Sciencemadness Discussion Board

How to control the temperature of an oil bath & PID controller equipment basics

pen1234567 - 27-5-2011 at 19:35

To sum up my question: Using air-free techniques (a vacuum-nitrogen gas manifold), I am solvent stripping a solution of a metalorganic (oxygen-sensitive) product dissolved in toluene+hexane contained in a 5L stainless steel vessel under vacuum in order to remove the solvents yielding a dry solid (the metalorganic product) in the 5L vessel. Currently, I am manually adjusting an analog hot plate under the oil bath to maintain a temperature in the oil bath of 37C +/- 2C. Because of the air-sensitive nature of the product, I must externally heat the 5L vessel; I can not expose the product+hexane+toluene solution to oxygen. I would like to integrate an electronic controller (I'm thinking a PID controller would work) to automatically maintain the oil bath temperature at 37C. What equipment/heating element could I use with a PID controller to maintain the temperature of an oil bath at 37C +/- 2C. How does the controller interface with the heating element? Could anyone give me specific examples (brand and model) of a controller and an oil heating element that would be useful in this situation? I understand there are alternatives to heating the 5L vessel, for example using a heating jacket rather than a hot oil bath, but I would like to avoid spending too much money on a heating jacket if I could effectively control the hot oil bath already being used. I am willing to spend the money on a PID controller/oil heating element.

More details about the process and previous attempts at control below...

The current solvent stripping process & equipment:
I am running a metalorganic air-free chemical process that involves solvent stripping in a 5L stainless steel vessel. The 5L vessel contains a solid metalorganic product dissolved in hexane and toluene. The 5L vessel is half-submerged in oil in a stainless steel bath on top of an analog stirrer hot plate (Corning 10inch x 10inch stirrer hot plate model PC-620). I use a thermocouple submerged in the oil (attached to a simple thermometer) to measure the temperature of the oil in the bath. The maximum target temperature of the oil bath is 40C (much above 40C will cause product decomposition). The 5L is under vacuum (using a vacuum-gas manifold). The 5L vessel is attached to a collection Schlenk cooled with liquid nitrogen, so as the solvent evaporates, it is brought into a collection Schlenk (by the vacuum) cooled with liquid nitrogen, where it condenses and later is disposed of.

Currently, I am manually adjusting the hot plate temperature knob to achieve a 35-39C temperature in the oil bath. This requires constant adjustment: I turn up the temperature knob until the oil bath temp. is ~37C, then lower the temperature knob (the temperature reaches ~39C, then slowly decreases down to ~37C), then I turn up the temperature knob again (the temperature decreases to ~35C, then it slowly rises to ~37C), and repeat. It is not possible to maintain a steady, constant temperature of the oil bath using the analog hot plate at a constant setting.


I would like to use temperature controller equipment to automate the process of heating the oil bath and maintaining the temperature around 37C (+/- 2C). One option is to buy a Corning accessory for the PC-620 hot plate: a simple (not tunable) temperature controller with thermocouple. The temperature controller plugs in to the back of the hot plate and the thermocouple is placed in the oil bath. You would set the desired temperature of the oil on the controller and the thermocouple/controller accessory will automatically adjust the hot plate heat to maintain the oil bath temperature. This is simple and exactly what I need, however, a very experienced/knowledgeable colleague of mine has used this temperature controller accessory and hot plate combo and says "it is junk" because it needs to be re-calibrated consistently and in general is not a good controller (my colleague said that someone using this controller/hot plate combo consistently over-heated his solution, resulting in product decomposition). In other words, this particular hot plate controller can not control the oil bath temperature effectively within a +/- 2C range.

My understanding is that a better option is to use a controller, such as a PID controller, that will allow me to tune P, I, and D gains and such for a much more effective control. Because the hot plate I am using has a custom input port, it is only compatible with the specific Corning controller accessory I mentioned above which I do not want to use.

