Hardware Insights Forum

It uses a single-transistor forward design with a now rather charred ST Microelectronics W9NK90Z MOSFET as the switching transistor, rated for up to 8A at 25°C and 5A at 100°C.

Odd, that's actually higher rated than the FQAF11N90C used in Hipro HP-D3057F3H. Any other explanation for why it blew??? The one I can think of is that, as with some of their integrated circuits, maybe ST = Suicidal Trash and their parts aren't as reliable as their competitors'.

Also, the +3.3V is rated for 28A, not 19A, so the rectifier for it isn't over-specified.

The ST W9NK90Z has a slightly higher RDS(on) than the switcher in the Hipro, so it's less efficient and will generate more heat. The Bestec also has a primary smaller heatsink than the Hipro does.

LongRunner wrote:

It uses a single-transistor forward design with a now rather charred ST Microelectronics W9NK90Z MOSFET as the switching transistor, rated for up to 8A at 25°C and 5A at 100°C.

Odd, that's actually higher rated than the FQAF11N90C used in Hipro HP-D3057F3H. Any other explanation for why it blew??? The one I can think of is that, as with some of their integrated circuits, maybe ST = Suicidal Trash and their parts aren't as reliable as their competitors'.

Also, the +3.3V is rated for 28A, not 19A, so the rectifier for it isn't over-specified.

Actually, the FQAF11N90C doesn't even have a metal tab whilst the STW9NK90Z does, which is why the STW9NK90Z has a higher power dissipation limit than the FQAF11N90C does (160W vs. 120W), so I would say the STW9NK90Z is more capable. I think there are two reasons as to why the Hipro does better - one, the better primary heatsink, but two, a more aggressive fan controller. I don't know if it's the case of the ATX-300-12Z but I have a ATX0300D5WC with the same fan and even at relatively high loads (below 200W) the fan doesn't rev up much at all beyond its default speed (which I would estimate in my case to be between 1300RPM and 1325RPM) until I reach very high loads (200W-250W+) and even then it doesn't spin that fast relative to its full speed (3000RPM, it might max out at around 2000-2200RPM, or about a 70% duty cycle). That being said, it is extremely quiet for the amount of air it moves, probably because of its very effective blade design. Anyway, I guess the fan controller in these 300W Bestecs is just a little too slow in speeding up to compensate for the augmenting heat output.... thus why it exploded at 405W (the passive PFC in the reviewed unit might also hinder some airflow to the secondary, or my mistake, primary side). Could also be that the unhealthy amount of conductive glue splattered over the secondary side worsens its performance. The 16 gauge wires of the Hipro also help (compared to the Bestec's 18 gauge wires).

EDIT: I also noticed the Jamicon fan looks very dusty in the Bestec whereas the Superred fan in the Hipro does not. Perhaps that has something to do with it?

^
It would hinder airflow to the primary side, as the coil was installed behind the primary heat sink. That might be another reason why it blew the switcher.

The fan controller is another possible explanation. I don't remember much about how it behaved at high loads, but do recall the Hipro's fan becoming quite loud when I tested the poly-modded one.

Wester547 wrote:Could also be that the unhealthy amount of conductive glue splattered over the secondary side worsens its performance.

I would be less concerned about the glue on the secondary capacitors and coils and more about the portion that crosses the isolation barrier and has already deteriorated from the heat from a primary-side resistor and the +5VSB rectifier.

The 16 gauge wires of the Hipro also help (compared to the Bestec's 18 gauge wires).

Well, let's try that out, shall we?

The 24-pin main power cable has four +3.3V wires that, in the overload test, were loaded to, in total, what should have been 30A (but ended up somewhat lower due to the voltage falling). As 18AWG copper wire has a resistance of about 0.02Ω/metre, if we assume the cables are about 0.5m long (undoubtedly longer than what this unit actually has):

0.01Ω/wire ÷ 4 lines = 0.0025Ω overall
0.0025Ω × 30A = 0.075V drop

For +5V:

0.01Ω/wire ÷ 5 lines = 0.002Ω overall
0.002Ω × 30A = 0.06V drop

I don't know what proportion of the +12V load was through the main power cable. But the main cable only has 8 ground lines shared between all rails, so 60A in total is pushing the limits already. In that case:

0.01Ω/wire ÷ 8 lines = 0.00125Ω overall
0.00125Ω × 60A = 0.075V drop

So between the voltage and ground wires:

+3.3V loses 0.15V, or 4.5%
+5V loses 0.125V, or 2.5%
Overall loss of 3.3%

Now let's redo those calculations with 16AWG wiring, which is about 0.013Ω/metre:

0.0065Ω/wire ÷ 4 lines = 0.001625Ω overall
0.001625Ω × 30A = 0.049V drop

For +5V:

0.0065Ω/wire ÷ 5 lines = 0.0013Ω overall
0.0013Ω × 30A = 0.039V drop

For the grounds:

0.01Ω/wire ÷ 8 lines = 0.00081Ω overall
0.00081Ω × 60A = 0.049V drop

So between the voltage and ground wires:

+3.3V loses 0.0975V, or 2.95%
+5V loses 0.0878V, or 1.76%
Overall loss of 2.2%

So the overall effect the wire size difference has on efficiency is a rather negligible 1%. Current returning from +12V through those grounds would make some difference, but not a big one.

I have found another possible explanation for the kaboom, though. It's that passive PFC reduces the voltage on the primary side (compared to no PFC), therefore requiring the switcher to handle more current.

You're right with respect to the wire gauge. I've seen reviews on Hardware Secrets where Acbel/Coolermaster units have no trouble handling 400W (most of which is loaded on the +12V rails) with all 20 gauge wires. Still, I can't imagine the larger gauge causing problems, especially with mains cords. Very interesting that Passive PFC reduces the voltage. That said, a STW12NK90Z (two steps up from the STW9NK90Z in the Bestec review) that was the main switcher for a Delta Antec VP350 reviewed on HWS exploded at 475W, and that was with no PFC at all and an 120mm fan, but that particular unit didn't have optimal heatsinks and only 470uF primaries, maybe that had something to do with it. I think it also depends on how well the main switchers are driven.

I think a bigger factor in the Bestec review is the fact that the fan remained "relatively quiet" so it didn't really rev up as much as it should have in response to the intense load... if it were running at +12V at that point, I don't think it would have failed.