Student Question on using a PID as a Dimmer to Control a Light Bulb!!! This is a fun one!!!!!!

snipped-for-privacy@gmail.com wrote: (snip)

The response to error of the control system is set with the three constants that configure how much P, I and D terms. You can set them to provide the fastest possible response, but with decaying cycles of over and undershoot, or an extremely slow creep toward setpoint, or anything in between. Making the decisions about how to set the three gains is called loop tuning.

-- snip --

-- snip again --

I generally recommend that book every chance that I get, but I was refraining for just that reason. If you're motivated, and if you got through your first few weeks of calculus, it may be of help. The title is "Applied Control Theory for Embedded Systems", the publisher is Elsevier, the author is yours truly, the intended audience is embedded hardware/software engineers who only learned digital circuits or none at all, and the book doesn't assume that you know anything about control theory except that you think you need to use it. I did keep the theory to what I felt was a necessary minimum -- but there's still a good slug of mathematics in there to be gotten through.

Here's Elsevier's page on the book:

There's a blurb about it on my website, including this link: will let you buy the book from my favorite bookstore. Of course Amazon has it as well, and your local bookstore can probably get it.

As I mentioned in a recent posting, the book doesn't directly answer the question of reducing overshoot, although it gives you a pretty complete toolbox to do so. I've already marked up my development copy, so should the book ever get to a 2nd edition there will be something in there then, and something will be popping up on my website soon.

"What's your physical arrangement? Are you heating your thing directly? Through plastic? Glass? Is there a great need for the bulb to be compact? Does it have to heat with light? Does it have to heat with _visible_ light?"

The bulb is plugged into the corresponding base for that bulb type. It heats the object without touching it, from the underside. Just like a bunson burner. However, its not that simple, its got air running past it to heat by convection. Its a crazy application design, but its feasable because we tried using it to boil water, and the beaker heated perfectly, but we had no control over temperature, which brings me here. There is a definate need for the bulb to be compact and there is no way around that. Regardless, any sort of heating element will need the variable voltage or alternations like we are implementing into this project. I dont mind replacing a 3 dollar bulb every 6 months, so no worries about that.

So, whats the best plan here, get a PID with an linear analog (0-10v) voltage output and hook it up to a (blank?) and then to a dimmer circuit.

The big question you have to answer is what is the thermal time constant of the object being heated. This will determine whether a 2 second pulse cycle will produce acceptable temperature variations or not.

If not, then you can use the lowest cost PID controllers that have a low voltage (usually 10 to 24 volt) solid state output and an ordinary solid state relay that can control the lamp power in on/off fashion.

If the thermal time constant is so short that a significant thermal swing will occur in a two second power cycle, then you will have to use a much faster cycle time, like 1/120th second you get with phase control. This will require a controller with a 0-10 volt analog output and a much more expensive dimmer type solid state relay that will accept the analog control signal and turn the power on for a proportional fraction of each half line cycle. At this pulse rate, the filament temperature will vary very little and the object being heated will almost certainly be unaware of the high frequency pulses.

I saw an omega controller for $195 that has the thermocouple input and analog output, but I am fairly confident you might find one made in Japan or China that would cost about half that. The phase control relays cost most of $100, but if you want to play around with an old lamp dimmer and a photo cell (to act as the variable resistor normally connected to the knob) driven by an LED from the analog output, you might make one that works well enough for 10 or 20 dollars.

By the way, I haven't seen what voltage bulb you intend to use.

The solid state relays work best with 120 volt bulbs.

John Popelish wrote: (snip)

By the way, the pulse cycle output type controllers are, by far, the most common type, and the lowest cost. (for example:)

vendor has similar units for sale, all the time.

Solid state relays that convert the pulse to power on off control, are also common:

So you want to be sure you really need non pulsing power before you go from these cheap kind of parts to something several times as expensive just to get rid of a little power ripple.

Ok, so to recap, we have a few different ways of going about this.

I break them down into 2 catagories, continuous and interval,

where the continuous method keeps the lamp on the entire time and just dims

it, and the interval method turns it on and off to increase or

reduce average heat.

(thanks John!)

1. Variable Voltage

thermocouple -> Basic Stamp -> PWM (built into stamp) -> MOSFET ->

lamp (thanks Tim!)

(need a triac somewhere if it is AC...) (Thanks Tim and John and Jerry!)

How does this look? Do i have the right idea?? What do you think of the different options?

Will this PID work?

Good summary. You can get AC output solid-state relays (probably with internal triacs).

I believe that controller would work -- in fact, if you can live with a long slow time, you could use the relay output and dispense with the SSR output. Of course, if you want to go faster (and follow their advice) you would use an external SSR. Overshoot may be an issue, but with careful juggling of the differential term and the integrator limit you should be able to reduce it. If overshoot is absolutely critical this controller may not work, but you won't know until you try.

SSR is a triac with an isolated DC interface that lets a logic type signal turn the triac on and off. Or you can roll your own, but SSRs are cheap and give you guaranteed isolation between the control smarts and the power line.

As I said before, the slow pulse method is standard for many heating applications, because many systems have quite a bit of thermal mass to average the pulses. The continuous control (phase control or actual variable voltage) are more general and often used to control the position of valves or other mechanisms, besides operating heaters. For this reason, PID controllers with continuously variable analog outputs (0-10 volts or 4-20 mA) are often called process controllers (for their versatility), instead of temperature controllers, even if they have an input option for thermocouples.

