[ARC5] Cool New DC-DC Converter

Tom Lee tomlee at ee.stanford.edu
Sun Mar 18 16:19:28 EDT 2018


They seem to go *very* dead short. They're the closest things to 
room-temp superconductors we'll likely encounter.

HP gear is hardly unique. Pretty much any vintage gear that has 
tantalums is awaiting an unscheduled power supply crowbar test. I have 
lost track of how many 400-series Tek scopes I've brought back to life 
by recapping. When a friend brings one to me to fix, that's pretty much 
the first thing I check now. I'd say that something like two out of 
three have shorted tants. On rare occasion, I'll find a tant that 
auto-heals after being off for some time, but most go short and stay 
that way.

Tom

-- 
Prof. Thomas H. Lee
Allen Bldg., CIS-205
420 Via Palou Mall
Stanford University
Stanford, CA 94305-4070
http://www-smirc.stanford.edu
650-725-3383 (public fax; no confidential information, please)

On 3/18/2018 1:05 PM, Scott Robinson wrote:
> Wehn Tantalums fail, they internally whisker and go dead short, even 
> catch fire sometimes when used as bypass caps.
> If I'm fixing HP grad test gear from teh 1970-80s, first thing I look 
> for is shorted tantalum caps.
>
> /scott robinson
>
> On 3/18/18 10:35 AM, Bob kb8tq wrote:
>> Hi
>>
>> Like I said, it *really* surprised us when he popped out the papers 
>> and the data.
>> None of us were aware of an issue. The driving mechanism seems to be 
>> the way
>> the insulator “grows” in a polar capacitor. If you don’t put “enough” 
>> voltage on it,
>> the process gets confused. That’s about as well as I can remember it, 
>> several
>> decades later ….
>>
>> The really bad news, I’d run into the same thing on tantalum’s *much* 
>> (like 6 months
>> out of school)  earlier. The world of SMD parts was new and 
>> wonderful. We
>> were setting up a screening / life test process to qualify some 
>> parts. I ran through
>> what we intended to do ( more or less, max voltage / max temp for a 
>> month,
>> while monitoring leakage) with the vendor. He smiled and said “that 
>> will be fine with us”.
>> Later at lunch he pointed out that lack of bias (and re-forming after 
>> that situation)
>> was the real issue for leakage (and possible failure).
>>
>> Bob
>>
>>> On Mar 18, 2018, at 1:08 PM, Dennis Monticelli 
>>> <dennis.monticelli at gmail.com <mailto:dennis.monticelli at gmail.com>> 
>>> wrote:
>>>
>>> Hmmm.  Don't recall seeing the lifetime vs voltage appear as a 
>>> bathtub curve. Life vs Time, yes, but not vs voltage.  Do you know 
>>> why? Seems odd to me.  The only thing I can think of that would 
>>> explain it has to do with operating at a much lower voltage for a 
>>> long time and then expecting the cap to be OK right away once 
>>> returned to operating at a much higher voltage.  All 'lytics pass 
>>> current (we call it leakage) as part of their natural ongoing 
>>> process of oxide decomposition and reformation.  Operate a 450V cap 
>>> at 50V for a long period of time and it will reach a new 
>>> equilibrium, eventually becoming a 50V cap.  To use that cap at much 
>>> higher than 50V safely would require slowly reforming 
>>> (re-thickening) of the now very thin oxide layer until an new (high 
>>> voltage) equilibrium is reached.
>>>
>>> We see a version of this situation whenever we power up an old piece 
>>> of equipment whose lytics have been dormant for years. If they 
>>> haven't dried out they will properly reform if given the chance. 
>>> That is why we are careful to severely limit the current into those 
>>> caps while we allow the oxides to slowly reform.  Slow growth 
>>> creates a better quality of oxide and insures that the negative 
>>> affects of self-heating are keep at a minimum.
>>>
>>> Dennis AE6C
>>>
>>> On Sun, Mar 18, 2018 at 9:44 AM, Bob kb8tq <kb8tq at n1k.org 
>>> <mailto:kb8tq at n1k.org>> wrote:
>>>
>>>     Hi
>>>
>>>     The whole “less is better” approach is what we had been using
>>>     across the board on parts.
>>>     50% max was our magic number. Went into the design review and
>>>     started going through
>>>     everything. Made it to parts and one guy’s eyes lit up. He reached
>>>     into a folder and pulled
>>>     out a coupe of studies. (Odd that he just *happened* to have them
>>>     along with him …).
