Hardware Insights Forum

My suggestions:

• Surely there should be some penalty for using non-safety-rated capacitors in the EMI filter? What about for other safety issues, such as improper earth connections, or inadequately insulated mains wires?
• Could cross-loading tests be done after finishing the main tests? What about testing for power-on transients?
• How about commenting on how easy (or difficult) repairs can be expected to be? The ideal here is to use connectors both for the mains input and for the fan, and to use only standard-sized capacitors; it's not that much to ask, really. It's also nice if the fuse is mounted in a holder rather than soldered to the PCB (it's not that unusual for fuses to fail simply from fatigue even when there's no actual fault).
• And on the topic of fuses, they are available with either glass or ceramic cases. The ceramic types have a sand filler that helps absorb the arc created when the fuse blows; glass types do not (as you can see by looking at one), and have been known to explode if the fault current exceeds what the fuse is designed for (usually only 10× the operating current limit; ceramic fuses, on the other hand, can take as much as 1.5kA, which would take a very beefy mains supply to exceed). The NTC thermistor may or may not have enough resistance for a glass fuse to fail safely; if it doesn't (as is probable when it's hot), all bets are off.
• Would it be a realistic option to obtain a variable autotransformer (a.k.a. Variac™) and use it to vary the mains voltage into the PSU to see how it reacts (after the main tests have passed)? I suggest one rated for at least 5A (ideally more), and the unit should take at least from 185~255VAC or so without issue to be considered a "pass" in this respect.

Oh, and while you're at it, could you finally update those outdated articles I asked about last year?

LongRunner wrote:Surely there should be some penalty for using non-safety-rated capacitors in the EMI filter? What about for other safety issues, such as improper earth connections, or inadequately insulated mains wires?

Yup, and I will be knocking a point off for it in the el-cheapo PSU roundup this year (as well as modifying the tests to suit my newer load tester, and one other small surprise...)

LongRunner wrote:Could cross-loading tests be done after finishing the main tests? What about testing for power-on transients?

I've actually been considering that one myself. I'll probably start doing that after the next review or two.

LongRunner wrote:How about commenting on how easy (or difficult) repairs can be expected to be? The ideal here is to use connectors both for the mains input and for the fan, and to use only standard-sized capacitors; it's not that much to ask, really. It's also nice if the fuse is mounted in a holder rather than soldered to the PCB (it's not that unusual for fuses to fail simply from fatigue even when there's no actual fault).

I dunno. Us who actually repair PSUs are a minority. 9 users out of 10 couldn't care less about that. Besides, you can figure most of that out just from the internal shots.

LongRunner wrote:Would it be a realistic option to obtain a variable autotransformer (a.k.a. Variac™) and use it to vary the mains voltage into the PSU to see how it reacts (after the main tests have passed)?

Maybe one day, but it's out of the budget at the moment.

LongRunner wrote:Oh, and while you're at it, could you finally update those outdated articles I asked about last year?

When I have time. I'm still not totally done with the load tester yet. There are some modifications which I still want to make to it, such as a screwed-on hot box, and some ammeters that don't fail every time a PSU blows a secondary rectifier.

c_hegge wrote:

LongRunner wrote:Surely there should be some penalty for using non-safety-rated capacitors in the EMI filter? What about for other safety issues, such as improper earth connections, or inadequately insulated mains wires?

Yup, and I will be knocking a point off for it in the el-cheapo PSU roundup this year (as well as modifying the tests to suit my newer load tester, and one other small surprise...)

On that note, I should also add a point of clarification about the mains capacitors.

The colour and texture of the coating of the ceramic capacitors means nothing in particular. Indeed, there are many plain 2kVDC ceramic capacitors that are the same light blue colour as most Y-class units; I can identify those in many of the¹ cheap units reviewed² here (each³ word links to one).

