Heat Treat Oven Build + Kanthal Wire Calculator

[Brandon Bearden]

I am going to build a heat treat oven. It will be semi based on the design from Jen-Ken. It should be able to handle knifes with a 25" blade and 7" handle.

 

I am going to use 2" thick ceramic fiber board wrapped with another 2" of wool contained with sheet metal. I am going to bond the board with Blakebond Refractory Mortar. The top and bottom will both be 4" thick with 2" recessed into the chamber. This has been purchased already and arrived today, (except the metal). It will 36" tall without the stand. 

 

This will be about a 12,000 watt system. I have been banging my head trying to understand how to size Kanthal wire and think I finally have it figured out. I have a spreadsheet that I built that anyone could use and input their values to get the lengths of wire they need to buy. I wanted to run my calculation sheet by someone else. If anyone finds it useful, feel free to take it and do whatever you want with it. There are only a few fields to enter.

 

I have most of the materials, including a controller, already purchased. I will be using an Omron E5AC-QX4D5M-010 with the software, (CX-THERMO VERSION 4), for heat treatment. I will use another, differently calibrated for lower temps, controller and temp sensor for tempering. (Omron E5CC-RX3A5M-001).

The relay will be Omron as well, G3NA-490B-UTU-2 AC100-240.

 

It will probably take me a couple of months to get this thing built based on life circumstances at the moment. I didn't want to post until I had more to show than a partial drawing and some specs, but I need help ordering the correct wire. If anyone has thoughts on the design too, I am all ears. 

 

The opening inside the baffle is about 9x15"

 

k1 wire calc sheet.xlsx

Edited October 7, 2022 by Brandon Bearden

[Joël Mercier]

12000W means around 55A at 220V. You might want to rethink that...

[Alan Longmire]

True.  Evenheat kilns max out at 20A 250V, something may be off.  

[Brian Dougherty]

I'm sipping whisky in a nice restaurant, so I'm not inclined to get into the calculations at the moment, but look up Tim Gunn's posts.  He covers a lot of it.

 

Buy yeah, 12kW is off the charts.

[Joël Mercier]

8 minutes ago, Brian Dougherty said:

I'm sipping whisky

Which one? 

I'm currently finishing a bottle of Glenrothes Select Reserve. 

 

 

[Brian Dougherty]

Michter's American tonight.  However, I have to admit it's in an old fashioned.  This is somewhat shameful for someone who prefers their whisky neat

 

But damn, Michter's makes such a good old fashioned!

[Brandon Bearden]

I have a 60 amp breaker and 85% load is 51amps and just over 12k on 240.

 

I want it to heat up very quickly.

 

Take a look at my excel sheet.

Edited October 8, 2022 by Brandon Bearden

[Joël Mercier]

6 hours ago, Brandon Bearden said:

I have a 60 amp breaker and 85% load is 51amps and just over 12k on 240.

 

I want it to heat up very quickly.

 

Take a look at my excel sheet.

Okay, looks like you did your homework. I just wanted to make sure you knew you'd need a high power dedicated circuit. 

 

I am currently on the process of building an horizontal kiln and I bought the kanthal A1 wire. When you mentioned K1 It made we wonder what was the difference so I went to the Kanthal official website and couldn't find K1. It looks like it is discontinued. You did not mention if you had bought the element already but if not, you may need to recalculate the power/length with A1 specs. 

Edited October 8, 2022 by Joël Mercier

[Brandon Bearden]

I have no idea what prompted me to put K1, it is A1 that I was looking at in the Kanthal Handbook.

 

Sourcing this wire is something I have not done yet and is a major PITA really. I was thinking of just contacting Kanthal directly. Their webstore is being rebuilt apparently. 

 

I should probably embed this into the Excel Sheet as a reference tab.

 

 

[Brandon Bearden]

Here is the updated, properly named sheet with the above reference embedded.

 

I am pretty sure this would be a good tool for someone else. It took a while to build this and understand what was going on.

Kanthal A-1 Wire Size Calculator.xlsx

[Joël Mercier]

1 hour ago, Brandon Bearden said:

Sourcing this wire is something I have not done yet and is a major PITA really.

I am in Canada and I found mine on a pottery supply store. It might be worth looking for something like that in your area. But finding such a large gauge wire may be harder. I went for the 1.2mm since I'm just shooting for around 3-3.5kW on 220V to be able to use in on a standard household circuit. 

