Articles: Cases/PSU

Bookmark and Share

Pages: [ 1 | 2 | 3 | 4 | 5 ]

Every chip in this PSU is marked as "FSP", so the "chipset" can indeed be viewed as FSP's original. As far as I know, FSP not just rebrands a standard controller but changes something in its circuitry, putting some of the elements that are usually installed outside into the controller chip. Producing such custom-made chips is justified by FSP's large production volumes. The expense of ordering them is covered by the lower manufacturing price of resulting PSUs (thanks to the transfer of some components into the chip) and also by the inability of the mainland China competitors to copy the circuit and produce cheap clones. Knowing the PSU circuit design is useless unless you have FSP chips. This is an insurmountable obstacle to typical Chinese factories that do not have experienced engineers capable of adapting a PSU circuit to their manufacturing capabilities, let alone developing a new circuit from scratch.

The emptiness of the PCB when viewed from above is made up for by the bottom view: most of the smaller components (resistors, low-rating capacitors, low-power diodes and transistors, optocouplers, etc) are simply surface-mounted and populate densely the bottom of the PCB.

We can also see some of the power transistors of the output circuitry here. I mean the four rather large squares of the field transistors in the bottom right. Thanks to their high efficiency and low heat dissipation they don't need a heatsink but can be cooled well enough through the PCB interconnects they are soldered to. The same can be seen in the Seasonic X series as well as some other products with DC-DC converters at the output.

The quality of soldering is immaculate. The top right corner (as viewed in the photograph) with wires from the mains connector is somewhat smudged with soldering flux, but that’s not a problem at all. Every surface-mounted element is secured with glue as is typical of PCBs with dual-sided mounting that are assembled in three steps. First, the bottom-surface components are glued to the PCB, then the PCB is turned upside down and gets its top-surface components, and then it goes into the machine for wave soldering. Without the glue the process would be more difficult. The components would have to be soldered in two passes with such methods that the components on the bottom side of the PCB didn’t fall off.

Although we’ve read about Japan-made capacitors on the product packaging, the PSU proves to have mostly Taiwan-made ones. It is the KF series from CapXon. I wouldn’t call them bad, yet I wouldn’t call them Japanese, either. Well, on a second thought, the packaging mentions a Japanese capacitor in the singular and you can indeed find one in the PSU. It is the high-voltage thing standing next to the PFC device. By the way, it is the capacitor that has got special protection according to the same text on the box.

The PSU is cooled by a Protechnic Electric MGA12012HF-A25 fan that has a fluid dynamic bearing and a rated speed of 2400 RPM. The fan looks pretty with its rounded-off shapes but the noisiness of the PSU is going to depend on speed regulation since the max speed is quite high.

The PSU is equipped with the following cables and connectors:

  • One mainboard cable with a 20+4-pin connector (55 centimeters long)
  • One CPU cable with one 8-pin and one 4+4-pin connector (55+15 cm)
  • Two graphics card cables with two 6+2-pin connectors on each (55+10 cm)
  • One cable with four SATA power connectors (55+15+15+15 cm)
  • One cable with two SATA and two PATA power connectors (55+15+15+15 cm)
  • One cable with one SATA, two PATA and one floppy-drive connector (55+15+15+15 cm)

We've got a sufficient selection of connectors here, especially as the CPU cable is as long as 70 centimeters (from the PSU case to the connector on its end) and you can easily connect it in a full-tower system case with a bottom PSU bay and a hidden compartment for cables. Unfortunately, some makers still produce PSUs with a CPU cable of only 55 to 60 centimeters long, which may be insufficient. Take note that the Aurum series models with a wattage rating below 700 watts do not have the second connector on the CPU cable, so the latter is only 55 centimeters long with them.

Every cable has a nylon sleeve.

The manufacturer says that the PSU has four “virtual” +12V output lines, each for a max current of 18 amperes. Indeed, the individual +12V outputs are connected on the PCB via current-measuring shunts. However, I could not overload any single line during my tests. For example, the CPU power line (12V2) worked normally at a load of 35 amperes irrespective of whether the load would increase from 0 to 35 amperes smoothly or instantaneously. So, the PSU seems to be protected against overload on the +12V lines combined, but not against overload on the individual lines.

