Thermal Camera for PWBA assembly inspection

[schmitt trigger]

We produce an automotive board assembly, to which a connector, with significant thermal mass, is soldered to after all the other components are in place.
This is an electronic throttle control, and thus considered a safety assembly.

This is quite a large volume assembly, 500K/year, and it is processed via a robotic soldering machine with a single preheating stage. 
The solder is SAC305. Thus a camera with a +350C rating should be sufficient.

Whoever chose the robot before my time here, did not take into account the quite different thermal masses on each soldered joint. The result is uneven solder joints. Bad news for for a safety related assembly. We have played around with temperatures and solder tip times, but are essentially just guessing,

We have tweaked and tweaked the profiles, but cannot get reliable joints all the time.

Rather than keep guessing at possible solutions (which could include another robot with a different profile from the first one), my thinking is that an IR camera would help me visualize my problem.


The board itself is about 4cm by 7 cm.
Does my idea make sense? Any camera suggestions?

[Spirit532]

An IR camera would definitely help with visualising the issue, but I would not consider installing one permanently, as it is quite an expensive device(machine vision thermal starts at ~$1-2k and goes up ad infinitum).
I suggest renting a good mid to high-end camera for a few days to find out what your problem might be - I suspect it's uneven heat distribution or wrong heating time. A FLIR E60 with the 0-650C range option, or a FLIR T640, with the same 0-650C range should be okay for this job. They both focus relatively close and let you stream(record) video to PC over USB(with a free 30-day trial of FLIR ResearchIR software - I recommend it over FLIR Tools+) for later analysis.
Record a dozen boards being soldered, mark which ones were bad and which ones were good and compare the two on thermal.

If you're able, posting the videos here could also help, it's an electronics engineering forum after all.

[Fraser]

Please be aware that solder, being very shiny metal, has very poor Emissivity. As such imaging it can be challenging and is not normally recommended if accuracy is required. Thermal Imaging of the PCB substraight may be possible if the Emissivity of that material is reasonable. FR4 with a solder mask should be OK.

Back to the solder Emissivity..... normally we would apply a paint or tape to a shiny metal surface to increase its Emissivity. Molten solder presents a challenge. The cameras Emissivity compensation can be set to very small numbers but it is generally inadvisable to trust readings taken on very low Emissivity targets if any sort of accuracy is required. The Emissivity of shiny lead is 0.08. In a case such as yours where a process needs to be monitored in order to set the correct pre-heat and solder tip temperature, it would be advisable to install suitable thermocouple sensors and real time data logger on a test board for a batch run. That board is sacrificial but you can keep testing it until you are satisfied with the soldering. The data collected by the thermocouples may then be used to understand the pre-heating and soldering stages. The thermocouples only need to be employed again if the PCB type or materials change.

Thermal profiling an automated soldering process is core to reliable production so surely other electronics fabricators suffer the same challenges ? Is there no organisation or company that can offer you advice on this production challenge ?

Thermal cameras are great (I am a great fan of them) but sometimes they are limited by the laws of physics. This appears to be such an occasion. They are very useful in thermal profiling of a populated and running PCB in order to determine potential hot spots or components that require greater heat-sinking. That is very different to monitoring a soldering process however.

Fraser

[Vipitis]

You want to visualize the problem to understand it better.

Best option is to rent a camera for a few days, or even better. Hire a professional with a camera who can use it and help you analyze the data. I am not sure if some are trained for soldiering processes, but chances are - that others had a similar issue and might be able to share.

A thermal camera on for computer vision and automation will have a hard time working with shiny liquid metals.

[Cat]

The cat got curious and gave it a try on a difficult to solder PCB (to justify spending more $ in thermal imagers  ).
Some observations:
epsilon = 0,08 for molten solder gets a perfect reading of my soldering temperature, but molten solder on top does not show the quality of the joint.
The solder mask is of course reflecting the hot heating element.

One possible solution could be monitoring the temperature of the FR4 (or other Substrate) near the solder joint. Like the others said, probably a test run with thermal images of every PCB is needed to get a rough estimation what readings are needed for a good solder joint.

I don't know about your robots, but it might be easier to use a different tip, temperatures and solder time to get a reliable result.

[Gyro]

That's a TS100 isn't it? 

[Cat]

That's a TS100 isn't it? 

It isn't 
Its a JBC with a C245 series cartridge.

[schmitt trigger]

Wow, thanks for the prompt and informative replies. Very useful info here.
It seems that there are plenty of cat lovers, so I'm sharing a photo of my cat relaxing on freshly cut grass 

Anyways, back to the problem being discussed:
-I was aware of the significant emmissivity differences between the substrate (ceramic) and the solder itself. But you have properly alerted me on their huge differences.
-Anyways, this test is to be qualitative more than quantitative; I would like to see how the heat flows from the tip to the piece being soldered. As mentioned, there are some points which always solder fine....I would like to compare those against the others that give difficulties.
-I like the idea of a sacrificial board with thermocouples. My concern would be that the thermocouple itself would drain some heat of the surfaces being soldered and thus provide a wrong result.
-Love the idea of renting a camera. If I were to spend over $3k, I would have to write and justify a capital spending request, which could be daunting and take a long time.

I believe that I now have new fresh ideas where to steer my investigation. I thank everyone who responded, and certainly after I perform my experiments, I will share some results.