Need some advice on hardening and testing a thermocouple

M

MKarish

Joined
Jun 18, 2016
Messages
34
New guy/fairly new knife maker here. I know there's a huge amount of knowledge on this forum so I'm hoping you guys can help me with some advice on process and setting my gear up.I'm trying to get better control over my heat treating processes rather than just take past magnetic and quench. Apologies for the first post wall of text but I figured I should give you all the info I could.

I've bought myself a small top opening 110v electric kiln and tried to harden two blades in it last night. I'm using 0.156" 1084 for one and 0.125" 15n20 for the other. Both steels from NJ steel baron. I did stock removal only on both, and did not bother to normalize. Both blades did harden some, but neither of them made it to "full hardness" as gauged by scratching them/scratching things with them. I don't have a set of hardness files or easy access to a Rockwell tester at the moment.

My process was to heat to 1500 deg F and quench in 130 deg F canola oil. I brought the kiln up to 1500 which I believe should be in a reasonable range for both steels. I loaded the blades into holders (not wrapped or clayed) and lost about 200 deg in doing so. I let them come back up to temp which probably took 10 or 15 min. I had two friends waiting on me to go to dinner so I did not give them an extended soak and quenched as soon as the thermocouple gauge read 1500. I did the 1084 blade first (just dropped it into quench bucket) then pulled the 15n20 blade and dropped it in as well. I do not know how much temp may have been lost in the second blade while doing the first, but it was probably in the kiln with the lid off for less than 30s after the first one went into the oil. Both blades looked a little on the darker side of red to me by eye, but I honestly do not have enough experience to say I can judge temp by sight, and it's quite possible I've overheated the few blades I've done in the past.

the 1084 blade was soft enough that it could be filed, not easily, but filed none the less. The 15n20 was harder, but still soft compared other blades I know to be in the 58-62 range that I would like to hit.

I fired the kiln up again this morning and thinking that the old thermocouple may not be reading accurately tried to track the voltage it was putting out and comparing it to a thermocouple chart. As it came up to temp it seemed to be holding pretty close to where it should both on the gauge, and with the voltage readings I was getting. HOWEVER when I bought the kiln the thermocouple was pretty well burned out at the end so I cut it back and TiG welded the end back together once I had got to clean wire. I also found that I was getting different readings of voltage at the leads of the thermocouple, and at the gauge. When it was coming up to temp these were both pretty close, but once I got up to temp and the thermocouple itself started bleeding heat out of the leads I was getting as much as a milivolt difference. By the chart this could be as much as 40-50 degrees difference. I rehardened the blades one at a time, same temps, but let them soak a few minutes each and got better results. I still feel like the 15n20 is softer than it should be though. The 1084 is ok, by my very unscientific tests.

So what I'm wondering is do you see anything terribly wrong with my process? And have any of you guys tried testing a thermocouple's voltage, have any experience with rewelding/repairing the probes, or have advice on how to go about taking a reading for the output voltage so as to get an accurate reading?

Thanks in advance-MK
 
Thermocouples should be spot welded, melted together. Adding a dissimilar metal can effect the accuracy. I've spot welded them in the past by twisting the end together and shorting the wires to a capacitor bank.
 
Navin-
Thank you, I should have clarified that. I did not add filler rod to the weld, I used a TiG torch to melt the ends of the thermocouple back until they flowed into each other. Then I flipped it so that the welded tip was down, and hit it with the torch again to form an even ball on the end. Does this sound more reasonable?
 
Yes, it does sound more reasonable. However, for a thermocouple to function properly the composition needs to be maintained tight. If your thermocouple was old to start with, you should throw it away. The extension cable as well.

Check this brief intro: http://www.omega.com/temperature/z/pdf/z021-032.pdf

What I would do is replace the thermocouple/cable assembly for a new one, and buy a cheap thermocouple reader. Use the stand alone reader to validate the main TC measurement from time to time.

It's also useful to have realistic expectations on the accuracy of the measurement: The voltmeter will have 2F of error, the cold compensation another 2, and a (good) tc another 2. This means that when new, the whole setup will have at least ±6 F of error. If you monitor and change the tc every 50hs or so, I would expect it to age no more than 4F.

In the end, if you need an accuracy better than ±10 F, you need to start dealing with regular calibrations against a standard cell.
 
gammarad; you're tossing out exact numbers of degrees F, will this be the same 2ºF at 200, as it would be at 2,000 ºF? OR - will it be larger at high temps? You're right, TC's do age and change calibration with age.
 
These where examples at around 1800F.
Some of the errors will be proportional to the temperature, some will not.
- The TC/cable error is proportional to the temperature.
- The cold reference error is constant.
- The voltmeter error will have a small constant bias, and a larger proportional part.

YMMV.
 
Thanks gammarad, maybe I should just dump the old TC and get a new one like you suggest. I don't need extreme precision out of the temps, but without knowing how far out of whack it is to begin with there's no telling just how far off my reading are. I also realize I have no idea if the wire leads are the correct type of wire, and they have crimp on loop connectors on both ends which are probably each adding some small degree of error as well. I suppose I can test it against some other known values like boiling H2O or melt some lead and see where its at but it might be better in the long run to just junk it. I may buy a PID controller anyway so I'll probably just factor that into the upgrade cost.
I'm sure in the end this will be one of those cases where a super cheap craigslist score will end up costing me twice as much in parts and upgrades as I paid for it, and then I'll be building or buying a proper oven in the end anyway.
 
