more than adequate. Alternatively, several smaller wires can be be run in parallel. I use a Cat-5 cable, with 8 wires, each 24 gauge. 3 wires carry current out, 3 wires carry current back, and the remaining 2 wires are used for sense lines.

The voltmeter across the sense leads displays the measured voltage. I used one capable of displaying between 0 and 3 volts, DC.

In the upper left, A1, R5, and C2 together comprise an integrator. The integrator compares the voltage reference to the measured voltage and accumulates the error over time. Thus, if the measured voltage is low, the accumulated error (a voltage) will slowly increase, causing the current booster to force more current through the plug, eventually raising the measured voltage to eliminate the error. Similarly, if the measured voltage is high, the error will be negative and the accumulated error will slowly decrease, causing the driver transistors to reduce the amount of current going through the plug. Note that when I say "slowly", it's a relative term; the voltage stabilizes in a fraction of a second.

A1 is actually 1/4 of an LM324 quad op-amp chip (A2 and A3 use 2 more of the op-amps). Since each of the op-amps can only drive a small amount of current, we need a current booster to achieve the levels required for the glow plug.

We use a cascaded pair of power transistors (I use an MJE3055 for Q1 (overkill) and an MJE2955 for Q2 (about right)) to boost the current through the plug. In the section above, the voltage at X is controlled by an op amp and varies between 0 and 2.5 volts, relative to ground. As this voltage increases above about 0.7 volts, current begins to flow from X through R7, into the base of Q1, out the emitter, and through R8 to ground. We'll call this current Ibe (current from base to emitter).

Since Q1 and Q2 are power transistors, the gain is about 50x; that is, the current from the collector to the emitter (Ice) will be about 50 times Ibe (the exact amount doesn't matter). Since a silicon NPN transistor will keep its base voltage about +0.7 volts relative to the emitter voltage, we can solve some simultaneous equations to find that

      X - 0.7
Ibe = -------
        660

and

       X - 0.7
Ice ~= -------
          13

So, at its limit, when X is 2.5 volts, Ibe will be less than 3 mA, well within the capability of the op amp, and Ice will be about 140 mA.

Now, as the current into Q1's collector increases, it will come from the base of Q2. The current out of the base of Q2 is multiplied again, permitting about 50 times as much to flow out of the collector into Y (which connects to one end of the glow plug). Therefore, we see a maximum current of almost 7 amps through the glow plug, which is enough to hold it at 1 volt even its resistance should fall to 0.14 ohms. We use an ammeter (0 to 5 Amps, DC) in series with the glow plug to display the current.

Of course, the transistors and resistors aren't perfectly predictable devices. In particular, the gains of the transistors will vary depending on temperature and current. For the transistors I used, the data sheets suggest that Q1 will have a gain of close to 100, whereas Q2 may be closer to 30.

Note well that Q2 will get hot! Dropping perhaps 2 volts at around 4 Amps implies 8 watts. It needs to have a good heat sink.

While none of the values in the schematic are too particular, truth in advertising forces me to admit that my first efforts oscillated and I had to use an oscilloscope and consult with experts (see above) to understand what was happening. Capacitors C3 and C4 were added to avoid oscillation. C3 was suggested by Bob Carver and avoids a relatively high-frequency oscillation. C4 is apparently needed to control inductive surges (which sometimes cause oscillation) from the coil in the ammeter (suggested by Bob Pease in Troubleshooting Analog Circuits). I used ceramic disc capacitors for C1, C2, and C3, and an electrolytic capacitor for C4.

The foldover protects the battery against a short circuit. The idea is to compare the voltages on each side of the clip. If they are too close, we assume they are shorted together and pull down the voltage into the booster.

We actually divide the upper voltage by 2 before comparing. Thus, when the lower voltage is greater than 1/2 the upper voltage, the comparator A3 will turn on Q3 (an MPS2222A), pulling down the voltage going into the booster.

The op-amp A2 is configured as a voltage follower (the output voltage will equal the input voltage) and serves as a buffer, isolating the foldover and the booster.

You may wonder about cost and complexity. The electronic parts are almost free these days. Unfortunately, buying the meters and batteries new can run into money. Depending on your finances and inclinations, you may want to look for these things on the surplus market.

It's possible to simplify the circuit, saving parts but reducing functionality. For example, leaving out the foldover will save 4 resistors (R6, R10, R11, R12) and 1 transistor, Q3. It's also practical to leave out the voltmeter and even the adjustable voltage reference. I suppose the ammeter, together with C4, could be left out, but this seems a drastic loss of useful feedback. Finally, feel free to substitute different batteries. Lower voltages may not work, but 4.5 or 6 volts should be no problem. Higher voltages will probably destroy Q2.