Linksys WRT54GL v1.1 bad CapXon caps

[Per Hansson]

Friend called me about a non-booting WRT54GL the LED for power in the front was just flashing.
As I opened it up I found it really hot, there was also a very audible hissing noise from the coils...



It has 4x CapXon ST 220uF 16v caps, I recapped it with 3x used Samxon ULR 220uF 16v solid polymer caps, and one Rubycon ZLH 220uF 16v for the input.
This caused the hissing to stop but the router still would not boot.
Diode DA7 & DA6 become really hot, I was not able to find a schematic though so I gave up.
The router LED's for Ethernet connections would come on one by one when it was plugged in but without cables connected...
And with a cable connected the computer never registered a link as up, always showing cable disconnected... (It was like this before the recap too)
I also tried the "30-30-30" reset but it did nothing at all

Two of the CapXon's where around 1.X in ESR one had 4.0 but the incoming cap had 50ESR when hot and infinite when cold!



As he has another router just like this one I opened it up too. (It is working fine).
It had the exact same caps, and also a little hissing noise from the coils...
I desoldered the CapXon's: it was the same, only not (yet) as bad.
The cap for the incoming shows 3.5, rest where around 1.2
So I moved the caps I had put in the other router to this one, the hissing magically disappeared in this one now

Funnily I have already recapped this one's AC/DC adapter some months ago, bad caps there too of course!

[ben7]

Re: WRT54GL v1.1 bad CapXon caps

The diodes might be shorted, or they might be overloaded from some other IC that is shorted.

[Per Hansson]

Re: WRT54GL v1.1 bad CapXon caps

I measured them out of circuit & they seemed ok, I'm no expert at this though...
Might solder on a serial port just to see if it says anything...

[momaka]

Re: WRT54GL v1.1 bad CapXon caps

My sister's WRT54G also went out the same way just two weeks ago. I went to visit her and she mentioned that there was a "running water" noise coming from the router/modem box in the hallway. The internet was still running fine at that point, though. Since I wasn't sure if it was the router or the modem that was faulty (and I highly suspected bad caps), I cut the power on both and powered them ON one by one. Now that was a bad move! The router never started again and did the flashing power LED light just like yours + the hissing noise again. I opened it and looked at the caps, but they looked fine.

Long story short (I will make a thread and post some pics of the router when I get back home), the buck regulator was overshooting its output voltage badly due to a bad cap (actually the only cap) on the input (high) side. Although I haven't done a thorough investigation yet due to lack of SMD tools, I suspect the voltage overshoot has taken out at the very least the wireless IC and also possibly either the RAM or flash IC... or both.

Moral of the story: don't ignore crappy cap brands just because they look "okay" - especially if those caps are in a buck regulator circuit (i.e. near a coil). I used to think that only the output (low) side caps were important on a buck regulator and that the input (high) side caps couldn't cause any catastrophic issue if they failed. Clearly, this is no longer the case.

[ben7]

Re: WRT54GL v1.1 bad CapXon caps

My sister's WRT54G also went out the same way just two weeks ago. I went to visit her and she mentioned that there was a "running water" noise coming from the router/modem box in the hallway. The internet was still running fine at that point, though. Since I wasn't sure if it was the router or the modem that was faulty (and I highly suspected bad caps), I cut the power on both and powered them ON one by one. Now that was a bad move! The router never started again and did the flashing power LED light just like yours + the hissing noise again. I opened it and looked at the caps, but they looked fine.

Long story short (I will make a thread and post some pics of the router when I get back home), the buck regulator was overshooting its output voltage badly due to a bad cap (actually the only cap) on the input (high) side. Although I haven't done a thorough investigation yet due to lack of SMD tools, I suspect the voltage overshoot has taken out at the very least the wireless IC and also possibly either the RAM or flash IC... or both.

Moral of the story: don't ignore crappy cap brands just because they look "okay" - especially if those caps are in a buck regulator circuit (i.e. near a coil). I used to think that only the output (low) side caps were important on a buck regulator and that the input (high) side caps couldn't cause any catastrophic issue if they failed. Clearly, this is no longer the case.

There's a reason why the datasheets for the SMPS ICs say you need an input bypass capacitor. :P
If I recall, it isn't under nearly as much stress as the output cap(s) though.

In a boost converter, the input caps are under more stress from more ripple current.

[Per Hansson]

Re: WRT54GL v1.1 bad CapXon caps

Yea momaka I can easily see how that would cause it to die.
Because as I said that capacitor you mention showed 50 ESR when it was hot.
But as it cooled down my meter could not even take a reading anymore...

His other router which was working had the same behavior.
I mentioned the incoming cap was at ESR 3.5 but that was cold.
When I measured it in-circuit and hot it was 1.0 which is better than what this CrapCon is rated for according to the datasheet...

I did hook up a serial port but it output nothing.
Unfortunately I must have been asleep when plugging it in because I had forgot that I desoldered the diodes.
So now mine is more thoroughly fried than before, not that I think there was any chance of reviving it anyway...

[Per Hansson]

Re: WRT54GL v1.1 bad CapXon caps

His other router which was working had the same behavior.
I mentioned the incoming cap was at ESR 3.5 but that was cold.
When I measured it in-circuit and hot it was 1.0 which is better than what this CrapCon is rated for according to the datasheet...

I have another WRT54GL that I had discarded since it's Ethernet ports did not come up.
They are after all of limited use with their 4MB flash and 16MB RAM.

