Thermocouple PCB attachment method

[alank2]

Hi Everyone,

A friend of mine who has baked more boards than I ever will told me that he just has the end of a thermocouple wire twisted together and just finds a via to stick the wire end into for temperature sensing.  This sounds like a great way to do it as I can see the tension from having the wire hold the twisted end at an angle holding it in the hole, so no attachment to be concerned with.  I just haven't seen this method in any of the documents some have posted here as an attachment method.

One other issue I have is that I am using a metal tray in my oven that came with it and I'd probably have to stop using that and put boards right on the bars like he does so the wire could hang through the hole and go below the bottom of the board surface.

How do you guys attach the thermocouple?

Do you have a metal pan on the bottom or have your boards right over the elements?  I could put the rack in the middle so it is perfectly between the elements.

Thanks,

Alan

[mjkuwp]

I documented something like that here:

http://mjkuwp94.tumblr.com/Soldering

A difference being that I added some extra-high temp solder to secure it.  If you get the connection tight the soldering would not be necessary.

I have to say it is a pretty big pain to get the wire secured without interfering with the components.  Or if you do it the reverse way... pretty hard to apply paste and components with a thermocouple attached to the board.


what I have done more recently is to connect the TC to a scrap PCB and leave it next to the target board.  I think this is a pretty common method.

I have always placed my boards on the metal bars and used the bars to help strain-relieve the wires.  You want to have really fine thermocouple wire... 30 AWG?  I don't remember exactly what I used.

[alank2]

Where do you guys get high temp solder?  Is it a certain type or advertised that way?

[mjkuwp]

I used silver solder wire.  98% tin 2% solder.  I got it at the local hardware store (I think Ace Hardware in the US).

You may want to pay attention to what kind of flux is comes with the wire or is in the wire.  The kit that I bought had flux with it that said it was not for electronics.  I didn't care about that since I was only making one temporary joint and I  cleaned the board.... and it was a board for my own personal use.

melting temp stated to be 450F [232C].  I use lead solder that melts much lower than this. 

[Alexei.Polkhanov]

I think best solder for thermocouples is Cd70Sn30 Which is 70% Cadmium 30% Tin - it reduces chances of creating parasitic thermocouples.
I was not able to get this type of solder from regular suppliers like DigiKey when I needed it so I made it myself using 99.9% Cadmium and Tin in bars.

[Richard Crowley]

Yikes! Twisting, or even soldering seems like a risky to try to make a reliable thermocouple.
The normally acceptable methods of joining thermocouple wires are brazing, welding, or crimping.

[mjkuwp]

Yikes! Twisting, or even soldering seems like a risky to try to make a reliable thermocouple.
The normally acceptable methods of joining thermocouple wires are brazing, welding, or crimping.

true, but it only has to work once. 

[mzzj]

I think best solder for thermocouples is Cd70Sn30 Which is 70% Cadmium 30% Tin - it reduces chances of creating parasitic thermocouples.
I was not able to get this type of solder from regular suppliers like DigiKey when I needed it so I made it myself using 99.9% Cadmium and Tin in bars.

I cant see why cadmium solder would be any better for thermocouple wires.
(Yes, its (theoretically)better for copper wires when you want to avoid parasitic thermocouples)

[StrangeloveMD]

Hey alank2, nice to see more reflow going on over there!

"IPC-7530: Guidelines for Temperature Profiling for Mass Soldering Processes (Reflow & Wave)"
Costs money. If only a google search yielded a free PDF somebody posted online. 

Note that most "thermocouple wire" is actually "low grade extension wire", and is /not/ of the same grade as the wire used to make proper thermocouples.  Thermocouples are made from alloys precise enough to yield a Seebeck voltage that is accurate to a defined point, over a defined range.  The thermocouple extension wire is (usually) spec'd accurate over a -50*C to 150*C range.  It's intended for long runs between TCs and test gear, where you don't want to introduce a seebeck junction (and thus must use the "same" alloy as the TC wires), but also don't need alloys precise enough to give the supersexy version of the Seebeck curve.

The main thing I don't like about the described method is its repeatability.  Just chucking some wires into a via is going to give sloppy results.  (I know it seems like it's making a decent, consistent thermal connection, but it isn't.)  In hobbyist reflow precision is arguably more important than accuracy.

If your buddy has reflowed more boards than you will in a lifetime, ask him to estimate the number.  Proper thermocouples, even pretty good ones, are cheap.  Calculate how much money you'd save over your entire life, even if you did manage to reflow as many boards as him.  Divide it out over the years, for context.  Is it still  worth the savings?

