Bad-caps-syndrom

From Electriki
Jump to navigationJump to search

Admitting it

We have all been there, having a piece of hardware that would sometimes not start, we just said to ourself "its okay, I know how to work around it" and went on powering on and off a few times untill it started. At some point, this stoped working or we might become afraid to turn it off, in case she would not come back on. - Admit it, it is a problem that needs to be fixed


ACTUNG: this only works for electrolytic capacitors.

Diagnose

Signs of bad caps:

  • Visible damage to cap.
  • It runs fine, but have start problems
  • Can be started when pre-heated with a hairdryer or soldering iron on top of caps.
  • It is more than 3 years old and build after 1990.

Taking it apart

BEFORE you start randomly desoldering capacitors, you must note the following of each:

  • Capacitors place on board, usually Cxxx but some boards might not have a silkscreen showing parts name.
  • Rating of cap, electrical: Capacity, voltage.
  • Rating of cap, mechanical: Diameter, hight and pin-spacing

Parts sourcing

Based on the above note info, searching for replacement can begin. Be sure to get low-ESR rated capacitors for anything incorperating a switchmode, this includes but is not limited to, computer motherboards. For some popular computer motherboards, there is allready kits out there with capacitors needed. If you do not trust there quality, you can always copy the info and get your own :)

Do not hesitate to go from 85 C to 105 C capacitors, the price differance is almost 0.

Your local pusher might now have LOW-ESR caps, if he gets a funny look in his eyes when you say "low esr", walk away slowly.

Farnell / RS-components probably have what you need, but they also are damn expensive.

Ebay have it too, but it might be capacitors made of cardboard and vinigar, you never know..


Re-assembly

Rule no. 1: LOOK TWICE ON THE POLARITY OF THE CAP!

Rule no. 2: look a 3rd time, yes you have fucked up.

Rule no. 3: DO NOT assume the silkscreen is correct! look at the cap you take out.


Now, put it all back together, a series of pictures of you taking it apart would be a good idea to have now. oh, Should I have mention that in the start?


Practical example

Ze defective hardware
Ze PSU
Heatsinks removed
Packed up nicely, waiting for new caps


Time for a practical example, a Cisco Catalyst 3500 switch, that would not start after a power-down for +1h. guess what, bad caps!

Diagnose: it was proved that she would start after heating it with a hairdryer, yay!

Taking it apart: pics or it didn't happen. most caps were hidden under heatsinks, thiese were de-soldered from the board, including the semiconductors they hold. One COULD start testing with a ESR-test to tell what caps were bad, but as it is very visible on the board that this lives a HOT life, might as well change all or it will die again within the next year or so.

List of capacitors:

C202: 1000 uF, 35 V, 13 mm / 5 mm (diameter, pin-spacing)

C203: 680 uF, 16 V, 10 mm / 5 mm

C217: 680 uF, 16 V, 10 mm / 5 mm

C219: 10 uF, 50 V, 5 mm / 3.5 mm

C214: 1200 uF, 16 V, 13 mm / 5 mm

C206: 3900 uF, 6.3 V, 13 mm / 5 mm

C207, C208, C209: same as C206

C309: 120 uF, 16 V, 7 mm / 3.5 mm

C310: 120 uF, 16 V, 7 mm / 3.5 mm

C301: 10 uF, 50 V, 6.3 mm / 3.5 mm

C112: 100 uF, 63 V, 8 mm / 4 mm

Max hight for all: 25 mm

So, 11 caps, thats not too bad, they were all 105°C, but we know nothing about the quality - hopefully we can do better.

Elfa.se came up with the following:

specs Elfa.se part no. price, DKR - Conrad.com part no. price, DKR amount needed
1000 uF 35 V 67-229-87 9,48 - 442608 3.87 1
680 uF, 16 V 67-233-99 3,14 - 442638 1.12 2
10 uF, 50 V 67-193-63 1,04 - 446459 0.74 2
1200 uF 16 V 67-181-67 6,35 - 442634 3.05 1
3900 uF, 6.3 V 67-232-90 7,02 - 442652 1.64* 4
120 uF, 16 V 67-233-65 1,42 - 442601 1.34 2
100 uF 63 V 67-153-20 1,87 - 445400 1.12 1

Elfa, sum totalis: 57.72 DKR ( 7.65 Euro ) + shipping + VAT + Sillyness. Not bad.

Conrad, sum totalis: 21.00 DKR ( 2.82 EUR) VAT included but + shipping. Shipping is pretty much the same, around 9 EUR. borh for Elfa and Conrad. Much better than not bad :)

  • 3900uF caps was not fund in the right size at Conrad, went for 3300uF instead, hopefully it will work :) The 4 of them sits closely together and measures 13 mm, the ones I can find at Conrad is 16 mm @ 3900uF, the 3300uF are 13mm.


Rejoiced over those low prices compaired to Elfa, that it turns out I cannot shop at, not easy anyway, the 6 capacitors on the switch/router board itself, an extra DC-stepdown converter after the main psu, was deimed for replacement as well, 4 x 100uF 10V, the 35V found for the main PSU will do, so 4 more of those, then 2 x 470uF 25V. Conrad part no. 442625.

---

Capacitors ready, set, GO!
All new caps mounted

2012-01-14 - Judgement-day

During the replacement of the caps, a measurement of ESR was done on each old and each new capacitor:

- OLD NEW
pos. Capacity[uF] / Voltage ESR[ohm] Capacity[uF] / Voltage ESR[ohm]
C217 680 / 16 0.06 (0.08) 680 / 16 0.03
C203 680 / 16 0.11 (0.15) 680 / 16 0.07
C202 1000 / 35 0.09 (0.16) 1000 / 35 0.02
C214 1200 / 16 0.05 (0.05) 1200 / 16 0.02
C209 3900 / 6.3 0.00 (0.04) 3300 / 6.3 0.00
C208 3900 / 6.3 0.00 (0.05) 3300 / 6.3 0.00
C207 3900 / 6.3 0.00 (0.05) 3300 / 6.3 0.00
C206 3900 / 6.3 0.00 (0.05) 3300 / 6.3 0.00
C309 120 / 16 0.18 (0.26) 120 / 50 0.07
C310 120 / 16 0.20 (0.26) 120 / 50 0.05
C219 10 / 50 0.54 (0.56) 10 / 63 0.61
C301 10 / 50 0.52 (0.88) 10 / 63 0.41
C112 100 / 63 0.17 (0.22) 100 / 63 0.18

All replaced, heatsinks+semiconductors reassembled and she starts the first time, exelent.

What can we learn from this? well to start with, no ESR value was much higher on old caps vs. new caps, but some of the caps, the 120 uF ones, did leak a bit from the bottom when desoldered.

ESR can tell us a lot of the state of the capacitor, but perhaps this set were just not that far gone yet.

UPDATE: Made the rookie mistake, desolder cap, measure it == cap being measured in hot state == more likely to work (that is why the blowjob/hairdryer trick works) Table above is updated with ESR-measurements on old caps, number in () are cold values. We now see the 3900 uF caps give a reading, 0.05 vs max 0.005 (assumed threshold of rounding to least significant digit on display). Also the 120 uF were somewhat high relative to the new ones, but still well within "normal" limits according to the table on the ESR meter.