Disassembling an Ultra X-Connect X-ULT500P

Will be updated ...



I got an e-mail from a rather angry chap named Raven asking me if I can take a look at his PSU as it was dead. Upon quizzing him a bit regarding the unit, he mentioned that it was an Ultra X-Connect. Well, I've never fixed an older Ultra before, so I was curious on what I can find about this unit. Ultra had copped a bad reputation for making miserable power units. Newer Ultra products (OEM Andyson Electronics) are vastly better and the difference is like night and day.

Reputation is a finicky thing. It only takes one batch of defective units to create an avalanche of bad press. As time goes by it becomes hard to filter fact from fiction, as negative press is magnified by second hand and third hand accounts.

The question is, are these units really garbage? Let us open it up to see!

Primary caps are Sus'con 1000uF 200V cans. Suscon on motherboards are crap and I would not consider them a good brand. Notice the slight discoloration of one of the primary cans close to the bridge rectifier. The rectifier is not sinked and the discoloration is probably due to the heat produced by the rectifier. There is a basic input surge protection stage as evinced by the yellow spike absorbing capacitors and the Blue capacitors.

The layout is clean and the unit is group regulated. I'm troubled by the size and nature of the heatsinks. It looks all flashy, but does it do the job? I really doubt if this unit can put up 500W as the lable claims. Ofcourse, to be sure, I'd have to check the datasheets of the switching transistor and this will be done shortly.

The secondary side is a disaster. It is full of Fuhjyyu capacitors. These are among the worst caps you can put in a power supply unit. I don't care what anybody says, but I've replaced one Fuhjyyu too many to think otherwise.

This unit was purchased by Raven this year and has been in operation for a since 2005 It powered a modest AMD S939 setup, FX-xx CPU +6600GT. Nothing that would require a nuclear reactor to power it!

Poor chap was "refused an RMA" by Ultra (in his words) because he opened up the unit. While this is entirely reasonable from a policy perspective, I would have expected a company that is looking to re-establish itself as a major player in the market to be a bit more considerate.

Feast on these pictures while I update this post with more analysis and pictures.



Now let us look at what this unit can do in theory and why exactly did it fail?

• Power Specifications:
• 115V, 230V switchable power supply
• +3.3V, 28A
• +5V, 30A
• +12V, 34A
• -12V, 0.8A
• -5V, 0.3A
• +5Vsb, 2.0A

On the primary side we have three 2SC3220 fast switching transistors made by Fuji Electric. According to the datasheet given HERE, the maximum continuous collector current this transistor can handle is 15A at 25C. I couldn't find information about peak values. This is a two transistor forward configuration. THIS may prove an interesting read. It seems to be some kind of a "Half-bridge" converter, but I'm not sure.

There are 2 switching transistors on the PWM side are manufactured by ST Microelectronics and are rated at 15A continuous, 400V at 25C. Since this transistor is going to work in pulsed mode, the only relevant quantity is the peak current per pulse. I had a really hard time finding that information from the datasheets. Please contact me if you know how to decipher that. Note that usually a set of diodes is used on this side after the transformer, here we see the use of power transistors.

There is also a (K)SC5207 Switching transistor made by Fairchild Semiconductor rated at 10A in pulsed mode.

You can see that the 3.3 and 5V lines are tied together and use the same transformer, with the 12V line. The transformer next to it may be an isolation transformer. I'll have to check into that though. The third transformer is independent of the other circuitry and it provides the standby voltage.

Here are a few more pictures of this unit disassembled...





Three diode banks being used are MOSPEC S30D40C Schottky barrier rectifiers. There are 3 such diode banks. The datasheet rates them at 30A per pack. The label shows that the maximum independent draw is 28A on the 3.3V and 30A on the 5V rail, yet gives a maximum combined rating of 200W (as opposed to a maximum of 242.4W if they were independent), telling us that the 3.3V and 5V rails are tied together.

If you look at the datasheet, the 30A forward current limit is maintained upto a case temperature of 80C, thereafter it begins to decrease rather rapidly.

Even if we assume that this is not true, the on-paper maximum current that can be drawn would be provided by the two diode packs for the 3.3V and 5V rails. So we are safe on paper but really skirting the edge. The third pack works for the 12V rail which is worrying rated at a max of 34A! So, under ideal conditions, the 12V rail can only supply a max 360W as opposed to the claim of 408W. However, assuming that the 3.3V and 5V numbers are correct (200W), the 12V rail has been under-rated to deliver 23.3A, which is below the component limit of 30A.

To conclude, on paper, everything looks fine although the numbers are a little too close to the theoretical maximum of the diode packs which is 3.3 x30 +5 x30 +12 x 30 = 602.4 W. Most modern PSU's have diode packs at almost double the rated output capacity. No worries if things are kept cool.

A few more pictures:



In one of the pictures you see a resistor on the secondary side almost charred, so clearly heat was an issue here. Therefore the preliminary diagnosis of heat related failure is further reinforced.

Notice the secondary coil I ripped out. It has 4 windings instead of the usual two. I would speculate that this is indicative of a 3.3V+5V group regulated scheme. I am a bit puzzled by the difference in thickness between the 4 strands. I'll have to check on this to see why they are different.