Thermocouple deadlock

[Faringdon]

Hi,
Can anyone break the deadlock on thermocouples for electronics temp measurement?
Every place you go uses thermocouple wires which are stripped back, the wires twisted together over some 3mm, then soldered together.....rightly or wrongly.
Have you details on the rectitude of this, as opposed to spot welding the wires together?

I even know places who solder thermocouple wires together, then solder that to a DPAK tab...and they claim to get good results.

I must admit every time ive soldered them, and checked it....they work ok......but many www docs say its bad to solder them.

Recognise the theory says "its the junction of 2 metals"....bla bla bla....but nowhere spot welds.


...the EEVBLOG vid on this seems to have gone(?)
[EDIT...oh its back...it was just me in a railway tunnel]
is this true?...

You can form a junction by joining a pair of thermocouple wires together with a third metal. As long as both junctions are at the same temperature, and at the temperature of the thing you are sensing, the voltage you read will be the same as if you had a simple junction between the two metals.

I find twisted thermocouples can be soldered as long as its with tin/lead solder with flux....the wires must first be degreased with IPA.

For our temperature tests, we only need some +/-4degC accuracy....obviously if its 10 degrees inaccurate, but only reads 50degc, then thats also fine....since even if it were really 60degc, no fet is going to melt at 60degc.

[nfmax]

In principle, the presence of a third metal at the thermocouple junction makes no difference. The thermoelectric potential is actually generated along that part of the thermocouple wire which has a temperature gradient. The voltage/temperature relation of the two metals in the couple is different, so there is a net output voltage. That said, a welded junction is mechanically stronger, and may be less sensitive to corrosion. Also, not all thermocouple alloys are solderable.

[SeanB]

Yes, I have brazed and silver soldered thermocouple wires to make a temporary sensor, but the best is to weld them together. For the higher temperature couples, welding, either electric arc, plasma or laser weld, is the only way to join them, as they operate at temperatures well past the melting point of most other metals. You can even just bolt the 2 bare wires together, and it will work for a while. Biggest issue it with extending them, you need to take care to have the joints to the extension cables be isothermal with each other, at every point, and also to not add in a mix of metals, so if you have a terminal pin on the junction, and the wire breaks off, you have to replace both of them, using preferably bare copper pins, from the same bag, so they are the same copper pour batch. Same for a wire, you change one, you change both.

[jpanhalt]

Presumably, you are referring to type K.  The TC amplifier (Maxim) I am using is very sensitive to any disturbance in the voltage.  I was able to make spot welds with a battery tab welder, but they were not reliable.  Same with twisting.  I did not try solder, as the metals are difficult to solder.  Brazing would be my second choice.  I use a TIG welder.  It's very fast and burn back is trivial.  Attached are some pictures.  I strip the wires with the stripper shown, put about one good twist, wedge into the slot on the copper plate, strike the TIG torch off to the side and pass it over the exposed wire.  Bingo, I have a complete TC (shown).  Welder setting is very low.  My TC stock has Teflon as its outer insulation. 

[Faringdon]

Thanks, Yes, i have got  a big  drip of solder below a weller iron tip with temp regulation....i have the twisted pair of the K thermocouple end, and i hold it into the solder drip....i then cool it down to 100degc so i dont have to keep holding it...the thermocouple does not touch the actual solder tip....it just holds in the solder drip........the  TC-08 thermocouple reader does not read 100degs...it reads 86degs C....and if i yank the thermocouple wires about, then  the reading can sometimes change a few degrees either way.

Taking it up to 207degs.......and holding the thermocouple wires in the drip, the thermocouple reader reads just 160degs C.

Soldering thermocouples doesnt seem great...but then again....maybe the weller temperature meter is wrong.

We wouldnt be able to afford a TIG welder.

i redid one, and soldered a seemingly better joint...and i get agreement with both weller and TC-08 showing 76 degrees..but as you can tell, this  seems a bit hit and miss.

[jpanhalt]

If you get different readings when wiggling the wires, I suspect your joint is not soldered but rather encased in solder.

[Ian.M]

Déjà vu anyone? https://www.eevblog.com/forum/testgear/twist-and-solder-k-type-thermocouple-wires/
Proverbs 26:11 applies! 