How could I use a PID controller to maintain/control the temperature of the oil in the bath? I have fairly extensive knowledge of PID control theory (e.g. transfer functions, control block diagrams, tuning rules with Cohen-Coon method and others, etc.), but I have no experience with integrating controller equipment in a real-world laboratory setup. What equipment would I need to control the oil bath, and what method could I use to heat the oil bath? I am assuming I will not be able to use the analog hot plate with an advanced temperature controller, so I will probably have to do away with the hot plate. Could I use a temperature controller with a submersible heater in the oil bath? Is this an effective method? Can I directly interface the PID controller with the submersible heater and a thermocouple? Is there another option other than using a submersible heater in an oil bath? Could someone give me a specific example of a PID controller and heater equipment (e.g. brand and model) that would be useful for maintaining the temperature of the oil bath?

I appreciate any responses helping me understand appropriate PID controller equipment and the integration of the controller in the laboratory setup.

Thank you.

Magpie - 27-5-2011 at 19:43

This has been thoroughly described and discussed several times already. Please use the search engine first.

BromicAcid - 27-5-2011 at 22:29

I was always a fan of the J-Kem

http://www.jkem.com/temperature-controllers/oil-bath-control...

They learn as they go so if you have to hold a temperature for a long time they will zone in on exactly what they need to do and as the conditions change they will change right along with them. In the case of a water bath this really is to their fault because once you top off the water in the bath everything changes all at once and they cause some major fluctuation trying to get their footing again. Sometimes you will find a J-Kem cheap on eBay.

peach - 27-5-2011 at 23:19

You can buy PID controllers from China on the UK eBay for about £18



On the side you can see the layout for the screw terminals. 1 and 2 are the mains, 5 and 7 connect to an alarm if you want to add one. 8,9 and 10 are for connecting the feedback. Bridge 8 or 9 to 10 with an RTD (Resistance Temperature Device, usually a platinum wire) or 9 and 10 with a thermocouple.

4 and 5 are the output terminals. I think this one has a mechanical relay built in, but precisely what it's rating is I'm not sure from the details. Either way, it can be scaled up by having it switch a second relay. Scans twice per second.

PV stands for Process Voltage (the thermometers reading) and SV is Set Voltage (your target).



There are tons of these slower scan rate versions on there, here's a more compact one for £20.99. Some of them include the thermocouple and I think I've seen one with a universal mains input as well. Use as is or put it in a fancy box.





[Edited on 28-5-2011 by peach]

Magpie - 28-5-2011 at 06:43

Quote: Originally posted by peach  

PV stands for Process Voltage (the thermometers reading) and SV is Set Voltage (your target).


Don't you mean process value and set value?

piracetam - 28-5-2011 at 09:07

process variable/set variable
and the I/O measurement is usually ..?
voltage.

I use a multifunction DAQ, interfaced to LabView 2011 for measurement and control
but Arduino PIDs are popular and considerably cheaper. the software is free/open source



[Edited on 28-5-2011 by piracetam]

peach - 28-5-2011 at 10:55

Quote: Originally posted by Magpie  
Quote: Originally posted by peach  

PV stands for Process Voltage (the thermometers reading) and SV is Set Voltage (your target).


Don't you mean process value and set value?


Something like that :D;)
Quote: Originally posted by piracetam  
process variable/set variable
and the I/O measurement is usually ..?
voltage.

I use a multifunction DAQ, interfaced to LabView 2011 for measurement and control
but Arduino PIDs are popular and considerably cheaper. the software is free/open source


Have you posted about this in more detail on here and if not, could you?

It's something I was wondering about but I can't program much beyond some simple HTML and I can't code microprocessors, so I was put off by it possibly being too much for me to get functioning. I have an MSP430 board but should have started with something more widespread I expect, like the Arduino.

piracetam - 28-5-2011 at 11:24

http://www.arduino.cc/playground/Code/PIDLibrary
http://arduino.cc/forum/
Arduino has a lot of support, the coding is done in C++, I believe
but support for these boards is abundant on their forum, and people who are code-saavy
always post control drivers.

it's an incredible project, no doubt

Beagleboard is another decent opensource project, dealing with ARM-based processors.

[Edited on 28-5-2011 by piracetam]

Magpie - 28-5-2011 at 13:00

Quote: Originally posted by piracetam  
process variable/set variable
and the I/O measurement is usually ..?
voltage.


