Capacitor lifespan and audio quality

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Jack (in NJ),
yeah, its tricky to do. but it is achievable. the CREE lamps i worked on did this by ensuring the caps didnt get too hot - IOW excellent thermal design, as well as proper cap lifetime control (and good caps).

I fixed a friends huge plasma telly a couple of months ago. He'd been quoted some stupid price. He helped me pull the back off (its huge), and sure enough - bulging, swollen caps. I forget the vendor, but something asian (Samwha?). replaced with Rubycon caps that I had lying around. the desoldering gun meant it took a lot longer to remove then replace the back panel (40 or so screws) than the repair itself.
 
indeed. the temperature rating alone isnt enough to base a purchasing decision upon. And although it too is necessary, I maintain that high rated lifetime isnt enough either.

there are a lot of really crappy cap manufacturers. there are very few that make excellent quality caps. I've been designing industrial SMPS and motor controllers for 20 years, and will only use the following manufacturers:
Nippon Chemicon, Nichicon, Rubycon, Panasonic
there are some other high quality manufacturers as well (eg Sanyo) but the above four have the best parts - longest rated lifetime, highest ripple current, lowest ESR
In particular I would NEVER buy caps from a chinese manufacturer - and I have made plenty of money by solving problems caused by doing this. Samxon for example make lousy caps, and I dont believe anything their datasheets say wrt lifetime -
something to be aware of also is specmanship - crappy caps often spec end-of-life at 300% leakage current, 30% capacitance drop. the good ones use 200% & 20%. the difference can be huge - the 300/30 points are a lot further along the time axis, making the spec'd lifetime appear larger.
My preferred suppliers, newark (farnell) and mouser, do not carry samxon brand. They mostly carry the four premium brands you specify, so my choice is limited by my vendor choice. However, a friend just bought some 1000 hour panasonic 105 deg caps from mouser, so buyer beware. They also stock a 3000 hour Panasonic cap that costs more.
Newark does carry CDE brand, which I have found to be Chinese sourced. In the specmanship game , CDE life specs on one line I looked at have a worse end of life ripple spec than your premium 4 brands,IMHO. I do have some CDE caps installed, and they are okay going into the 3rd year.
The Nippon Chemicon caps that have come in have been Chinese manufactured. Perhaps that company maintains a home office QA person on site at the factory. If that person actually performs unscheduled audits Saturday morning at 3AM, it might explain their quality reputation in your experience.
 
Last edited:
My preferred suppliers, newark (farnell) and mouser, do not carry samxon brand. They mostly carry the four premium brands you specify, so my choice is limited by my vendor choice. However, a friend just bought some 1000 hour panasonic 105 deg caps from mouser, so buyer beware. They also stock a 3000 hour Panasonic cap that costs more.
Newark does carry CDE brand, which I have found to be Chinese sourced. In the specmanship game , CDE life specs on one line I looked at have a worse end of life ripple spec than your premium 4 brands,IMHO. I do have some CDE caps installed, and they are okay going into the 3rd year.
The Nippon Chemicon caps that have come in have been Chinese manufactured. Perhaps that company maintains a home office QA person on site at the factory. If that person actually performs unscheduled audits Saturday morning at 3AM, it might explain their quality reputation in your experience.

Good post.

just because its 105C doesnt mean its necessarily long life. all caps are not created equal - even from the good manufacturers. If you aim for "low impedance" and "long life" cap ranges you should get what you are after. Manufacturer selector charts are really handy for this - the alternative is to download every single datasheet, read the lot of them and pick the best. tedious, but I've done it before.

Another good trick is to look at the ripple current specs (and the ripple current multiplier chart). caps for 50/60Hz rectification have ripple current specd at 100/120Hz. And the ripple current multiplier is 1 at LF, and gets smaller as frequency increases.

Whereas SMPS caps have ripple current spec'd at 100kHz - and the ripple current multiplier is < 1 at LF, and rises to 1 at 100kHz.

The manufacturers use different dielectrics to achieve this. it allows them to have great specs at 100/120Hz as long as you dont care about HF performance, and vice-versa.

I've never done the relevant calcs, but an argument could be made that a C-R-C or C-L-C filter, with the rectifer bank of Caps being rectifier caps, and the amplifier bank of Caps being SMPS caps (layout whinge included) might give the best overall result. dont be surprised if the production manager pees in your coffee pot if you suggest it though.

What I have done is make C-L-C filters (all SMPS caps) to decouple the rectifier current from the smps current, for acheiving desired lifetime and getting a pi filter for 3rd-order EMI filtering. That came about because I needed 2 caps to get the desired lifetime. then thought "hang on a minute...." and found the space for a small iron powder toroid (about $0.20). it didnt do much for the lifetime (1 cap way too small, 2 caps plenty big enough), but did wonders for conducted emissions, as the resultant pi filter gave an extra 40dB/decade rolloff.


Not sure what you mean by "worse end of life spec" - I assume its 300% tan(delta), as opposed to 200%?

I just noticed that the end-of-life cap change isnt as I'd recalled, and varies even between the good brands. For SMPS thats not really an issue, as ripple current and hence ESR, or tan(delta), is the governing factor. In the last 20 years every single SMPS cap design I've done (input & output) I've calculated a desired capacitance for some allowable dQ ripple, calculated the desired ESR for some allowable I*ESR ripple (these two are often made equal in textbooks), calculated the actual ripple current, then picked enough caps to give me the required lifetime. And every single time I've ended up with much more capacitance than my Cmin calcs, and lower ripple voltage than my target value.

Or in other words: output ripple may be dominated by ESR ripple (or maybe dQ ripple), but getting a decent lifetime over an industrial temperature range invariably governs the actual amount of capacitance required.

After the 3rd or 4th design, I started doing the lifetime calcs first, then calculating the output ripple and seeing what I ended up with. Havent had too much ripple yet. But to be fair I do pick a very high operating temperature... 70C to 80C.

AFAIK all the good caps are made in China too. so are iPhones. turns out not all chinese manufacturers are the same. as companies get bigger they are forced to have better warehouse control etc - and if they have customers with strict standards, penalty clauses, batch testing and the like, you can be pretty comfortable with what they make. Small manufacturers though - they can be astonishingly bad.

One technique we've (for various "we") used a lot is to build the testers ourselves - complete with real-time comms. that way you can be sure the testers work (ahem. eventually), are accurate, and get real-time test data.

and the corollary: some of the manufacturers are literally horrific. A production manager came back from a casting factory once, labelled it "Dantes Inferno" and ensured they would never, under any circumstances, get any work. his description of the processes were frightening - every single thing done by hand, using ancient equipment and NO safety gear at all. he said he was terrified just walking through the place.

what did we do? used a different factory, one filled with the latest NC machinery that wouldnt look out of place in a toyota plant. and it was a few doors down the road.
 
I trust these. almost all of which came from ebay.
 

Attachments

  • my lab.jpg
    my lab.jpg
    481.7 KB · Views: 345
Well, test equipment is harder physically to counterfeit than a silk screen logo on an aluminum can. Some large part of what test equipment you buy may be original based on the picture.
Interesting about the electrolytic caps specified for ripple for line frequency and the ones specified for SMPS 100khz. Since everything I repair has a line frequency power supply and response of 20-15000 hz (<7000 hz on organs) I'll have to start reading datasheets to make sure I don't buy any SMPS capacitors for audio equipment.
Except the stack of dead SMPS power supplies I have, PCAT ones. They blow about every 18 months and I am beginning to think of attempting repair on one. Funny, none of the brands of leaky capacitors I find in those, Tepo in particular, is sold by Newark(farnell USA) or mouser.
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.