Re-forming High Value Capacitors

[AB2EZ]

I have 16 Sprague electrolytic capcacitors that I purchased (NOS) about 5 year ago:

900uF, 450VDC, +85^oC, codes: 32D901F450DF2B, and 9232L

I used 8 of them, in series, with 10,000 ohm balancing resistors, to build the HV supply for my plate modulated amplifier

The others have been sitting in a box in my basement.

I decided to do some preventive maintanence on them by re-forming them, one at a time, with the circuit shown in the attachment.

Note: There is no internal capacitor in the box; and the peak voltage out of the rectifier, with nothing connected to the output of the box, is (nominally) 120 x 1.4 volts = 170 volts.

The 3000 ohm resistor limits the current to 60mA.

As I connected the capacitors, one by one, to re-form them, I observed the following:

Each capacitor would initially charge to a voltage of around 140-150 volts with the expected time constant of around 900uF x 3000 ohms = 2.7 seconds

Then, over a period of around 30 minutes, the voltage would slowly rise to around 163 volts... and level off.

It appears that until re-formed, each capacitor had a leakage current that corresponded to a leakage resistance of around (as low as) 3000 ohms x 140 volts / 20 volts = 21,000 Ohms.

In my HV power supplies, I use 10,000 ohm (or larger) balancing resistors... so it is clear that re-forming these capacitors before use is a good thing to do.

I wonder if re-forming these types of high value capacitors every few years... while not in use... as a preventative maintanence routine... is a worthwhile investment of time?

Stu

[The Slab Bacon]

If those caps are only 5 years old, they shouldn't need reforming. If they do, they are trash can fodder. I have NOS caps that are over 20 years old and have never had problems with them. 5 years should be nothing, especially for modern caps which are far far better than stuff from 30 or 40 years ago.

I have reformed caps in the past just simply using an adjustable bench supply set for the full operating voltage of the cap and a high value resistor in series to limit the current. (somewhere between say 470K and 1 Meg) Hook it up, walk away and go do something else for a while. After an hour or so, check the voltage on both sides of the resistor, if it is equal and the cap is not warm, it is reformed.

But....................I have noticed that caps that needed to be reformed usually have
a tremendously diminished life expectancy and will most often fail soon after being put back into normal service. Basically a big waste of your time, You would be better off replacing them in the beginning.

But for some, "Idle hands are the devil's playthings"   

[WD8BIL]

Hi Stu,

Well, in one time constant the cap should charge to 63% of the voltage applied. That should be around 107V. it should be around 147V after 2 time constants or 5.4 seconds. It'll reach 99.3% after 5 time constants.

Now that's textbook stuff so any variance from that would indicated a real world variable is present. But to reach that far beyond 110V at the first TC is highly unusual.

http://www.bowdenshobbycircuits.info/rc.htm

So running your numbers thru the calculator shows 140V at 2700ms thru 3K gives an effective capacitance of 518uf.

[AB2EZ]

Bacon

I wonder if your observations about the longevity of capacitors that require reforming are valid for capacitors with very high values of capacitance and very high voltage ratings. The high value of capacitance would imply that the dielectric oxide is very thin, and/or the surface area is very large. The eight capacitors I used in my existing HV supply (out of the total of 16 that I had acquired) have held up fine for around 5 years. I was primarily concerned with properly storing/preserving the other 8 capacitors for use in a different/future HV supply

Bud

To clarify... what I meant to say was:

The capacitors would initially charge toward 140-150 volts with the expected time constant of 2.7 seconds. I.e. they would reach 63% of 140-150 volts in 2.7 seconds and 95% of 140-150 volts in 3 x 2.7 seconds, etc.

Then, the rate of voltage increase across the capacitor would slow dramatically. It would take around 30 minutes for the voltage across the capacitor to increase to a value that was close to the peak value at the voltage at the output of the bridge rectifier.

