Hardware Insights Forum

c_hegge (in the review) wrote:The input filtering consists of one X-Capacitor, one coil, two Y-Capacitors and one MOV. This is fewer components than recommended – there should be an extra coil and an extra X-Capacitor. Another problem is that the X-Capacitor is not UL certified…

Actually, I'm pretty sure what you're referring to is the film capacitor used after the rectifier bridge in any half-decent active PFC circuit. That capacitor is not used for low-frequency ripple smoothing (if it was large enough to do that, the PF would be bad again); it just filters out the high-frequency current pulses produced by the booster, and increasing its size can partially compensate for using less AC-side filtering. As far as I know, there is no requirement to safety rate it. The bad news, of course, is that there are zero capacitors operating on live-to-neutral in this unit.

If you take a close look at the PCB, you will see that there is an unused position for an X2 capacitor labelled CX1, but the capacitor mentioned is installed at C101 (no X).

Oh, by the way: Where are the datasheets for Yang-Chun capacitors? A Google search returned nothing very useful [there's too much unrelated crap in the way], there's no folder here, and I don't see fit at all to outright return to Badcaps.net itself after all of Topcat's disrespect. If you're staying there that's fine by me, but I don't see fit to actively encourage (whether explicitly or by implication) bad behaviour.

I fixed that bit about the X-Cap

As for Yang Chun, I don't know. They don't have a website, so my guess is that they don't publish their datasheets at all.

Nice review, c_hegge.

It's definitely Solytech. The Yang Chun capacitors, and that transformer has the same exact layout and font as the new Allied units.

A few things impress me about this unit, besides the silicon. I can't believe that toroid can handle the load you asked from it. Also, even though it went way out of spec, I'm amazed that two 1000µF caps and a coil kept the ripple down that much. I bet some 1500µF nichicon PW parts there would solve the ripple issue on the 12V.

What's the rating of the bridge rectifier? Are you going to recap it? If not, I want it

I did recap it. For the 12V, I used a 1000uF 16V Nich PW for the cap before the coil, and a 2200uF 16V Panny FK after. It was only about 10mV lower on the 12V rail, and no better at all on the other rails. I also polymodded the VRMs with 330uF 16V OS-CON SEP, and 1000uF 6.3v Fujitsu RE caps.

All that thing really needs is just 50W knocked off the rating and the extra PCI-E connector. It would have at least got a score above the half-mark.

And here's some pictures to show for it

For what it's worth, my preferred strategy in the absence of documentation for the original capacitors would be to go with the lowest ESR (or maybe the ripple current rating itself in some cases) attainable without sacrificing the endurance rating or capacitance, and then test it to ensure that it still behaves — as I have yet to see truly robust proof of lower ESR being a substantial risk, but there is no doubt whatsoever that too high ESR will accelerate eventual failure, even if the ultimate ripple voltage is below the half-line at the start. (In a full pi filter, which is classed as "second-order", doubling the ESR of both capacitors will approximately quadruple the ripple voltage. As the specifications usually permit a doubling of ESR (tripling for some series) within the endurance span, the best practice is to design so that, when new, the ripple voltage will be below a "quarter-limit". Keeping their temperature lower will compensate somewhat, though.)

So the factor by which you need to stay below the absolute ripple voltage limit, to run the capacitors to their full endurance rating, is approximately:

(specified ESR multiplier)^(filter order)

Of course, if you're mixing capacitors with different multipliers, it will get somewhat more complicated — you will have to take the multipliers for each capacitor (or group of parallel caps to be more precise) and multiply them all together. But the simplified option will usually work when you are using the same series throughout the filter in question.

Note, though that filter orders as applied here are based around ESR's dominance of electrolytic capacitors' impedance at the switching frequencies. With ceramic capacitors, for example, the formulae would be quite a bit different…

Nice! It looks good recapped. Did you solder on another PCI-e connector?

I could be wrong here, but is it possible that the (maybe) same specced 1000µF nichicon PW has a similar ripple current rating as the Yang Chun LE, not improving some of the ripple? We have no way of knowing because there are no datasheets for YC caps, like you mentioned. Maybe that's too low of a capacitance for the cap before the coil. Maybe a 2200µF cap would improve the ripple further

What series is that Rubycon cap next to the transformer in the 3rd pic? (Just curious)

Also, did you replace the PFC filtering cap as well?

Pentium wrote:Nice! It looks good recapped. Did you solder on another PCI-e connector?

No. I can't imagine myself putting it in a PC which needs two connectors. There are only a hundful of such PCs in this area, and I know that those who own them would be willing to spend their money on something better, like a Seasonic S12II Bronze or G series.

Pentium wrote:I could be wrong here, but is it possible that the (maybe) same specced 1000µF nichicon PW has a similar ripple current rating, not improving some of the ripple? We have no way of knowing because there are no datasheets for YC caps, like you mentioned. Maybe that's too low of a capacitance for the cap before the coil. Maybe a 2200µF cap would improve the ripple further

Possibly. I used them mainly because I had a few spare ones on hand. I don't really want to play with it too much, when I know that it's still quite capable of delivering 500W in spec, and I doubt if it will ever be pushed harder than 200W.

Pentium wrote:What series is that Rubycon cap next to the transformer in the 3rd pic? (Just curious)

YXG series

Pentium wrote:Also, did you replace the PFC input cap as well?

No. I don't see enough failures there to make me feel the need. It's 450V rated too, so it has more headroom for the PFC to misbehave than some (even if it is only an 85*C part).

No. I can't imagine myself putting it in a PC which needs two connectors. There are only a hundful of such PCs in this area, and I know that those who own them would be willing to spend their money on something better, like a Seasonic S12II Bronze or G series.

Good point.

Possibly. I used them mainly because I had a few spare ones on hand. I don't really want to play with it too much, when I know that it's still quite capable of delivering 500W in spec, and I doubt if it will ever be pushed harder than 200W.

Yeah, 500W is definitely good enough. I was also mentioning it for the sake of discussion, since we don't know the specs of the YC cap

No. I don't see enough failures there to make me feel the need. It's 450V rated too, so it has more headroom for the PFC to misbehave than some (even if it is only an 85*C part).

I agree, I was just curious. I've only ever seen one PFC filtering cap fail before, and it was a CapXon in an OCZ branded FSP (Big surprise!)

CapXon seem to be fine on the voltage doubler configuration though