PID Controller-Build or Buy?

Thread Starter

mandarpatil

Joined May 27, 2020
2
I am looking for everyone's opinions on this, and I know its going to vary wildly, but maybe someone can point me in the right direction or offer their advice. I am in the process of building a heat treating kiln for knife making. It will go up to about 2000F and use a K type thermocouple. The controller will be able to control the heating element. So here is where the "build vs buy" comes in.
I know I can get a good PID controller from Automation Direct for $34. Yes the Chinese ones from ebay work, but I wanted something a little nicer and with better documentation. So spending that amount of money is OK for me. My one major want though is to be able to ramp/soak. Im sure I could do this with a regular PID controller and a timer (OR...figure out how to use Modbus with a PIC/Arduino), but that's not very user friendly and would require me to keep going back and forth.
 

tindel

Joined Sep 16, 2012
936
If what you need is only $35 or so - why on earth would you build your own? These sort of things take at least a month+ to develop (depending on how low level you want to go with your 'build'). I've recently spent a month developing an application specific PID controller - and I'm not done yet. I've done this during nearly all of my waking hours too.

I strongly recommend buy.
 
I wouldn't use a K type there. Your too close to the limit of 2300F. https://www.thermometricscorp.com/thertypk.html C is probably better. We used K whenever we could. We used C when melting copper. T for cryogenic and room temperature.
R or S for contamination issues (semiconductor). C is good to 4200F.

C is VERY rare.

You might consider a second monitoring thermocouple with a cut-out too.

When I was first employed everything was messy. We custom wound tantalum heaters and used Eurotherm controllers with "direct drive" dual SCR outputs running into a variac and separate thermometers. The controllers didn't have a display, just a deviation analog meter. I made sure that the SCR's were 25A units and had working current limit. I added a separate 3AG fuse because blowing a $25.00 USD 25A semiconductor fuse was expensive. We were operating in a vacuum, so heater shorts were a normal occurance.

Me and a co-worker wrote a 7 loop PID system with recipies, but no ramping, this is 1980's, before the PC existed. We had logging, but before spreadsheets were common. We used DC power supplies and added an "unusual feature" of "Heat-up energy limit (in Watt-s). This effectively could protect against a shorted thermocouple on heat-up. It also verified placement.

We effectively could control P, I, V and T. Really one was controlled and the rest were limits.
dp/dt was added. p was the output voltage. Power supplies were about 300W. Most were like 40V, 35A or so.
I thought it was "way cool". It used a mini-computer, a DEC PDP 11/23+. Loop time was high, a few seconds.
It was written in FORTRAN with a few assembly language instructions and took 2 of us 9 months to write. Hardware budget was like $25,000 USD.

I added status and stability criteria like heater open/shorted, stable and conditionally stable, what limit the system was in.
The terminals was a standard Vt100 terminal.

We were pretty well maxxed out on features. One problem was you could only use R or S or J or K in group of 4. No room to add software tables. No money for hardware conditioners. Comma separated spreadsheet data could have been added if there was room. It had "device drivers" basically. We did not implement auto-tuning.

The DAQ was a 9600 baud serial link and the computer was located rooms away with 8 high and 8 low level isolated analog

inputs and 8 thermocouple inputs and some digital I/O. The "extra" 2 analog in's was for a vacuum gauge and a thickness monitor. The digital I/O was for shutter OPEN/CLOSE and state of OPEN/CLOSED/unknown and to reset the thickness monitor.

When that technology died, we went to Eurotherm ramp/soak controllers with an isolated analog output controlling a power supply. Temperature time logging was done using RS485. Management would not allow any other output like isolated 0-5 out for V and I (Added about $500). We had V and I limits and could read V, I and compute P manually. The power supplies were like $1500.00 each and we needed 7.

The later Eurotherm SCR units took 4-20 mA in and controlled power with current limiting and phase angle firing. That was our new standard for other stuff. The 0-100% control signal should either control power or be proportional to V squared.

Phase angle firing gives you the finest resolution. Another method is pulses of 1/2 cycles, time on time off and bang/bang.

That case had the ability to:
1. Be wired for 120 or 240
2. It contained a relay to disconnect the load because of operating in a vacuum system with exposed wires,
3. May contain an over temperature limit This was actually mounted on the rear of the case.
4. Contained internal/external setpoint and On/Zero switches.
5. Contained a LED if the contactor was engaged.
6. Front panel Fuse(s).
7. Power fail dropout or not was front panel enabled.

The controllers used for this and the DC power supply gig was the Eurotherm 818P. The setup manual is 49 pages long. buphy.bu.edu/~stein/etherm/818eng.pdf The operating manual is separate.
 

Thread Starter

mandarpatil

Joined May 27, 2020
2
I am looking for everyone's opinions on this, and I know its going to vary wildly, but maybe someone can point me in the right direction or offer their advice. I am in the process of building a heat treating kiln for knife making. It will go up to about 2000F and use a K type thermocouple. The controller will be able to control the heating element. So here is where the "build vs buy" comes in.
I know I can get a good PID controller from Automation Direct for $34. Yes the Chinese ones from ebay work, but I wanted something a little nicer and with better documentation. So spending that amount of money is OK for me. My one major want though is to be able to ramp/soak. Im sure I could do this with a regular PID controller and a timer (OR...figure out how to use Modbus with a PIC/Arduino), but that's not very user friendly and would require me to keep going back and forth.
Any Updates??
_______________________________________________________________________________________________________________________
real estate digital marketing company
 

Reloadron

Joined Jan 15, 2015
7,501
I got nothing to add. We made our own K thermocouples used for furnace surveys. Using AWG 10 gauge wire they did fine at 2,000 F but as KISS suggested you can consider another type. For our needs the fabricated K types worked fine.

Things went a little like this:

Bare Wire
Image12.png

Any TC supplier will have the ceramic insulators.
Image13.png

Using Oxy Acetylene we would weld the tips.The wire used was also certified and calibrated checking start mid point and end of roll.
Image14.png

Image15.png

Anyway, I would buy verse build and if ramp soak is what you want just make sure it does it. Thermocouple? Really per KISS whatever works. :)

Ron
 
They would not have worked in our case though. We typically had spring loaded thermocouple probe in tantalum wells in a vacuum. They were controlling Boron Nitride evaporation sources with a bottle and a nozzle heating element. Big difference.
Other times we were doing surface measurements of thin films, cryogenic (vacuum) or ambient, so one junction.

In general, I would use a grounded pre-made probe if possible.
 

svkadam

Joined Oct 3, 2009
1
If you need very high accuracy of control then build it. if an error of +-1% in control of temperature then use available controllers.
PID algorithm is simple,but to develop tuning algorithm for high accuracy is difficult.
 
I did I wierd varient once. Had an environmental chanber with it's own heat/cool controller, BUT wanted to control surface temperature. So, I use the proportional band of the chamber controller as -100 to 100%. That was P-out. The chamber was <10 degress off. Range was -90 C to 200 C. So, if the setpoint was 100, the PB was -90 to 120. It also turned out the PB was 10 degrees. The surface temp thermometer was an RTD and read separately. The chamber and thermometer were IEEE-488,
 
Top