Back in 2013 I was given a very nice rackmount RAID server, with 8 1TB drives. It has an Areca ARC-1220 8-port PCI-Express SATA RAID Controller, which turned out to be completely dead. Bummer.
It took a while to find a replacement identical card. Eventually one turned up on ebay, claimed to be working. I bought it. It arrived, and turned out to have one of the SATA connectors broken. Fortunately once you knew what to look for in the seller's listing pics, it could be seen to be broken there. Subtle but clear enough to demand a refund, which I got. And also got to keep the broken card.
The obvious solution was to swap a connector from my dead card, to the new one, which did run.
Only problem was, the connectors have two rows of surface-mount solder joins to the PCB, and one of the rows is well underneath the connector. It was obviously assembled in a vapor phase reflow process. Presenting quite an obstacle to hand repair. So I didn't tackle it for a long while, just pushed it down the projects queue.
It finally bubbled up to the top again, and now it's done and works. Here's how.
The beast.
The bought card, as it arrived. Broken connector looks obvious, but the seller's pic was small and had no light reflecting off the gold pins.
Closeup of the connector. For a while I considered trying graft on a bit of plastic, with superglue. But I knew it wouldn't last.
Dead card on left, bad connector one on right. First step, remove the piezo buzzers that were in the way. I figured I had four goes at getting a connector off the dead one, in usable condition.
First one off, after a LOT of struggle. It didn't go well, and the connector body was too melted and distorted to use. Part of the problem is that power plane two of the pins attach too.
First attempt on left. Ruined. You can't see, but the SATA connector tabs are distorted.
2nd try on the right. Hmm... I can work with this.
Btw, does anyone know who makes these connectors, and where I can buy a few?
I might still be able to get the dead card going, if it turns out to be just a lost data fault in the flash ROM.
I found I could pull out the pins on the outside row, and get them back in again. Making it possible to solder the inner row of pins with the outer row removed. Yay.
The first attempt to do that using an iron and solder wire failed. Too tight a space to work in. So I went and bought some solder paste. Not the first time, I just didn't have any atm.
Solder pasting. Probably too much.
Inner row done. This is after a lot of cleanup, since there were a million small solder balls adhering around the joints, especially behind them. Some reachable with a dental pick, but mostly just blasting them with a syringe full of IPA, over a tray. The tape in this pic was to try and avoid washing the balls into crevices around other components. I cycled between IPA jet soaking, scrubbing with a pointy stiff bristle paint brush, and blowing out with compressed air, then trying to peer into the gap under the socket with a binocular inspection microscope (really poor visibility in there.)
First three of the outer contacts back in place. These are the 'long' ones, or were supposed to be. SATA has very slightly longer contacts for the 3 ground pins, compared to the 4 signal pins. By eye it's hard to tell the difference, and in fact I'd goofed, in that the leftmost of these turns out to be a 'short' one. Despite carefully comparing them and keeping in two separate tiny ziplock bags... Oh well.
Done. After soldering the pins I'd noticed that 2nd from the left one, that won't push in all the way. Because it's a 'long' one, in a plastic channel for a short one. Doesn't matter, the different lengths are to make ground contact before signals, for hot plugging. And I'm never going to hot plug these internal connectors.
After they are in place, and another cleaning out of solder balls, I added a little super glue since the barbs that held them in the plastic originally have lost much of their their grip. After the glue dried I made sure to clean off the contact faces again since superglue can deposit a film on nearby surfaces through evaporation and condensation.
So then, big moment. Did the board work now? It's had a LOT of handling while doing this, and did I avoid static zapping it, or lodging solder bits between anything?
Plugged it in, started the system. The RAID card inserts its initialization stage into the BIOS startup, and it has its own status display, plus good/bad LEDs on the 8 drive trays.
It started up, scans the drives (further than I'd had the machine before) then complains there's a problem with drive 3 and 4. "Disconnected"...
Dammit, 3 & 4 are on the SATA connector next to the one I replaced. It was late in the evening. I tried some swapping - drives, cables, the two whole drive bays, and just managed to confuse myself. The problem was intermittent and variable, and I wasn't able to isolate it to a particular module. It didn't seem likely to be the card, but....
Getting depressed, I gave up and went to bed.
Next morning was better. A few more swaps and I was sure the problem was with the left hand bay of 4 drives, and specifically the bottom two slots. But symptoms were variable.
So I pulled the whole bay out, to have a look at the rear PCB for anything obvious.
And obvious it was. Sometime in this machine's history, someone had removed the 4 screws I've marked in the pic with red dots. Result: the board flexed enough that the bottom two drive SATA and power connectors made intermittent contact. I replaced the screws and reassembled the machine.
Now it works! All drives present, and after some grumbling the RAID BIOS presents four RAID arrays of a bit under 2TB each. I don't care what they are, as I'll be restructuring and formatting them. And a few other hacks to the machine too.
This machine will be archival storage, and USB-3 server for the nice microscope video camera I bought earlier this year.