Pulse-width modulated heater?

[riceman0]

Hello, I wanted to get a sanity check on an idea... we are controlling temperature by turning on and off a metal heating plate. Our controller has a digital output, and we are considering getting analog-ish control of our temperature by generating a PWM signal to the plate. The theory is that toggling our heater on/off with (say a 5Hz) PWM signal with 40% duty cycle will produce 40% of the heat as a constantly-on heat plate.

Is that a sound approach?

Thanks very much in advance!

 

[lownox]

Yup.

See here...

http://www.indeecocontrols.com/

 

[MintJulep]

Yes. However, depending on power levels, power source, and whatever else your heater may be around, you may have EMI issues with nearby equipment - although at 5 Hz probably not.

 

[IRstuff]

"Producing" at one level is not always the same as what's transferred to the next level. If you are heating something sitting on the hot plate, then during the off periods, that something will be losing heat to the hotplate, which wouldn't necessarily occur if you were continually supplying 40% of the heat.

If this is only to be an approximation, then there should be no problems.

TTFN

FAQ731-376

 

[kevindurette]

As far as electrical power supplies are concerned, efficiencies don't get much better than PWM; that's a fantastic idea. As long as you're assuming that heat transfer is linearly proportional to the temperature gradient, you're fine, but be aware that radiation is important when high temperature differences are involved. (It scales with the fourth power of temperature.) Changes in material conductivity with respect to temperature may also play a role, but intuitively I'm going to guess that they're probably not significant enough to matter in most applications.

If you're looking to get a little more involved... look into the TRIAC/DIAC circuits used in light dimmers. It's very similar to pulse-width modulation, but instead of regulating the duty cycle of a square wave, these actually use part of the AC sine wave as a switching mechanism, regulating the power by a chosen cutoff point. (They recommend SCRs for higher loads. I don't know how high yours is.)

http://home.howstuffworks.com/dimmer-switch3.htm

http://en.wikipedia.org/wiki/Light_dimmer

 

[kevindurette]

"'Producing' at one level is not always the same as what's transferred to the next level. If you are heating something sitting on the hot plate, then during the off periods, that something will be losing heat to the hotplate, which wouldn't necessarily occur if you were continually supplying 40% of the heat."

Fourier's Law disagrees with the "losing heat" premise; the hot plate will always be as hot or hotter than the device sitting atop. I'm assuming a linear conduction model with constant conductivities and negligible change in the heat transfer coefficient of the freely convecting air. If one heater is set at 40% power and the other is set at a 40% duty cycle, then the time-averaged temperatures of the heaters, the plates, and the containers will be respectively identical. Work is the integral of power over time; the precision of the pulse width approximation is only dependent on the driving frequency.

 

[IRstuff]

When you're not powering the hotplate, especially if it's off 60% of the time, the hotplate can cool down below the object it's heating, particularly if the bottom of the hotplate is exposed to the air.


TTFN

FAQ731-376

 

[moltenmetal]

It's doubtful that the time constant of your hotplate is small enough to warrant a 5 kHz PWM signal to control its temperature. An ordinary solid state relay run directly off your digital output with a ~2 second cycle time will do just fine with less complexity.

 

[kevindurette]

"When you're not powering the hotplate, especially if it's off 60% of the time, the hotplate can cool down below the object it's heating, particularly if the bottom of the hotplate is exposed to the air."

So what you're telling me is that the extremities of the system, which form a heat sink to the air, will be at a lower temperature, for some of the time, in the switched system. This will mean that it'll lose less power during that time (due to a lower temperature gradient), but at the same time, the other end of the cycle will require such parts to get hotter for a while, which will result in more loss to the ambient. The two effects ought to cancel, again provided we're ignoring the effects of surface temperature on natural air convection. (Such convection becomes more efficient at higher gradients.) Right?