Voltage? Well, certainly, that's what the controller sees. But the operator reads and sets in process values like degrees C, and that is what is displayed.

dann2 - 3-4-2012 at 19:21

Hello,
Went to look into buying a PID controller but I have a problem (as I see it!)

The outputs on all the cheap (and not so cheap) controllers cannot drive a large load (say one or two kWatts heater) unless my simply turning on and off the output relay. ie. the heating device is either fully on or fully off. Its a bit like going back to 'bang bang' control (as it is called). I do accept that since the controller is a PID then the control will be much better than a complete 'bang bang' controller.

Some of the PID controllers have a 4 to 20 mA output which could be used to control some sort of (fast, like a power controller) switching device to control the large heater.
Am I missing something here?
Does the PID controllers with a solid state relay (SSR) actually switch on and off (like a triac) many times a second so that a large (say 2kW) heater will:
run at 500w if the temperature is just below the set point
run at 1000W if the temperature is alot below the set point
run at full blast if the temperature is way below the set point
etc etc

I would imagine the 'ordinary' relay output will not switch on and off rapidly (like a power controlling switching circuit) like I am suggesting the SSR does.

I cannot seem to figure this out by googling. No circuit turns up that shows me how to convert the 4 to 20mA output to a (say) 0 to 2 KW (using a , say, 230V mains powered heater) power controlling circuit ('switching a triac' type circuit)
I am of course presuming that the 4 to 20mA output gives an output that goes something like:
Low mA's out-----> temperature way below set point -----> heater on full blast
More mA's out ----> temperature somewhat below set pont -----> heater on quite a bit
midrange mA's out ---> temperature somewhere around set point ----> damm all power to heater
high mA's out ----> temperature way above setpoint ------> turn on cooler (if you have one!)

This is turning into one very long winded post.

I cannot believe that having gone to the trouble of implementing a PID solution (from inputs measured) that the output is just a crude turn-on-heater-full or shut-off-heater with a period of a few seconds(using either a SSR or an ordinary relay)?
The controller should give varying power to the heater, 10% on , 20% on 100% on etc

Something like this;
http://www.omega.com/ppt/pptsc_lg.asp?ref=PCM3&Nav=
I would like to see in the output of a PID



Dann2

[Edited on 4-4-2012 by dann2]

[Edited on 4-4-2012 by dann2]

chemrox - 3-4-2012 at 21:00

I love equipment but I also love cheap. Is there anyway you can size down to 3L? If you can, you can get a deep fat fryer with 1°C tolerance. Vacuum pump oil works well for most purposes. I'm copying notes on all the good suggestions you got here too. As I said, I love equipment. One more idea; I haven't tried this, but one of the big bird oil immersion roasters might be as sensitive and stable as the fryer. It might handle the 5L RB.

Magpie - 3-4-2012 at 21:15

Quote: Originally posted by dann2  

Something like this;
http://www.omega.com/ppt/pptsc_lg.asp?ref=PCM3&Nav=
I would like to see in the output of a PID


Unless I am reading the Omega description incorrectly that unit still just gives a time proportional full blast or full off. This would not be like the output of a triac, which chops every sine wave of the AC output.

I have built a triac (actually a quadrac) power supply for my tube furnace (described elsewhere on this forum), as did garage chemist. But it is only manually set, not PID controlled.

watson.fawkes - 3-4-2012 at 21:29

Quote: Originally posted by dann2  
Does the PID controllers with a solid state relay (SSR) actually switch on and off (like a triac) many times a second so that a large (say 2kW) heater will:
run at 500w if the temperature is just below the set point
run at 1000W if the temperature is alot below the set point
run at full blast if the temperature is way below the set point
[...]
The controller should give varying power to the heater, 10% on , 20% on 100% on etc
The typical use of a standard 1/16 DIN form PID controller is indeed to run the control output into an SSR, as you've surmised.

The output of the PID is a pulse-width modulated (PWM) waveform, but the typical period of the PWM is in second, not milliseconds. The reason for this is that almost always the heaters these control (1) have large thermal mass and (2) run from 60 Hz AC power.