¹Can't actually read their markings from the angle of the shot, but they do look too thin to be the real thing — and with the "bonus" of the exploded glass fuse, I think some penalty for using glass fuses is certainly in order.
²This one also had 400VDC film capacitors instead of X2 types. I suppose you weren't looking too closely when you reviewed it, were you? (I mentioned it in the discussion thread for the round-up, but perhaps you glossed over it.)
³Well, in this one they're only rated for 1kV, which is even worse.

Similarly, non-safety-rated capacitors in rectangular cases exist as well as (less commonly) coated X-class units. The markings are the only reliable indicator, though even they can be faked by counterfeiters that pay enough attention to detail.

And repair may not be in fashion now, but eventually the whole world will have to face the fact that waste is not an option, and replacing a few bad capacitors or a seized fan is certainly less wasteful than discarding the entire unit (so long as the repairer doesn't zap themselves, anyway). Of course, at the moment waste is just out of control — I've heard of computers that are disposed of without even being used. Whatever it takes to make a fair world — one where new stuff is bought only as actually needed, where whatever can be repaired is repaired, where quality takes priority over price, where corporations aren't so overpowered, where the rich don't rip off the poor, etc. — we just have to do it.

I'm curious, by the way, if there's anything that can result in a PSU getting a negative score (and presumably an Epic Fail Award ); presumably extreme safety hazards would be it.

LongRunner wrote:On that note, I should also add a point of clarification about the mains capacitors.

The colour and texture of the coating of the ceramic capacitors means nothing in particular
[snip]
Similarly, non-safety-rated capacitors in rectangular cases exist as well as (less commonly) coated X-class units. The markings are the only reliable indicator, though even they can be faked by counterfeiters that pay enough attention to detail.

I know. You really have to look for the approval logos (UL, etc)

LongRunner wrote:And repair may not be in fashion now, but eventually the whole world will have to face the fact that waste is not an option, and replacing a few bad capacitors or a seized fan is certainly less wasteful than discarding the entire unit (so long as the repairer doesn't zap themselves, anyway). Of course, at the moment waste is just out of control — I've heard of computers that are disposed of without even being used. Whatever it takes to make a fair world — one where new stuff is bought only as actually needed, where whatever can be repaired is repaired, where quality takes priority over price, where corporations aren't so overpowered, where the rich don't rip off the poor, etc. — we just have to do it.

I doubt very much if will happen. As long as wages and standards of living remain high, it's always gonna cost more to fix a PSU (once you pay someone for their time to do it), than to buy a new one. Besides, I'd have to knock points off 99% of PSUs for not being as easy as they could be to fix. Personally, I think that having the AC wires soldered to the PCB is nothing compared to using a dual-layer PCB. Those things are virtually impossible to take heatsinks off without damaging. With a single layer PCB, even a 40W soldering iron with some braid is quite capable of desoldering any component.

LongRunner wrote:I'm curious, by the way, if there's anything that can result in a PSU getting a negative score (and presumably an Epic Fail Award ); presumably extreme safety hazards would be it.

Nope, I don't ever go below 0. See http://hardwareinsights.com/the-2013 ... roundup/7/. I took all 10 points off just for the high ripple, and continued to find problems.

c_hegge wrote:I doubt very much if will happen. As long as wages and standards of living remain high, it's always gonna cost more to fix a PSU (once you pay someone for their time to do it), than to buy a new one.

Maybe wages and standards of living are high for us, but not for factory workers. In a fair world, of course, people worldwide would be paid equally, and presumably then repair would be more economical.

Besides, I'd have to knock points off 99% of PSUs for not being as easy as they could be to fix.

I'm not saying you'd have to do that, just that it would be worth mentioning (at least what isn't obvious from the photos). Even I wouldn't knock more than 1 point out of 10 at most for a unit being difficult to repair, so a unit that would otherwise achieve a perfect score would still get a silver award.

Personally, I think that having the AC wires soldered to the PCB is nothing compared to using a dual-layer PCB. Those things are virtually impossible to take heatsinks off without damaging. With a single layer PCB, even a 40W soldering iron with some braid is quite capable of desoldering any component.