[jpanhalt]

C'est Faringdon

[Ian.M]

Thoughts on crimping it?   I *think* a nickel plated copper crimp sleeve would go a long way towards making a reliable gas-tight joint if a well fitting ratchet, hydraulic or screw-down crimper is used.

[jonpaul]

Most TC wire is not solderabe and not bonded by twist, crimp due to oxidation of the material.

Type K, Chromel-Alumel, Iron-Constantan all WILL NOT tin and make a good solid connectioon.

Solder will contaminate the junction.

As you move and heat the TC, even a seeming good TC will fail.

The ONLY reliable solution is braze or weld.

We got 1000' Omega Type K Kynar #32 wire, setup a simple tack welder and fabed  hundreds (1080s) for hugh rack T1 testing with dozens of points.

Omega also sells the extension wire and plugs, sockets and finished TC with weld and plugs.

https://www.omega.com/en-us/temperature-measurement
Enjoy,

Jon

 

[David Hess]

Where the thermocouple wires are solderable, it works just fine, or is at least no worse, because if you follow the wires back to the cold junction compensation and amplifier, there are lots of other junctions which are in series with the thermocouple junction and soldered together.

So in series you have a thermocouple, some wire connections to a plug, the plug and socket connection, the socket connections soldered to printed circuit board traces, printed circuit traces soldered to parts for some signal conditioning, soldered traces to the pins of the lead frame of the amplifier package, and then bond wires from the lead frame to the semiconductors.

So any additional error from soldering the thermocouple junction itself is minor.

[MarkT]

Thermocouples are used for measuring high temperatures typically, so a soldered one is basically a chocolate teapot...  For low temperatures there are much cheaper and more accurate options.

[David Hess]

Thermocouples are used for measuring high temperatures typically, so a soldered one is basically a chocolate teapot...  For low temperatures there are much cheaper and more accurate options.

For test instruments it is difficult to get better than thermocouples simply because they are so common, which comes from their wide operating range.  I would love to have something better for the freezing to boiling temperature range, but thermocouples are the least expensive.

I do wish thermocouples were more accurate though.  I have started doing single point calibration on mine which helps, but that is not available on most thermocouple meters.  I would like to find a good RTD meter but so far they all seem to be junk.

[Faringdon]

I must admit i wish there were really tiny 10k NTC's, (10k 25degc)  that could  be superglued (and then  "activiated") to FETs as easily as a thermocouple end can be.

Going to find a good vid on how to weld the ends together with eg a discharging 10mF 100V capacitor.

[David Hess]

I must admit i wish there were really tiny 10k NTC's, (10k 25degc)  that could  be superglued (and then  "activiated") to FETs as easily as a thermocouple end can be.

Going to find a good vid on how to weld the ends together with eg a discharging 10mF 100V capacitor.

Bead NTCs are common and inexpensive, so how much smaller did you want?  I have seen them as 0603 and smaller soldered down onto a strip of flexible printed circuit board; my motherboard uses one as an auxiliary temperature sensor that can be placed anywhere.

I have been thinking about using very accuracy NTCs as a rough temperature standard in place of an RTD which is more difficult to reliably read accurately unless you can buy a set of 3 precision resistors to do a 3-point calibration.

[jpanhalt]

I suspect his client will insist on a 201 size or smaller. 

[David Hess]

I suspect his client will insist on a 201 size or smaller.

I am not sure if any are smaller than 0201:

https://www.mouser.com/c/circuit-protection/thermistors/ntc-thermistors/?package%20%2F%20case=0201%20%280603%20metric%29&resistance=10%20kOhms&tolerance=1%20%25&instock=y&sort=pricing

[Gyro]

If you're gluing to a FET package then size isn't particularly important, it's heat conducted away from the temperature sensor by the wires that it - working against the thermal resistance between the package and sensor. Thermocouples aren't great at that, the wires are of the same or greater csa than the junction (which is rounded). Coiling part of the wire in close proximity to the heat source helps.

The SMD thermistor and flexible pcb that David mentions is far superior in responsiveness. Even a small bead thermistor soldered onto fine wires would be better, both in responsiveness and not drawing excessive heat away from the item being measured.