This indicates a relatively high value of leakage current... corresponding to a leakage resistance of around 21,000 ohms... until the re-forming process was (essentially) completed.

Stu

  

[Opcom]

But....................I have noticed that caps that needed to be reformed usually have
a tremendously diminished life expectancy and will most often fail soon after being put back into normal service.

This has happened several times here. It's very annoying when the equipment is racked and the spewing cap is inaccessible. A way to life-test after reformation would be interesting.

[WD8BIL]

AHHH. That puts the right color on the horse. Now we're on the same page.

[The Slab Bacon]

Bacon

I wonder if your observations about the longevity of capacitors that require reforming are valid for capacitors with very high values of capacitance and very high voltage ratings. The high value of capacitance would imply that the dielectric oxide is very thin, and/or the surface area is very large. The eight capacitors I used in my existing HV supply (out of the total of 16 that I had acquired) have held up fine for around 5 years. I was primarily concerned with properly storing/preserving the other 8 capacitors for use in a different/future HV supply

Stu,
       Most of the caps that I have had this happen to have been 450v rated (although not necessarily running at that high of a voltage) Some have, some haven't. Most were in the 100uF or lower capacitance area, and most were quite old. I have seen little or no failure rate with the newer caps. (stuff made in the last 20 years or so) I have seen no need to reform caps periodically just to preserve them for parts storage. If it makes you feel better,
It probably wouldn't hurt anything to give them a little reforming action just prior to installation. The best preservation for caps allready installed into something would be to just periodically fire it up and run it for a few minutes.

I deal with a lot of vintage receivers and antique broadcast receivers. I have a saying regarding the electrolytic filter caps in them: When in doubt just whip it out!" I usually just replace them as standard order of procedure so I dont have to worry about a shorted one taking something else out. (like a power transformer) I have seen them start out OK, then after a short period of time start to get warm and then go to short. If you're lucky it will only take out the rectifier tube. If not, bye, bye power transformer...............

But with modern caps, I have never (at least not yet) noticed this to be a problem.

There have been guys at the antique radio club that I belong to who have too much time on their hands and are too cheap to replace the filter caps in old radios. they go through great pains to reform electrolytic caps, one of them even going so fat to build a device that "automatically" reforms caps, slowly cycling them up and down. They have presentations about their wonderful machines and I just sit in the audience and heckle them as "idiots that are too f'ing cheap to buy new caps! !"   

[WA1GFZ]

I have a stash of computer grade caps for solid state applications. The colt 45 can size. The price of these things are quite high so I always reform them with a variac on first power up. I do agree with Frank. I have not blown one up in many years but as a cheap ham I do reform expensive parts. 

[KM1H]

I wonder if your observations about the longevity of capacitors that require reforming are valid for capacitors with very high values of capacitance and very high voltage ratings.

Since youre caps dont meet either criteria I wouldnt worry about it. They are 19 years old and of modern design and the only thing I would do is use a step start circuit to limit inrush current. Those resistors are fine as shack heaters; I use 75K 7W across a string of 1200uF @ 500V for a 5200V supply since I had them; in most supplies I use 100K 3W.

Id be more concerned about energy storage and how you plan to handle short circuit fault current.

Carl

[AB2EZ]

Carl

Thanks for the additional inputs. I assume, from your comments that the code: 9232L means week 32 of 1992.

On your 5200V supply... have you ever measured the voltages across the individual capacitors? The reaason I ask is that 75k ohm balancing resistors just seem a little high to me.

As an example:

Suppose you have 12 capacitors, rated for 500V each, stacked in series. Suppose the balancing resistors are 75k ohms each. Suppose six (6) of the 12 capacitors each have a low enough leakage current that one can neglect their equivalent parallel leakage resistance compared to 75k ohms. Suppose the other six (6) of the twelve capacitors have more leakage... and they each have (for simplicity of calculation in this example) an equivalent parallel leakage resistance of 75k ohms.

