# Why do consumer ovens use thermostats instead of PID + PWM?

I measured the temp stability of my oven, and it's a really big swing of 10-15c in a saw tooth pattern. Aren't PI(D) controllers with variable output meant to dampen this type of oscillation? I imagine it would even make thermal recovery faster when the door is open and closed.

• Comments are not for extended discussion; this conversation has been moved to chat. – Voltage Spike Mar 24 at 22:20

Mechanical thermostats are capable of switching high currents with very long life in that application. They are inexpensive and accurate enough. They don't mind the high ambient temperatures and don't require cooling fans or heat sinks. The technology is mature, and can be produced by automated lines. They don't tend to fail 'on' and cause damage or fires. The short-term variations in the air temperature are not a hindrance to competent cooking.

Yes, you could replace the thermostat (or slow-switching relays controlled by electronics as in our oven) with more sophisticated controls and phase-controlled heaters or something like that and it would keep short term variations in air temperature down, but why would you if it doesn't matter?

And it would not aid recovery- might slow it if the controller is tuned to avoid overshoot at the expense of rise time. Best-case recovery is related to control authority, not the controller. When the temperature drops with an on/off controller, the heater(s) go on 100%, you can't get any better than that.

That's typically the limitation in real control systems- a PID controller could in theory turn a battleship on a dime, but the amount of power required of the engines just isn't there to make that happen.