An 0603 pkg or smaller diode or diode connected transistor, glued to the the device to be monitored, again connected by fine wires, also works well. It is cheaper than a thermistor and probably has better linearity.


EDIT:

I suspect his client will insist on a 201 size or smaller. 

His 'client' is very cheap! 

[ConKbot]

I must admit i wish there were really tiny 10k NTC's, (10k 25degc)  that could  be superglued (and then  "activiated") to FETs as easily as a thermocouple end can be.

Going to find a good vid on how to weld the ends together with eg a discharging 10mF 100V capacitor.

Bead NTCs are common and inexpensive, so how much smaller did you want?  I have seen them as 0603 and smaller soldered down onto a strip of flexible printed circuit board; my motherboard uses one as an auxiliary temperature sensor that can be placed anywhere.

I have been thinking about using very accuracy NTCs as a rough temperature standard in place of an RTD which is more difficult to reliably read accurately unless you can buy a set of 3 precision resistors to do a 3-point calibration.

I want thin film ones with the material laid down between kapton with termination that fits in a ffc connector so I can slip them between cells in a lithium polymer battery pack without worrying about the corners of an 0603/0402 package on a board, or the dimple from a bead leading to thermal events if the pack takes a bit of rough handling.

[jpanhalt]

You could probably roll or hammer the ends of the chromel and alumel wires into thin sheets and spot or TiG weld them.  Then grind off any protuberance.

[David Hess]

You could probably roll or hammer the ends of the chromel and alumel wires into thin sheets and spot or TiG weld them.  Then grind off any protuberance.

I have seen thermocouples made exactly like that for the thinnest possible profile.

Something else to keep in mind if it was done with thermocouples is that a pair of thermocouples in series will measure temperature difference without any cold junction compensation required.  Or if you had multiple thermocouples, perhaps multiplexed, then cold junction compensation only has to be implemented once.

I want thin film ones with the material laid down between kapton with termination that fits in a ffc connector so I can slip them between cells in a lithium polymer battery pack without worrying about the corners of an 0603/0402 package on a board, or the dimple from a bead leading to thermal events if the pack takes a bit of rough handling.

If I understand NTC thermisters correctly, they are all thick film construction or made like a composition resistor.

It seems however that what you describe does exist:

https://atcsemitec.co.uk/product/semitec-jt-ultra-thin-film-thermistors/
https://www.adafruit.com/product/4890?gad_source=1

[ConKbot]

You could probably roll or hammer the ends of the chromel and alumel wires into thin sheets and spot or TiG weld them.  Then grind off any protuberance.

I have seen thermocouples made exactly like that for the thinnest possible profile.

Something else to keep in mind if it was done with thermocouples is that a pair of thermocouples in series will measure temperature difference without any cold junction compensation required.  Or if you had multiple thermocouples, perhaps multiplexed, then cold junction compensation only has to be implemented once.

I want thin film ones with the material laid down between kapton with termination that fits in a ffc connector so I can slip them between cells in a lithium polymer battery pack without worrying about the corners of an 0603/0402 package on a board, or the dimple from a bead leading to thermal events if the pack takes a bit of rough handling.

If I understand NTC thermisters correctly, they are all thick film construction or made like a composition resistor.

It seems however that what you describe does exist:

https://atcsemitec.co.uk/product/semitec-jt-ultra-thin-film-thermistors/
https://www.adafruit.com/product/4890?gad_source=1

I looked at digikey after posting my comment, and some 0201 thermistors are 0.2mm maximum thickness, which may fit into a pocket on some FFC stiffener, to get below the 0.5mm max of those. I may evaluate that more in the future if the need arises again.

You could probably roll or hammer the ends of the chromel and alumel wires into thin sheets and spot or TiG weld them.  Then grind off any protuberance.

Oh the super thin application is certainly better suited to thermocouples, or RTD, etc.  The application I had in mind was battery monitoring/integrated chargers where the chip has a temp input that's suited for the big voltage change of NTCs, and an extra thermocouple converter, or RTD amp messes up compact space claims or ruin ultra low current requirements. Pretty niche, in that in most cases, just throw the bead on the side of the battery or something covers the needs.