Then, in effect, 1/3 of the total voltage will appeaer across the capacitors with ther higher leakage, and 2/3 of the total voltage will appear across the capacitors with the lower leakage.

For that reason, I used 10k ohm blancing resistors with my (beer-can-sized) 900uF 450V capacitors. I (carefully) checked the voltage drop across each capacitor, at full supply voltage, to verify that the total supply voltage was split very close to evenly among the 8 capacitors in my supply.

Stu

[KM1H]

Thanks for the additional inputs. I assume, from your comments that the code: 9232L means week 32 of 1992.

Correct

On your 5200V supply... have you ever measured the voltages across the individual capacitors? The reaason I ask is that 75k ohm balancing resistors just seem a little high to me.

Absolutely. The value is high if the caps were of the old style -50, +80% tolerance. OTOH they are rated at 20% and measured in the 6-8% range.

Suppose you have 12 capacitors, rated for 500V each, stacked in series. Suppose the balancing resistors are 75k ohms each. Suppose six (6) of the 12 capacitors each have a low enough leakage current that one can neglect their equivalent parallel leakage resistance compared to 75k ohms. Suppose the other six (6) of the twelve capacitors have more leakage... and they each have (for simplicity of calculation in this example) an equivalent parallel leakage resistance of 75k ohms.

ESR goes down as the C goes up, all were first tested on a Sprague Model 16 at 500V and then as a group of 12 at several Variac controlled voltages with a Fluke DVM and HV probe.  A LOT of time and care was spent with the chicken stick ending in a 100K 225W bleeder and then ending with 25K for the final discharge. That supply and others quite similar are in their owners happy hands

Having spent some working years around 100-500KV supplies I take it seriously

There are some big amp builders using 5600 and 6800uF @ 450V caps and Ive read several reports of rather scary failures during testing and debugging. That is some serious energy storage and they werent prepared for it dissipating in a few milliseconds! Fortunately nobody was hurt as the PS was in another room as the tests were going on.

Carl

[KA2DZT]

I agree with Stu about the value of the equalizing resistors.  Too high a value for each resistor along with the different values of leakage current that can occur with each cap can cause the voltage to vary greatly across each.

Fred

[W3GMS]

Hi Stu,

Aluminum electrolytic caps tend to "form" at the applied voltage that is across them.  When I was working for a living we were required to do MTBF (Mean Time Before Failure) analysis on our designs.  Aluminum Electrolytic caps ideally like to stay at the voltage you put across them.  Doing so, you can expect the normally life to be realized as long as things like ripple current and overall ambient temperature are not exceeded.  As an example, a 450V Electrolytic cap with 350V applied will meet the reliability expectations.  The same is true for Electrolytics that have close to their maximum voltage rating across them due to the forming at that voltage.  What they don't like is to have the voltage increased beyond the nominal or forming operating voltage.  As an example, in a product they are sitting at 350V and then for whatever reason the voltage goes up to 400V.  The capacitor will fail sooner than if the voltage was at 400V from the beginning.  If you do any leakage current test, I would use a source voltage very close to the maximum rating of the capacitor.   I was educated about the forming issue from either Sprague or Mallory back some years.

Joe, W3GMS   

       

[The Slab Bacon]

Aluminum electrolytic caps tend to "form" at the applied voltage that is across them.  When I was working for a living we were required to do MTBF (Mean Time Before Failure) analysis on our designs.  Aluminum Electrolytic caps ideally like to stay at the voltage you put across them.  Doing so, you can expect the normally life to be realized as long as things like ripple current and overall ambient temperature are not exceeded.  As an example, a 450V Electrolytic cap with 350V applied will meet the reliability expectations.  The same is true for Electrolytics that have close to their maximum voltage rating across them due to the forming at that voltage.  What they don't like is to have the voltage increased beyond the nominal or forming operating voltage.  As an example, in a product they are sitting at 350V and then for whatever reason the voltage goes up to 400V.  The capacitor will fail sooner than if the voltage was at 400V from the beginning.  If you do any leakage current test, I would use a source voltage very close to the maximum rating of the capacitor.   I was educated about the forming issue from either Sprague or Mallory back some years.
Joe, W3GMS    

Joe,
      that is EXACTLY the same way it was explained to me some years back. It was very important to get the forming voltage up to the voltage that the cap either was rated at or was going to be working at while limiting the current draw and cap heating.

