Hipro HP-D3057F3H Review

[Wester547]

Aside from having Japanese capacitors on the standby output and Taiwanese ones on the main outputs, there's another way the main output capacitors could fail first - leaving the PC on most of the time, as opposed to shutting it down when you're not using it.

Shutting it down when you're not using it would lead +5VSB to fail sooner so the PSU would have failed sooner anyway (assuming you don't remove AC power). I noticed a minimum load resistor around or between the +5V and +12V rail, those don't help (even if they didn't cause discoloration to the PCB they do tend to run hot). Also, by your own admission, at light loads, that Hipro's fan spins so slowly that it doesn't push much air (though I'm sure at higher loads that it's quicker to spin faster, quicker than the Bestec 300Ws for sure, though they spin faster at lower loads...)... that doesn't help those Teapos either. The reason why I say this is that most people actually don't draw more than 150W from all such PSUs unless they're heavy gamers or do heavy video editing, or professional audio, or something like...

As the VIPer22A used switches at 60kHz nominal and the 470µF 10V Teapo SC used after the coil has a nominal impedance of 0.12Ω, the divison ratio with my chosen inductor is approximately 32:1. This means that the ripple voltage on the 1000µF capacitor before the coil could be >200mV while keeping the output ripple in spec, stressing the already unreliable original capacitor and causing it to fail in short order.

More ripple current -through- the capacitor would lead it to fail sooner than high ripple voltage (the ripple that "escapes" the output filtering and ends up at the load). Unless, of course, you're making reference to the ripple current to begin with and not the end ripple voltage by saying the ripple "before" the output. Flyback topology results in much higher ripple current through the capacitors to begin with. Though, in essence, what you're saying is if they used a +5VSB switch that worked at a higher switching frequency, that it would be easier on the output?

[LongRunner]

More ripple current -through- the capacitor would lead it to fail sooner than high ripple voltage (the ripple that "escapes" the output filtering and ends up at the load).

I wasn't saying the unit forces that much ripple voltage on the capacitor, I was saying that you could have that much ripple voltage (so long as it's the switching noise) going into the coil and still keep the ripple voltage after the coil in ATX spec.

With the 1000µF Teapo SC used, having, say 100mV RMS (as opposed to peak-to-peak) on it (which, if it's a sine wave, will result in 283mV peak-to-peak - but I know ripple is never in the form of a sine wave) will push over an amp through it - and, therefore, it's finished. But if the filter following it works as it should, the outgoing ripple will not show any sign of a problem (for however long - or short - the overstressed cap survives).

Installing the Panasonic FR will cause the ripple voltage to drop, of course, and it will last longer, as the ripple current will stay about the same (more on that below).

Given the low-ish ripple current rating of that Teapo, using it in that position on a flyback meant to output 2A is a dubious design decision.

Though, in essence, what you're saying is if they used a +5VSB switch that worked at a higher switching frequency, that it would be easier on the output?

It would allow a smaller coil, but it wouldn't reduce the load on the capacitor before the coil.

As for ESR vs. ripple current flow...

Short version: Outside of the multiple-caps-in-parallel scenario, whoever told you that changing a cap for one with lower ESR will put it under more stress is wrong.
Long version: In flyback topology, the capacitor before the coil has to take a relatively fixed ripple current of up to half (?) the output current.
In forward and traditional half-bridge (LLC resonant is more complicated) configurations, the reactance of the big mag-amp coils is monstrous compared to the ESR of the caps, so changing the specs of the caps has little effect on the ripple current through them - it's the ripple voltage that changes with them. Any increase in ripple current would be more than compensated by the higher current handling of the higher-grade capacitors.
The capacitors after the ferrite coils (pi refers to how the schematic arrangement of the caps on both sides of the coil, along with the coil itself, looks, not to the type of coil) in either arrangement are relatively unstressed - until the capacitors before the coils fail, forcing those after to take the load. If, of course, the capacitor before the coil is fine and that after it is bad...well, that tells a lot about its quality, doesn't it???
Aside from the switching noise, a degree of double-mains-frequency ripple often appears on some of the outputs - you can see it in the scope shots for +12V, +5V and -12V in this review. That passes "straight through" the secondary-side filtering and doesn't stress the caps (but the connected hardware won't take so well to it). It's the control loop's task to compensate for it, but only one output per transformer can truly be clear of it (in this unit, +3.3V). That's why the ripple voltage doesn't drop much with polys. Using buck converters for +3.3V and +5V, along with a linear regulator or buck-boost converter for -12V (as I suggested in the dream PSU thread), you could eliminate it. But no such fancy features on this unit.
It's barely possible that replacing caps with ones with too low ESR will result in an inadequately damped filter that will "ring" at its resonant frequency, and there are some other issues that could pop up. But the "lower ESR = more ripple current" suggestion is not it.