• Also, most of the thermal mass of the system is in the food itself (and whatever container it is in). Tuning a PID controller to behave correctly over the whole range of possible thermal masses (from empty oven to one with a giant stuffed turkey in it) is going to be lots of work. And when all is said and done, it's much more important to make the temperature consistent everywhere without significant hotspots than to hit a setpoint to within a degree. No amount of sophistication in the controller helps with the former. – TooTea Mar 22 at 8:17
• Uh... the only thing that needs to be in the hot oven is a thermocouple, and those are available for peanuts and can tolerate temperatures in the thousands of whatever degrees you care to use, so the arguments about cooling fans and heat sinks seems to miss the mark. Phase firing is for reactive loads - an oven would work fine with zero-cross SCR control, so that too seems to miss the mark. A simple PID controller can be had for a few dollars, but the SCR/thyristor unit would probably come in around $100 for the OEM, so it's really about cost - and not a big one at that. – J... Mar 22 at 17:56 • @SpehroPefhany Range top? Aren't we talking about the oven? A bake element is only 2600W - a thyristor control for that power level absolutely does not require a fan.$100 I just threw out there - you can make them cheaper,surely, especially for a mass produced consumer device. For a $2500 appliance, it's certainly getting into achievable territory is my point - I agree that cost is and has been the major showstopper, but I think that argument is rapidly losing steam as the cost of automation equipment comes way down. – J... Mar 22 at 18:06 • @J... Well, it's been obvious and quite practical to do for more than 50 years (since triacs and SCRs and op-amps became very affordable). Why do you think this is not a huge thing, even in gadget-happy places like Japan and China? – Spehro Pefhany Mar 22 at 18:13 • @SpehroPefhany$3 (1971) = $20 (2021). Now add that to the thyristor cost. And the rest of the electronics you need besides that single op-amp. Do you honestly mean to argue that the real cost of electronics has not dropped by orders of magnitude over that time? – J... Mar 22 at 18:23 A thermostat is already PWM, just slow. The element can be on or off, and the only thing that allows your oven to be set to 500°F or 250°F is varying the duty cycle. That's PWM. You could reduce the ripple by increasing the frequency. Would that be of any benefit for ordinary cooking applications? I doubt it. And while you could add a more sophisticated PID controller, an oven does not exhibit significant overshoot or undershoot on the time scale of the PWM switching time, so there's no need. You would not improve recovery time, because the thermostat already turns on the element 100% when the door is opened. It's not the controller that limits recovery time, but the power of the heating element. You could increase this of course, but you'd then have to upgrade the electrical supply. There would also be significantly increase safety concerns in the case the element gets stuck on, necessitating additional complexity and expense. You'd also be increasing the cost of the electronics. Replacing a relay for a solid-state device isn't too expensive, nor is the controller. But if you add a high-power phase-controlled load in every home you will probably attract some regulatory attention. Done cheaply, the power factor of a phase-controlled oven will be terrible. This will decrease power distribution efficiency and spew electromagnetic noise. Computer power supplies are required to have active power factor correction and undergo EMC testing for these reasons, and it's likely you'd need to do the same if you started manufacturing these ovens in significant quantity. So ovens stick with the simple thermostat design not because an oven is not a machine for maintaining a maximally stable temperature, but rather a machine for baking food. Baking food does not have especially stringent temperature stability requirements. A thermostat is the simplest, cheapest, most robust way to meet those requirements. • +1 for the first half of your answer, but the part about phase control is an unnecessary strawman. Given the thermal inertia of the system, there's exactly zero need to switch elsewhere than during zero crossings. PWM with a 1-second frequency would be plenty fast for any oven and still give you a 1% duty cycle resolution. – TooTea Mar 24 at 15:59 • If the element isn't already on when you open the oven door, it takes some time for the temperature to drop enough for the thermostat to trip, and then for the element to heat up. A really smart controller would sense the door being open, or sense the temp headed down quickly, and start heating the element several seconds sooner. Does this matter? Probably not, but it would slightly improve recovery time. – Peter Cordes Mar 24 at 21:36 • @TooTea Sure, synchronous switching would improve the power factor and reduce the noise. But you'd still have to develop that, including the engineering time, components, and compliance testing. That's my point: it's not that it's an unsolvable problem, it's that it's a non-zero cost for a near-zero benefit, because the thermostat is already good enough. – Phil Frost Mar 29 at 13:34 Why do consumer ovens use thermostats instead of PID + PWM? The answer is the word 'Consumer', which means cheap and low specification. Professional ovens for (say) resin curing will use PID control. To cook food, a 15°C sawtooth is acceptable, and much cheaper to provide than a more tightly controlled temperature. The time constant for heat absorption into the centre of the food lasts many sawtooth cycles. Not many foods have a thin surface layer that will be damaged by a few more degrees of heat. I've added a big N/C relay in series to the heater/thermostat circuit of my kitchen oven (don't tell the warranty people, and I didn't tell my wife until she caught me using it one day). With nothing connected to it, it just looks and behaves like a normal domestic oven. It cooks fine, and has a sawtooth temperature swing of 10 to 15°C. When I want to heat plastic for forming, or to cure resin, then I hook up a cheap PID temperature controller to it. The 10 s cycle time results in a swing not really measurable with my cheap thermocouple meter. Leaving the oven's thermostat in circuit provides an emergency max temperature limit should anything fail on the PID • Here is a big part of the answer, baking food in the oven doesn't require more than a dozen degrees of accuracy. so there is no need to provide that to consumer units. – ratchet freak Mar 24 at 9:29 • Most foods with a thin surface layer actually benefit from high heat on that (for example starting bread at 240°C then turning down to get a nicely browned crust, roasting a chicken when you want the skin crispy...) – Chris H Mar 24 at 11:21 • When you are baking, roasting, etc, you are typically in ranges between 120 and 150c. There are massive differences in the development of myard development over this range. Cook times, can vary as much as 50% over a 15c temp in this range. – Justin Alexander Jun 3 at 8:17 The more meaningful way to measure temperature variation would be with a probe inserted just below the surface of some baked goods. I don't see any advantage unless you are broiling very close to the element. Although my digital sensor has undershoot and never changes more than a degree so it must be sensing with mass damping. Obviously, if you sense anywhere within view of the infrared or near the element can get attenuated variations from an element > 2000'C. But the mass stabilizes it and cooking with a pyrex tray of water underneath also stabilizes it and adds moisture content to reduce drying out. I have seen 1 Relay replaced on our LG stove within 4 yrs and none for the next 5. The thermal mass of the baking and high thermal resistance of air means the cycle time of 1 minute or so makes little difference to the average temperature of the baked goods. The air temperature gradients will be reduced with force air flow which speeds up cooking and with 5p% humidity added and regulated allows commercial baking to be done in far less time with more precise temperature control. So given the life span of inexpensive 15A relay contacts having this hysteresis increases the longevity of the Relay to 10 years more or less with moderate to low effects on baking. But as every baker will tell you , every brand of oven is different so temperatures need to calibrated from results to prevent burning as the time constant to reach internal optimum temps on ranges depends on hrs/per kg. You might be able to design an IGBT proportional oven or stove top but the benefits are less than you might think. My stovetop cycles every 10 to 30 seconds and 50k cycles is typical loaded for a 500k mechanical wear lifespan. I think you might be able to offer competing sales advantage with this but keep in mind cost of goods is about 10x the BOM cost so you are competing with cost reduction pressures of using a 1$ relay and sensor vs your design.