 

And I was lucky and found an industrial supplier for the insulated bricks and got them real cheap. 

Edited October 8, 2022 by Joël Mercier

[Brian Dougherty]

Sounds like you are aware of what you are in for from a power requirement.

 

As much as I would like to see a 12kw oven built, I do feel compelled to mention that you might be creating a control issue.

 

PID controllers will control the power to the element in order to intelligently ramp the temperature to a setpoint without a lot of overshoot.  What most people don't understand is that the math behind tuning a PID for a system is very complex.  The controllers we all buy are "tuned" for a "typical" system.  What typical means is up to the engineers who designed the controller.

 

Where you might have trouble is of the input power to thermal load ratio is way off of what the engineers designed the controller for.  If this happens, you may end up with a lot of overshoot, and have to wait an hour for the oven to cool back down to the desired temp.

 

It might work out fine.  However, when I try to explain this to the engineers I am training, I tell them you want to stay away from scenerios where you are either trying to heat up a bathtub with a tea candle, or trying to heat up a thimbal with a blow torch.

[Brandon Bearden]

@Brian Dougherty Thanks for that bit of info. I was thinking about that. The controller I purchased is not a basic one, it is purpose built for scenarios like this. However, I will give Omron tech support another call and go over my spec requirements in more detail and make sure the models I have chosen will work for the intended purposes. I also think the chamber volume I have will be about right too for the specified heat input and target ramp rate.

 

If you have any interest, or input better yet, the exact models I chose are in the first post.

[Dan Hertzson]

Brian,

Sounds like you are talking about the PID constants for the feedback loop.  I would expect most decent controllers would have the capability to manually set those to tune to equipment design or have autotune.  Certainly the ones I've used have that capability (of course I built most of mine from old lab controllers from Kodak that I picked up for a song from a liquidator (1/4 DIN Honeywell and West programmable units).  Somewhere I even have an algorithm for tuning a PID heating system that gets things close with only a couple of iterations.  Really wasn't that hard to do. 

[Brian Dougherty]

@Dan Hertzson, that is mostly right and I certainly didn't mean to make more of this than necessary.  However, some systems simply are not tunable even if you do have access to all the parameters.  Since this is a simple heating system, it is likely that turning down the Proportional (sometimes referred to as "Gain") constant will make the system behave.  However, this has the same effect as using less power in the first place.

 

 

If you are interested in more geeky stuff that is poorly explained, read on:

 

Back in the 90's I spent several years developing control systems.  This was the olden days when you had to do everything manually in PLC code.  However, even with that I'm not really qualified to explain what makes one system stable, while another one in inherently unstable.  This is the most clear article I could find, and it triggers bad flashbacks of my controls systems classes:  https://web.mit.edu/2.14/www/Handouts/PoleZero.pdf

 

How I simplify this so I can hold it in my mind is like this: 

 

Every system can be described with a transfer function that models how the input (electrical current in this case) flows through the system and eventually becomes the output (temperature in this case).  It is quite easy to end up with a system that oscillates, or ends up in a run-away situation.  If you look at the MIT article, this happens when the poles end up in the right-hand plane.  (This concept is so often quoted in control system circles that is spawned an old chestnut of a joke about a how to stop an airplane from crashing.  I'll let you look that one up if you are brave.)

 

Let's think about the oven in question:  When the solid state relay (ssr) closes, current starts to flow.  This causes the wire to heat up, but it takes time for this to happen.  As the wire heats up it starts to heat up the walls of the oven through all 3 forms of heat transfer. (Conducted, convection and radiation).  It also heats up the air in the oven through convection and radiation.  All of this takes time, but the rates of temperature rise are all different.

 

As the walls of the oven heat up, they also start to heat up the air in the oven through convection and radiation.

 

Eventually, the air in the oven starts to heat up the temperature sensor, which follows its own temperature vs. time plot.  This temperature sensor is the only direct information the PID has to work with in order to control the system.

 

When the PID algorithm decides that it is time to turn off the ssr, all of these elements start to cool down at their own rates.  All of these plots of temperature change over time are what define the overall transfer function that the MIT paper describes.

 

I like to visualize all of these elements as springs (ok, I think of Slinkies actually) that are all tied end to end.  At one end you are pushing back and forth to control the position of the far end.  If you move real slow, this is easy.  However, the faster you try to move the end you are holding onto, the harder it is to hit your position target at the other end.  The more springs (elements in the system) the more complex it gets.

 

PID loops are an approach to controlling the desired set point by continually looking at the error between what you want the setpoint to be (temperature in our case), and the actual value.  They do this by combining 3 factors to determine the new power level to apply.  The P (Proportional) factor simply uses the error as part of the calculation.  The larger the error, the more power you should apply.  The I (Integral) factor allows the effect of the error to grow over time.  A small error may not generate enough P influence to ever close the gap.  Integrating this over time allows that small error to start to have significant impact on the dictated power. The D (Differential) factor looks at how fast the setpoint is changing over time.  If the value is really changing fast, it acts negatively on the power output to try to slow things down so you don't overshoot.  The controller continuously monitors the temperature, and makes new calculations for what to set the power level to.  This power level becomes the ssr duty cycle. 

 

You get oscillation when the PID algorithm makes a power change that then causes the temperature to change in such a way that the values dictate an equal but opposite change in the next cycle.

 

PID loops are very clever things, but they cannot overcome a system that is inherently unstable.  Having too much power is one of the easiest ways to create an unstable system.

 

 

 

 

[Brandon Bearden]

@Brian Dougherty Like when your 454 squirts the drive train - sort of, not really. Thanks for the detailed information. It totally makes sense to me. I talked to Omron tech support again and told them what I was doing and they think it will work just fine, however, what else do you expect them to say really. I can always size down if this fails I suppose.

 

--

 

On another note, I realized I ordered the wrong PID controller, I wanted a E5AC-T model. They are not cheap and hard to find. However, with Omron's CS Thermo software, and even without, it would be a powerful controller that should be rather easy to setup and use. Since it has ramp/soak features and 8 programs with auto-tune, I can set this up to perform temper cycles without any issues and also have several different metals pre-programmed. With the software, I can update the programs very easily.

 

I decided to give Omega a call, and they say their top of the line controller is the CNPT Series. Their documentation is much less professional but the product seems sold, as described. They are also cheaper and more available. They said they make the CNPT models in house and lead time is max two weeks. It is also not clear to me which model will support driving a SSR - though I can call them again for that info.

 

I hoping that someone has spent time with either Omega or Omron PID controllers to provide some feedback.

 

/cc @Dan Hertzson

[Brian Dougherty]

Either brand will probably work well as advertised. 

 

Omega is a huge name in temperature control, but to my knowledge, they primarily sell products designed by others.  They either license or private label them.  (The company my dad worked for in the 70's and 80's was one of these)  However, they have been doing this for decades, and have a good reputation.  Just expect that there may be a weak link if you ask them super detailed tech questions as they may have to reach out to the firm that designed it.

 

Omron makes a lot of quality control systems products.  They don't specialize in temperature control, but all of their stuff I have used has been quite good.  I would generally consider Omron to be the more refined of the two, but that is my impression only.

 

Edit to add:

BTW, in spite of my hesitations, I am really looking forward to seeing this beast come to life

Edited October 10, 2022 by Brian Dougherty

[Brandon Bearden]

Yeah, I was thinking along similar lines. The tech support at Omron was way better just on the initial calls. On the Omega side, apparently this is a house brand unit, which is why the docs are so lacking but also why it is so fully featured and less expensive. 

 

Edit... your hesitations are helpful and thanks  

Edited October 10, 2022 by Brandon Bearden

[Brandon Bearden]

So, I also considered a Novus. However, I am definitely sticking with Omron. Their manuals are 10x better, and the software is nice. The Omega has terrible reviews and Novus is just a step down and their manuals are horrible too.

[Dan Hertzson]

@Brian Dougherty  Sorry, I thought that modern SSR allowed fast enough response to provide modulating heating load allowing better system tuning.  I built a heat treat oven that was only around 8 x 6 x 14 with salvaged coils from an old duct heater, mercury contactors, and an older, programmable, PID controller.  Pulled over 40 A at 220V and was still able to maintain stability without hysteresis at temperatures between 400 and 1500 deg. F.   

 

Been a very long time since my System Control classes (over 40 years), but thanks for brining some of it back. 

