Measuring Temperature of PCB Via

[willDavidson]

TL;DR Summary

I would like to measure the temperature of a PCB via

I am interested in measuring the temperature of a PCB via. Since vias are so small, I'm not sure how to go about this. Are there any methods for doing this?

If I use a thermocouple, I would need to adhere it to the via wall somehow. Let's just say my via is 0.5 mm? The via would probably be filled the past or whatever I used. I'd prefer not to use a thermocouple unless there's some clever way to do it which will allow me to accurately measure the via temperature.

The other way I thought about was with a thermal camera. I'm not too sure about this option either due to the small via size and the emissivity. I don't think I'll get a good reading inside of the via but maybe someone knows a good way to?

At the end of the day, I would like to compare the measured results with simulated and analytical results.

 

[berkeman]

Is the via tented with soldermask, or is it open?

We have had good luck with the FLIR ONE IR camera adapter for smartphones:

https://www.flir.com/flir-one/

We were mostly looking for hot spots, though, rather than using it for absolute temperature measurements. Have a look through the datasheet to see how close you can get to the object and what kind of accuracy you have when you select a region of the image (your via).

 

[Averagesupernova]

I don't see a good way except thermal imaging. Attaching a thermocouple to a small via on a circuit board will certainly draw enough heat away to cause an error in the reading. Are you concerned about heating due to the signal current through the via or are there external effects.

 

[willDavidson]

Is the via tented with soldermask, or is it open?

We have had good luck with the FLIR ONE IR camera adapter for smartphones:

https://www.flir.com/flir-one/

We were mostly looking for hot spots, though, rather than using it for absolute temperature measurements. Have a look through the datasheet to see how close you can get to the object and what kind of accuracy you have when you select a region of the image (your via).

The via is open.

Thanks for the suggestion on the FLIR ONE IR camera. I've tried the same camera as well as a handheld E-series from FLIR and I'm not able to get a good reading. To get an accurate reading, one of hte stipulations is that the object needs to be larger than the spot in the crosshairs. The minimum distance the camera can be from the object is 1.6 feet (0.5 m). At 1.6 ft, the via is a lot smaller than the spot in the crosshairs.

 

[willDavidson]

I don't see a good way except thermal imaging. Attaching a thermocouple to a small via on a circuit board will certainly draw enough heat away to cause an error in the reading. Are you concerned about heating due to the signal current through the via or are there external effects.

I guess you could say I'm worried about the heat due to the signal current in a sense but it's for power. It's used on a power plane to move current from one layer to another. I have multiple VIP and wanted to measure how changing different parameters affected the temperature. Parameters such as board thickness, via diameter, etc. I know that there are some general rule of thumbs out there. I could even do a basic calculation for heat transfer but I suspect those may be off from what's actually happening under certain circumstances since those are average values and do not consider current density.

 

[jrmichler]

You might be able to use fine wire thermocouples. As an undergrad about 1977, I spent some time hanging out in a physics lab. One grad student was attaching fine wire thermocouples to steel plates. My recollection is that he used a stereo microscope to place the wires. He charged up a capacitor connected to the wire being attached, then then touched the wire down. A microscopic ZAP and the wire was welded to the steel plate. The two wires of each thermocouple were placed as close together as he could put them. I think that approach has a good chance of making a good electrical connection to solder, but don't know about connecting to copper. Mechanical support needs to be carefully considered - those connections will not survive loose dangling wires.

Voltage drop across the via will be a problem that can be dealt with by using a data acquisition system to record temperature immediately after removing power. Omega sells thermocouple wire in sizes down to 0.003" diameter: https://assets.omega.com/pdf/cable-...nd-rtd-wire-and-cable/TFIR_CH_CI_CC_CY_AL.pdf. At that size, thermal conductivity should not measurably affect via temperature. Wire resistance needs to be allowed for.

 

[berkeman]

I've tried the same camera as well as a handheld E-series from FLIR and I'm not able to get a good reading. To get an accurate reading, one of hte stipulations is that the object needs to be larger than the spot in the crosshairs. The minimum distance the camera can be from the object is 1.6 feet (0.5 m). At 1.6 ft, the via is a lot smaller than the spot in the crosshairs.

Hmm. How big is the spot in the middle of the crosshairs at the minimum distance of 1.6 feet? If it's not too big, you may be able to place an IR magnifying lens between your PCB and the camera to enlarge the image.

https://www.google.com/search?client=firefox-b-1-d&q=IR+transparent+magnifying+glass

Another option might be to use a low-mass thermocouple and an oscilloscope to capture the contact transient when the thermocouple first makes contact with the via. That should give you a reasonable idea of the temperature right before contact.

