Bad new caps

[Conrad Hoffman]

I was building some boards for someone, with the parts supplied from the well known very large US supplier. The caps were brand name electrolytics in the 47-200 uF range, 10-50 V. Being the anal person I tend to be, I quickly checked a few on the bridge and did a double take. These brand new caps had dissipation factors as bad as 1.0 at 1kHz. When I'm servicing old equipment I get nervous when I see anything up over.5 or so for small caps. 1.0 would head straight to the trash. Not that they won't work, especially at 60Hz, but I've never purchased new brand name parts that were this bad. Do you really want a cap with a -53 degrees phase shift? It barely deserves to be called a capacitor. I'm not naming any names until I check out the data sheets for the specific parts, but I'll continue my suspicious ways and test every cap I'm about to install. One more time, just for the record, get a cap meter that reads DF, or use a high frequency esr meter. The simple DVM cap meters can't see this stuff.

CH

[rdf]

A brand new 470u 250V Fuhjyyu recently went into the 70 vdc bias supply of a project amp. It showed visible end bulging within a month. They're still around.

[tiefbassuebertr]

if you use the keyword "bad caps" by google, you get many informations e. g. this
bad caps - Google Search

have a look also about
Best electrolytic capacitors

[LAJ]

from an electronics supplier in Atlanta. They were wide open. They looked like old stock which is why I checked them. I went to the owner and he said that I could bring them back for exchange. But all he had was more of the same. He supplies to Georgia Tech also. Unbelievable!

[Conrad Hoffman]

Well, here's what I think. Low voltage electrolytics are inherently not very good at anything above line frequency. Some are probably worse than others (like these), but all should be avoided. It's said that you shouldn't run an electrolytic way under its rated voltage, but all my experience says it's an old wives tale. Any slight loss of performance over the years by using say a 63V cap in a 5V circuit, is negligible compared to the overall poor performance of a 10V (or lower) cap to begin with. I've also seen a much higher failure rate over the years with low voltage caps. When I rebuild test equipment, it's always the LV caps that go first. The case size of caps keeps getting smaller and that's not a good thing either.

The cap in question was a Panasonic 220uF, 10V part, EEU-EB1A221S (Digikey P13456-ND). At 120 Hz it looked ok, 211.6 uF with D=0.178 (series measurement) At 1kHz it was 197.0 uF with D=1.19 (series measurement). That's equivalent to a parallel combination of 1.64 ohms and 81.5 uF, giving a total Z of 1.2 ohms. Yes, it'll work, but don't expect great HF noise filtering.

Just for comparison, a NOS Rubycon YXF out of my junk box, also 220uF but 35V and a slightly larger case size, measured 196.6uF with D=0.194 (1kHz series measurement). That's equivalent to a parallel combination of 4.33 ohms and 189 uF, giving a total Z of 825 milliohms. Remember, these are AC model equivalents- I'm not suggesting a short circuit!

Note that soaking the 10V cap at 10V didn't improve things any, in fact it got slightly worse.

Ignoring all the numbers, if you can use a 25V or greater electrolytic, do so, even in a low or no voltage circuit.

Ok, these are garden variety caps. You want something really good? How about an Illinois KXM series, 220uf/35V? 1kHz is 202.1 uF with D=0.079 (1kHz series measurement). That's equivalent to a parallel combination of 10 ohms and 201 uF, giving a total Z of 789 milliohms. That's pretty much the Z of the reactance alone. More importantly, it maintains good performance at much higher frequencies, and is a perfect part for switching supplies because it won't overheat and fail. For a 60 Hz supply, I'd still use the Rubycon.

CH

[a.wayne]

Originally Posted by Conrad Hoffman

Well, here's what I think. Low voltage electrolytics are inherently not very good at anything above line frequency. Some are probably worse than others (like these), but all should be avoided. It's said that you shouldn't run an electrolytic way under its rated voltage, but all my experience says it's an old wives tale. Any slight loss of performance over the years by using say a 63V cap in a 5V circuit, is negligible compared to the overall poor performance of a 10V (or lower) cap to begin with. I've also seen a much higher failure rate over the years with low voltage caps. When I rebuild test equipment, it's always the LV caps that go first. The case size of caps keeps getting smaller and that's not a good thing either.

The cap in question was a Panasonic 220uF, 10V part, EEU-EB1A221S (Digikey P13456-ND). At 120 Hz it looked ok, 211.6 uF with D=0.178 (series measurement) At 1kHz it was 197.0 uF with D=1.19 (series measurement). That's equivalent to a parallel combination of 1.64 ohms and 81.5 uF, giving a total Z of 1.2 ohms. Yes, it'll work, but don't expect great HF noise filtering.

Just for comparison, a NOS Rubycon YXF out of my junk box, also 220uF but 35V and a slightly larger case size, measured 196.6uF with D=0.194 (1kHz series measurement). That's equivalent to a parallel combination of 4.33 ohms and 189 uF, giving a total Z of 825 milliohms. Remember, these are AC model equivalents- I'm not suggesting a short circuit!