The effect of (1) is that there's no practical reason to run the PWM faster, that is, unless you need temperature stability less that 0.1°, and to do that you generally need a fancier thermometer, more care with signal conditioning, etc. Even 2 L of water counts as a "large" thermal mass for a 500 W heater. Furthermore, the meaning of "large" is really "large relative to the heating element", which means you can use such long PWM periods as long as don't horribly overrate the heater.

The effect of (2) is that you don't want to use a fast PWM if what you're doing is just switching an AC line. You'd have large electromagnetic interference (EMI) problems to deal with them, which you have to deal with or else they interfere with the PID controller itself. Expect to spend lots of money on chokes and ferrites cleaning up the EMI mess. You could avoid EMI by switching only at zero-crossings, so if there's a way to program your controller to use a PWM period at 120 Hz (the frequency of zero-crossings) and use an SSR that switches only at zero-crossings, you'll have a system that works. Even then you'll have some EMI to deal with, just not nearly as much. The third and fifth harmonics are smack in the middle of the human audio range, so if your heating elements have any vibration ability, you'll hear it.

Alternately (and cheaper), it would be possible to use a fast PWM signal to modulate a switching power supply (it would replace the error amplifier part of the circuitry). In that case, your output could indeed be properly proportional. It would be DC and not AC, which matters little in most heating applications. On the other hand, I've never seen such gear in ordinary commercial trade, though I'd guess it exists for specialized purposes.

The short version is that you can just use the SSR and you'll be fine.

peach - 4-4-2012 at 02:27

Quote:
I cannot believe that having gone to the trouble of implementing a PID solution (from inputs measured) that the output is just a crude turn-on-heater-full or shut-off-heater with a period of a few seconds(using either a SSR or an ordinary relay)?


That is pulse width modulation (digital) for you, and it's everywhere.

My IKA controller does precisely the same thing. It just switches the element on and off for varying bursts to modulate the temperature. It becomes 'analogue' because the thermal mass does not respond directly to pulses (even when they're occuring on a second long basis), it accumulates them (or the lack of) and averages them over time. E.g. the pulses might last 1 to 10 seconds, but it takes about 30 minutes to an hour to warm something up; so each pulse accounts for... 1,000th to 4,000ths of the total energy needed to produce that temperature. I know the pulses are lasting that long with the IKA because it says on / off on the display, so I can watch it flicking the element without having to take anything apart.

Got to remember that it takes quite a long time just for heat to move from the heating element to the sensor in the thing being heated, seconds to minutes for a lot of regular sized things. The PID (disregarding it's switching frequency) has to average for a similar length of time for it to pick up what's changed since it last applied a pulse.

A huge amount of electronics (nearly everything in newer electronics) uses the same concept. If an analogue style input or output is required, capacitors and / or inductors are added to do the same thing as the thermal mass, but electronically. In switch mode power supplies, the output is a series of high frequency pulses. There are capacitors and inductors on the output, with their values tuned to the frequency of the pulses, such that the voltage / current after them looks more like smooth DC. Many of the high end laboratory power supplies use the idea - so do computer PSUs.

It's also used in audio - PWM encoding. And normal power supplies, where the pulses from the rectifier are smoothed by capacitors.

A cinema is a perfect analogy. The picture is not continuously changing, it's strobing. It's just happening too quickly for your eyes to see it.

The PID algorithm isn't just switching the element on and off whenever it's at or off the set temperature (like a normal thermostat), it's using the data from the feedback to calculate how to long to apply the next pulse for.

The seemingly long duration of the pulses is because the thermal mass is so large it wouldn't really benefit from going any faster. Going faster means worrying about radio frequency noise, which is something that electronics is officially (legally) regulated by. High frequency switching devices need designing and testing to pass as compliant, or they'll spew side band radio noise out that disrupts other devices that need a specific radio frequency clear (e.g. wireless adaptors for computers). That's why a lot of power supplies are now housed inside their own little cage. You'll see the same same cages in radio gear and mobile phones for similar reasons.