P.S. One of the app notes I posted in the other reflow topic mentions the high temp solder alloy they prefer for TC attachment, as I recall.  (290*C liquidus iirc)  Don't remember which one... I'm thinking OSRAM, or the 90's article. 

[Alexei.Polkhanov]

Well if you go to Wikipedia and look up solders you'll see  something like that:

Cd70Sn30 140/160[15]Cd no   Cd70, thermal-free solder. Produces low thermal EMF joints in copper, does not form parasitic thermocouples. Used in low-temperature physics.[34].

That is the one of two solders listed in that table that has this property.
Since I have major in Process Control I was suppose to know that, but then I forgot all tiny details since I don't work in this industry. Year ago I needed precise temp measurement for chemical reaction I was playing with so I looked at my old university metrology book - and there it was.

Quick Googling gives lots of results like this article by popular on this forum Jim Williams: http://m.eet.com/media/1142215/82758.pdf.

It is not that I don't have silver - my soldering iron does not melt it  If use something like a welder then I would use silver of course.

[Nerull]

It doesn't form parasitic thermocouples with copper, what does it do with the alloys in your thermocouple wire? They aren't copper, so the chart doesn't apply.

That solder also has a very low melting point - it will fall apart before your board reflows.

[Alexei.Polkhanov]

Huh? You are thinking of putting the board with thermocouple into reflow oven?!

I assume you solder thermocouple wires by hand and then screw the wires inside the thermocouple assembly when you install it. Thermocouple wire is made from material close to thermocouple itself or same so it does not form any parasitic thermocouples, but then you have to attach it to copper on the board and that is where "copper/solder" problem comes in which can be mitigated by using proper solder.

[alank2]

Hey alank2, nice to see more reflow going on over there!

I've only reflowed 10 boards, but other than my shaky hands in positioning them, I sure like it more than through hole.

Here is a screen shot of he software I've been working on:

[mjkuwp]

Huh? You are thinking of putting the board with thermocouple into reflow oven?!

...

not just thinking of it, I have done it and it has worked fine.  I put two TC wires into a via and twisted them up. Then soldered in place with silver solder.  I had to use my butane-powered iron to reach the temperature.  The joint did not melt in the reflow process and with this method as I recall the solder melted just as the display showed the melting temperature.  It was a better correlation than I have obtained with other methods.

Maybe more than one junction is created but in the loop they sum correctly anyway.   Whatever happens... it just seemed to work.

However... this was too much work and more recently I have been putting aluminum foil tape over the thermocouple and kapton tape in a few places to keep the wire in place.  Ideally there would be a plated hole in the ground plane to put the head of the thermocouple in to - or a similar feature on an area of interest such as a sensitive part.

[StrangeloveMD]

alank2, that puts the command line interface I'm working on to shame!  Is it scratch coded, or a modification/application of existing software?

Guys, also be wary of charts that list "41uV/*C" or "13uV/K", and such things.  Remember that the Seebeck voltage does /not/ rise linearly with temperature!  If you are looking for accuracy, precision, or useful resolution, you will need to employ a well stocked lookup table, or derive the temperature using the appropriate equation.

There's also still some talk of the twisted TC wire being used as an actual thermocouple, hopefully we can all be a little more clear.   What is an alloy?  It's a  mix of metals; this seems very simple.
Let's look at an alloy of 60% lead (Pb) and 40% tin (Sn).  In fact, lets pretend we are making it.  We have a 60Kg block  of lead, and a 40Kg block of tin.  If we put them in a large pot (a crucible), we still just have two chunks of metal.  As we heat the pot, the metals melt.  First the lead starts to flow, and eventually there is a pool of molten lead around the solid tin.  Still not an alloy, still just two metals in a pot.  But then the tin melts, and a bit like oil and water, it doesn't quite mix with the lead.  Still two metals in a pot.

But then we begin to stir.   Lets say we stir ten times.  The metals are mixed together, into an alloy.  60/40 Pb/Sn, 100Kg.   But say we dipped in a spoon, and pulled out 100 grams.  0.1% of our alloy.  We do a test, and find it is pure lead.  We did not mix far enough, so we try again.  Next sample is 80% tin.  Mix mix mix.  A sample comes out 66% lead, 34% tin.  We sample again without mixing; 46% tin, 54% lead.  Mix mix mix.
Finally we take 1,000 individual 100g samples.  We test the entire mix.  Each sample is a 60/40 mix.  A perfect alloy.  We go home, tired and satisfied.