BTW, have you done a sanity check yet with your fingertip? With a little experience you can get a reasonable idea of the temperature when you first touch it, and what it stabilizes too as you hold your fingertip on it. I use a similar technique for a quick check of transistor temperatures -- if I can't hold on to the case/heatsink, the transistor is running too hot. (Of course we use thermocouples for such measurements too).

Caveat -- keep in mind that some power transistor circuits involve AC Mains voltages and higher voltages, so you need to understand how they are connected in the circuit before feeling them to see if they are running hot. Don't ask me how I know this...

 

[Baluncore]

I am interested in measuring the temperature of a PCB via.

What rise in temperature do you expect?
Is there a critical temperature?
You could measure the voltage dropped across the via while monitoring the via circuit current. That will give you the resistance of the via, which is proportional to absolute temperature.
Calibration is not critical because you know the initial temperature of the via and board and the change is proportional.
If the initial temperature is 20°C, that is 273 + 20 = 293 kelvin.
If the resistance rises by 10%, that is (0.10 * 293) = 29.3°C rise = 49.3°C

 

[DaveE]

There are thermal indicating paints of various sorts. Some change color once at some threshold, others you illuminate and take a picture of (like fluorescence). But they sound like a pain to use to me.

 

[willDavidson]

You might be able to use fine wire thermocouples. As an undergrad about 1977, I spent some time hanging out in a physics lab. One grad student was attaching fine wire thermocouples to steel plates. My recollection is that he used a stereo microscope to place the wires. He charged up a capacitor connected to the wire being attached, then then touched the wire down. A microscopic ZAP and the wire was welded to the steel plate. The two wires of each thermocouple were placed as close together as he could put them. I think that approach has a good chance of making a good electrical connection to solder, but don't know about connecting to copper. Mechanical support needs to be carefully considered - those connections will not survive loose dangling wires.

Voltage drop across the via will be a problem that can be dealt with by using a data acquisition system to record temperature immediately after removing power. Omega sells thermocouple wire in sizes down to 0.003" diameter: https://assets.omega.com/pdf/cable-...nd-rtd-wire-and-cable/TFIR_CH_CI_CC_CY_AL.pdf. At that size, thermal conductivity should not measurably affect via temperature. Wire resistance needs to be allowed for.

Those are really thin! I'll order a few and see what I can do with those. I'm not sure I'll be able to attach it in a small via but I may just have to change my approach.

 

[willDavidson]

Hmm. How big is the spot in the middle of the crosshairs at the minimum distance of 1.6 feet? If it's not too big, you may be able to place an IR magnifying lens between your PCB and the camera to enlarge the image.

https://www.google.com/search?client=firefox-b-1-d&q=IR+transparent+magnifying+glass

Another option might be to use a low-mass thermocouple and an oscilloscope to capture the contact transient when the thermocouple first makes contact with the via. That should give you a reasonable idea of the temperature right before contact.

The via appears to be about 1/3rd the size of the dot. The IR magnifying lens is an interesting idea. I'll give Edmund Optics a call tomorrow to get more information so I can hopefully place an order.

jrmichler posted a link to some low-mass thermocouples above. I'm going to give that a shot too. The tricky part is going to be figuring out how to attach it to a via wall but it's worth a shot. What is a contact transient? Are you suggesting letting the via heat up and just breifly touch the thermocouple to the via wall and the initial reading would be the contact transient and would represent the temperature?

 

[berkeman]

What is a contact transient? Are you suggesting letting the via heat up and just breifly touch the thermocouple to the via wall and the initial reading would be the contact transient and would represent the temperature?

Yes, and the lower the mass of the thermocouple wires, the better. So maybe combine JR's references to very thin thermocouple wires with my contact transient capture to get a good idea of the temperature right at the contact instant. Especially if you can correlate the magnified FLIR image measurement with the contact 'scope capture, that will increase your confidence in the numbers.

Very interesting challenge.

 

[willDavidson]

What rise in temperature do you expect?
Is there a critical temperature?
You could measure the voltage dropped across the via while monitoring the via circuit current. That will give you the resistance of the via, which is proportional to absolute temperature.
Calibration is not critical because you know the initial temperature of the via and board and the change is proportional.
If the initial temperature is 20°C, that is 273 + 20 = 293 kelvin.
If the resistance rises by 10%, that is (0.10 * 293) = 29.3°C rise = 49.3°C

I'm planning to test different via configurations at different current levels so the temperature will vary. Measuring the via resistance might be an option. I guess I could measure the resistance of a via at room temperature (record temperature), then measure the temperature under different current levels and try to extract some information. I'll have a lot of data and I'll be trying to search for the one piece that sticks out, but hey it may show me something. Again, I'm mainly interested in finding localized hotspots along the via wall not necessarily the average temperature. But I think I may be able to extract something this way. At the very least, it gives me something to do lol.

