What will furnace pid values actually look like?

[coppercone2]

So i have a decent mechanical and thermal design of a furnace ready along with a high power supply that accepts simple analog control.

The idea is to get to like the limits of kanthal a1 temperature in a box thats like 8x8 inches and possibly poorly isolated by design, with optional removable insulation.

The amount of power and heating element is pretty heavy overkill for tge size of the furnace.

Practically do you think i can get away with analog pid control? Will i get ridiculous values for a d term or something. The furnace is like 4kw where a typical one would be 1.5 for performing with poor insulation so it can do cycling. Its ramp would obviously be limited so it does not self destruct.

[ajb]

Depends on what your time constants look like.  How much thermal mass vs how much thermal power, rate of heat loss, how much input and sensor lag you have, etc.  Small systems with reasonable power to mass ratios and minimal lag between control input and feedback should be pretty simple.  If you don't need really tight control you don't even need PID.  A simple on/off control could do the job, either hysteretic or constant on-time once the system is characterized.

[coppercone2]

so not enough information right now to ball park it with 8x8x8 inch chamber, up to 4000W of heating from coils evenly distributed on the walls and firebrick?

Reason I am asking is because I don't have the capability to finish that job at the moment, mainly due to vacuum system design problems, but I can play around with a PID on a bread board or even make a PCB, but I would like to know if the control problem is unrealistic for analog components.

I don't need numbers, just a sanity test. If it turns out I need some garbage can sized capacitor or w/e for small signals, then I could investigate digital control or using a off the shelf PID (I would like to make everrything myself though).

Say a big load would be 2kg of copper for a vacuum brazing heat exchanger/heat sink job or a couple of hundred grams of ferrite for things like sintering inductors or w/e. I figure copper is the worst case.

Theoretically since the chamber is 8x8x8 or so, you could fit like 18kg of copper in there (5x5x5 inch cube).

I thought since those controllers are all over the place someone might have a feel for what will happen.

[floobydust]

How is the thermocouple sampling load temperature verses chamber temperature? It's in a vacuum too?

Do you have proportional control for the 4kW heating element, like an SSR that can do x# powerline cycles?

I would consider a REX-C100 PID controller for under $10 (chinese clone of RCK Instruments Japan).
It may be less time and hassle.

[coppercone2]

i have a high power DC supply that will be used. Maybe two so the heaters are not all in series to make it more uniform to maintain a more even temperature. It will segway into building multizone tube furnaces eventually. dont want to deal with weird ass ac shit like triac, ssr

not sure on thermocouple placement thats a good point

wanna do it myself, if analog, if digital then I have some kinda mid-range shit I can mess with but I don't want to use someone else's thing for this job.

I would say I am not impressed with the quality of (electronic) stuff from Omega (at ALL, but I have no issues that I know of from their sensors) and I defiantly won't  be using some 10$ Chinese shit. For simple system I would go with a PIC that just has a setpoint based on a 10 turn pot value so you don't need to mess with interface controls etc. And it would probably be more reliable then omega lol. But I would like to avoid any digital code.

For two thermomemters and a PID I thought that maybe one thermometer can dampen the pid response based on gradient but it seems complicated to put a buncha variable resistors or whatever to some how slave it.

Whats the algorithm/relationship between having two therometers in a kiln?

[rstofer]

It would not be unusual to not even have a 'd' term.
There are web sites all over the Internet that deal with tuning PID controls.
Real PID controllers come with "Auto Tune".
Then there is this Arduino code for an auto tune PID:
https://github.com/br3ttb/Arduino-PID-AutoTune-Library

[coppercone2]

I think I tuned one manually before but never at such high temperatures.

again prefer to not code the project loses alot of magic if it has to go there. I kind of want it to work because if you consider putting a bunch of silver, electrical copper and other 'precious' metals by our standards I begin to desire analog so some shitty arduino does not wreck hours and hours of work if I am brazing a bunch of shit together. If I use a good filtered low noise supply, good op-amps, quality soldering and good filtering and ground loop considerations + shielding it should work when configured without code mysteries.

Considering how much labor goes into making those coils, vacuum sealing a big box and cutting the bricks and everything its just not worth some code fuckup to melt things.