You could smooth out the pulses to a heating element by sticking a capacitor across it. When the relay flicks on, it'll charge. Off, it'll discharge. The capacitor will smother the pulsing. Although, there's not really much need with a lot of normal sized gear. It'd be more important for something like a melting point apparatus, where the sample is tiny and the measurements need to be very precise.

Technically, you are correct. That by pulsing from on to off and nothing in between, the element, in close contact with the glass, could raise the temperature of the glass quite a bit higher than it needs to be, scorching things in contact with the surface. But it's a case of practicality versus perfection.

[Edited on 4-4-2012 by peach]

Twospoons - 4-4-2012 at 15:55

The Omron controllers I use at work have an adjustable cycle time - down to 0.5 seconds. Unless you are putting kW into something tiny this is plenty fast enough. Example: 1000ml of water with 1kW of heating will rise at just 0.24 degrees per second.

The controllers do not have high power outputs because it is uneconomic to make different controllers for all the possible loads people might want to control. Hence the usual method is to pick a controller and add a suitable power switching device. So one controller model can be used for 1W loads or 1MW loads.

A big advantage of using an SSR on the output is lifetime - relays with big loads suffer contact burning if not properly sized and snubbed, and the contacts may eventually weld shut.

dann2 - 4-4-2012 at 16:15


Thanks for replys.
I agree with the fact that the inertia of the (heating) system will smooth out things greatly.

"The PID algorithm isn't just switching the element on and off whenever it's at or off the set temperature (like a normal thermostat), it's using the data from the feedback to calculate how to long to apply the next pulse for."

If the PID controller is actually doing that then it is (almost) doing what a 'real' proportional output would be doing.

The SSR on these cheap (and not so cheap) PID seem to be simple on/off relays. Reading up on SSR's there are different types (see attachement). The Proportional Control Solid-State Relays are the type I would like to see on my PID (if I may!). You simply feed in an analogue signal and you get out GENUINE analogue power control, not just switch on/off.
But as pointed out, in the most cases there is not need or point for this added cost.
I managed to find just one of these types of SSR here
Looks like they are rather an unusual item though the attached file says they are common enough.

Does the 4 to 20mA output of a PID (when it's present) actually give out a signal that is proportional to the calculated error. The error being calculated from the 'P', the 'I' and the 'D'.
Or is this 4 to 20mA output just another switcher, ie. 5mA out if error is minus and 19mA if error is positive?

Example of cheap PID with 4 to 20mA out (an option anyways) here

High end PID's have just the switching outputs (ordinary relays and SSR) and sometimes the current output. They have an analogue output which is a function of the input measured parameter. They do not state if this output is a PID function of the measured parameter but I suppose it is?

I was thinking about buying a PID but if I could get one that puts out an analogue signal that is a function of the measured P + I + D as opposed to an on off signal then it would be much more versitile and could be used to do thing other that control systems with large time constants.
Dann2

Think perhaps I need to read up on PID stuff.



[Edited on 5-4-2012 by dann2]

Attachment: Solid Statements - SSRs switching types[1].pdf (152kB)
This file has been downloaded 718 times

watson.fawkes - 4-4-2012 at 18:13

Quote: Originally posted by dann2  
Does the 4 to 20mA output of a PID (when it's present) actually give out a signal that is proportional to the calculated error. The error being calculated from the 'P', the 'I' and the 'D'.
Or is this 4 to 20mA output just another switcher, ie. 5mA out if error is minus and 19mA if error is positive?
Please verify this, but my recollection (a bit hazy) is that 4 mA means OFF and 20 mA means ON. In other words, it's just another way to deliver the PWM signal.

If you want a controller that works exactly like you want it to, look at osPID. It's an open source project that's designed to be hacked.

Twospoons - 4-4-2012 at 18:48

Dann2, you are over thinking it. The PID algorithm does control the on:off timing ratio of the output. The 1/2 second or greater period is plenty fast enough for most heating applications. The complexity and expense of a true analogue power control is simply not justified.

Watson.fawkes is correct about 4-20mA: its a digital comms standard, not an analogue one.
Some of the controllers can output an analogue control signal and /or an analogue temperature signal. But they cost more.
Get an ordinary PID controller, attach an ordinary SSR, put the thing through its auto-tune cycle, and all will be well.

