Interesting Capacitor Observations

[CatalinaWOW]

Had some spare time and started organizing my junk bins of electrolytic capacitors.  These are mostly pulls and floor sweepings, with a few project leftovers and the like.  Used a C meter to evaluate each one.  A couple of hundred capacitors (yeah I needed a break from thinking or doing anything really useful).

1st observation - roughly 25% were duds.  No surprise here.  Electrolytics are notorious for drying up.

2nd observation - 50% were 1000 uF.  Not expected, but not a huge surprise.  These things are usually used for general filtering, and are selected based on intuition, price and other things rather than a calculated frequency response.

3rd observation - 20% were 100 uF.   Similar to 1000 uF, but second choice in the winner pool.  There were no other standout values which ranged from 1 uF to 20000 uF.

Now comes the really interesting thing.  Because they make so many of them, they apparently get really good at the 1000 uF capacitors.  The junk box had several brands and several voltage values, but there were no duds in this value.  Not only were there no duds, but 90% of these were within 5% of nominal value, and none over 10% out.  Much better than you would expect from electrolytics,  particularly since they were probably used for general filtering.

The 100 uF capacitors were not as good, but still better than average electrolytics.  One dud, and two that were 200% of nominal value. 

My conclusion from all of this is that if you want reliability in your circuit, and need to use electrolytic capacitors, select one of the magic values.  The sampling process used to draw this conclusion is not statistically valid so YMMV.

[SeanB]

Good observation, and probably right about the values most commonly used. What voltages were you finding there in the 1000uF and 100uF sizes, I would guess 16 and 25V were the most common, with 35V being next, then 10V. Most unreliable ones of course will be 10V and lower, as there they really trade off oxide thickness for volume. Above 63V I have found they get more reliable, and in many telecoms applications I often have seen a 5V rail with 100uF 100V electrolytics used for decoupling purposes, and there they last for decades with power applied.

[CatalinaWOW]

I didn't do statistics, but they ranged from 6.3 V up to 25 V.  Most common seemed to be 10 V. 

I am in the midst of another box of 100 or so, and the dud rate in this box is much higher, more like 50%, including one 1000 uF unit. 

[Someone]

Now comes the really interesting thing.  Because they make so many of them, they apparently get really good at the 1000 uF capacitors.  The junk box had several brands and several voltage values, but there were no duds in this value.  Not only were there no duds, but 90% of these were within 5% of nominal value, and none over 10% out.  Much better than you would expect from electrolytics,  particularly since they were probably used for general filtering.

The 100 uF capacitors were not as good, but still better than average electrolytics.  One dud, and two that were 200% of nominal value. 

My conclusion from all of this is that if you want reliability in your circuit, and need to use electrolytic capacitors, select one of the magic values.  The sampling process used to draw this conclusion is not statistically valid so YMMV.

It may also be the case that the 1000uF caps were available cheaply in bulk (or already in the parts list/system for the company) and the designers took the easy way out and put in a cap that was overrated for the task by voltage and/or current rating. But when they had free choice the rating margins were lower to show to their boss how they saved a few % cost here and there.

[CatalinaWOW]

Now comes the really interesting thing.  Because they make so many of them, they apparently get really good at the 1000 uF capacitors.  The junk box had several brands and several voltage values, but there were no duds in this value.  Not only were there no duds, but 90% of these were within 5% of nominal value, and none over 10% out.  Much better than you would expect from electrolytics,  particularly since they were probably used for general filtering.

The 100 uF capacitors were not as good, but still better than average electrolytics.  One dud, and two that were 200% of nominal value. 

My conclusion from all of this is that if you want reliability in your circuit, and need to use electrolytic capacitors, select one of the magic values.  The sampling process used to draw this conclusion is not statistically valid so YMMV.

It may also be the case that the 1000uF caps were available cheaply in bulk (or already in the parts list/system for the company) and the designers took the easy way out and put in a cap that was overrated for the task by voltage and/or current rating. But when they had free choice the rating margins were lower to show to their boss how they saved a few % cost here and there.

Good point.  But that doesn't explain the tight tolerance observed unless the manufacturers hold better on this value.  These were made by several companies, and came out of several different companies product/stock.  Light duty could explain lower dud rate, and also would tend to reduce the number on the low side of tolerance, but not to make them cluster tightly.  On other values I am observing tolerances up to +300%, and then of course the very low values that come from drying electrolytes.  For my purposes anything below -20% is being treated as a dud regardless of the spec, unlikely to perform over time and I am actually thinking of moving that to -10%.