It will perform the slow pulse output through either the built in relay contact output (to control low wattage heaters, directly or drive larger electromechanical relays) or through the DC SSR driver output (that can control heater power limited only by the voltage and current rating of the SSR). It will not produce an analog output for either the phase control or any other continuously variable voltage control scheme.

Its instructions must be translated from a foreign language, because they translated fuzzy logic as fussy logic. :-)

Page 5 says that the pulse period (control period) can be set for any integral number of seconds between 2 and 199. So your worst case thermal ripple your system would have to put up with would occur with a 50% output and a 2 second cycle (1 second on and 1 second off). The filament in the lamp will soak up a lot of that pulse.

snipped-for-privacy@gmail.com wrote: (snip)

By the way, I have a controller very similar to this operating on my kitchen table with a solid state relay driving a nichrome heating element. I have been holding some ferrite at 500C for the last week.

-- snip --

-- snip --

Maybe not -- I've designed some systems with pretty darn fussy logic.

Yes, they are either very dim fully on

That's fine then.

Jerry

...

An excellent and simple way to control overshoot with an arrangement of this type is locating the sensor where it receives the heat earliest, In this case, attached to the work on the heater side. (In an electric furnace heated from the outside, I usually retract the thermocouple into the refractory lining a distance determined by experiment.) This has the same effect an an anticipator in a thermostat, except that it is linear. Although you still want the controller to have a derivative term for dealing with load disturbances, exposing the sensor to part of the heater can supply all of the derivative needed to stabilize the basic loop and avoid overshoot altogether when integrator windup is prevented.

Jerry

By the way, I have a controller very similar to this operating on my kitchen table with a solid state relay driving a nichrome heating element. I have been holding some ferrite at 500C for the last week.

John, you must have one hell of a kitchen. I have to thank all of you guys for your support here. I am a psychology major at a B grade university. So this is nuts trying to figure out something so complex and specialized. I am starting to get this, but I need to re-learn power oscillations and how sine waves affect power output.

I was told that power is varied to the load by changing the percentage of the sign wave that is applied. A controller with a proportioning DC output which drives a Phase Fired Power controller, varies the light / power from about 2% to about 98% in a reasonably linear manner. No flashing.

I am still strugiling to understand this. But once I do, I think everything will fall in place. There are a few process controllers (thanks Tim) that I am considering on eBay.

That is certainly one way to do it, and is one of the ways that John touched on. It would indeed have no flashing, but I question the 'reasonability' of the linearity. If you were in psychiatric instead of psychology you could modify that with drugs, of course (not to change the physics, but to change the threshold of what's 'reasonable').

(snip)

Maybe so, but the oven is a test tube wrapped with a fiber glass tape with nichrome ribbon would through it. That it is wrapped in a blanket of fiberfrax ceramic wool, which is a very good insulator. The whole oven is about the size if the toaster.

Functionally, they all seem reasonable. Don't bid more than $50.

Jerry

That is certainly one way to do it, and is one of the ways that John touched on. It would indeed have no flashing, but I question the 'reasonability' of the linearity. If you were in psychiatric instead of psychology you could modify that with drugs, of course (not to change the physics, but to change the threshold of what's 'reasonable').

...snip :)

Tom, In terms of psychology, the flashing lights could trigger an epileptic seizure and may irritate your optic nerve causing temporary disorientation. But that is pretty extreme. As for drugs, that's the only thing that allows a psych major to expand himself to the realm of applied electrical engineering. haha. just kidding. but seriously, flashing lights get way too annoying when your trying to focus on an experiment. you could say i have a compulsive obsession with continuity, or maybe im just a stubborn teen.

Jerry,

$50 is basically all i have, so that works out.

(...snip) Maybe so, but the oven is a test tube wrapped with a fiber glass tape with nichrome ribbon would through it. That it is wrapped in a blanket of fiberfrax ceramic wool, which is a very good insulator. The whole oven is about the size if the toaster. (...snip)

John,

Can you cook pizza bagels? :)

(snip) I saw an omega controller for $195 that has the thermocouple input and analog output, but I am fairly confident you might find one made in Japan or China that would cost about half that. The phase control relays cost most of $100, but if you want to play around with an old lamp dimmer and a photo cell (to act as the variable resistor normally connected to the knob) driven by an LED from the analog output, you might make one that works well enough for 10 or 20 dollars. (snip)

How can I find a cheap controller from Japan or China??? Does anybody know? What is/are the various technical names for a phase control relay and where can i find a cheap version? I am a little hesitant to use a photocell and a dimmer because I am an accuracy freak. But it was a neat idea and ill probably end up playing with it.

I am considering building my own PID from scratch. I bought an 2x16 lcd today with the standard hitachi interface. What type of mcu should i use? the basic stamp is a bit pricey, and seems more hobby oriented than industrial. I need an I/O for my thermocouple, 3 push buttons, and the phase out. I think thats 8 (since theres 2 wires each), but im not sure. Anyway what do you suggest? and what language do they use? is it called asm? re are there different languages, like asm, C, and basic, for each mcu? I know theres a bunch out there like eprom and flash, but which has a good enough speed to sample over 10hz?