[Edited on 5-4-2012 by Twospoons]

arsphenamine - 4-4-2012 at 19:02

If you want to burn up a lot of time instead of fractionating your solvents,
feel free to hand-cruft something from open source.

Speaking only for myself, if the total solution was $50
and needed only a screwdriver for setup, I'd grab it.

This is a 1st Law Thermo corollary I call
the "Principle of Maximum Laziness."

Amazon, eBay, and LightObject will be happy to sell you
an inexpensive controller+temp sensor+SSR for under $50.

Worst case accuracy for the JLD612 set point and temperature measurement is 0.2%,
but it's nominally much better if you go for a Class B (cheep) Pt100 RTD immersion sensor.

...and it learns. ...and you can tweak the PID constants.

peach - 5-4-2012 at 01:00

There are many different ways to skin a cat, and control something electronically.

Variac

A variac is by far the simplest, it gives a constant, variable output. There is also no feedback (unless you build some epic mechanical device to turn the dial).

Phase fired control

You can achieve something similar, direct from the mains (which is rather dangerous if not correctly insulated), with phase fired control. A solid state switch is placed on the mains, but it will only switch on at a specific point in each AC cycle, then back off again when the wave passes through zero. By varying the position at which it flicks on, a varying amount of each wave is delivered, and so, a varying amount of energy per cycle. That is one of the simplest variable AC controls to implement (if feedback is not required) because it only requires a couple of capacitors, resistors and the switch. It is common in furnaces. The thing that controls the firing position is primarily a variable resistor (which can be replaced with digital variable resistors to more easily implement digital control). It does, however, mean working with the unisolated mains and AC. Unisolated meaning that the live is still referenced to earth and neutral, and poses a severe shock risk if the live ever manages to conduct through anything bar the element (e.g. the casing).

Linear regulator

Linear regulators are one of the simplest integrated circuits. They have three pins. In / Out / Adjust. Their output can be ultra smooth / almost noise free DC. They're controlled by a variable resistor on the adjust pin (which can be swapped for some other variable resistance device for feedback control).

They are also wasteful of power. If you want 16V out and the supply is giving 24V to the regulator, it will output the 16 by dumping the other 8 across it's self. Same for current. Result, the regulator 'burns off' the unwanted power as heat. The chip gets HOT and usually needs heat sinking. I have two industrial linear supplies. They weigh about 30kg each versus about 5kg for a similarly rated switching supply.

Computer power supplies

They'd be good for supplying a linear regulator, but something more interesting is that the output voltage / current can actually be varied by inserting a variable resistor across two points in their own feedback circuit. They are high frequency supplies that output rather smooth DC (thanks to the filtering on their output). They also happen to be rated at similar power figures to hotplates and mantles - e.g. few hundred watts to a kilowatt. If you buy / build a PID that outputs a proportional error signal, you'd only need couple the error signal to a solid state resistance and connect that across the computer PSU's own feedback circuit (or, feed the error signal directly into the computer PSU's own feedback pin if they use the same logic levels) to have the PID varying a compact, cheap, powerful DC power supply. You may even be able to pull the error signal out of the PID by finding it on the board inside, rather than using the contacts on the outside; it is in there, somewhere (hopefully present on an actual track or pin and not incorporated inside the IC's themselves).

Conclusion

You gets what you pays for.

If you want really precise, compact lightweight control, you either pay more for it or have to be able to build it yourself.

The easiest method, by a long way, to get rid of the pulsing output on a 1-10 second pulse from a PID would be to dump a lot of capacitance or inductance across it. The capacitor / inductor will average the pulses towards one value, depending on their duration and it's own size. This can also add a lot of lag to the feedback, and some devices do not like switching large capacitances or inductances on and off. You'd also need to think about surge control as, when it applies the full power to the output, a capacitor will initially look like a near perfect short circuit, and an inductor a huge resistance.

If you are heating a decent sized oil bath, that becomes your capacitor / inductor to an extent. The base of the bath might go through pulses of being too hot and too cold, but the oil it's self won't allow vast temperature differentials to exist within it's self.

[Edited on 5-4-2012 by peach]