Well, as I don’t have to point out, it’s been a while. This time it’s actually legitimate however, as I really haven’t had a chance to do anything with the project at all since my last post.

Since then however, there have been issues with the intended design of the heater core, and I’ve been advised that my approach would not be effective — that in fact, it possibly would not work at all.

So, I’ve had to seriously rethink everything, and I’ve been worried that not only would I have to change my control plan for the heater, but that I wouldn’t end up getting the level of control I’d wanted. Instead of smooth electronic control like a light dimmer, I’d instead have to deal with using relays to selectively switch elements on and off to select from a discrete set of output values.

Why though? What was the problem with my plan? As I’ve mentioned before, the ceramic elements I’m using are called PTC elements, which stands for Positive Temperature Coefficient. This means that as temperature rises, so does resistance. This makes these elements safe since they’re inherently self-limiting. However, according to those who have warned me about my plan, this characteristic is also part of the problem.

Since resistance rises with temperature, this also means it lowers as the device cools. The result is that at lower heater settings, the device will want to draw very high currents, and this will mean that the semiconductor(s) necessary to control power will have to be very large and therefore very expensive. Neeless to say, I was less than pleased at this warning, but decided to completely rule it out before giving up. That was last month, before an avalanche of outside concerns basically left me with no time to devote to the project. This weekend, I finally sat down to do some experimenting.

With the elements cool, I’d decided from the warnings I’d received that even at low voltages they’d draw a large, unmanageable amount of current. I started with one volt across each layer of ceramic material, by putting a high-current 12V AGM battery across all twelve rows of wafers, in series. Though I left it connected for several minutes, I got no detectable heating — not what I’d expected. Maybe it was warming up slightly, but more from the heat of my hand than from the ceramic wafers themselves. Measuring the current, I saw 6.3ma, which was a LOT less than I was expecting. Clearly the cold resistance was high, which was surprising but encouraging. Gradually putting the battery’s voltage across fewer wafers in series until finally putting all I2V across a single chip, I still detected no heating and negigible current draw. I measured the cold (70F) resistance across a single element and found nearly 200 ohms.

I’ll stop for a moment and explain something. To call these ceramic elements “PTC” is not telling the entire story. This characteristic is accurate for part of their operating range, in which the resistance does increase fairly evenly (and quickly) with small increases in temperature. It’s commonly understood that powering a 120V element with 240V will yield about the same amount of heating, since the resistance rises so sharply with temperature, the element gets a little hotter but not much. However, at lower temperatures, the relationship becomes highly nonlinear. In fact, it’s a curve with a valley, and where this valley is located was the detail I did not know. I had thought it was at a fairly cool temperature. I was wrong. As I would soon discover, the valley in the curve occurs at a pretty high temperature. I don’t know what this temperature is exactly (I need to obtain something that will allow me to accurately measure it) but it is well into the useful heat range of the device.

I kept adding batteries, one by one. First 24V, still no detectable heat. I did however notice that the current was starting to rise. Very, very slowly, the resistance was dropping in response to the undetectable increase in temperature. This meant that the material was not acting with a PTC characteristic, it was instead acting as an NTC material. So I had not yet reached the valley in the temperature/resistance curve.

I tried 36V. I could now barely detect warmth in the device, and the current rose faster. Still, even with the power flowing for a couple minutes, it never became more than a little warm. Then 48V. It started to get very warm at this point, maybe around 100F. Current climbed, gradually, showing that the ceramic material remained in a Negative Temperature Coefficient mode.

I added another battery, for a total of 60 volts. As expected, the element heated up much more quickly, and current rose more quickly as well. However, after maybe around 30 seconds, the current reached a maximum, and finally began to fall. I’d reached the valley temperature, and as temperature continued to rise, the material changed from NTC to PTC. At the lowest resistance, current never reached more than approximately 2.7 amps. Adding another battery for a total of 72V, current peaked at about 3.3 amps. I’d found the minimum resistance, which works out to roughly 20 ohms per wafer.

Trying ten batteries in a string for a total of 120V, I hooked up the power and found that no matter what I did I never drew more than approximately 5.6 amps or so. Multiplying this by my 12 elements wired in parallel, and we’re looking at less than 70 amps. A 500V, 150A transistor or set combined to make these values would be more than sufficient for controlling this amount of current. The going price for a 20A 500V MOSFET is about $3.50, so with 8 of these in parallel, I’m looking at $28 in power transistors. Not cheap at all, but not beyond sanity.

However, there’s a catch. Electronic current control typically works by a process called Pulse Width Modulation. By rapidly turning the current on and off and by varying the proportion of the time during which the current is on versus how long it’s off, you can create an effective voltage that’s anywhere between zero and the total input voltage (minus semiconductor losses). The problem is that with a purely resistive load, there’s nothing to smooth out the current “ripple” — during the “on” portion of the phase (though it may be a tiny fraction of a second), the resistor sees the full pack voltage. In my case that’s 348 nominal, rising to well over 400V at top of charge.

What’s needed is some kind of inductor to smooth out the current, so the heater sees the average 120V instead of pack voltage. And that is my next endeavor — with a given PWM frequency and a given input voltage and maximum current consumption, how should I smooth out the current flow? If I use an inductor, much inductance do I need? If based primarily on capacitance, how should I size the capacitor? This gets into topics in analog electronics that are elementary to any electrical engineer, but of which I have little existing knowledge. Time to crack some books, and study a bit of basic theory.