Note that soaking the 10V cap at 10V didn't improve things any, in fact it got slightly worse.

Ignoring all the numbers, if you can use a 25V or greater electrolytic, do so, even in a low or no voltage circuit.

Ok, these are garden variety caps. You want something really good? How about an Illinois KXM series, 220uf/35V? 1kHz is 202.1 uF with D=0.079 (1kHz series measurement). That's equivalent to a parallel combination of 10 ohms and 201 uF, giving a total Z of 789 milliohms. That's pretty much the Z of the reactance alone. More importantly, it maintains good performance at much higher frequencies, and is a perfect part for switching supplies because it won't overheat and fail. For a 60 Hz supply, I'd still use the Rubycon.

CH

Would this be the same for PS caps .... say 35 v rails would be better off with 100V PS caps ...

FWIW , i do remember being told this years ago about electrolytic or caps in general to use voltage ratings @ least double what is needed ..

[Conrad Hoffman]

Once you get up to 25-35V I don't think it matters much. The really low voltage parts just seem to be iffy on their losses. For safety factor I try to estimate what the worst case high line voltage might be, and how high the supply might go under that condition and with no load on it. Then I add maybe 10-20%. Caps already have a surge rating that's usually higher than the continuous rating, and I think doubling the voltage might increase the size and cost excessively for what's gained. A typical audio supply running at 40VDC with 50VDC caps seems about right to me, though I've no doubt someone can make a case for 75 or 100V caps. I'm just not that paranoid, and you should always design with the assumption that the caps could fail at some point- choose fuses and such accordingly.

CH

[h_a]

Hi Conrad,

thanks a lot for your numbers, that's great and rare information!

I'm really surprised the Panasonic EB perform so poorly. While I have never used the EB, I use massively the NHG for small current applications and TS-UP or TSHA for the cans. To rebuild SMPS I use FC or FM and honestly I never had a single problem (overheating, failure...) with any of the Panasonic parts.

Thanks also for mentioning good ones!

Now another difficulty is to get high quality axial caps, I have a signal generator here that's full of it. Nearly no choice in parts available for axials.

Have fun, Hannes

[djQUAN]

Originally Posted by Conrad Hoffman

Well, here's what I think. Low voltage electrolytics are inherently not very good at anything above line frequency. Some are probably worse than others (like these), but all should be avoided. It's said that you shouldn't run an electrolytic way under its rated voltage, but all my experience says it's an old wives tale. Any slight loss of performance over the years by using say a 63V cap in a 5V circuit, is negligible compared to the overall poor performance of a 10V (or lower) cap to begin with. I've also seen a much higher failure rate over the years with low voltage caps. When I rebuild test equipment, it's always the LV caps that go first. The case size of caps keeps getting smaller and that's not a good thing either.

The cap in question was a Panasonic 220uF, 10V part, EEU-EB1A221S (Digikey P13456-ND). At 120 Hz it looked ok, 211.6 uF with D=0.178 (series measurement) At 1kHz it was 197.0 uF with D=1.19 (series measurement). That's equivalent to a parallel combination of 1.64 ohms and 81.5 uF, giving a total Z of 1.2 ohms. Yes, it'll work, but don't expect great HF noise filtering.

Just for comparison, a NOS Rubycon YXF out of my junk box, also 220uF but 35V and a slightly larger case size, measured 196.6uF with D=0.194 (1kHz series measurement). That's equivalent to a parallel combination of 4.33 ohms and 189 uF, giving a total Z of 825 milliohms. Remember, these are AC model equivalents- I'm not suggesting a short circuit!

Note that soaking the 10V cap at 10V didn't improve things any, in fact it got slightly worse.

Ignoring all the numbers, if you can use a 25V or greater electrolytic, do so, even in a low or no voltage circuit.

Ok, these are garden variety caps. You want something really good? How about an Illinois KXM series, 220uf/35V? 1kHz is 202.1 uF with D=0.079 (1kHz series measurement). That's equivalent to a parallel combination of 10 ohms and 201 uF, giving a total Z of 789 milliohms. That's pretty much the Z of the reactance alone. More importantly, it maintains good performance at much higher frequencies, and is a perfect part for switching supplies because it won't overheat and fail. For a 60 Hz supply, I'd still use the Rubycon.

CH

in that case, I think my ESR meter (100kHz test freq) should show common problems easily?

[Conrad Hoffman]

Yes, those caps should look terrible on a high frequency esr meter! IMO, if one doesn't have a bridge that gives DF, an esr meter is almost essential to troubleshooting. Just knowing a capacitance value isn't enough.

For axial lead caps, I've found a huge number on the surplus market in years past. In a pinch, I've certainly soldered a wire to one lead and folded it back over to make a radial fit where an axial went, but I'm never happy about it.

CH