But I was a little curious and connected up a serial port to it to see if it was really dead.
Infact it was not, it was just sitting at the CFE!
I pressed CTRL+C to get into the CFE and issued a "nvram clear" followed by "reboot"
Then it booted right up into the DD-WRT (Eko) image that was on it and the Ethernet ports lit up!

[COLOR="DarkOrange"]<SNIP>[/COLOR]
Boot version: v3.7
The boot is CFE

mac_init(): Find mac [00:25:9C:XX:XX:XX] in location 0
Nothing...

No eou key find
Reading :: Failed.: [COLOR="Red"]Error[/COLOR]
CFE>
CFE> [COLOR="red"]nvram erase[/COLOR]
*** command status = 0

CFE> [COLOR="red"]reboot[/COLOR]

[COLOR="DarkOrange"]<SNIP>[/COLOR]
Committing NVRAM...done
[COLOR="DarkOrange"]<SNIP>[/COLOR]
Boot version: v3.7
The boot is CFE

mac_init(): Find mac [00:25:9C:XX:XX:XX] in location 0
Nothing...

No eou key find
Device eth0: hwaddr 00-25-9C-XX-XX-XX, ipaddr 192.168.1.1, mask 255.255.255.0
Reading :: Failed.: [COLOR="Green"]Timeout occured[/COLOR]
    gateway not set, nameserver not set
Loader:raw Filesys:raw Dev:flash0.os File: Options:(null)
Loading: .. 3916 bytes read
Entry at 0x80001000
Closing network.
Starting program at 0x80001000
CPU ProcId is: 0x00029008, options: 0x0000004d
Linux version 2.4.37 (eko@dd-wrt) (gcc version 3.4.6 (OpenWrt-2.0)) #5009 Sat Apr 7 11:37:51 CEST 2012

I had already recapped this router.
Since the original caps where CapXon ST 220uF 16v I had decided to use something cheap.
(Instead of poly modding it like in the first post above).
As the Capxon ST are general purpose caps I had recapped it with UCC KMG 220uF 16v caps.
But I noticed allot of noise on the incoming bulk capacitor fed from a linear power adapter:


I swapped the UCC on the input for a lightly used Suncon WX Low ESR 25v 470uF cap:

-As you can see the difference is quite remarkable:
All the high-frequency noise that was interposed on the waveform is gone!

Later I also swapped the rest for Rubycon ZLH 220uF 16v caps.
At position CX10 (5v) I also noticed allot of ripple, which was reduced to about 1/10:th of what it was before.
So just because a manufacturer has (cheaped out) and put in GP caps does not mean that is what's best for the device!

[stj]

Re: Linksys WRT54GL v1.1 bad CapXon caps

crap caps are standard.

i just re-capped a netgear DG834GT - the things are full of Teapo and one perticular one always domes and causes the wifi to drop.
no need to say what i replaced them with.

[Per Hansson]

Re: Linksys WRT54GL v1.1 bad CapXon caps

Here is an after shot, after as in after recapping the second time and then cleaning the PCB that was absolutely filthy with acetone.

crap caps are standard.

Yea but special "low profile" GP caps rated 1000 hours in a case that can accomodate 20mm Z height?
Sounds a bit overboard on the planned obsolescence part...

i just re-capped a netgear DG834GT - the things are full of Teapo and one perticular one always domes and causes the wifi to drop.
no need to say what i replaced them with.

I know that router all too well
https://www.badcaps.net/forum/showthread.php?t=7475

[momaka]

Re: WRT54GL v1.1 bad CapXon caps

Later I also swapped the rest for Rubycon ZLH 220uF 16v caps.
At position CX10 (5v) I also noticed allot of ripple, which was reduced to about 1/10:th of what it was before.
So just because a manufacturer has (cheaped out) and put in GP caps does not mean that is what's best for the device!

Yup. Especially when you are dealing with a buck regulator - in this case, looks like three on the board, going by the number of toroid inductors. Speaking of inductors, looks like those have a fairly large number of turns. That means the buck circuits are probably not running that high of a frequency. So in that case, going up in capacitance may be a good idea. 220 uF is a joke. I would have gone with 470-1000 uF caps for everything, with the outputs getting 6.3V or 10V caps. You can see the board can accommodate up to 8 mm caps. Might as well take advantage of that. Also, 470-1000 uF caps rated for 6.3 or 10V won't cost that much more than 220 uF caps.

Awesome save again!

[Per Hansson]

Re: Linksys WRT54GL v1.1 bad CapXon caps

Yea the switching frequency is 150KHz for the 5v and 3.3v supplies.
Controlled by a AP1509-50 and AP1509-33 respectively
The 1.8v supply for the processor runs at 300KHz controlled by a AP1605


This picture is from a WRT54GS v1.1 that I recapped yesterday, but they are pretty much identical.
On this one I used Rubycon ZLJ 470uF 16v caps, they fit very nicely on the PCB.
For input I used a Suncon WX 1000uF 16v cap.

I tested a Sanyo SEPC 560uF 4v cap for the processor but it resulted in worse noise.
So the switching frequency is too low for ultra low ESR capacitors. (0.007 in this case)
I actually did not have a scope when I recapped the WRT54GL in the first post in this thread.
Would be interesting to see if the ESR of 0.020 was too low with the Samxon ULR 220uF 16v caps...
At least it tolerates the 470uF Rubycon ZLJ caps well, their ESR is 0.075
And the 220uF ZLH is no problem at 0.094