In the first part of this Design Idea (DI), we looked at simple ways of keeping critical components at a constant temperature using a linear approach. In this second part, we’ll investigate something PWM-based, which should be more controllable and hence give better results.

Wow the engineering world with your unique design: Design Ideas Submission Guide

Adding PWM to the oven

As before, this starts with a module based on a TO-220 package, the tab of which makes a decent hotplate on which our target component(s) can be mounted. Figure 1 shows this new circuit, which compares the voltage from a thermistor/resistor pair with a tri-wave and uses the result to vary the duty cycle of the heating current. Varying the amplitude and level of that tri-wave lets us tune the circuit’s performance.

This looks too simple and perhaps obvious to be completely original, but a quick search found nothing very similar. At least this was designed from scratch.

Figure 1 A tri-wave oscillator, a thermistor, and a comparator work together to pulse-width modulate the current through R7, the main heating element. Q1 switches that current and also helps with the heating.

U1a forms a conventional oscillator running at around 1 kHz. Neither the frequency nor the exact wave-shape on C1 is critical. R1 and R2+R3 determine the tri-wave’s offset, and R4 its amplitude. U1b compares the voltage across the thermistor with the tri-wave, as shown in Figure 2. When the temperature is low so that voltage is higher than any part of the tri-wave, U1b’s output will be solidly low, turning on Q1 to heat up R7 as fast as possible.

As the temperature rises, the voltages start to overlap and proportional control kicks in, progressively reducing the on-time so that the heat input is proportional to the difference between the actual and target temperatures. By the time the set-point has been reached, the on-time is down to ~18%. This scheme minimizes or even eliminates overshoot. (Thermal time-constants—ignored for the moment—can upset this a little.)

Figure 2 Oscilloscope captures showing the operation of Figure 1’s circuit.

Once the circuit is stable, Th1 will have the same resistance as R6, or 3.36 kΩ at our nominal target of 50°C (or 50.03007…°C, assuming perfect components), so Figure 1’s point B will be at half-rail. To keep that balance, the tri-wave must be offset upwards so that slicing gives our 18% figure at the set-point. Setting R3 to 1k0 achieved that. The performance after starting can be seen in Figure 3. (The first 40 seconds or so is omitted because it’s boring.)

Figure 3 From cold, Figure 1’s circuit stabilizes in two to three minutes. The upper trace is U1b’s output, heavily filtered. Also shown are Th1’s temperature (magenta) and that of the hotplate as measured by an external thermistor probe (cyan).

The use of Q1 as an over-driven emitter follower needs some explanation. First thoughts were to use an NPN Darlington or an n-MOSFET as a switch (with U1b’s inputs swapped), but that meant that the collector or drain—which we want to use as a hotplate—would be flapping up and down at the switching frequency.

While the edges are slowish, they could still couple capacitively to a target device: potentially bad news. With a PNP Darlington, the collector can be at ground, give or take a handful of millivolts. (The fine copper wire used to connect the module to the outside world has a resistance of about 1 Ω per meter.) Q1 drops ~1.3 V and so provides about a third of the heating, rather like the corresponding device in Part 1. This is a good reason to stay with the idea of using a TO-220’s tab as that hotplate—at least for the moment. Q1 could be a p-MOSFET, but R7 would then need to be adjusted to suit its (highly variable) V_GS(on): fiddly and unrealistic.

LED1 starts to turn on once the set-point is near and becomes brighter as the duty cycle falls. This worked as well in practice as the long-tailed pair approach used in Part 1’s Figure 4.

The duty cycle is given as 18%, but where does that figure come from? It’s the proportion of the input heat that leaks out once the circuit has stabilized, and that depends on how well the module is thermally insulated and how thin the lead-out wires are. With a maximum heating current of 120 mA (600 mW in), practical tests gave that 18% figure, implying that ~108 mW is being lost. With a temperature differential of ~30°C, that corresponds to an overall thermal resistance of ~280°C/W. (Many DIL ICs are quoted as around 100°C/W.)

Some more assembly required

The final build is mechanically quite different and uses a custom-built hotplate instead of a TO-220’s tab. It’s shown in Figure 4.

Figure 4 Our new hotplate is a scrap of copper sheet with the heater resistors glued to it symmetrically, with Th1 on one side and room for the target component(s) on the other. The third picture shows it fixed to the lower block of insulating foam, with fine wires meandered and ready for terminating. Not shown: an extra wire to ground the copper. Please excuse the blobby epoxy. I’d never get a job on a production line.

R7 now comprises four -33 Ω resistors in series/parallel, which are epoxied towards the ends of a piece of copper, two on each side, with Th1 centered on one side. The other side becomes our hotplate area, with a sweet spot directly above the thermistor. Thermally, it is symmetrical, so that—all other things being equal, which they rarely are—our target component will be heated exactly like Th1.

The drive circuit is a variant on Figure 1, the main difference being Q1, which can now be a small but low-R_ON n-MOSFET as it’s no longer intended to dissipate any power. R3 and R4 are changed to give a tri-wave amplitude of ~500 mV pk–pk at a frequency of ~500 Hz to optimize the proportional control. Figure 5 and Figure 6 show the schematic and its performance. It now stabilizes within a degree after one minute and perhaps a tenth after two, with decent tracking between the internal (Th1) and hotplate temperatures. The duty cycle is higher, largely owing to the different construction; more (and bulkier) insulation would have reduced it, improving efficiency.

Figure 5 The driving circuit for the new hotplate.

Figure 6 How Figure 5’s circuit performs.

The intro to Part 1 touched on my original oven, which needed to stabilize the operation of a logarithmically tuned oscillator. It used a circuit similar to Part 1’s Figure 5 but had a separate power transistor, whose dissipation was wasted. The logging diode was surrounded by a thermally-insulated cradle of heating resistors and the control thermistor.

It worked well and still does, but these circuits improve on it. Time for a rebuild? If so, I’ll probably go for the simplest, Part 1/Figure 1 approach. For higher-power use, Figure 5 (above) could probably be scaled to use different heating resistors fed from a separate and larger voltage. Time for some more experimental fun, anyway.

—Nick Cornford built his first crystal set at 10, and since then has designed professional audio equipment, many datacomm products, and technical security kit. He has at last retired. Mostly. Sort of.

Related Content

• 5-V ovens (some assembly required)—part 1
• Fixing a fundamental flaw of self-sensing transistor thermostats
• Self-heated ∆Vbe transistor thermostat needs no calibration
• Take-back-half thermostat uses ∆Vbe transistor sensor
• Dropping a PRTD into a thermistor slot—impossible?

The post 5-V ovens (some assembly required)—part 2 appeared first on EDN.