Problem with USB connector on bladeRF part 2: The Fix

In the previous post I described the problem with the USB connector on my bladeRF. Now it's time to describe how I attempted to fix it.The first attempt was successful only partially, I had more luck the second time, but I will describe it in another post.

 I had two options of bringing the board back to life: To request RMA, send the board back to Nuand and wait several weeks till it will be fixed, or try to fix it myself. Having all the necessary tools for SMD rework on hand, and experience from previous projects, I decided to do the job myself. Since it is quite likely that others will also experience similar problems with USB3 connectors, I decided to document the process and describe on the blog. Recognizing that not every hardware hacker will have on the bench specialized tools for SMD rework, I wanted to use only basic ones, bringing in more advanced instruments only when it was absolutely necessary. I started with the absolute minimum: temperature controlled soldering iron with chisel tip and tweezers.

Replacement of the USB3 connector is not a simple task. If you have never worked with complex SMD components, or do not feel fully confident that you can carry the job yourself, contact Nuand to request repair, or find someone who will do it for you. Paying someone will be cheaper than the new board!

I have been informed that Nuand no longer offers  connector replacement service. Have fun!

The first step in fixing the board was to order new connector. Although I don't mind buying parts on ebay, for a  job where quality of the part is critical I prefer to place the order at a more reputable distributors, what usually means that I shop on Digikey.
After a long time of browsing the inventory and reading datasheets, I decided to use a USB3.1 connector. It is mechanically compatible with USB3.0, but should be much more robust: instead of small indents in the metal frame, the plastic insert is held in place by metal flaps shielding the contacts from the back.

The fist step in fixing the board was to remove the metal frame from the board. The operation seemed to be very simple: heat up the "wings" on the sides of the frame, lift it with tweezers and done. Unfortunately the life turned out to be more complicated. The frame is held on the board by 6 solder pads: two on the sides, under the metal "wings", and 4 under the frame. This made the removal a bit more difficult: I had to heat up the entire frame and keep the solder underneath melted until the frame would be removed. I ended up partially filling the inside with solder and keeping it melted until I could remove the frame.

Solder pads under the connector.

Metal frame partially filled with solder.

Next, the pads were cleaned form the old solder using solder wick and prepared for soldering. At this point it was obvious that  the soldering iron would not be sufficient to properly install the connector, so I reached for my hot-air station. There are many techniques of soldering SMD components. I'm pretty sure that you, as an experienced hardware hacker, also have your own. My favorite method of soldering is to leave small blobs of solder on the pads and then heat the board with a hot air gun. CuriousInventor did an excellent job documenting the process in one of his videos:

Since the hot air can heat the adjacent components and desolder them, I insulated the outside of the working area with aluminum foil. Additionally, I affixed a small thermocouple in the back of the board and closely monitored the temperature during the entire process.

Board protected with aluminum foil.

Thermocouple in the back of the board to monitor temperature.

The pads were cleaned from solder with solder wick, and then the remainings of flux I used to desolder the frame were removed with isopropyl  alcohol. The board was ready for the new connector.

The next step was to prepare the connector for soldering. To make sure that the solder on the board will stick to the frame. And here came the first problem. While on the "wings" molten solder had no problem to wet the metal, the middle of the frame was a completely different story. It took me a lot of time and a lot of flux to tin the entire bottom of the frame. The difference between the part which accepted the solder, and the part difficult to tin was readily visible; The "wings" were of yellowish color, while the rest of the frame looked silver. It must have been plated with some kind of metal, which did not accept solder almost at all (my guess is that it was nickel).  The picture below shows the connector partially tinned. As you can see, the wing on the right is fully covered with solder, while the rest of the frame is not touched. This shows very well the boundary between the part which was activated for soldering, and the part which was not.
As strange as it is, only one of five connectors I've got from Digikey had this problem. The other 4 had the entire bottom side prepared for soldering.

Partially solder-coated bottom side of the new connector.

After the entire bottom part of the frame was tinned, I soaked the entire part in isopropyl alcohol to wash off excess of the flux. Now I could proceed to mount the connector on the board.
As instructed in the video, I left a small "bump" of solder on each of the pads, applied fresh flux to the board, and placed the connector aligning the leads to the pads.

Soldering the connector to the board turned out to be more challenging than I expected. To create a proper bond between the board and the part, the solder on both surfaces must be melted. This means that I had to heat up the board and the bottom of the part to solder melting temperature. While heating up the board does not look difficult, heating the part is more complicated. Since the solder is at the bottom surface and the hot air blows on top, I had to heat up the top of the part to much higher temperature than necessary to melt the solder.
I wrote that heating up the board did not look difficult, because there was one big problem hiding inside the board, making rework bigger challenge that it seemed to be: power planes on the inner layers of the PCB. The planes are layers of solid copper, delivering power to the components, and providing RF shielding for the circuits and components installed on both sides of the board. Copper conducts heat very well, and the planes acted as heat sinks, distributing all over the board the heat delivered to the solder point.
In the end the entire operation of soldering the connector took over 2 minutes. During this time I had to make sure that the connector and the board were heating up evenly, and keep the connector hot.

Even though I tried to remove the flux from the connector, it was difficult to clean it completely. In addition to that, the solder left on the frame, and the solder from the pads lifted the connector above the surface of the board enough, that the leads did not make contact with the solder pads.
After adding more solder to the leads the board was able to communicate over USB2, but USB3 connection still did not work. I suspect that heating the connector while tinning, and then during reflow distorted the plastic frame, and the connection of the USB3 lines did not contact well enough to provide stable ultra-speed connection.