>>>     They pretty well showed that on some caps, you get a bathtub curve
>>>     for reliability vs voltage.
>>>     Go to low and the reliability starts to degrade again …… surprised
>>>     the heck out of all of
>>>     us (to say the least).
>>>
>>>     Bob
>>>
>>>>     On Mar 18, 2018, at 12:31 PM, Dennis Monticelli
>>>>     <dennis.monticelli at gmail.com
>>>>     <mailto:dennis.monticelli at gmail.com>> wrote:
>>>>
>>>>     The failure of an electrolytic cap vs how close to its rating it
>>>>     is run is a soft function.  It's not like the sudden failure of a
>>>>     solid dielectric cap.  It takes a long time.
>>>>
>>>>     To get a decent lifetime a 'lytic should always be run under its
>>>>     rating, just how much depends upon how much reliability you are
>>>>     after.  HP was conservative; their design guidelines were 60%. 
>>>>     Most other makers were less conservative in order to keep cost
>>>>     and size down.  My vintage ham gear seems to apply roughly 80%
>>>>     and sometimes more.  They also don't respect ripple heating
>>>>     enough, mounted high wattage bleeders close by and increasing
>>>>     moved toward ESR hostile cap input filters with solid state
>>>>     rectification.
>>>>
>>>>     The other factor is dry-out due to age and heating.  The higher
>>>>     the ESR and the higher the ripple, the more self heating.  The
>>>>     more heating, the more drying. The effect is a very slow runaway
>>>>     condition that eventually manifests in cap failure.
>>>>
>>>>     I you buy a modern 105C-rated low-ESR cap for replacement and
>>>>     don't try to squeeze the last volt out of it, it's going to
>>>>     reward you with long life.  They are built to survive in abusive
>>>>     PWM circuits.  Our circuits are a cakewalk by comparison.
>>>>
>>>>     Dennis AE6C
>>>>
>>>>
>>>>
>>>>     On Sun, Mar 18, 2018 at 9:12 AM, Bob kb8tq <kb8tq at n1k.org
>>>>     <mailto:kb8tq at n1k.org>> wrote:
>>>>
>>>>         Hi
>>>>
>>>>         The real point is that a “250V” cap may or may not have a max
>>>>         rating of exactly 250.0V.
>>>>         It may indeed have an actual max voltage on that sample a bit
>>>>         higher than 250V. These
>>>>         days it’s very much a “who knows” sort of thing.
>>>>
>>>>         Oddly enough the same rule that says “don’t use the last 10 
>>>>            %” also says “don’t use less
>>>>         than 60%” on an electrolytic cap. Learned that one in a
>>>>         design review …. on product headed
>>>>         for space no less …. gulp ….
>>>>
>>>>         Bob
>>>>
>>>>         > On Mar 18, 2018, at 12:06 PM, Scott Robinson
>>>>         <spr at earthlink.net <mailto:spr at earthlink.net>> wrote:
>>>>         >
>>>>         > Well, a switching power supply designer who is even older
>>>>         than I am told me that if you avoid the using the last 10% or
>>>>         so of an electrolytic cap's voltage rating that it will last
>>>>         longer.
>>>>         >
>>>>         > FWIW,
>>>>         >
>>>>         > Scott Robinson
>>>>         >
>>>>         > On 3/18/18 7:33 AM, Bob kb8tq wrote:
>>>>         >> Hi
>>>>         >>
>>>>         >> In some cases, indeed a 250V cap is quite happy for a very
>>>>         long time at 300V. It’s not
>>>>         >> all “great balls of fire” sort of stuff ….
>>>>         >> Bob
>>>>
>>>> ______________________________________________________________
>>>>         ARC5 mailing list
>>>>         Home: http://mailman.qth.net/mailman/listinfo/arc5
>>>>         <http://mailman.qth.net/mailman/listinfo/arc5>
>>>>         Help: http://mailman.qth.net/mmfaq.htm
>>>>         <http://mailman.qth.net/mmfaq.htm>
>>>>         Post: mailto:ARC5 at mailman.qth.net 
>>>> <mailto:ARC5 at mailman.qth.net>
>>>>
>>>>         This list hosted by: http://www.qsl.net <http://www.qsl.net/>
>>>>         Please help support this email list:
>>>>         http://www.qsl.net/donate.html
>>>>
>>>>
>>>
>>>
>>
> ______________________________________________________________
> ARC5 mailing list
> Home: http://mailman.qth.net/mailman/listinfo/arc5
> Help: http://mailman.qth.net/mmfaq.htm
> Post: mailto:ARC5 at mailman.qth.net
>
> This list hosted by: http://www.qsl.net
> Please help support this email list: http://www.qsl.net/donate.html




More information about the ARC5 mailing list