Why no gold on some BGAs on Iphone PCB?

[mikeselectricstuff]

I got this Iphone PCB from Scotty (the "I built my own iphone guy") at Supercon - does anyone know why some BGA footprints are OSP and not gold ?
By best guess is it's to do with the metallurgy of the solder balls used on some of the devices

[Howardlong]

Anything to do with the underfill process?

[Howardlong]

I keep seeing a phenomenon known as “black pad syndrome” coming up.

[trophosphere]

Don't know if it applies but I know that ENIG has a higher power loss per unit length compared to other surface finishes in high frequency traces especially when such traces are used as differential pairs. This can result in significant signal distortion.

Surface Finish Effects on High-Speed Signal Degradation

[Someone]

Its for ultimate metallurgical compatibility with specific lead free alloys, increased mechanical strength too.

[EEVblog]

Gold can in some cases cause embrittlement when used in combination with certain alloy pins/balls.
If the device has that issue for whatever reason then you would not gold flash those pads.

[coppice]

Getting rid of gold, and avoiding potential embrittlement issues, is generally a good thing. What I find odd about the board is that having used an OSP process, they didn't apply it to most of the parts, especially the other substantial BGA package.

[Dago]

Gold can in some cases cause embrittlement when used in combination with certain alloy pins/balls.
If the device has that issue for whatever reason then you would not gold flash those pads.

Gold embrittlement requires quite a lot of gold in the joint to happen and ENIG doesn't have nearly enough gold (the coating is extremely thin), but seems it is not quite so simple: http://www.circuitnet.com/experts/87487.html

[coppice]

Gold can in some cases cause embrittlement when used in combination with certain alloy pins/balls.
If the device has that issue for whatever reason then you would not gold flash those pads.

Gold embrittlement requires quite a lot of gold in the joint to happen and ENIG doesn't have nearly enough gold (the coating is extremely thin), but seems it is not quite so simple: http://www.circuitnet.com/experts/87487.html

The common used figure is you are OK if there is less than 3% gold mixed with the solder. However, you need to consider whether the solder mixes well, or not. When I had experience of one gold embrittlement related reliability issue with a large surface mounted device, a metallurgist sectioned some example joints. The gold had mostly stayed in one part of the joint, and a crack had formed through that part of the joint. So, checking for the gold being greater than 3% of the size of the solder blob doesn't tell you the whole story. One of the notes on the page you referenced touches on this issue of how well things mix, but only in terms of having sufficient turbulence inside through holes to allow full mixing of the gold.

I am always amazed that BGAs stay put on their solder joints as well as they do. It requires a really good match in expansion rates between the package of a large device and the PCB to avoid massive stresses as the temperature varies, especially if its a high power device that might heat up much faster than the PCB it is on.

[David Hess]

I keep seeing a phenomenon known as “black pad syndrome” coming up.

That may be exactly what I witnessed in one Tektronix 7A26 vertical amplifier which I repaired.

Tin based solder will become brittle if the gold content exceeds about 5%.  If the gold plating is thin enough, and it is deliberately made so, then it will completely dissolve into the solder while keeping the percentage of gold low.  But if the gold does not completely dissolve, then the interface between the gold and tin based solder will diffuse over time increasing the concentration of gold in the solder making the solder brittle along the interface.

This particular Tektronix 7A26 had heavy gold plating and several solder joints had failed.  Removing the solder revealed a dark purplish-brown discoloration which I took for Purple of Cassius which is gold suspended in tin dioxide.  New solder did not adhere to the surfaces with "black pad syndrome".  My solution was to use my soldering iron at lower temperature to protect the substrate, which I believe was polysulfone, to add and remove tin-lead solder until the gold was dissolved and removed which is the same procedure recommended by NASA for soldering gold plated substrates and leads.

[HalFET]

Depends on how the chip is bonded to the PCB, not all finishes are compatible with all processes.

Also OSP is actually pretty damn good these days, we started using it as standard for prototyping, additionally you need no obnoxious chemicals other than a short micro-etch pre-dip after resist stripping. The main issue you see with a lot of OSPs is that you can't really put them in a spray installation (they do exist though), but you need an "immersion" system instead. But overall ENIG has a few issues, the nickel plating is quite porous and doesn't really protect the copper underneath that well. And the thin gold flash basically does nothing, and if you have any strong acids in your plating bath (and yes, you do) these can corrode the nickel and copper underneath. If you want to see how "good" the ENIG plating is, dump a piece of scrap PCB in some sulphuric acid and make a cut-through afterwards, you'll see some interesting things happening. But BGA balls are rather small, so that thin 100 nm gold flash might contribute a considerable portion. (9 micron copper +  1 micron nickel + 0.1 micron gold) If you then consider the solder will normally leach away all the gold on the pads things ain't pretty.  Additionally you get to play with cyanide to do ENIG plating, so there's also that little problem.

The main reason to still use ENIG is that it's very flat and very cheap because everyone in Shenzen and their mother does it, OSP often involves manual processing due to the above caveats. Many board houses don't get enough OSP volume to invest in a line for it, so it's still done in trays or even beakers. ENIG on the other hand it's a case of stick the panel in the line and wait until it comes out. Process control is also an important factor to making ENIG work, any mistake and any impurity in the plating solutions will cause solder joint defects later on. And lets just say, you haven't lived until you've been in a backroom PCB manufacturing line in rural China; and if you have been there it'll shorten said life by a few months most likely. I recommend a good pair of acid resistant safety shoes/boots if you do wish to experience it, not that it's worth recommending in the first place.

But you can find many overviews on these things, I think there's a pretty good description in Coombs' Printed Circuit Handbook. And enough overview tables online, like this one preaching ENEPIG: http://www.epectec.com/pcb/enepig-boards.html The main reason to like ENEPIG is that you can pronounce it like "and a pig" during conferences if you butcher it a bit.

[NF6X]

I'm curious about how selective ENIG/OSP like this is accomplished. I was under the impression that both are applied with all-or-none submerged processes. I guess that additional masking steps are needed to do this?

[HalFET]

I'm curious about how selective ENIG/OSP like this is accomplished. I was under the impression that both are applied with all-or-none submerged processes. I guess that additional masking steps are needed to do this?

I suspect the OSP is on top of the ENIG as well, no point in trying to mask for OSP. The pre-dip/micro-etch to remove the oxide isn't that aggressive, so ENIG should protect it.

But to protect pads during plating: peelable soldermask or photo resist are the common methods. Depends a bit on the exact plating solution, feature size, and the required exposure time.  Peelable soldermask is screen printed and has a somewhat limited resolution as a result, dry film resist on the other hand tends to be a bit more fragile and is useless against caustic substances. (Usually stripped using NaOH or NaCO3.)

Peelable soldermask looks like this:


*EDIT*
If you want to see what a manufacturer really did you pretty much need to see the panel border, which they rarely - if ever - send to customers. The reason for this is quite simple, the panel border is sort of the "overflow" or clamping area depending on the processes used. It also contains the verification structures and so on. I have a few complete panels here in the office, so to show you why they will rarely if every send it along unless you're a "good" customer see the attachments.