• Adding more sophisticated temperature control would not only allow for improved temperature stability, but if the stove and oven used a common controller it might be practical to arrange things so that the amount of power fed to the oven would adjust itself according to what burners were in use, and possibly also allow it to be jumpered according to the capacity of the feed wires (so that if a range is on a 40A circuit and the stove isn't being used, the amount of power available to the oven would be much greater than on a 30A circuit when all of the burners are on "high"). – supercat Mar 21 at 22:29
• @supercat You don't need a smart controller for that, it could work just fine with simple bang-bang thermostats plus primitive NAND logic (implemented by a handful of relays if you wish) to ensure they can't all switch on at the same time. – TooTea Mar 22 at 8:22
• @TooTea something similar is actually done in some portable stoves, which is an improvement on some older designs that completely prevented use of hotplates with the oven on and vice versa. – Chris H Mar 22 at 12:55
• Load sharing is an idea that defeats the purpose of self-cleaning ovens – Tony Stewart EE75 Mar 22 at 13:05
• @TonyStewartSunnyskyguyEE75: If one wants to allow maximum-speed preheating under different conditions, one may want the ability to switch on heating elements that produce different amounts of power. Once one has that ability, the marginal cost of adding a temperature controller that would try to e.g. cycle the oven between using 15A and 20A once it was up to temperature would be relatively slight. – supercat Mar 22 at 15:13

Elaborating on one bit of Spehro Pefhany's answer: there are three relevant quantities here. Primarily, we have the power of the heating element and the thermal mass of the oven. We observe that the thermal mass is pretty large compared to the heating power; it takes a long time to change the temperature of the oven significantly, something like 30 seconds per 10°C. So there's no use doing PWM at hundreds or thousands of Hertz to get "analog" control; we can have a switching time of quite a few seconds and get basically the same result, because the temperature response of the oven is our low-pass filter, and because as long as you don't go too far out of whack the final result of cooking depends on integrated power over time anyway.

Secondarily, there is the response time of the heating element itself, governed by its thermal mass. I would rate this as "mediumish", going from dull red to yellow-white hot in about 10 seconds, and the same in reverse. This quantity is relevant if you want to prevent overshoot, but if that turns out to be an issue, engineers have known how to control overshoot with conventional thermostats for a long time. It can be done using something called an "anticipator", which is just a very small heater located right next to the temperature-sensing element. When the heat is on, the anticipator warms up, which effectively lowers the thermostat setpoint, resulting in an earlier turn-off. When the heat is off, the anticipator quickly cools down, and the setpoint rises again so that the turn-on isn't delayed. With correct design, an anticipator acts very much like a "D" term and greatly decreases overshoot.

I don't know if ovens commonly have anticipators (I found a patent for one dated 1964), but they could be used, and if they're not it indicates that the oven works well enough for everyday use even without one, so there was no reason to spend money on something no one would notice — a thought that could also be applied to the PWM concept.

(Incidentally, I suspect that an oven element PWM'd at any frequency in the human audible range would hum or whine quite noticeably.)

Unlike mobile phones, ovens are used for 20 years, sometimes more. As a result the market for new ovens is relatively small, and any design change increases the price rather significantly.

Now, if you were to buy a new oven, how much extra would you pay to have a PID in it? Knowing that it won't really improve the cooking results or reduce power consumption. I suppose it will be not much.

The necessary change, on the other hand, will be far from trivial to implement properly. Electronics sitting next to the oven need to deal with high temperatures, and probably have good cooling. Cooling is noisy, and a quiet cooling fan able to last for 20 years is expensive by itself. And then you will need a DC power supply for your PID. And then you won't be able to use a conventional relay for power switching, so you'll need SSDs or thyristors. And if any of this fails within 20 years, you'll get the reputation of a poor quality brand. Do you still think it's worth the trouble?

• Modern ovens are very well insulated, so the electronics cooling wouldn't be too bad. That said there's often a cavity fan already, to pass air through a gap outside of the insulation, as the next surface is likely to be wood. Route this airflow over the electronics first and the cooling is done. DC power supplies are often in there for LED clocks and cooking timers anyway. Overall I agree it's unnecessary extra complexity and would be expensive, but those specific points aren't the worst ones to deal with. – Chris H Mar 24 at 11:17
• ...Reliability and unusual behaviour in edge cases are major factors, The PID must work when the oven is absolutely rammed full, but also when running at minimum temperature with the door propped open and next to no thermal mass of food (e.g. drying herbs; I have a food dehydrator but often use the oven when I can't be bothered to get the dehydrator down from the attic) – Chris H Mar 24 at 11:19
• @ChrisH I agree with out points regarding cooling. The DC supply for a LED clock cannot be used as is for the controller: if the clock dies or resets while the over is working, nothing dramatic will happen. Powering down the controller will be far worse. And I don't see the problem with PID covering all use cases, sure, you have to consider them all, but it's not rocket science. – Dmitry Grigoryev Mar 24 at 12:22
• If the PWM is designed to fail off should the controller crash, the failure mode is un(der)cooked food, as it would be for the common function of timed cooking. I agree that sensible limits should be able to handle safety of all physically possible use cases - doing so satisfactorily will be harder, i.e. more expensive) – Chris H Mar 24 at 12:30
• I don't know if anyone would understand "PWM/PID" on the spec list, but I do see something like "precision temperature control for consistent baking" in my \$ Wolf oven. – JDługosz Mar 24 at 14:32

