Hardware Insights Forum

If the only limitations were available technology, your own sanity, and optionally the ATX form factor, what would you put in a PC PSU??? Power output can be anything you want, but preferably something you could at least run a current mid-range PC from.

I would go for (ATX form factor in mind):

• Input OVP up to 500VAC (consisting of a small circuit with a normally-closed relay in one side of the mains input to the PSU circuitry that activates the relay if the voltage goes out of range, preventing damage to downstream circuitry)
• CAPZero
• Thermistor shorting relay (already a feature of high-end units)
• HiperPFS-2 (in a small unit - for larger units, a suitable discrete APFC solution)
• HiperLCS (small to medium units) or otherwise a suitable discrete LLC resonant solution for the main supply
• TinySwitch-4 for the standby supply
• Synchronous rectification (of course)
• The highest quality capacitors available e.g. Panasonic ED, EE or XB on the primary and FR on the secondary, and where suitable, solid polys or even ceramic capacitors
• Buck converters for +3.3V and +5V, and buck-boost for -12V. Cross-loading issues - gone.
• Protection against just about everything that can go wrong: Outputs shorted to ground or together (you may want a rather robust protection diode for -12V), excess current draw from any of the outputs (-12V included), overheating, fan failure, excessive ripple (I'm sure you've been waiting for that protection), you name it.
• Continual fan operation at low speed in standby mode to improve reliability. (Why don't more PSUs do this???)
• As for the fan itself, an 80mm fluid-dynamic bearing (or failing that, dual ball bearing) model, rear-mounted with a 20mm gap in front to minimise noise. The heatsinks are as large as desired and the front vent is a honeycomb type, covering the entire panel (excluding any space used by modular connectors).
• Two bi-colour LEDs on the rear panel for diagnostic purposes - see below for coding.

LED 1 ("power"):
Green - operating
Orange - standby mode
Off - wait for power good

LED 2 ("protection"):
Green - normal
Orange - SCP, OCP, OPP, OVP, UVP tripped
Orange flashing - overheated, wait to cool down
Red - capacitor or fan failure

(Orange is obtained by mixing red and green.)

Note that gold plated connectors are not on the list. This is because gold and tin don't mix. Well, at least mixing them doesn't improve reliability over tin/tin. Only connectors that are gold plated as standard should be.

Also notably absent is the usual large fan. There are three reasons for this.

LongRunner wrote:[*]Continual fan operation at low speed in standby mode to improve reliability. (Why don't more PSUs do this???)

The accumulation of dust. However, an additional intake fan for +5VSB operation, spinning at slow speeds, might help that lots (to my knowledge, exhaust fans are quicker to accumulate dust). The real problem with +5VSB circuits is the use of flyback topology. I wish they would gravitate away from that... and reliability would be less of a problem with good brands of capacitors.

excessive ripple

This would be nice, but it really depends on what one considers excess ripple to be. I don't consider ripple voltages bordering upon the limits acceptable at all, but some might as they might be the sort of folk who upgrade their system every 2 to 3 years. Fan bearings, a well lubricated sleeve bearing fan should last about as long as a dual ball bearing fan would with better shock resistance. I don't actually have that much of a problem with current power supplies from decent manufactures (the decent brands have always built fairly good units), I just wish the choice of capacitors (and sometimes fans and heatsinks) were better.

Wester547 wrote:However, an additional intake fan for +5VSB operation, spinning at slow speeds, might help that lots (to my knowledge, exhaust fans are quicker to accumulate dust).

How would an intake fan pick up less dust???

The real problem with +5VSB circuits is the use of flyback topology.

They use it because it requires the fewest parts.

Flyback: 1 switcher + transformer + 1 rectifier per output + filtering
Forward: 1 or 2 switchers + transformer + 2 rectifers per output + mag-amps + filtering
Half-bridge (old style): 2 switchers + film cap + transformer + 2 rectifiers per output + mag-amps + filtering
LLC resonant: 2 switchers + resonant inductor and film cap + transformer + 2 rectifiers per output + filtering

The capacitors are as large as they are because the output is much harder to filter than with other topologies. I think polys would be good for flybacks, providing the low ESR needed without having to use an oversized component.

I don't consider ripple voltages bordering upon the limits acceptable at all...

Maybe not in a new PSU, as if it's borderline already it will only get worse. Over-engineering the output filters is done to keep it in spec despite capacitor aging (to a point - low quality capacitors will still push it above the limit).

For reference, half-bridge PSUs usually produced -12V from the same transformer windings used for +12V, but with the opposite diode polarity, sharing the mag-amp used for +5V and +12V. -5V was produced either by doing the same thing with the windings otherwise used for +5V, or with a 7905 drawing from -12V. Forward converters instead use the mag-amp as a flyback converter of sorts, with one end of the -12V winding grounded and the other connected to the rectifier diode. These arrangements often result in poor -12V regulation - compare, for example, the H305P-01 (which uses a 7912) with the HP-D3057F3H (which has no -12V regulator).