If you reform a 450v cap at say, 200v, you should not use it above that voltage without first reforming it to the higher voltage.

If you do something that really heats up the cap, it's pretty much a guarentee that it will not last long after that.

Pretty much anything over 1kv I much prefer oil caps if you have the room for them!!
They are MUCH more reliable than stacks of aluminum electrolytics and equalizing resistors.
What's a few PCBs among friends! !

[W3GMS]

Hi Frank,

We had some pretty stringent de-rating curves on components and it was not unusual to use a .6 max factor of the maximum rated, but Aluminum Electrolytics can be used as explained in my prior post and will have rated life values.  In the switching power supply business the ESR and ESL become a variable in the overall loop stability and cap manufactures made caps that were suited for the output filters of such supplies.  We noticed that the voltage headroom on those caps was not what we thought we wanted based on our de-rating guidelines.  Though discussions with the manufactures of the caps,  is how we learned about the forming issues as related to reliability.  Ripple current along with the associate temperature are the main killers on premature cap failure.  We did a lot of accelerated life testing at the rated capacitor voltage and test confirmed the rated life.  When buying new caps, try and obtain the 120C ones if you can. 



Joe, W3GMS

[KM1H]

That forming info is ancient history and does not apply to modern production since the 60's at least..  If it did then every time you turned an amp on in the CW or Tune position and then later went to SSB the caps would suffer. Nothing is further from the truth; many show no excessive leakage even after 25-30 years since the days of using junk in commercial amps is long gone.

The requirement for such low value equalizing resistors is another old wives tale that never was used plus you are wasting around 100W in heat. The lowest Ive ever seen is 30K in Heath SB amps since they purchased the lowest price caps they could find; Harbach replacements use 100K which saves 25W in an already marginal SB-220 transformer for example. Others used 50K in the 60's and switched to 100K in the mid 70's in commercial ham amps.

Using 390 to 560uF as replacements in existing amps (the form factor is often identical to the original 150-220uF caps) the voltage variance with 100K is typically 10-12V. Anything over 330-390uF should use a step start.

Standard temperature ratings are 85 and 105*C, I buy 105's for long term reliability. Its cheap insurance.

Carl

[W3GMS]

Carl,

This was well into the 90's when this issue was discussed with both Sprague and Mallory while I was the manager of the Power Engineering group at Unisys at our East Coast Development Center.

I would imagine that there is an interdependency on how long you sit at a voltage higher above what the capacitor has been formed to before the reliability curve degrades.  The formed voltage is defined as the normal operating voltage of a capacitor as operated in a given circuit.  Raising above that on a steady state basis and staying at that voltage showed a reduction in MTBF as compared to that voltage being at the higher voltage from the start of its life.  No, the capacitor will not instantly fail but its MTBF curve will be degraded according to the capacitor manufactures we talked with.  I meant to type 105C! 

I don't intend to debate this but simply wanted to share so past experience. 
 
Joe, W3GMS

[WA1GFZ]

Joe,
I think we are derating the crap surface mount caps at 33%. We have tracked field failures and find they track manufactures junk ratings. We now derate tuffer to get better yield.

[W3GMS]

Hi Frank,
We used .6 for ceramic caps.  Our calculated MTBF was around 500K using Mil Handbook 217 and the demonstrated was about 1.5M hrs.  We were in the 24 / 7 never go off the air with redundancy everywhere in order to meet the expectations.  Tantalums really need some pretty steep de-rating curves.   