Then, the next morning, we awake with a start.  We run out and test a 1 gram sample.  0.001% of the mass of the alloy we made.  It is 59.9% lead, and 40.1% tin.  We didn't mix long enough, and we were not looking close enough to tell.  If we made a thermocouple from this alloy (and another alloy for the other wire), we wouldn't know if the tiny little Seebeck junction was 60/40, 59.9/40.1, or any other combination.  So, we could only call it "60/40, +/- .1%", and we could only ever assume it had a Seebeck coefficient of an alloy /within that range/. 

A thermocouple extension wire will be made from the "same alloy" as the wires used in proper thermocouples, but only in 100g samples, as it were.  By the time you make a fine wire out of them, the slight differences in composition mean they will perform unreliably, with variation between any two "thermocouples" you make.  Because Seebeck voltages do not rise /linearly/ with temperature, this is more significant at the extremes of operating temperature.  This is why thermocouple extension wire, while made from the "same" alloy as the wire in thermocouples, is only rated accurate over -50*C  to 150*C (generally).

A fun fact that demonstrates how serious this gets:
When manufacturing high quality thermocouples, many samples of alloy A will be tested, and many samples of alloy B tested.   The company will then match batches of wires, based on the performance difference caused by inaccuracies in manufacture, to create the best A/B pair they can, that conforms as tightly as possible to an ideal curve.

[alank2]

alank2, that puts the command line interface I'm working on to shame!  Is it scratch coded, or a modification/application of existing software?

Thanks.  I've made it from scratch along with the avr atmega32u2 based ic controller.  Been working on it for awhile, but it is starting to come out the way I was hoping it would.  It really came together when I added the graph, until then I was exporting csv to excel and making a graph in excel which was a pain.

[StrangeloveMD]

alank2, that puts the command line interface I'm working on to shame!  Is it scratch coded, or a modification/application of existing software?

Thanks.  I've made it from scratch along with the avr atmega32u2 based ic controller.  Been working on it for awhile, but it is starting to come out the way I was hoping it would.  It really came together when I added the graph, until then I was exporting csv to excel and making a graph in excel which was a pain.

To what extent are you logging data, and what is the measure/calculate/set cycle like?   (In terms of sequence and functions performed)  Actually, wanna just write a topic on your controller when you're done with it/get a chance? 

[alank2]

To what extent are you logging data, and what is the measure/calculate/set cycle like?   (In terms of sequence and functions performed)  Actually, wanna just write a topic on your controller when you're done with it/get a chance? 

You set the time loop parameter for how often you want the logging events (and pid loop) to run, in my case I am using 5s although measurements are taken from a max6675 every 0.250s, so I have 20 measurements averaged into a single pid PV.  You can set the loop time as low as 0.250s.  The CSV format is:

time,pv,mode,sp,co,fan
where mode is 0=manual (co), 1=auto (sp pid), 2=thermocouple error, 3=threshold error, 4=inactivity error

The design is that the controller controls the co output (1khz pwm for co of 0.0% to 100.0%) and a fan output (off or on).  It takes measurements from the max6675 and makes the immediate measurement available and averages it into a better loop time measurment as well.  It runs the PID and supports bumpless transfer from manual to auto mode.  It stores up to 32 records of each loop time event (such as 0s, 5s, 10s, 15s (intervals of looptime)) and the PC requests them and keeps the buffer from overflowing so no records are lost.  It has a USB interface and the PC software uses commands like GET_INFO, SET_CONFIG, GET_STATE, GET_RECORD, SET_AUTO, etc. to control it.

The PC software isn't done yet, but it has a step execution engine that supports:

Code: [Select]

    {SET_AUTO,        "Optional",                         SP,           MaxChange,    Fan           },
    {SET_MANUAL,      "Optional",                         CO,           MaxChange,    Fan           },
    {DELAY,           "Optional",                         Seconds                                   },
    {DELAY_THRESHOLD, "Optional",                         0(<=) 1(>=),  PV       ,    Seconds       },
    {PAUSE,           "Optional",                                                                   },
    {END,             "Optional",                                                                   },

You can ramp up or down to a SP or CO using the MaxChange option, for example set SP->140 and move there at 3 C/s.

The delay threshold allows you to wait until the PV is >= or <= a value and you can specify how long it needs to be (such as 30s) to satisfy the delay.  It looks into history to determine this so, if when this command is executed, you have already been above the PV for 25s and you asked for 30s, it will only delay 5 more seconds.

Still a lot to do as far as storing/loading/editing configurations, so that is next to work on.

I may try my hand at selling it together with a pid controller board.  The controller board would not have any power components like ssr's on it, but have outputs for co and fan that you can feed to a ssr.

Thanks,

Alan