Edited October 11, 2022 by Dan Hertzson

[Brian Dougherty]

14 minutes ago, Dan Hertzson said:

...I thought that modern SSR allowed fast enough response to provide modulating heating load allowing better system tuning. 

 

You are right about the SSR switching speed, and it is something that I glossed over because I was being lazy.

 

Controllers can switch the SSR on and off several times a second if necessary so they don't work like a classic oven thermostat.  The way they control the power into the system is by controlling the duty cycle.  All duty cycle means is how much time is spent on as opposed to off.  For example: Lets say you are switching your relay on and off once a second.  If you leave it on for 5 seconds in a row and then turn it off for 5 seconds in a row, you have a 50% duty cycle over that 10 second period.  Similarly, if you turned it on and off every other second, you would also have a 50% duty cycle.

 

How often your system samples and refreshes its output calculations impacts how well it can control a system.  You can even model this as part of the transfer function, but I can't explain that.  It involves Z and S tranforms, and I had pretty much given up by the time I got to that part of the class work.  

 

One of the limiting factors is that the controller can only adjust the power based on what it can see happening.  If several seconds go by between increasing the input power and seeing a change on the output, you will have been pumping energy into the system blindly that whole time.  There is no way to take it back out so all you can do is turn the power off and wait to see what happens.  This is where the auto-tuning came in and made these little PID brick so useful for everyone.  The better ones have the ability to learn a particular system so that they know how long to blindly pump in power before they back off even if the feedback signal isn't timely enough.  The better units used to learn a particular system, and get better every time.  I think that is still a thing, but I am not 100% sure about that as I haven't built a system form scratch in a decade or so.

 

BTW, I don't think the cheaper ones that most of us buy are truly auto tuning.  I think they are just tuned for a middle of the road solution and make the best of it.  They seem to follow the same profile each time with no learning happening.

 

I haven't looked closely at the data sheets for the controllers Brandon is considering.  The ability to have profiles will make them more expensive than the $30 units we all buy on FleaBay these days so it would be difficult to judge by price alone, but it is possible that they are truly auto-tuning as well.

 

I'm terrible with names, and just realized @Brandon Bearden's name is familiar.  Were you the guy I was talking to at the Bowie Hammer In this weekend?  If so, I am face-palming right now as we could have talked all this through in person 

 

 

 

[Brandon Bearden]

@Brian Dougherty - $30 ehh. Multiply that by 10, then by 3 and then you can buy the E5AC-T. For that much, it better darn well have true auto-tuning. I was only able to find a model without event inputs which stinks, because that means I have to build the door switch off a relay. I am still trying to find out how long it will take to get one with the event inputs because then I could just wire a cheaper tactile switch to the input and pause the program whenever the lid opens. Maybe I should consider one of the cheaper units, but this one seems sooo nice.

 

Also, you are off the hook. Different guy. I live in the Colorado mountains. 

[Brian Dougherty]

Yeah, for near a grand I'd hope it was pretty capable.

 

I just pulled up the data sheet, and it claims to have multiple levels of algorithm training so I suspect it is true auto-tuning.  I was shocked that there isn't a means for a remote run/stop switch as that would seem pretty standard for something at that price point.  However, as best I can tell from my quick glance, you are right in that the event inputs are the only way to get digital I/O into the thing.

 

Are you trying to buy through Digikey? 

[Brandon Bearden]

Yeah, digikey. I placed an order for a -000 model which the only had one version quantity one in stock. I also placed an order for a -019 model which is actually cheaper. I am waiting to see what the lead time will be. 

 

I just found some models on ebay - E5AC-TQX4ASM-060, but I am not sure what is gong on because the "S" should be a "5" and the -060 isn't a valid option on the current datasheet. Google is of no help, I did find the part number on their website search, but it only pulls up a strange page. I am wondering if they are non-usa models or something. I tried calling Omron again but was kicked to VM, I will call back soon and find out. 

Edited October 11, 2022 by Brandon Bearden

[Brian Dougherty]

I happened to have a meeting with my EE guys this afternoon so I asked them where they are sourcing Omron components from these days.  They've had some luck with Allied when Digikey was out of stock.

 

Getting anything electronic has been a struggle.  We've made more design decisions based on what was available rather than on what was ideal in the last year.