 

[willDavidson]

Yes, and the lower the mass of the thermocouple wires, the better. So maybe combine JR's references to very thin thermocouple wires with my contact transient capture to get a good idea of the temperature right at the contact instant. Especially if you can correlate the magnified FLIR image measurement with the contact 'scope capture, that will increase your confidence in the numbers.

Very interesting challenge.

Just to clarify, when you say "contact transient", are you suggesting there may be a voltage transient that appears on my scope and that transient may correlate to the voltage? I just want to make sure I'm understanding the terms you're using and their context.

 

[berkeman]

So the voltage output of a thermocouple is proportional the temperature (I'm sure you know this already):

https://en.wikipedia.org/wiki/Thermocouple

And it seems like if the mass of the thermocouple wires is low (very thin wires), the voltage transient will rise to the approximate temperature of the via soon after contact, and will fall after that as the wires conduct some heat away from the via. But given the very thin wires that JR posted, that may be minimized so the peak temperature/voltage you read should be close to the actual temperature right before mechanical contact.

But please keep in mind that I haven't tried this, so that's why it would be good to correlate this measurement with the magnified IR camera measurement. Well, those and the finger thing too...

 

[Baluncore]

Again, I'm mainly interested in finding localized hotspots along the via wall not necessarily the average temperature.

The via is so small and the thermal time constant so short that I doubt you will need more than an average via temperature.

By passing a thin insulated wire through the via and soldering it to the back surface you can measure the voltage dropped along the via, referenced to the front surface.

Following a step change in current, you can also measure the thermal time constant of the via.

 

[eq1]

What is the expected temp, accuracy and what bandwidth does the measurement require?

For me, the cheapest and simplest way is one of the hand held IR thermometers from fluke. They're usually tuned to the emissivity of metal too.

https://www.fluke.com/en-us/learn/b...et-great-results-with-an-infrared-thermometer

If it has a 12:1 ratio then you could hold it 6mm above the 0.5mm via to get it's temperature. It will probably achieve like +-5C accuracy over 0C to 150C range.

 

[DaveE]

If you do use thermocouples that are touching the via, you may have trouble with emi in the meter they are attached to if there is a big , or fast AC signal on that trace. One of my favorite old TC meters back in the day was the one no one else liked because it was big, old, and kind of slow. But it gave the most reliable readings for SMPS temperatures. You'll know immediately when the temperature jumps when you turn the circuit on.

 

[berkeman]

For me, the cheapest and simplest way is one of the hand held IR thermometers from fluke. They're usually tuned to the emissivity of metal too.

https://www.fluke.com/en-us/learn/b...et-great-results-with-an-infrared-thermometer

If it has a 12:1 ratio then you could hold it 6mm above the 0.5mm via to get it's temperature. It will probably achieve like +-5C accuracy over 0C to 150C range.

We use that too. Keep in mind that the projected spot is not very well aligned at short range with the spot being monitored. For our Fluke IR thermometer, the spot being measured is a bit below the projected red dot. You can usually just scan in x and y directions slowly to zero in on getting the measuring spot over the hot spot.

 

[Rive]

I wonder if it could be done with some thermometer//bridge configuration.
It does not require high current so it'll not affect the measured value too much, but you need three other similar VIAs which are free.
And some extra work with calibrating the setup.

 

[willDavidson]

Thanks for the suggestions everyone. I'm going to order my test boards with different via sizes as well as low-mass thermocouples. The lenses are a little pricier, but I'm going to speak to the company and see if they think it may work for this application.

Here are a few images that I pulled from google which sort of show what I'm trying to view investigate. On the first image, they're not really showing current density but just imagine they are. That's the thermal calculation that most are using. The assumption is that current is uniform throughout the trace and the via. The second image is showing the actual current density of the via. If we average current over an area then we can estimate temperature as a function of average current, but it would be more accurately explained as a function of current density. That's why poor electrical connections lead to higher temperatures, right? So what I would like to do is view the difference in temperature throughout the via if possible. At the very minimum, it would be nice to look at several via sizes and see how the temperature changes.

I may have to go with larger vias for testing, get a good fit with simulation results, then rely on the validated sims for more complex designs which would include smaller vias.

Baluncore
I'm interested in more than the average temp because if PCBs degrade due to temperature, wouldn't minimizing local hotspots increase the PCB lifetime (only considering the thermal side of things)? I think that only considering average temperatures may be a shortcoming. I know it's easier and more practical though.

eq1
The PCB's Tg=180 C so I'm expecting this to be below 200 C. For the Fluke, wouldn't the laser have to be the same size or smaller than the via? I've only used those several feet away but it seems like the laser was quite a bit larger. Do you know if it's can get that small? I think this is going to be quite the challenge so I'm open to trying different methods. At the end of the day, whichever method words best and matches up closest with sims may be the one I have to keep.

DaveE
That's good to know about the EMI issue for future references. This will be DC so at least that's one less challenge I may have.

 

[willDavidson]

I wonder if it could be done with some thermometer//bridge configuration.
It does not require high current so it'll not affect the measured value too much, but you need three other similar VIAs which are free.
And some extra work with calibrating the setup.

Do you have more information on this? I can definitely have free vias. Actually, my goal at first is to just use a single via so the task is simplified and I can actually get something from this analytically. Also, so my simulations don't take a month to complete before it tells me its out of memory (exaggerating). I do plan to an array of vias and would like to measure the difference between them.

 

[Rive]

Do you have more information on this?

Not really, and I'm actually not familiar with bridges either. But as far as I know it's quite a common solution.

If you have several similar VIAs then you can make a bridge out of them (with independent voltage domain from the actual current of the VIA) and sense the resistance change caused by the higher temperature.

 

[eq1]

Whoops. This is the same solution proposed by Baluncore in post #8. Obviously, I totally endorse this method. :)Maybe you can measure it indirectly by looking at the change of voltage on the via itself for a constant current. An instrumentation amplifier can likely achieve this without much trouble.

So for example, I usually use the rule of thumb of 1.5mΩ per plated through hole via. [1] Put current through it before it heats and at room temp and there should be a 1.5mV/Adc drop on the via, totally measurable for currents bigger than say, 100mAdc.

Now let it heat in the application. Resistance should increase 0.386%/C [2] (assuming copper plated via, etc. etc.) If you expect it to heat to 180C (which is really hot for a typical PCB) then the via resistance should increase as Rf = Rn(1+a∆T)=1.5mΩ(1+(180-20)3.86E-3)=2.54mΩ or ~2.5mV/A and that delta should be very measurable.

Not the most accurate way to get temperature, and of course you'll have to watch for things like the common mode on the instrumentation amp (or float it), but it is cheap, easy and fast, (basically one afternoon of work) and could potentially be done in application.

[1] https://www.dataweek.co.za/59129n (a more complete calculation which shows all the various things one needs to consider, but since you're looking at the change in resistance you basically only care about what metal the via is made from)
[2] http://hyperphysics.phy-astr.gsu.edu/hbase/Tables/rstiv.html

 

[willDavidson]

Not really, and I'm actually not familiar with bridges either. But as far as I know it's quite a common solution.

If you have several similar VIAs then you can make a bridge out of them (with independent voltage domain from the actual current of the VIA) and sense the resistance change caused by the higher temperature.

Ok I'll look into this too. I wasn't sure if there was an easy (commonly used) method to do exactly what I'm trying to do so it was worth a shot. The more options the better.

 

[berkeman]

With the wonders of Internet cookies and Amazon and Facebook, an advertisement for this just showed up in my Facebook feed (presumably from a combination of my recent searches):

https://www.fluke.com/en-us/product...IMpIWFQu5VQnXhpHW0hdm9wKsfk9HR1XcdyZdhJp3f0ws

 

[Averagesupernova]

A bit off topic, but that panel looks like someone like me that is slightly OCD wired it. Those are some very neatly arranged wires and connections.

 

[Tom.G]

A bit off topic, but that panel looks like someone like me that is slightly OCD wired it. Those are some very neatly arranged wires and connections.

Not terribly unusual in industrial equipment, it makes maintenance A LOT easier; house wiring, yeah, to the point of non-existence.

 

[Averagesupernova]

Yeah I've noticed that industrial is always more neatly done. I've seen some built up control panels that look like art. And the types of panels I'm referring to are one offs. Not like someone had several hundred of the same to practice on.

 

[DaveE]

A bit off topic, but that panel looks like someone like me that is slightly OCD wired it. Those are some very neatly arranged wires and connections.

Not OCD in our equipment. If you were a technician in our Lab, back in the day, this is what I would want you to do. It's more inspectable and servicable, and it isn't really any harder to do it right once you are used to doing it.

 

[Rive]

With the wonders of Internet cookies and Amazon and Facebook, an advertisement for this just showed up in my Facebook feed...

They are so nice with you. Guess what started to come up regularly (without any previous search) when I passed forty

Yeah I've noticed that industrial is always more neatly done.

It's not about industry. Any tech who thinks he will have a second time with that panel will do it like this.
Well, it's a bit contradictory since because they do it like this it's unlikely to have a second time within a decade or so...