When I get a little time, I like to go through the “donation” pile and do some testing. As it turns out, I was gifted a set of four Cooler Master Xcraft External Hard Drive Enclosures along with matching power supplies and cables. These enclosures are good for 3.5″ IDE drives, and seeing as IDE ports won’t be any easier to come across in the future, are useful if just to have another method of accessing IDE devices.
Of course, in the case of hard drives, using an IDE to SATA adapter and a modern bridge or SATA port is theoretically possible, although in my experience, such methods are a little finnicky for a few reasons – one of which is the need to match power plugs for the drive and adapter, and another to ensure the polarity of the plugs are correct. Even then, you might encounter strange timing bugs which means that it just won’t detect. That method certainly does not work for ATAPI devices such as optical drives.
Of course, if you still have a computer with a PCI slot, an IDE controller card is still an option, and older machines with physical motherboard-integrated IDE ports an even better option.
Unfortunately, it turned out these units were troublesome. A drive could be fitted, but nothing happened. The problem? A classic. Bad power supplies.
Of course, with four enclosures, there were four power supplies. Testing the power supplies with the Tektronix PA1000 showed a healthy, if slightly thirsty, draw of about 1-2W at idle on three of the four supplies. The final supply had absolutely zero draw from the mains, so it’s probably a little more further-gone than the others.
Plugging the three supplies which worked into enclosures with drives and powering them on resulted in the front light coming on red, but otherwise no other action. No spin-up, no detection. However, close observation of the power supply showed a slight periodic dipping of the green power LED and a very quiet chirp/squeak. This is a good sign that the power supply is “collapsing under load” and restarting. Chief culprit? Bad capacitors. Again. You know how much I enjoy these …
Power Supply 1 and 2 in the Operating Theatre
I faced my first hurdle as soon as I decided to try repairing the units – there were no screws and no obvious way in. The case would have to be cracked open. Using a technique I had previously used, I shoved a flat-blade screwdriver in one side of the moulded rubber strain relief of the cable, and gave it a good prying until the seam started to crack in stages. Before long, two supplies had opened up, although a split did develop in one of the cases.
The power supply itself is marked SK Power Supply, model SKA-0512-20B. For convenience, the pin-out is listed on the rear – which is very helpful in case you want to splice in a different one for a replacement or just check compatibility. Other vendors used similar connectors with different allocations (e.g. Lacie) and damage is a high possibility where the pins are exchanged.
The power supply, as opened up. This was the first unit that was completely dead, and while it had no odour, the primary side capacitor was slightly bloated suggesting it may have failed. The second unit was one that couldn’t hold-up the load, and again, was clean inside but with no obvious capacitor failures. The transformer is labelled SKY, and the brand of the power supply is SK – these both refer to Skycable – the company that produced the power supply and the enclosure on behalf of Cooler Master.
Views from the side show copious use of insulation tape, and show the trauma on the primary side capacitor a little more clearly. The supply has a lot of adhesive goop over components as a form of vibration-support, but it has gone brittle and brown over the years.
A look at either end shows that there is a primary side fuse near the input socket, wrapped in heatshrink. On the unit reporting a zero load, the fuse had opened. The secondary side shows a cluster of capacitors.
This is where something odd was observed. While it’s not unusual to see Chinese capacitors in devices, often I see the brand in question spelled ChongX. In the case of this unit, it was ChengX and Cheng (primary side capacitor). So in other words, they used a counterfeit copy of a questionable quality Chinese capacitor. That can’t be good. Apparently, they also come spelled as ChangX.
The underside is covered with a single sided PCB with three holes (two used) as a shield. A pretty interesting way of making a shield, I thought.
Taking an inventory of parts, I needed:
T2A PCB Fuse Heatshrink 47uF 400V Capacitor 47uF 35V Capacitor 2x 1000uF 10V Capacitors 2x 1000uF 16V Capacitors
Rummaging through my parts box and after a quick order through element14, I had the necessary parts. Of note is that they could have used just all 16V rated 1000uF capacitors, but I think they decided to use the other rating to save money. Because I did have the voltage rating as per the original, I decided to use it anyway.
The surgery begins on the completely dead unit. The shield is to be desoldered. In the past, I used copious amounts of solder (to dilute the mess that is the lead-free solder) and quality desoldering braid (Goot or Chemtools branded stuff). Now, I’ve taken a liking to using a good desoldering bulb. While I’ve never had much luck with the solder-sucker spring-loaded pens, the desoldering bulb seems to do a great job, even if it is a Jaycar product. Much cheaper than wasting a mile of braid.
