Lead free component reflowed with leaded solder paste

[Niklas]

Presently i'm investigating an issue with broken soldering joints on a PCB that has been thermo cycle tested. I think I have found a possible root cause to the issue, but I lack someone to ask for a second opinion and look with "fresh eyes and mind".

Automotive application
-40'C to +85'C ambient and locally up to 110'C on the PCB surface
Lead free components, mostly with matte Tin finish on the solder terminals
Leaded solder paste, SnPb 60/40 or SnPbAg 60/37/3 (Demand from some customers)
SMD packages 0402, 0603, 1206, SMB, SMC, SOT23, QFP44, pin header 2x2
Dual side assembly, ie two runs through the assembly and soldering line
6 layer PCB, FR4, gold plating on all solder pads, RoHS compatible
All copper plane solder pads have thermal relief
Peak temperature during reflow is just reaching 230'C for a single panel

My idea, so far, is that the peak temperature during reflow is a bit too low. As pure Tin melts at 232'C, the surface material on the components is not melted by the solder paste, thus they will not form a fully mixed soldering joint. The two different metal alloys in the joints will have different thermal properties and cause mechanical stresses and cracks during thermal cycling.

We have also seen signs of poor wetting on SMC diode packages where in one case only half of the soldering terminal had a nice solder slope. Oxidation due to unproper storage conditions has been discussed but the actual shelf time at the assembly house is less than a month and in a sealed plastic bag. The SMC package is on of the larger components on the PCB, so it could be related to temperature.

Both Murata and Diodes points out in datasheets and appnotes that Matte Tin finished components must be soldered with a peak temperature above the melting point of Tin (232'C). Our PCB assembler claims that they can use approx 225'C and compensate the lower temperature with longer time above the liquidous temperature of the solder paste (183'C). Personally I doubt that and my gut feeling is that they are clueless and are in a state of total denial.

All kinds of input, both personal reflections and/or links to relevant documents are highly appreciated. Is soldering temperature the most likely explanation or is there something else I have missed totally and need to check up?

[JoeyP]

There are many variables to consider here. Which component types had the fractured joints?

Temperature profile could be an  issue, but I'm skeptical that you need to get above melting point of tin for proper soldering. I've done hundreds of lead-free BGA packages using leaded solder. The profile only got to about 220C peak (necessary to protect some temp-sensitive components on the PCB). We've had zero failures resulting from fractured joints. Granted, the product isn't exposed to the extreme temp cycling that you describe.

It's also worth looking at how they temp profiled the board. Which components did they measure? In general, smaller components will have a significantly higher temp rise, so if they profiled larger components to 225C, the smaller components would likely reach temperatures well above 232C. If those smaller components are having fractures, then look somewhere besides temperature.

The poor wetting would make me suspect the solder paste (assuming you could be absolutely sure the components weren't oxidized before assembly). I would try a different brand.  BTW, hope you're not using no-clean for this application, as it has the least aggressive flux, and so is most prone to wetting issues. Water soluble types work best in my experience.

[calexanian]

Generally the leaded solder will have better stress characteristics and be less picky about flow temp. That being said any shop should be able to make a satisfactory joint with any temp range solder, or I might be looking for a new assembly house, or that particular solder is not applicable to your use. As far as adding leaded solder to a previously lead free joint you should have no problems. Just remember to deflux. You will still have your ROHS compliance issues though. You gotta get it right to begin with. Also avoid any Chip Quick or rework type solders as they are notoriously brittle. We just had a meeting about that here this morning that we are going to refrain from letting boards go out with chip quick remaining. We are going to wick it all off before re installing the parts and do the new joint with our standard solder.

[Niklas]

Thanks for the fast responses.
JoeyP: The fractures are mainly on a pin header, ie one of the larger components on the PCB. But we have also noticed that some ferrite beads and other small components (<=0805 package) can be peeled of using little force. So far the testing has only involved thermo shock, dry heat etc and no vibrations. One of the solder pastes, Multicore something if I remember correct, has brownish flux residues. The other one has transparent residues. Both are considered no-clean.

calexanian: RoHS compliance is not an issue as we have an exempt, but we are probably moving towards lead free soon. The joints look ok visually from the outside, all with the shiny and smooth appearance. However, inside the cracked solderings on the connectors the color is more dull grey with a somewhat crystalline structure. If you compare the color and appearance to a tinned multistranded wire that is cut, there is a large difference. We don't see the same homogenous appearance.