Joe, W3GMS

[KM1H]

I was referring to computer grade or better, form factor style, electrolytics.
Ive also been using CDE Snap-In style for about 20 years with no reported failures. Initially I was concerned about their ripple current performance in a big amp PS but they seem fine and run cool in the up to 5KW range.

[AB2EZ]

Joe (W3GMS):

Very interesdting and useful inputs! Thanks

Stu

As an aside, I would have assumed that capacitors from a given manufacturer, and from a given product family would have a leakage current that is roughly proportional to the capacitance value. I.e., to some degree of approximation, a 900uF / 450VDC capacitor would have 9x the surface area of a 100uF / 450VDC capacitor; and its leakage current would be roughly 9x as great. This is essentially what you would expect with nine (9) 100uF capacitors in parallel.

However, leakage current is somewhat of a statistical phenomenon... and there are also reasons why you might make a 900uF capacitor out of a slightly different material sandwich (metal-oxide-electrolyte), versus a 100 uF capacitor in the same product family... for example, to ensure that the failure rate (in FITs) does not increase linearly with the capacitance.

Cornell Dubilier specifies a relationship between leakage current and capacitance in its family of aluminum electrolytic capacitors, that is proportional to the square root of the capacitance.

http://www.cde.com/catalogs/DCMC.pdf

To do the same job, the size of the balancing resistors has to be inversely proportional to the leakage current.

If a string of 100uF capacitors in series requires 100k ohm balancing resistors, then a string of 900uF capacitors would require (using the square root formula from the Cornell Dubilier specification sheet) balancing resistors of value 100k ohms /3 = 33k ohms.

If you assume (more conservatively than the Cornell Dubilier specification sheet) that a 900uF capacitor has a leakage current that is 9x that of a 100uF capacitor of the same product type... and if 100k ohms is the right balancing resistor value for 100uF capacitors in series... then one should use 100k ohms/9 = 11k ohms as the value for the balancing resistors in a series string of 900uF capacitors.

[KM1H]

Forever making assumptions never gets a product out the door, I prefer to go by what has been proven over decades of actual use by product engineers....not bookworms

No offense meant, I spent a lifetime in industry watching the bookworms never adjusting to a real job of getting a product out the door. 

[K1DEU]

   I just love antique things. Especially If I Don't have to use them too much!

One day I was sitting by my Johnson Valiant which was in the vertical (up on end) position and had my head nearby !

Was testing a 100 @ 450 axial Sprague that reformed and showed good and had run cool for several days. It was a little slow to show capacitance when repairing ? it.


Suddenly ### Bang ***  as aluminum foil unraveled and shot across the room, just missing my eye/head  Yes I ducked !

Who me gunshy ? Why I never Flinch I only Duck.  I just love to save money and especially old things.

Added   Perhaps I should consider my self the oldest in my entire collection of Junk ?

[Tom WA3KLR]

F.Y.I.  There is a new family of high capacitance film capacitors on the market, called DC link capacitors.   This new component offering is driven by the very high power switching supplies that exist today (for solar conversion and electric cars I presume).

Cornell Dubilier has 2 families of these polypropylene film caps.  I don't know what other manufacturers have these also, if any, yet.  (Checking the Mouser website, there are several other companies in this arena also.) These 2 families cover 33 uF to 1500 uF at 800 to 1400 Volts dc.  The one family is the physically larger ones for the higher capacitance.  The smaller one's (944U) packages look like the old D'Arsonval meter case.

They range from $50 to $100 as I recall, in small quantities.  They have extremely low ESR so care would need to be taken to not short them after being charged.  You would probably want to have some additional series resistance for tube circuits.

This is an alternative for the series strings of lower voltage aluminum electrolytics, but at a higher cost.   No possibility of electrolyte out-gassing, which is the reason for long-term failure of aluminum electrolytic capacitors.

Data sheets below: