# Help with closed loop heat controller (for a non engineer)

I am interested in setting up a controller which can modulate the output of 2 hotpads based on input from an infrared thermometer. The hotpads have their own dials, but I am considering controlling them by voltage instead. So my question is, what is the simplest method for modulating voltage (or rotating dials) based on temperature? One problem in particular that I am unsure about is how to account for the gradual propagation of heat through my system, since I am heating a large (2' diameter, 1" thick) rotating aluminum disk. I have a laptop at my disposal, and minimal funding.

I would prefer to avoid wiring my own sensors, so any recommendations you have on how to achieve this would be great. I could handle a breadboard at best with my experience, and I have only an introductory level understanding of electromagnetism... But I can code in C, Java, and Matlab, and I have experience assembling PCs, so I am hoping this is within my abilities.

• How tight does the temperature control need to be? +/- 1C? 5C? or??? Jul 8 '15 at 16:28
• I would hope for +/- 2C at most
– Ryan
Jul 8 '15 at 21:57

Since you seem to not be too familiar with control systems, electronics, and wiring things, I recommend a simple threshold detector that controls a relay. The relay then switches the heating element on or off.

Unless the IR thermometer has a output signal for this purpose, it won't be useful as the sensing element. However, there are various devices that produce a electronic signal as a function of temperature. For your temperature range, thermocouples or platinum RTDs is what you should look at. Thermocouples need more complicated circuitry to produce a useable signal, whereas RTDs are much simpler to use. Thermocouples go to higher temperatures though. I'd go with a RTD unless the temperature is too high.

• Ok so this would work like most conventional ovens do... I am a little concerned about accuracy, since the delay in heat propagation is on the scale of 10 minutes, and I am using this equipment for with the eventual goal scientific publication. My target temperature is about 80'C, and I don't see any reason why I would need run it above a max of 100'C. Would you recommend RTDs for this purpose? I should also remind you that my disk is rotating (about once per hour) so it may be difficult to implement contact thermometers.
– Ryan
Jul 8 '15 at 21:53
• @Ryan: You've got a long heat propagation delay and the thing you're measuring is rotating. This is not a simple problem. A basic on/off controller will probably overshoot and undershoot too much. At this point, get someone that understands this stuff to do this for you. It sounds like you'll need a real controller, and sensing the temperature won't be easy. About the only good thing is that your temperature is low. This is over your head. Jul 8 '15 at 22:06
• I appreciate your concern, and I will likely seek help in person, but I have basically unlimited time to do this, and I am willing to learn whatever it takes to finish the task. I am trying to get some perspective as to what needs to be done, and what my options are before I approach someone with my project.
– Ryan
Jul 8 '15 at 22:27

In my experience getting data from a rotating component is tricky without resorting to some kind of radio transmitter or a passing the data through a slip ring, which would probably require amplification of the data signal due to the noise generated by the slip ring... it can be done though.

I'd be cautious about using an IR thermometer for anything other than a reference measurement, i.e., about how hot is that thing. There are too many factors affecting the measurement, mainly the emissivity of the surface. Maybe others have better experiences, but I've never had much success.

I have seen reasonably good results that might fit your needs and budget using a magnet and a Hall effect sensor. Place the magnet on the wheel and since the strength of the magnet will decline as the temperature increases, so will the voltage from the HE sensor. You'll only get the measurement once/rev though so you'll have to look at the peak values and calibrate from there. Another benefit is that you'll also have a tachometer so you can measure the rotational speed if you need to. This shouldn't cost more than a few , but might be more complicated than what you're looking for.

As to understanding how heat the disk will heat up, you could do some modeling using the software of your choice...I like Ansys, but if you don't have access to something like that, it's probably not in your budget. There are other, less costly solutions. If you have enough symmetry, you could probably even do it in Excel or a similar program.

You could also take a number of temperature measurements over the surface of the disk to see how it heats up and cools down.

In the end, you want to have a good idea of how hot the disk is an a given location when you only know how hot it is in one location.

• Ok thanks, this is an interesting suggestion. I've looked a bit into IR sensors, and a simple workaround for variable emissivity is to stick a piece of masking tape onto your surface and use that as a reference to calibrate the sensor to account for emissivity. I do also have a digital thermal couple to calibrate the IR with, but I've found that it doesn't make enough contact with our flat metal disk to be accurate, since it is basically just a blunt-ended wire that rests on the aluminum. Would a magnet suffer from deterioration over time?
– Ryan
Jul 8 '15 at 21:44
• Also: my disk rotates very slowly, only about once per hour, so would noise with a slip ring still be an issue at this rate?
– Ryan
Jul 8 '15 at 21:56
• I've never used slip rings at speeds that low, typical speeds of 400rpm up to 20k rpm., so, I honestly don't know. I would expect that the noise issues would be low, but that's not based on direct experience at those low speeds. I've used a dab of epoxy to attach thermocouples to metallic and ceramic surfaces. The epoxy provides good enough thermal contact for wjhat I was doing, much better than just placing the TC onto the surface. Another approach I've used is to drill a hole into the piece, just big enough for the end of the TC and inserting the TC. Jul 9 '15 at 16:44
• Unless you only want data every hour, I'd abandon the magnet idea. While you could use a magnetic strip or a series of magnets, there will be variation in the magnet's strength along the strip and variation between each magnet that you'd need to account for in calibration. Based on what you've said, and your I would start with a TC or RTD, a slip ring and a cheap PID controller from eBay for your control system. Jul 9 '15 at 16:51
• Ok, I also thought this would be the simplest, but I've been looking into it some more, and Thermocouples produce very weak voltage (10 mV for type K TCs in the <100'C range) and most slip rings I've looked at provide a noise rating of <10milliOhms, which, if I am thinking about this correctly, would be enough resistance to seriously reduce the current. Given these details, would you still go ahead with this method? Also in case you are interested I found some flat (albeit expensive) TCs: omega.com/pptst/SA1.html but drilling a hole in the disk sounds like a better solution.
– Ryan
Jul 9 '15 at 21:46

The feedback topology for thermal systems is fairly easy when the heat source is continuously adjustable, which I believe yours is. This is because the transfer function of simple thermal systems has a single pole, just like a standard RC circuit. In this case, a stable closed loop controller can be made by simply connecting the sensor directly to the actuator (assuming both are linear) with a variable gain stage in between. After doing so, the gain can be increased until just below where the loop starts to oscillate.

For long-term tracking you should also consider adding a switchable integrator before the gain stage in your feedback electronics. It is important that it be before the gain stage so that the DC portion of the feedback signal doesn't get hit with infinite gain when the integrator is switched on.

All of this can be implemented in Matlab if you have the right packages and are capable of getting the signals into and out-of the computer. It is probably easier to build the rather simple circuitry yourself though. Alternatively, you could spend the money on a commercial PID controller and set the D gain to zero.

• Ok, I apologize for my lack of fluency with EE terms, but if you have the time, could you explain what you mean by the (a) transfer function (b) pole, (c) actuator, and (d) what the benefit of using an integrator would be?
– Ryan
Jul 8 '15 at 22:07
• @Ryan This isn't really a large enough space to explain a) and b), suffice it to say that the concepts are pretty fundamental to understanding complex feedback systems. Your system is simple enough that you don't need to fret too much about understanding them. For c), I simply mean the heating element, i.e. the thing that actuates on the state of the system as opposed to the sensor which senses the state of the system. For d), without an integrator the temperature of the system will never actually reach the set point, even if the system is undisturbed because the gain at DC will be finite. Jul 8 '15 at 23:30
• This is PID Control. If you can read the IR thermometer temperature into a computer, a PID algorithm could output a value used to drive the heaters. The goal of PID is quick, accurate response (from a sensor to a setpoint) with minimization of error. PID's need to be tuned to work properly, so some experimentation will be necessary. Jul 13 '15 at 2:46