Two other thoughts. Regulatory authorities have confidence in the present controls being adequately safe. Its hard to justify spending development money on testing and convincing to get a new approach certified.

Also, perhaps the controllers are the same as for gas fired ovens? If so, minimum on and minimum off times are in the minutes, so fast proportional control would not make sense.

And speaking as a baker, my intution is that the variation would make very little difference in baking performance. Having correctly calibrated setpoints would make a difference, though for touchy recipes. Its widely assumed that setpoints can be off by 50F, and need to be checked. Some old dial thermostats oven even let you adjust the offset by turning the control with respect to the dial graduations. This does not instill confidence in the accuracy.

However, I did go through the entire introduction of PIDs in High-End Coffee, which does give me hands on experience in industrial application of PIDs controls in consumer goods that used to only apply thermostat controllers.

For starters, you definitely should turn the question around. The more sane reasons are in application, not in rejection.

PIDs only got affordable very recently. The first digital ones i bought were in the thousand dollar range and that was just shy of two decades ago. In contrast, the cheapest thermostats never costed more than a few cents and are by all means reliable enough to toast bread for instance. That is a good reason to not choose a PID unless you have a clear and present need for it.

Besides that, PIDs come in a whole world of different kinds, varieties and qualities. Anything from a self adjusting aircraft auto pilot (MCAS) to a single stage analog contraption. That sheds light on another main reason to choose any kind of PID over a thermostat. The entire chain of whatever process you want to control has to allow the accuracy of a PID to have effect. A PID is just another link in that chain and will not contribute to its result unless all other links are at least equally accurate.

Thermostats are way more user friendly than even the most basic PIDs. That might also play a role, although I have seen use of PIDs where its sheer fanciness had to do the job, because they really didn't do the process any good.

Another thing is the choice where and how to apply a PID. We achieved stunning results in coffee machines that would permanently discourage you from ever going to Starbucks again, but demanded an operator proficiency in all other variables so exceptionally high, that any slip-up would immediately make that product's quality plummet. Considering your question is about consumer goods, that choice becomes an issue.

Finally, there is no end to how complicated a PID control can get. Currently I am working on a control system for a sample roaster that allows it to produce roasting profiles that can be directly copy pasted to any known large single batch roaster. For that purpose, the PIDs have to enable to adjust the roasting characteristics of that sample roaster to perfectly imitate those of any large roaster. That's like building a PID that imitates a thermostat. We have been working on this project for just over two years now. No prototype is available so far.

Nowadays you could buy a perfectly workable PID controller for less than 10 bucks. But then off course there are the more expensive contraptions you can simply plug in between the wall socket and the power plug of your oven or any other device you wish to control. Shove the sensor pin wherever you want a preset stabilized temperature. That PID will (learn to) do anything at all to keep that sensor signal stable, with an accuracy of +/- 0.07 degrees C and a response time of less than a tenth of a second, switching an SSR strong enough to lighten up the entire neighborhood. With that thing in your kitchen, your entire question becomes academic. I bought the lab equipment variety of those in 2012 for just over 800 Euros, but they must be available much cheaper now. Its a fun gadget and very practical if you want to do things that demand accurate temperature control, like working sugar candy or gently reducing delicate sauces, but also the making of ice cream or cucumber sorbet for instance.

If anyone would ask me how to get an oven with a PID control, that gadget is what i would advise him to buy, rather than look for a PID controlled oven, which in its turn probably provides the best answer to your question.

Someone who would have any real actual need for a PID controlled oven, would not be looking for any specific oven, but rather for a multi purpose PID. So why build a (high volume) consumer product that only few people could use and nobody is really looking for?

• Control theory meets culinary arts. Never thought PID theory could possibly come in handy to prepare cucumber sorbet... – P2000 Mar 24 at 17:25
• That and a zillion other things.What has never ceased to amaze me, is the world of things that have become easy to tackle or have even first entered human awareness with the help of PID theory and microprocessor technology. – Berend Mar 25 at 16:08