Bad Capacitors
How parts worth a penny bring down expensive equipment
Daniel Senie
Amaranth Networks Inc.
Originally Written 3 March 2003
Updated 15 February 2004
In the last few years, I've noticed damaged electrolytic capacitors in
several devices, resulting in various types of failures. In February 2003,
an
article in IEEE Spectrum explained the genesis of the situation.
It seems someone stole the formula for the electrolyte in electrolytic
capacitors, rather than doing the honorable and honest thing and licensing
the information. Problem is, they stole only part of the formula. Apparently
the missing part had the information on how to keep capacitors from
exploding.
The electrolytic capacitors look like little tin cans with leads out either
both ends, or just one end. All the failed units have had radial (out one end)
leads. the other end of the capacitor has a plus sign like marking. That end should be flat. When it bulges up, your capacitor is on its way out. When things get worse, goo starts emerging from the bottom of the capacitor, where the leads come out, and causes damage to other components and the circuit board.
So far, I've had some items that I've successfully repaired, and others that haven't been worth the effort.
Some manufacturers will own up to the problem if pressed. Others will not.
Though devices are out of warranty, manufacturers SHOULD be willing to fix problems resulting from this problem.
Items Under Investigation
- Promise Technologies ConnectStor-II
The ConnectStor-II is a network-attached-storage (NAS) device. It contains a small form-factor Pentium motherboard that appears to be based on one of Intel's reference designs. A Promise ATA disk controller chip two IDE interfaces are present to connect a pair of ATA drives into either a RAID-0 stripe, or a RAID-1 mirror.
The device recently was removed from service, fitted with larger hard disks and when we were about to place it back in service, the distinctive odor of burnt electronics was evident. Upon opening and inspection, it was clear the power regulation circuitry on the main board contained two bad capacitors. It's troubling that the circuit in question, the voltage regulator for the Pentium processor, contains many more such capacitors, though the rest have not (yet) started to explode.
An email support inquiry has been made to Promise Technologies. When they have responded, this web page will be updated.
Items I've Fixed
- Proxim/Orinoco AP-500
- (The Orinoco product line/division has changed hands a lot lately, product may be labelled Agere or Lucent. This unit was also re-sold by Avaya, and many other companies). The
AP-500 was a cost-reduced access point with a circuit board containing all
logic other than the actual 802.11b card. The design of the AP-500 was
a simple reduction from the AP-1000 and WaveLan2 products.
New in the AP-500 was
power-over-Ethernet (Active Ethernet in the Orinoco parlance).
Power is either derived over the Ethernet, or is supplied by a 9 volt wall wart.
-
- Power conditioning is done on the board.
The unit had begun emitting a high-pitched audio whine. While still functional,
the power supply circuit was not operating correctly. Replacing two clearly-damaged capacitors
resolved the problem.
-
- Orinoco service didn't admit to a fault, and wanted more
to fix the board than the cost of a new unit.
-
- Avocent/Cybex SwitchView OSD
- This is a rack-mountable 8-port KVM (keyboard, video, mouse) switch. It uses
a tiny switching power supply that's internal to the box. The supply accepts
110/220 AC, and produces +5 volts and -5 volts.
-
- Symptom of the failure is erratic behavior or complete lack of
functionality. Measuring the power supply output shows good -5V,
but the +5V rail shows 3.55 volts. Two 330uF electrolytic
capacitors were obviously damaged, oozing out the bottom, and
puffed out on top.
-
- Avocent offered to replace the power supply even though the unit was no
longer under warranty. I'd have had to ship the unit in,
though (they didn't seem to want to just send me a new power supply).
A previous unit that failed the same way cost over $100 to have fixed.
Though the problem was also caused by the same issue, they said they had
no way to refund the previous repair bill. Be sure to raise this issue
if your unit fails before paying.
-
- Replacing the two capacitors solved the problem.
Total cost to fix: $1, plus about an hour of time to
disassemble and reassemble the unit. It's an easy fix for anyone skilled
in soldering. Tools required: regulated temperature soldering station,
de-soldering tools (wick, pump, etc.), decent DMM.