Fair enough, but double-sided boards are often unavoidable, however difficult they are to work on. The most that would have to be compromised to use a connector for the mains wires would be using a slightly larger PCB.

The original Macron MPT series did get pretty close to ideal repairability, with a fuseholder and connectors for the mains and fan, and capacitors that are all easy to obtain replacements for; there aren't many units that are easier to repair (the trickiest part is replacing the two 10µF caps in the base driver, if you choose to which is optional), not that the series overall is great of course.

Nope, I don't ever go below 0. See http://hardwareinsights.com/the-2013 ... roundup/7/. I took all 10 points off just for the high ripple, and continued to find problems.

Sure, it's a major failure, but nothing that really defies belief. What would be worthy of a negative score (IMO) would be if they used an "earth" wire so thin that it melts if the unit actually shorts to earth (thus defeating the purpose), or if they left so little clearance between the primary and secondary that a simple solder blob could bridge the gap (which, depressingly, has already manifested itself in the form of cheap-and-deadly plugpack supplies).

On some PSUs (such as In Win/Power Man IP-P410Q3-2), the rear panel flexes noticeably under the force required to connect/disconnect the power cord — so the rigidity of the power inlet mounting is another thing we should test. I suggest going through your stash of cords (presuming that, like me, you have such a stash), and choosing one of the ones with tighter socket contacts for the routine.

For that matter, many cheap-and-nasty PSUs have cheaply made power inlets (and pass-through outlets if equipped) with only Chinese approval (if any). Those inlets have not the proper solid pins with nice chamfered points, but pins made from a folded brass strip (still nickel-plated, but I've never seen an IEC inlet with pins that aren't nickel-plated; let's set aside the gold-plated audiophool stuff), resulting in rounded "points". That may require more force to connect/disconnect the cord, and I don't know if bad things could result from doing it too often. (The cords with an IEC-type plug on one end make a convenient reference for the proper pin shape.)

Actually, the cable I use normally is a bit of a tight squeeze. You get a little bit of flex on 99% of PSUs. There were these PSUs that occasionally showed up in PC repairs branded "TT" (not Thermaltake), which were usually low end Solytech builds, and they had exceptionally flimsy cases. I haven't seen one for a while now, but I did see a couple which were permanently bent inwards at the back. I would absolutely score against that.

As for the IEC socket itself, I've never ever seen a damaged one. Nor have I heard any widespread complaints from others about it happening. I'm sure some may not be as solid or well made as others, but I don't really like scoring against things which I've never experienced problems with.

I have in plan to obtain (or even wind myself as I have lot of big cores from bad UPSes) fixed uninsulated transformer for testing line-interactive UPSes with under- and overvoltage. I am also thinking about winding it with deviation route in the middle for 115 V AC, even though I can actually get one such transformer (+-2 kVA) right away. So I can possibly test even PSUs with that. There's a possibility I'll supply c_hegge with reviews so your wish may come true sooner than you thought

ADD// I went through your testing methodolgy C_h and you have one big problem there, you do not comply with Intel's ATX methodology to measure ripple. You measure some ripple in there, but it is not the ripple you want to measure. You have to add decoupling capacitors and measure the ripple on them.

Actually, you can measure whatever you want, but as long as you do not even try to comply with ATX (ripple measurement is somewhat alchemical whatever you do, you should not only have those decoupling capacitors, but also differential probe etc.), it is somewhat stupid to condemn PSUs they do not comply either, right?

^
I know. It's a 1uF lytic and a 0.1uF ceramic required by ATX specifications, and they are present in the load tester. The way I have it is that there is one set of them for each rail, with the anode side soldered to an input pin of rail select switch, and the cathode side soldered to ground (and reversed for the -12V rail).

It is 10uF, Intel used tantalum cap so I picked up something close enough (low ESR aluminium). You also should measure on the caps themself, not far away from them. I first thought it does not matter, but it does.