[Wester547]

Given the low-ish ripple current rating of that Teapo, using it in that position on a flyback meant to output 2A is a dubious design decision.

It doesn't mean the power supply can't do 2A @ +5VSB. It just means the Teapo won't last long for the lack of quality.

Short version: Outside of the multiple-caps-in-parallel scenario, whoever told you that changing a cap for one with lower impedance will put it under more stress is wrong.

Yes and no. Lower ESR capacitors are better at suppressing ripple current and precluding that from getting to the ripple voltage (the output), which means more ripple current would have to be going -through- them to do that. That being said, like you stated, topologies outside of flyback having the magamps and ferrite coils doing huge parts of the filtering does significantly allay that problem. Also, lower ESR capacitors tend to be of higher grade and tend to have higher ripple ratings than higher ESR capacitors, so because of that they would indeed be under less stress. I think there's a reason why lower ESR CapXon KF and GL capacitors fail more often than the GP KM ones...

If, of course, the capacitor before the coil is fine and that after it is bad...well, that tells a lot about its quality, doesn't it???

That would depend on what components radiate heat in the proximity of the PI filter as a whole (the ferrite coils in between capacitors can get hot as well and heat is conducted through the PCB substrate and traces).

And, if you notice, the ripple voltage results in the Hipro of this review are much higher than that of the polymod test (which also had Teapo SCs on the output), so it can be concluded that at least some of the Teapos in this review weren't afar from failure.

That's why the ripple voltage doesn't drop much with polys.

It didn't seem to increase at all, either.

[LongRunner]

How hard can it be to explain...

...which means more ripple current would have to be going -through- them to do that.

The mag-amp coils keep the ripple current flow pretty much independent of the characteristics of the capacitors. Their reactance at the switching frequency is on the order of ohms, and the capacitors are a fraction of an ohm at most. Saying what you've been told is like saying that running an appliance with a shorter power cord (which, of course, has lower resistance than a long one) will cause more current to pass through the cord - it makes no sense. The voltage drop (analogous to ripple voltage in the PSU filter) is less with the shorter cord, and I think that's a good enough explanation.

I think there's a reason why lower ESR CapXon KF and GL capacitors fail more often than the GP KM ones...

Probably because low ESR capacitors are harder to manufacture reliably.

That would depend on what components radiate heat in the proximity of the PI filter as a whole (the ferrite coils in between capacitors can get hot as well and heat is conducted through the PCB substrate and traces).

Sure it would. I guess I forgot about that when saying it.

And, if you notice, the ripple voltage results in the Hipro of this review are much higher than that of the polymod test (which also had Teapo SCs on the output), so it can be concluded that at least some of the Teapos in this review weren't afar from failure.

Sure, there are larger amounts of high-frequency ripple here, but the secondary capacitors would struggle to suppress 100/120Hz ripple and the coils would be pretty much a dead short circuit to it, so there's a limit to how low you can go. c_hegge did try a brute-force approach in an attempt to fix a problem resulting in massive amounts of such ripple under certain loads on the -12V rail of the Thermaltake Evo Blue 2.0 750W, using two large Panasonic caps (a 3300µF FC and a 4700µF FR), and even that didn't get it into spec.

That's why the ripple voltage doesn't drop much with polys.

It didn't seem to increase at all, either.

Yes, because the secondary caps do virtually nothing to 100/120Hz ripple. It would also be why the poly-modding tests show significantly lower ripple on the +3.3V and +5VSB, than on the other rails (even though the caps are similar).

[Wester547]

The mag-amp coils keep the ripple current flow pretty much independent of the characteristics of the capacitors. Their reactance at the switching frequency is on the order of ohms, and the capacitors are a fraction of an ohm at most. Saying what you've been told is like saying that running an appliance with a shorter power cord (which, of course, has lower resistance than a long one) will cause more current to pass through the cord - it makes no sense. The voltage drop (analogous to ripple voltage in the PSU filter) is less with the shorter cord, and I think that's a good enough explanation.

What about the ferrite coils? Do those play a much bigger part in suppressing ripple than the filtering capacitors? Since those have even lower resistance than capacitors do (often close to 0, very low in terms of ohms), I would imagine not (but people on Badcaps always act like no ferrite coils = higher ripple voltage...).

