Hardware Insights Forum

(Get your resistors ready.)

I have seen - on Badcaps.net, no less - a "statement" that the output filters in PSUs are tuned to resonate at the switching frequency to remove ripple most effectively. But it's never going to work that way, for several reasons:

1. Electrolytic capacitors are nowhere near precise or stable enough to be used in a resonant circuit in the first place.
2. Nevermind tolerances and aging, their quality factor is simply far too low at the switching frequency to resonate even if you did get the value right.
3. And if you did make the filter resonate (for the purpose of this explanation, the component arrangement is what would form a series resonant circuit), you would get hundreds of volts of ripple and your entire PC would be instant ash.

No doubt there is such a thing as "too low" ESR/ESL, but not for "that" reason. In truth, the capacitors used are probably not the best available but instead chosen with cost in mind. I'm curious about the practical effect of using caps with different specifications. The reason for showing the ripple waveforms is to distinguish between 100/120Hz ripple that makes it through to the outputs and the actual switching noise (the former not being attentuated much by the filters).

Please stick this thread.

Well, I thought deviating from the ESR of the stock capacitors contorts what the control loop is "tuned" to (so the real problem would be oscillating).

Wouldn't the control loop's response be affected by the bulk capacitance, not ESR or ESL? If the poly-mod article is any indication:

Antec SP-400 - normal ripple results but couldn't boot the PC. +12V capacitance was probably too small - only 1 substantial cap after the ferrite coil, and a tiny one before it.
Hipro HP-D3057F3H - worked as normal, likely as it has 2 caps on +12V.
Antec NeoHE 550W - control loop knocked out. Not entirely sure whether the ESR or capacitance is at fault, but to hazard a guess, the latter.

The connected devices do, after all, have capacitors of their own, commonly poly (motherboard) or ceramic (e.g. HDD), and they obviously handle those just fine.

True. Actually, if the additional filtering on the motherboard and the video card even is good enough, the ripple voltage from the power supply can be reduced to one quarter of what it is by the additional filtering, by the time it reaches the motherboard, if I'm not mistaken. Of course, the capacitors on the motherboard have to work harder, especially the higher the ripple voltage, but even still. On the topic of transient response, while the Hipro did do very well both by way of the load tester and being tested in a computer after the polymodding, I do not think load testing at least is a true telltale indication of how voltage regulation is inside a computer because the load is constant and stable while it is not always consistent and very erratic inside a computer where the CPU usage could easily jump from 0% to 50% to 5% to 25% to 100% within the span of two seconds (I think that would be harder for voltage regulation, and that is where bulk capacitance in the output filter of a power supply and in the VRM circuits of a motherboard comes in handy).

If the control loop isn't the reason for ESR being too low, would it have something to do with the output filter itself?

Wester547 wrote:If the control loop isn't the reason for ESR being too low would it have something to do with the output filter itself?

I don't actually have a sure-fire answer to that.

I'm pretty sure it is the feedback loop that is affected when you drop the ESR too much. I remember one post on badcaps which more or less demonstrates this. The PSU was originally re-capped with Panny FM and afterward squealed when running (usually a telltale sign that the feedback loop is not happy, if it occurs after replacing parts). After replacing one of the caps with NCC LXZ, the squeal was gone. I actually believe that it was the ESR of the polies which caused the Antec NeoHE to go crazy after the poly mod.

c_hegge wrote:I'm pretty sure it is the feedback loop that is affected when you drop the ESR too much. I remember one post on badcaps which more or less demonstrates this. The PSU was originally re-capped with Panny FM and afterward squealed when running (usually a telltale sign that the feedback loop is not happy, if it occurs after replacing parts). After replacing one of the caps with NCC LXZ, the squeal was gone. I actually believe that it was the ESR of the polies which caused the Antec NeoHE to go crazy after the poly mod.

It's hard to say without further testing. goodpsusearch recapped an MPT-301 with Panasonic FM on +12V and FK elsewhere. It did whine at first, but eventually that subsided. The last I knew, there was no problem.

For what it's worth, FM uses aqueous electrolyte while LXZ is non-aqueous, and maybe there's a functional difference between the two categories that I don't know about. That, and maybe the FM mentioned there was brand new and the LXZ already used (I don't recognise the thread you're referring to).