Anytime you connect the thermocouple leads to another wire or terminal made of a different metal, you effectively create another thermocouple. The new "thermocouple" (junction) will create it's own voltage in the circuit and any change in it's temperature due to room temperature changes will affect your readings. Given of your furnace at 1500 degrees the effect of the room temperature going from 70 to 90 degrees will affect your absolute reading by 20 degrees, a small amount proportionally, but it all depends on how precise you are trying to be given your heat treat cycle. If you want to be precise, you should get a new thermocouple with long enough leads, and a meter specifically for that type of thermocouple, which will have terminals of an appropriate alloy.
 
As usual, it depends.

My brand new Evenheat oven uses brass terminals to connect the thermocouple to the extension wire, right out of the oven. I don't expect a significant error being added by it.

What happens when you split a TC cable, and interconnect with another metal is that you form two new TC: m1 to m2, and m2 to m1. As long as both intermediate tc's are at the same temperature one to the other, their effects cancel out. This is why you can simply solder the wires without introducing errors. Adding new metals to the circuit is unavoidable anyway: the measuring circuitry will be copper.
 
Well, except that your 2 thermocouple leads are 2 different alloys, so the voltages created by the 2 junctions do not completely cancel out. I speak from some experience.

For a few weeks during one of my summer jobs in college, I temperature profiled an experimental "see through" furnace to be used for studying the growth of Gallium Arsenide crystals. I used a K type thermocouple and an expensive new high precision digital meter designed specifically for the type K. All connectors were screw type and specific to a type K, wire was type K only, etc. No doubt the measuring circuit was eventually copper, but by controlling the alloys used in the attachment points the meter maker could compensate for the voltage effects.

Stick a random piece of wire off the workbench in the circuit, that's when you get problems.

When we got to the point of testing temperature stability, I hooked the thermocouple up to a strip chart voltage recorder, just wire nutting the thermocouple wires to the tinned copper leads of the cable to the recorder. As far as I knew, millivolts were millivolts, and it was what my boss told me to do. The tests showed a recurring 3 degree C swing in temperature, unacceptable given that the very expensive furnace controller was supposed to hold the furnace temperature constant to less than 1 degree C. There was much wailing and gnashing of teeth at higher pay grades, tweaking of adjustments and re-running of tests. A few days later, about the time I realized that the temperature swings occurred when the lab's HVAC kicked in, a senior engineer from another group visited our lab, pointed to the connection between the thermocouple and the chart recorder, and explained why that was a problem. I insulated the connections, and isolated them in an ice water bath (constant temperature), and indeed the furnace was stable to within 1 degree C. Obviously this was way more precision than is needed for heat treatment of steel.

Depending on the type of thermocouple you are using and the metal/s you are connecting it to this can introduce significant errors, something to consider if you've put a thermocouple circuit together from random parts and it's not behaving as you expected
 
One of the neat things in my workroom is a laboratory 8" temperature reading gauge from the 1930. It is diamond pivoted and reads in 1/10 degree. It has three ranges, 0-10, 0-100, 0-1000. Accuracy is stated at .1%. The terminals are gold plated and marked +/-. It is mounted in an oak case with hinged lid.
 
So I'm trying to follow along with you guys regarding junctions. I have K type wire running from the TC to the meter, which I have verified off a color coding chart (R/Y wire with brown jacket) as well as checking for magnetism on the negative lead. The wire appears to be thermocouple wire, not extension wire, but I don't see why this would make a difference since they're the same metal.
But at the TC and temp gauge junction there are zinc plated copper loop connectors. So both of those are now acting as TC's themselves. Would those second and third junctions be adding voltage? Or is this what you were referring to (gammarad) about them canceling out since they're both at nominally the same temperature? If they add voltage my gauge is reading higher than it should be and I'm not getting the steel hot enough which would explain why the blades are not hardening fully.
Stacy, you coming by with that test gauge? Beers are on me.
 
The screws in junction blocks and connectors at the PID are not of much concern. Technically, they do create a certain amount of voltage, but it is very tiny. It is also constant, as the temperature changes very little there. The wires at the TC junction are of a type that creates more voltage and are being exposed to the actual heat source.

Just use good wiring practices and use the correct type K wire and everything should be fine.

I am not sure what you tested "by magnetism", but a meter would be how to find the +/- wires. Did you mean "by polarity".
 
MKarish said:
(...)But at the TC and temp gauge junction there are zinc plated copper loop connectors. So both of those are now acting as TC's themselves. Would those second and third junctions be adding voltage? Or is this what you were referring to (gammarad) about them canceling out since they're both at nominally the same temperature?(...)
Click to expand...
Indeed. Those are the kind of unwanted TC junctions I consider of no importance for the precision we're aiming for. Stacy is right, cables are very important, as their length means the two ends won't be at the same temperature, and if not of the proper material they will create significant errors.
@stacy, I think he refers to alumel being magnetic, while chromel is not.
 
No matter what set up you start with you will need to develop a system for calibration. First, what accuracy and precision do you need?
Also, with your set up there are other problems that might occur besides temperature control. Radiative heating is one. With the oven on, the steel might heat up faster than the interior of the oven. So, second, are you correcting for other factors?
If your oven can hold at a given temperature, all you need to do is determine a correction factor to get within acceptable temperature values.
 
I want to thank you guys again for the input. I pulled the crimped on terminal loops off of the wire leads and hooked the wires directly to the thermocouple and to the temp gauge. I noticed that the temp gauge did drop a small amount just by rewiring it so I believe there may very well have been some error being caused by the zinc/copper connectors. There's also a chance that the reading changed because it was not level or something like that. I should have taken voltage readings before and after for the sake of science if nothing else. Playing a torch over the junction had minimal effect compared to hitting the TC itself so I think I should get more consistent and accurate temps.
So now that I have a somewhat (I hope) accurate and repeatable baseline to work from I can start doing some process testing to see what works well with this set up and steel.
 
You must log in or register to reply here.
Back
Top