OT PC trouble - ghost in the machine?

[J Tiers]

With maker and series number along with value and voltage, the design ESR can be determined. No need to guess, unless the parts are from a "china only" vendor with no english on the site (that was becoming more and more common, now that US companies are irrelevant).

That PWB looks "bottom of the barrel"......

[Mcgyver]

Am I doing this properly?

cap SMQ 100uf 25V.

http://www.chemi-con.co.jp/cgi-bin/C...=p&pdfname=smq

Xc for 100uf @ 120KHz = 0.0159
DF from datasheet = .16
theoretical ESR = Xc * DF = .002 ohms

I'm measuring around.046 ohms on one, and 3 1/2 ohms on three.

1) i'm using a calibrated meter with Kelvin probes so it should be accurate
2) measurements done in circuit
3) .002 ohms calculated seems really low compared to every cap I measure
4) 3.5 ohms seems really high for any good cap i measure - this is about the same as I got for the bad one in the PS
5) the fact that 3 are the same, and one is about 1/10 the ESR measure can't be explained by datasheet or calcs (or at least I can't explain it)
6) all of the measured ESR's differ from calculated by a factor of 10x or 100x approximate....so I'm thinking I'm not getting it. all my measurements were at 120 kHz

Thoughts?

[J Tiers]

My first thought is I thought I recalled that logo mark, but I'll have to look it up.

Yes, presumably SMQ series, with value and voltage, Nippon Chemi-Con.

Yes, found SMQ series http://www.chemi-con.co.jp/e/catalog...smq-e-2018.pdf

Part shows tan(d) of 0.16 at 120 Hz not KHz. I find that the ESR= DF * XC at 120 Hz is 2.1 ohms. 1/(6.28 x 120 x ,0001F) * 0.16 = 2.1 ohms. So the actual spec is 2.1 ohms ESR at 120Hz.

If you are measuring in-circuit, there are many things which can foul up the measurement, pkus you need to measure at the frequency that the spec is stated at. The spec value will not scale, since the component may or may not even be a capacitor at 120 kHz. All capacitors have self-resonant frequencies, they are capacitive below, and inductive above, plus they have an internal actual resistance.

[Glug]

Nice fix! Almost as good as your old tool restorations

[PStechPaul]

It seems the SMQ series are designed for line frequency (120 Hz) bulk filtering and energy storage, and not low ESR as needed for high frequency bypassing of logic circuits. Also, there are quite likely other capacitors, probably ceramic, in parallel, and in-circuit ESR measurement is not accurate. If one reads 0.046 ohms, it is probably not directly in parallel with the others, and it may have low ESR caps in parallel that you are reading with the LCR meter.

[macona]

The time for most of the bad caps has long past, yeah, you will still get one now an again like in the power supply, but the big issue was in the P4 era. I had to redo a single board computer for an industrial PC (P4) because of the caps, when they are bad the vents usually break open on the top. A couple won't do much but get 10-15 bad on a board and the thing will start acting weird. If the machine is running stable I would leave well enough alone, between RoHS solder and the heavy ground planes it can be difficult to change some caps and you risk messing up the board.

90c isn't too out of the ordinary for a modern processor running under a load, it is the top end though. It just really depends on the processor.

Yeah, machines have not gotten a whole lot faster. I am still running a 3rd gen i7 for my gaming machine and I rarely run into a game that will tax the cpu. The video card is another thing, I have a GTX1080 for that.

[macona]

Oh, and those caps are around the audio chip, so they are probably just there to provide decoupling on the audio output of the chip. I am betting if you looked at the data sheet you would find electrolytic caps in series on the audio lines.

[Mcgyver]

Part shows tan(d) of 0.16 at 120 Hz not KHz. I find that the ESR= DF * XC at 120 Hz is 2.1 ohms.

I see where the norm is stating the tan d at 120 Hz. Most/many cap spec sheet give the ESR @ 100kHz as I believe you use higher frequencies like that to somewhat reliably measure ESR in circuit. I missed that the tan d was associated with a frequency (not stated, but by convention).

It seems like, at least for this spec sheet, you do have to guess at what ESR to expect at 100 kHz? I know ESR's vary with different caps, but any 100uf cap I try is less than 0.5 ohms @ 100kHz, and I know there are other things in the circuit....but that could only make the ESR lower, not higher, right?

i'll put it back together and see what happens. Thanks again for all the help

[J Tiers]

The 120 Hz is not by convention, it is explicitly stated on the first page of the data sheet, by voltage value, and also at the top of the spec column in the table on p2. It will always be stated, as it is an essential part of the spec.

Basically, if the DF (tan (d)) is given at 120 Hz, there is NO information given about performance at 100 kHz. It just is not given, so the wise designer does not design-in that part for an SMPS operating up there, if there is anything critical about the ESR at that frequency. If it is not stated on the data sheet, you cannot count on it. Conversely, if it IS stated, it WILL meet that spec, you just have to be quite careful extrapolating to non-spec conditions.

You CAn guess a bit about the part by looking at the last table, the "ripple current multipliers", but it is better to stick with what is actually spec'd. Even a graph is not a spec, it is typical, and generally so stated. The spec is what is written and noted with value and conditions. New designers are often caught by things like that. It is a bit like legal papers in that way, and for good reason, as the spec may become an exhibit in a lawsuit, it must provide a clear description of what is guaranteed and what is not.

Low ESR parts are made differently, and typically cost a bit more, whether that price is due to construction, or just "because we can". For a high volume consumer product, the price is king, and it is worth spending time to select parts which are no better than they have to be, even of it means more part types to stock, since they are machine inserted, and come on reels.

If the associated chip is for audio, that would not typically need HF performance, although many audio systems are now PWM, even for small power. The caps may be PS bypass, or they may be in series with the output as DC blocking, and with 4, likely both.

[MattiJ]

agree, it could be the PS. Its Rockfish 700 watt not that old....doesn't mean it can't fail of course, but its suppose to a quality unit not the cheapest of the cheap. it also powers up instantly (by shorting the pins) and there is the correct voltage on all pins. The MB only starts sometimes (either from front panel or on board switch). Not conclusive, maybe hints its a MB problem?

https://www.hardwaresecrets.com/rock...ply-review/10/

Power supply doesn't look like worth a repair, craptastic generic junk. Hatefull amount of low quality caps and if you want to make it last you need to replace them ALL. And you still have mediocre PS after that.

Look for 80+ Platinum or 80+ Titanium energy efficiency class as it seems to weed out the weakest crap. https://www.newegg.com/Product/Produ...166&IsNodeId=1
My personal vote would go for Seasonic ie:
https://www.newegg.com/Product/Produ...82E16817151198
https://www.newegg.com/Product/Produ...82E16817151197