Dell H305P-01 Review

[LongRunner]

Why would you need to use 12AWG mains wires in this PSU???

Also, what voltage are they rated for??? If they're 300V, like the output wires, I think that's a bit close for comfort. If 600V or greater, they should be fine.

[c_hegge]

You probably don't need wires that thick, but I had plenty of spares, so I thought hey, why not? I got a heap of 1 and 2m long wire offcuts from a friend who is an electrician. They were left over from houses. I used a lot of them in the load tester, but I still had some spare

[LongRunner]

They were left over from houses.

Given that:

• In Australia, wouldn't they be measured in metric (e.g. 2.5mm², which is actually closest to 13AWG)?
• Wouldn't they be stiff solid-core or 7-strand wires, as opposed to the flexible ones with many strands?

The flexible cord the PSU is plugged in with would be 1.0mm² (≈17AWG) at most, and the fixed wiring up to the power outlet is probably 2.5mm² in Australia.

For reference:
0.5mm² - approx. 20AWG
0.75mm² - between 19AWG and 18AWG
1.0mm² - approx. 17AWG
1.5mm² - between 16AWG and 15AWG
2.5mm² - approx. 13AWG
4mm² - approx. 11AWG
6mm² - between 10AWG and 9AWG

LATE EDIT: I guess they're most likely 4mm² (or maybe old Imperial 7/.029″, which converts to 2.98mm² and is indeed close to 12AWG).

[c_hegge]

Yes, I know the size that leads up to power points. These were slightly thicker. 12AWG wasn't specifically written on them, but I'm pretty sure that's what they were going by the size.

Yes, they were 7 strand.

EDIT: Did I even say that they were 12AWG?

[LongRunner]

Did I even say that they were 12AWG?

Maybe not here, but you did so in the Hardware Secrets thread about this case-swap.

The problem with stiff wires is that they put more stress on the solder joints. You might get away with it, but it's not recommended.

[c_hegge]

Ah, that's right. Yes, I know you could put more stress on the joint, but I left the perfect amount of slack, so the current position is where they 'want' to be

[Wester547]

I can't help but notice that the +12V rail's ripple (and the +5V/+5VSB rail as well) in this PSU seems much lower even at relatively high current draws than that of the HP-D3057F3H in your review, yet the HP-D3057F3H also has two 2200uF 16V Teapo SCs of seemingly equal dimensions and a PI filter coil, as well as a comparable toroid. Any idea how Hipro pulled that off? Could it be the fact that this Hipro has a more overspec'd +5V and +5VSB rail in terms of output filtering and a linear regulated +3.3V and -12V rail which helps immensely with cross loading ripple suppression? Or perhaps the fact that the +5V and +3.3V rail were loaded more in the HP-D3057F3H review?

[c_hegge]

The toroid coil on the Dell is actually quite a bit larger than the one on the Hipro. If you compare the secondary side shots, you'll find that, on the Dell, the caps only come about half way up the coil. On the Hipro, they extend to about 5mm from the top. Both PSU's have similarly sized caps.

[LongRunner]

Isn't recapping with Panasonic FC before the +5V coil and Chemi-con KZE after it an odd combination??? Wouldn't using a higher grade cap in the more stressed position make sense, if you're going to deviate from the original specs??? (I see that the original 4700µF and 3300µF Teapos are the same physical size, but you could also use two 3900µF Panasonic FCs and the total capacitance would be close enough to original to not matter.)

At least the cap before the +5VSB coil is a good size - though I don't like the way they put that diode right next to it.

[LongRunner]

I'm surprised the power supply still worked given the discoloration caused by the hot load resistor next to that tiny Teapo.... if the PCB is really darkened, Chicony must have not chosen an appropriately spec'd resistor for the -12V rail. They have done that before in some old 200W Dell Hipros as documented by momaka a few years back on badcaps.

It's not for the −12V – this unit has a 7912 for that, and the resistor isn't anywhere near the 7912 anyway. It's for the +12V.

I also think the FYP1010DN is freewheeling for the +12V rail.

It's the other way around – the FYP1010DN handles the pulses from the transformer and the STPS20S100CT does freewheeling. With a typical (?) duty cycle of 33% (one third), it would indeed give 30A on the +12V.

Forward is also somewhat inefficient because the use of half-wave rectification on the output...

Actually, forward doesn't work that way. It's the isolated counterpart of the "buck" topology used in motherboard VRMs along with the +12V to +3.3V and +5V converters in current high-end PSUs. The mag-amp in forward topology does the same thing as the coils in a motherboard VRM, albeit at a higher power level and lower frequency. Basically all you're changing from buck topology is inserting a transformer and (by necessity) another diode.

I would guess that in a forward converter outputting +12V, the pulses from the transformer would be somewhere from +24V to +40V. A higher duty cycle is better as it reduces conduction losses in the switcher(s) and the pulse voltage from the transformer (therefore allowing lower voltage diodes on the secondary).

The 5V rail uses two Fairchild MBRP3045N Schottky rectifiers rated at 30A, so it should be capable of up to 60A.

Actually, because of how they're connected in this unit – which doesn't equally split the current between the two parts – it's 54A if the duty cycle at full load is a third…

I'm not agreeing with Hardware Secrets on this one - I think you can unevenly load the two diodes in a single part as long as the total dissipation of the two is within the safe limit, because they're on the same chip and will therefore stay at about the same temperature. But if you use multiple physical parts, you do need to be careful with the current split between them, as they can end up at very different temperatures. In this case, one diode out of the four is the pulse rectifier and the other three are freewheeling. With the one-third duty cycle:

(54*(1/3))+(18*(2/3)) = 30A average (the rectifier next to the transformer)
(18*(2/3))+(18*(2/3)) = 24A average (the rectifier on the other side of the heatsink, next to the mag-amp)
(18 being 54 / 3)

It's still 2.45x the label rating, though, and the one next to the transformer would have been more than enough all by itself.