I thought the dead unit may have just had the primary capacitor short out internally, resulting in the fuse blowing and a simple fix … I was wrong.
The sooty black mess shows that at least three of the SMD resistors had smoked quite nicely. Despite knowing the values owing to having a number of other units, I felt it wasn’t worth pursuing repair any further since the carbon deposits and high heat may cause safety issues in the future, and its proximity to the controller IC makes it impossible to rule out collateral damage. It wasn’t critical to have repaired this unit, so I gutted a few parts from it, and moved on to the second unit.
The first thing out was the primary side capacitor. It was secured by glue that easily crumbled as an attempt was made to remove it. With the desoldering bulb, and some patience, the pads were cleaned nicely without using any braid.
The replacement fit just fine on a diametric basis, and was soldered in with nice shiny joints thanks to the leaded solder. Unfortunately, it was about 3mm taller than the former capacitor – I thought that would be fine since they should be given a little room to vent, right?
I moved onto the 47uF capacitor on the side, which was also fairly easy as it came out without any protest. Again, the desoldering bulb fared just fine – heat up each pad alternately while rocking out the old capacitor, clean it out with the bulb, put a new capacitor in following orientation, solder and clip the leads.
The challenge was the crammed secondary side capacitors – the adhesive was still holding strong, and the capacitors were right up against each other, so rocking was not a possibility!
As a result, I resorted to my brute force technique. I think it’s pretty obvious what I did – grab a set of pliers, yank the capacitors off the board by force, then suck off the solder from each pad thus allowing the remaining leads and electrode fragments to fall out.
That was, very successful, despite the crudeness of the technique. The downside was that the capacitors were not saved for analysis (if that was desired). Given the awkward solder pattern and proximity of the joints, it would have taken ages to try and get it out any other way. On the plus side – still didn’t have to use any braid.
Unfortunately, this is where I get bitten in the arse by the fact that the original capacitors were of a smaller diameter than the replacements. It happens occasionally – the original capacitors may indeed be in such a small diameter that quality replacements of the same size just aren’t common or available. As a result, I tried to do my best by cutting down the LED and moving it down to the board, and squeezing it into the side. It is about 2-3mm proud of the heatsink. I thought that would be good enough.
I did have a thought as to whether it would be possible to omit capacitors, or substitute smaller values, and it seems this board uses a C-L-C pi filter to filter ripple, hence the two capacitors per rail. As a result, omitting capacitors would compromise the filtering, and changing the values may alter its effectiveness (although probably not as badly as failed capacitors would). So, it seems you can’t use a 2200uF capacitor in substitution, although maybe substituting 820uF in both positions may be sufficient at the cost of slightly increased ripple on the output.
A neat job overall underneath. The lack of braid usage also means a lack of braid flux residue – no black marks for once! The shield was soldered back, and then … I tried to reassemble the unit.
Rather embarrassingly, despite the hack to the LED, it turns out that the wedged capacitors on the secondary and the replacement primary side capacitor were just 3mm too tall and thus, the case could not be closed. The next best thing was to wrap the whole unit in tape – it’s “safe enough” for my own personal use … maybe I can even say the ventilation has been improved … *chuckle*
On testing the unit, it came up just as expected, and the drive spun up just fine. All was good again.
I even discovered an old IDE drive in my room that was part of my old FreeNAS installation, formatted in UFS2, that I managed to recover the data from via Ubuntu (running live, inside a VM). Note that the auto-mounter is not capable of handling the drive, so make a mount-point manually, and specify -t ufs -o ufstype=ufs2 when mounting. Support in Ubuntu is read-only, but that’s really all I needed.
It’s interesting to see that the fallout from bad capacitor electrolytes and just plain poor counterfeit components is still continuing on. While it is frustrating, at least where it is the only fault, it can make for a fairly simple repair. Leave it too late and the collateral damage may make repair a little complicated, and uncertain. In this case, I didn’t even attempt to resurrect the blown unit.
Unfortunately, while most repairs are mostly uncomplicated, close narrow spacing of components and Z-height restrictions can be a killer as in this case, as some of the cheap capacitors were providing high capacitance figures in small volumes unmatched by common high-quality capacitors from Japan. Getting things desoldered can be a challenge as well, as some aged lead-free solder is hardly co-operative. While the one repaired was not aesthetically saved, at least it is functionally repaired. Further challenges can be had in just getting the case open without destroying it entirely – the days of “screw-based” enclosures seems limited, thus repair will become more of a challenge.
The fate of the remaining two power supplies remains to be determined, but may be also repaired in much the same way should the need arise. However, I’ve grown to be fond of my desoldering bulb, which should save my desoldering braid to when it’s really needed.