Working with my UPS (APC SmartUPS SC 620 with non-sinusoidal output voltage), the power supply was stable at loads up to 395 W (when powered by the mains) and 350 W (when powered by the UPS’s batteries). This, the FSP Aurum shouldn’t have compatibility issues with uninterruptible power supplies.

We can see some red in the cross-load diagram but none of the voltages goes out of the permissible limits (5% from its nominal level) at any load. Moreover, the PSU is stable at near-zero as well as greatly misbalanced loads (for example, when there is a load of over 650 watts on the +12V rail and no load at all on the +5V and +3.3V rails). Thus, the FSP Aurum is going to be compatible with any PC configuration, including low-power-consumption ones (of course, using a 700W power supply for an Intel Atom machine wouldn't make any point, but we shouldn't forget that even top-end computers may have a very low power draw when idle).

The high-frequency output voltage ripple is quite conspicuous at full load, but its level is within the permissible limits.

The low-frequency voltage ripple (at the double frequency of the mains, i.e. 100 Hz in my location) could be observed on the +3.3V rail but never went beyond the permissible limit, either.

The PSU is almost 93% efficient at the peak and 90% efficient at full load. Even at a load of 50 watts (which is about one fifteenth of its full output power) the efficiency is but slightly lower than 80%. That's an excellent result.

It’s worse with noise. The fan rotates at about 800 RPM right after you turn the PSU on but accelerates to 1300 RPM in a few minutes even at a load of only 50 watts. However good a 120mm fan may be (and the PSU’s fan seems to be better than the older Protechnic Electric fans installed into the FSP Epsilon series), it cannot be silent at such a high speed. The hiss of the air is perfectly audible.

To make things even worse, the fan speed grows up linearly at higher loads (i.e. as the PSU gets hotter), reaching a maximum of 2230 RPM at only about 500 watts.

Thus, the FSP Aurum is below average in terms of noisiness. When it comes to acoustic comfort, it is no competitor not only to the quiet PSUs from Seasonic, Enermax and Enhance but even to the ordinary products from Channel Well Technology that sell under different brands (including Thermaltake, for example).

It is unclear what prevented FSP from implementing in this otherwise carefully designed and thought-through product the conventional algorithm of fan regulation: a constant low speed at 50% and lower loads and a linear growth of speed afterwards.

This algorithm is most logical. On one hand, there is a minimum voltage at which any sample of the fan is guaranteed to work. Let's suppose it is 5 volts. A fan with a rated speed of 2400 RPM will be rotating at 2400 / (12/5) = 1000 RPM if supplied this 5-volt voltage. However, working at 1000 RPM, the fan is quite capable of cooling the PSU at loads up to 450 watts, for example. We can’t supply a lower voltage at a lower load because we can’t be sure the fan will start up at all then. So, we can use a 4-pin fan with a larger guaranteed speed range or supply 5 volts to start the fan up at minimum load and keep this voltage until a PSU load of 450 watts. The second solution is implemented by almost all of the PSU makers as it allows to easily produce a PSU that will be rather quiet at low and medium loads and will not overheat at high loads.

It is unclear why FSP doesn’t use this algorithm but implements the outdated linear regulation method in its new product series. With the linear regulation, the minimum of speed is still determined by the minimum voltage of the fan, but the fan accelerates as soon as the PSU load increases although that's not really necessary. As you can see in the diagram, the PSU remains cold: at a load of 500 watts the difference in the temperature of the air going in and out the PSU is no larger than 3°C.

So, whatever the reason for that, the FSP Aurum is a noisy power supply.

The test of the standby source doesn’t produce any surprises. The device easily copes with the specified load of 3.5 amperes.

Pages: [ 1 | 2 | 3 | 4 | 5 ]


Comments currently: 15
Discussion started: 08/04/11 11:04:50 PM
Latest comment: 12/21/15 11:23:58 AM

View comments

Add your Comment