And, essentially, in the output of a SMPS that isn't flyback, are you saying that the output capacitors and ferrite coils don't do much to lower the ripple voltage? They're just there to further smooth things out, to speak rudimentarily?

Very true about the voltage drop. Longer cable means more conducting time, and thus the higher voltage drop.... and in other words, what I'm told is only true of output capacitors in flyback topology.

[LongRunner]

The output capacitors are still important - their ESR determines how much ripple voltage remains. (Let's not confuse voltage and current again...) The ferrite coils are indeed equivalent to a dead short at DC or 100/120Hz, but at high frequencies...see Wikipedia's article (I know they're not the ultimate source of information about everything but the principle is explained adequately). (Inductors operating at mains frequency have lots of turns, are built around massive laminated iron cores, are big, heavy and get hot.) If they did nothing, no-one would have installed them in the first place. PSU manufacturers know the role the components play.

The ripple current through the capacitor in a flyback is and can only be some portion of the output current.

(Late edit: Removed mention of ESL as it's not significant at PSU switching frequencies. It's a veteran member, no less, on Badcaps.net who told me there's "more to it" than ESR and that the value shown on the meters is an "approximation". Just goes to show that you have to be careful there.)

[Wester547]

I didn't mean to confuse ripple voltage and ripple current. Ripple voltage is measured in millivolts, ripple current in milliamps. Simple concept, really. In summation, though, you're saying that the toroidal coils ultimately do more to lower the ripple voltage than the output capacitors and ferrite coils do (despite them being important)?

[LongRunner]

The mag-amps take the lion's share of the pulsing voltage from the transformer and rectifiers, defining how much ripple current flows through the circuit, and the capacitors after them shunt the ripple current away, determining how much ripple voltage makes it out of the supply. Even using ceramic capacitors (which have the lowest ESR and ESL of all) won't increase the ripple current by much (of course they don't use ceramic capacitors in PSU output filters because of the expen$e). So the mag-amps control the ripple current and the capacitors control the ripple voltage. It's better to think of the ferrite coils, and capacitors following them, as "L" sections (well, rotating said L 180°), rather than combining the coil and capacitors on both sides of it into a "Π". However, if there's only one capacitor on a rail, it has to deal with the full ripple current from the mag-amp whether or not there's a ferrite coil before it.

Reliability-wise, Hipro would have been better off with a larger or higher grade capacitor before the ferrite coil on +5VSB, and a smaller capacitor after it.

[Wester547]

The mag-amps take the lion's share of the pulsing voltage from the transformer and rectifiers, defining how much ripple current flows through the circuit, and the capacitors after them shunt the ripple current away, determining how much ripple voltage makes it out of the supply. Even using ceramic capacitors (which have the lowest ESR and ESL of all) won't increase the ripple current by much (of course they don't use ceramic capacitors in PSU output filters because of the expen$e). So the mag-amps control the ripple current and the capacitors control the ripple voltage. It's better to think of the ferrite coils, and capacitors following them, as "L" sections (well, rotating said L 180°), rather than combining the coil and capacitors on both sides of it into a "Π". However, if there's only one capacitor on a rail, it has to deal with the full ripple current from the mag-amp whether or not there's a ferrite coil before it.

So what you're saying is, magamps control the amount of ripple current the capacitors and ferrite coils have to handle, and the capacitors and ferrite coils do more to keep the output ripple voltage low than the toroids do (or the toroids do the majority of the work by keeping the ripple current from the amount of pulses of the rectifiers and transformer in check)? Also, what if there's one capacitor and a ferrite coil after it on one rail, instead of one before? Sorry for all the questions, I just wonder what's worse for ripple voltage, an underspec'd magamp or an underspec'd output filter...

[LongRunner]

...and the capacitors and ferrite coils do more to keep the output ripple voltage low than the toroids do (or the toroids do the majority of the work by keeping the ripple current from the amount of pulses of the rectifiers and transformer in check)?

They're as important as each other for keeping ripple voltage under control, as using a smaller toroid that results in higher ripple current will result in proportionally higher ripple voltage, if you don't upgrade the caps to compensate.

Also, what if there's one capacitor and a ferrite coil after it on one rail, instead of one before?

Then you would be relying on the capacitors in the connected hardware to complete the filter. That may work, though, if they are good enough electrolytics, or (better) polys, or (better still) ceramic capacitors. But I don't think it would make sense as a design practice.