[calexanian]

Sounds like they are not letting it soak at flow temp long enough, or hot enough. Bad oven profile.

[woodchips]

Try a look at Optimising Quality in Electronics Assembly by Smith and Whitehall, lots of good stuff there.

My experience is rather out of date, and I only use lead solder forvreliability. But, are you mixing unleaded and leaded solder, apparently using an iron on leaded will ruin any future unleaded joints you make due to the lead percentage in these joints being too low. My guess about the peak temperature being higher than the tin melting point is simply to flow off any oxidation, all this unleaded solder and components seem to have a life measured in days. Why would a longer dwell time compensate for inadequate temperature? Seems daft to me, heat it for a week and use boiling water?

Stay leaded! Remember the people who fix their busted motherboards by cooking them in an oven to remakevall the soldered joints!

[JoeyP]

But we have also noticed that some ferrite beads and other small components (<=0805 package) can be peeled of using little force.

That's a really bad sign! Those parts would likely be the ones which reached the highest temps during reflow, so I would look elsewhere. Maybe the PCBs are crap. Maybe the solder is crap. Take a look at some bare boards if you have any left, and see what the plating looks like. See if you can scratch it off. Also, use a paste with water soluble flux (even if just for test).

There is another temp-related issue to consider besides peak. If the temp ramps up too slowly, it can cook the flux too much before it reaches a temperature where it's active. When it gets to the temperature where it's active, it's already "spent" so it doesn't do much. That's a bit counter-intuitive because from a component stress point of view there's the temptation to ramp slowly to protect the parts. Has the CM provided any measured temp profile graphs?

The thing to realize about no-clean flux, is that it's an engineering compromise between needing to be active enough to do the job, but not remain conductive after reflow. That's a big challenge for the flux, and the result is that they tend to be "just" active enough - and sometimes (as possibly in this case), just not active enough. Water soluble can be very active because it will be washed away after. We refuse to use no-clean in our assembly process.

[Niklas]

We have got some temperature probe measurements from a production panel, about 15x20 cm (6x8 inch), with peak readings around 225-228'C. The assembly house also mentioned that the actual soldering temperature can be about 5'C lower during a production run due to the increased mass to heat up. Unfortunately we are not located close to the assembly house otherwise I would already have gone there to have a look.

I have inspected a bare production panel, without any components, and the gold surface treatment look solid. It won't scrape of and the surface is not wrinkled.

Close up inspection of the cracked joint under the ferrite bead showed that the crack seems to be in the interface layer just between the solder and the component terminal. The copper pad is still on the PCB and most of the tin plating on the component still look intact.

I will ask my boss to buy the Optimising Quality book first thing in morning tomorrow, perhaps the weekend is "saved" :-)

[woodchips]

Have you heard of the intermetallic bond? This is layer between the solder and base metal formed at higher temperatures, ie those now generated by lead free solder. The layer is brittle and might explain the breaking away of the components.

Another gem from the book is that over 320C the solder will adhere to, but not wet, the components.

Sure the book was recommended by Bob Pease in one of his columns?

[kayvee]

I assume you have been given the complete reflow temperature profile from the board assembler.

Compare this against the recommended profile from the paste manufacturer for the specific paste used to verify it is within the recommended process limits along the graph.

[qno]

When its an automotive application it does not need to be ROHS.
All the big car manufacturers in Europe specify Leaded solder past.
Due to reliability.

[Niklas]

We have now received the soldering profile, datasheets for the solder pastes and the temperature calibration protocol for the reflow oven. There seems to be some issues about the temperatures in combination with the solder paste. Time above melting point for the solder paste is too long, about double that in the datasheet, with the risk of solderballing and evaporating the flux.

The reason that we have not converted fully to RoHS is fears among some of our customer. Personally I think that some customers actually don't have data that is up to date but instead relies on previous problems, even years back in time. Just mention lead free solder and they start to pack their stuff. Other customers, mainly those with own knowledge of things, are easier to discuss with because they can asses things more easily.