My dream PSU would simply be something like a Seasonic G-360 with an 80mm fan that was more affordable.

LongRunner wrote:

Wester547 wrote:However, an additional intake fan for +5VSB operation, spinning at slow speeds, might help that lots (to my knowledge, exhaust fans are quicker to accumulate dust).

How would an intake fan pick up less dust???

I might be mistaken (depending on how dusty the room in which the PSU resides is, or how dusty the case itself is)... but they could always add a dust filter.

The real problem with +5VSB circuits is the use of flyback topology.
They use it because it requires the fewest parts.

Flyback: 1 switcher + transformer + 1 rectifier per output + filtering
Forward: 1 or 2 switchers + transformer + 2 rectifers per output + mag-amps + filtering
Half-bridge (old style): 2 switchers + film cap + transformer + 2 rectifiers per output + mag-amps + filtering
LLC resonant: 2 switchers + film cap + transformer + 2 rectifiers per output + filtering

The capacitors are as large as they are because the output is much harder to filter than with other topologies. I think polys would be good for flybacks, providing the low ESR needed without having to use an oversized component.

Yes, the toroidal coils found in the switch mode section of the PSU do a huge part of the filtering, so capacitor input filters in flyback topology have much more ripple current going through them even at low loads, especially without any airflow.

Maybe not in a new PSU, as if it's borderline already it will only get worse. Over-engineering the output filters is done to keep it in spec despite capacitor aging (to a point - low quality capacitors will still push it above the limit).

High quality capacitors should not age so quickly that the increase in ripple voltage is noticeable or even significant.

For reference, half-bridge PSUs usually produced -12V from the same transformer windings used for +12V, but with the opposite diode polarity, sharing the mag-amp used for +5V and +12V. -5V was produced either by doing the same thing with the windings otherwise used for +5V, or with a 7905 drawing from -12V. Forward converters instead use the mag-amp as a flyback converter of sorts, with one end of the -12V winding grounded and the other connected to the rectifier diode. These arrangements often result in poor -12V regulation - compare, for example, the H305P-01 (which uses a 7912) with the HP-D3057F3H (which has no -12V regulator).

Fascinating. It would explain why Hardware Secrets found that some Huntkey power supplies (with half bridge topology) were capable of excellent -12V regulation even without a regulator, and just a 6.3mm or 8mm 220uF or 330uF 16V capacitor (usually either Nichicon PJ, PW, or PA grade in terms of ESR and ripple ratings, but from FCON or BH) and a ferrite coil. Do forward PSUs generate -5V and -12V from +5V and +12V as well, essentially?

EDIT: This also explains why there are combined -5V and -12V ratings with certain PSUs.

Wester547 wrote:It would explain why Hardware Secrets found that some Huntkey power supplies (with half bridge topology) were capable of excellent -12V regulation even without a regulator...

I haven't seen those for myself. But I have looked at reviews here of other half-bridge supplies, and those had similarly lame -12V regulation.

Although, the Power Man IP-S400EQ3-2 (a forward converter) has good -12V regulation without a regulator, so there's probably more to it than that.

...and just a 6.3mm or 8mm 220uF or 330uF 16V capacitor (usually either Nichicon PJ, PW, or PA grade in terms of ESR and ripple ratings, but from FCON or BH) and a ferrite coil...

That they got away with such a small capacitor doesn't surprise me.

Do forward PSUs generate -5V and -12V from +5V and +12V as well, essentially?

The aforementioned winding on the mag-amp does, if you like, "leech" power from both of those rails if the PSU uses group regulation. If the PSU uses independent regulation, it leeches from only the +12V.

I do have one PSU - Macron MPT-4012 - which generates an unregulated negative rail, fed into a 7912, using a separate transformer winding, a mag-amp winding and two rather beefy diodes (probably 3A rated). Too bad it's full of Fuhjyyu - it otherwise looks like a tank (and has OCP on all main rails). (Although at least I don't see any busted caps in it.)

As for -5V in forward converters, I don't see why they would use anything other than a 7905 drawing from the -12V.

Another note: Half-bridge designs often have an extra diode in line with -12V for the sole purpose of dropping the voltage a bit, and the -5V diodes will not be Schottky.

(Fun though mostly irrelevant fact: Many HDDs use buck-boost converters to generate a -5V rail for the head preamplifier.)

Dream PSU? Seasonic. Done. No point to reinvent the wheel.
I'd personally like a fully modular 600W 80Plus Platinum that's completely fanless

Nah, Delta. I love the HCP Platinums. Wolf's 850W one pulled some ridiculously tight voltage regulation and ripple numbers in the testing. It did even better than the Seasonic KM3 platform.