[ajb]

Just for a ballpark number, 2kg of copper means about 770J/K, so at 4kW that's a temp rise of 5K/s before accounting for losses and thermal resistance.  The furnace primarily resembles an RC system where R is the thermal resistance between the heating element and the load, C is the thermal capacity of the load, temperature is analogous to voltage, and thermal flux is analogous to current.  There is another RC system between the load and the temperature sensor (generally small C, but possibly substantial R).  Insulation losses represent parasitic resistance terms, primarily between the load and the outside world and the heater and the outside world.  Like in electronics, the complexity of the model is limited only by how much of the parasitic complexity you care to represent.  The C model gets a little trickier near melting/boiling points due to enthalpy of fusion/vaporization, sort of like the miller plateau of a FET.

For something as slow as a furnace, analog PID is only attractive for the learning experience.  Even an 8-bit micro will give you orders of magnitude more sophisticated control than is worth doing in analog, for a lot less headache.  Also, there are a lot of tempcos to think about in a sophisticated analog controller for something that's going to sit near a 4kW furnace 

[coppercone2]

trust me, when it comes to me writing code, for my psychological well being I would rather buy a bunch of foil resistors then open up a IDE. MCU piss me the fuck off and they always gave me some stupid problem.

Ok, I know the basic RC, it works for vacuum, and probably sealed air to a extent but how about when there is gas flow in it (i.e. argon flow that vents)? I honestly don't know much about what happens with convection at those temperatures.

I am telling you, I am gonna miss some bracket, name a variable wrong and the thing is gonna fucking melt on me if I go with a MCU here. I think I have all the equipment in my lab to fully test a analog PID look really nice, even do emissions tests and some basic suceptability tests to RF etc. I think if I do a by-the-book job it will work if I apply all the things I know, even if its near those big power supplies. Grounding and shielding will be critical I think.

Good call on the insulation though. The control board needs a over-ride shut down if it gets over a certain temperature to shut the process down early instead of doing it poorly.

If I do use a MCU, its just gonna be a temperature read-out that hooks into a analog output of the PID. I don't think I even wanna configure it with a MCU. Maybe it will have a set-point over-ride that you can plug something into through a BNC if you want it to ramp etc, or I will leave some room in the chassis for a external controller for later control.

'
Just for a ballpark number, 2kg of copper means about 770J/K, so at 4kW that's a temp rise of 5K/s before accounting for losses and thermal resistance.  '

Yea thats what I needed. I don't know why I did not think of this I did it for fluids before. The convection aspect messed with me too much for some reason. The thermal losses will also be nonlinear with temperature I think? Those bricks are gonna start conducting more at higher temperatures I think.



Q = mc∆T

2*386(Cu)*1000(K)=772000J

T=Q/P

772000/4000=~193 , close to your derivative .. ,my mind went out the window because its glowing hot

[coppercone2]

So if the process time is like 200 seconds, the integral constant should be larger then 200 right? Like 400?

That would mean 40meg resistor and 10uF foil cap.. then the D-constant would be like 5-25% of this?


I used the wrong formula. I thought the values were strange. I think its (1/RC)

Am I making a mistake with this assumption?

[SeanB]

This just begs for an off the shelf fully digital PIC industrial controller, from Omron, Braintech or such, that literally is programmed by selecting the ranges, the inputs and desired mode, then selecting "Auto Tune" which then ramps up the power in a controlled manner and sets all those parameters just by rate of rise initially then predicting the setpoint and adjusting to reach it with minimal overshoot. Simple DIN size panel cutout and you probably want a version with analogue output rather than a SSR or relay output, so you can control your power input using a 0-5V/10V/4-20mA or so signal, and in general they all come with input sensor failure detection, and in many there are auxillary relay connections that can be used for alarm, ready, fast heat or whatever control function based on either temperature or time or fault.

The DIN units are reasonably RF immune as well, and the power needs are well defined.

[coppercone2]

I really wanna try to do this myself.

Any idea what the PID terms would actually look like from something autotuned, realistically? I don't have a good process model, I have a feeling that process time based on the most basic equations is gonna be off significantly. But I am pretty sure I will be able to put values in a circuit that will have the right order of magnitude for tuning, which can then be zoomed in. I will just make a little more space, chassis height and larger pads for the integrator and differentiator op-amps in case I need to change capacitors etc.