cap replacement policies / was Re: VT101 8085 CPU Fault

Tothwolf tothwolf at concentric.net
Tue Feb 24 12:34:08 CST 2015


On Mon, 23 Feb 2015, Brent Hilpert wrote:
>>> From: Tothwolf
>
>>> I've seen pushback from some people in the vintage computing community 
>>> to wholesale replacement of aluminum electrolytics which are long past 
>>> their life expectancy and I just don't get it.

> For solid-state digital equipment: I don't think I've ever done a 
> blanket replacement of caps. The vast majority of my many dozens to 
> hundreds of calculators, digital test equipment, computers, etc., most 
> of which date from the 1960s and 1970s, are running with all their 
> original caps - including specifically electrolytic filter caps. I have 
> rarely, if ever, reformed a capacitor for a solid-state item. Keeping in 
> mind that a mildly leaky cap which could benefit from reforming will 
> reform during normal operation. (Also, those big screw-terminal filter 
> caps from the linear-reg days were called "computer grade" for a 
> reason.)
>
> So, in relation to computers and solid-state digital, I don't understand 
> why people get all concerned about caps. In my experience, blanket 
> replacement and reforming just hasn't been warranted.
>
> There are always exceptions - I well understand why the PDP-1 
> restoration team would choose to be careful with the large filter caps 
> there - age, historic significance, large amounts of energy involved. If 
> you're doing this as a business such as the mentioned arcade-machine 
> refurbishment, someone might choose to do blanket replacement as a 
> pragmatic measure of insurance to help avoid call-backs, etc.

To be clear, I'm not in the business of arcade machine refurbishment, 
however I know many people who are. When these machines are used in an 
arcade environment, they are often powered on 24/7/365, not really unlike 
a vintage computer coming out of a production environment (and they also 
suffer from the same sort of TTL logic chip failures we've discussed here 
on classiccmp). More often than not, when a "vintage" 80s/90s coin-op 
machine is being refurbished or having extensive maintenance done, all of 
the electrolytics in the monitor chassis and PSU (linear or switchmode) 
get replaced as a maintenance item. Some of those machines later end up 
back in the field earning revenue for operators, but a large portion of 
them end up resold to individuals too.

> Blanket replacement (and even reforming) have an element of 
> self-perpetuating myth to them, some people do it and recommend it, but 
> instances of it being done don't necessarily add to a body of evidence 
> as to whether it actually mattered.
>
> Some of it is a matter of personal preference/values: for example some 
> people are happy to do a blanket replacement on the probability of it 
> encompassing a fault, others (such as myself) prefer to diagnose a fault 
> down to the individual component level.

When you look at the construction and chemistry of aluminum electrolytics, 
the fact that they are as reliable as they are is quite amazing. The 
typical part is rated for 2000 hours at either 85C or 105C (even less for 
some older or specialty parts), at its maximum ripple current. The life of 
the part also doubles for every 10C below its rated temperature.

Say a nearly ubiquitous 85C part is used at very reasonable 40C (104F). 
That 2000 hour rated part then has an estimated lifespan of 48,000 hours 
at its maximum ripple current (more often than not, parts are not operated 
at anywhere near their maximum ripple current, so the expected life is 
going to be even longer). Just for simplicity, say the part is operated at 
a low enough ripple current where that 2000 hour part is then good for 
100,000 hours (which is of course being /very/ generous). That still only 
works out to about 11.5 years of expected service life. Larger more modern 
or high temperature rated parts may have a rated life of 3000 or 4000 
hours, but older parts made in the 80s and 90s are still largely going to 
be in that 2000 hour ballpark.

It is 2015, so just going with 2015 - 12 = 2003, I'd say the typical 
aluminum electrolytic operated at an average temperature present in a 
monitor, switch mode PSU, etc (higher than 40C) from the 1980s or 1990s 
which has seen daily or regular use is /probably/ going to be at or near 
its rated life expectancy.

Since a picture is worth a 1000 words, here are some photos of actual 
failed computer gear that was made in the 1990s. This gear was used by me 
in a home environment so it wasn't even subjected to the sort of heavy 
use/abuse found in a commercial setting.

This one is a photo of a 12V/5V DC power supply from an external SCSI 
drive enclosure made in 1992 where the Nippon Chemi-Con KMF series 
capacitors are leaking their electrolyte out of bottom around the rubber 
end seals:

http://strudel.ignorelist [dot] com/~tothwolf/classiccmp/Electrolytic_Capacitors/autec_powersupply.jpg

Here are the parts I'll be using when I rebuild it:

http://strudel.ignorelist [dot] com/~tothwolf/classiccmp/Electrolytic_Capacitors/autec_powersupply_parts.jpg

This is a photo of the board from a USRobotics Courier V.Everything modem. 
The three small radial Nichicon PL series capacitors are leaking 
electrolyte out the bottom around the end seals. Look near the empty U21 
location and you can see corrosion just beginning to form:

http://strudel.ignorelist [dot] com/~tothwolf/classiccmp/Electrolytic_Capacitors/usr_courier_veverything.jpg

The main filter capacitors are also out of spec, and the modem's speaker 
emits a constant digital "hiss" sound (which is an extremely common fault 
with these modems since the original capacitors run moderately hot due to 
passive cooling).

You'll also notice USR used 4 horizontal chip-type aluminum electrolytics 
on this particular board (there are several versions of these modem boards 
which use slightly different parts). Those particular chip-type 
electrolytics are no longer manufactured, and since those type are also 
known to be problematic and eventually leak, they will be replaced with 
common radial types stuck down to the board.

These are the replacement parts:

http://strudel.ignorelist [dot] com/~tothwolf/classiccmp/Electrolytic_Capacitors/usr_courier_veverything_parts.jpg

This particular modem was one of the best ever made, but every single one 
that I own (I have at least 4) has the same problem with the main filter 
capacitors being out of spec causing hiss from the speaker and the smaller 
6.3mm Nichicon PL series parts beginning to leak electrolyte.

These are just two out of about 250-300 projects currently in my queue 
(each has a parts bag just like the two I photographed). Both of these two 
are pre-"capacitor plague" with all the problems starting around 2000-2001 
with the cheaply made ultra-low ESR parts swelling up and popping their 
top vents.

-This- is one of the reasons why I'm for replacement of these sort of 
aluminum electrolytics as preventative maintenance in older gear. You 
wouldn't have expected either to fail, but they did. How often are you 
going to pop the top on something like an external modem and inspect, let 
alone test the electrolytics?


More information about the cctalk mailing list