Repairing A Router Plagued By Capacitors

[psgarcha]’s modem/router comes straight from his internet provider, is on 24/7, and is built with the cheapest components imaginable. Eventually, this will be a problem and for [psgarcha], this problem manifested itself sooner than expected. Fortunately, there was a soldering iron handy.

The problems began with a boot loop – starting the router up, watching the blinking LEDs, and watching these lights follow the same pattern forever. Initially thinking this would be a problem with the firmware, [psgarcha] did the only thing he could do – take it apart. Inside, he found some bulging capacitors. Unsheathing his iron and replacing the obviously faulty components, [psgarcha] plugged the router in and had everything work. Great. Until those caps failed again a few months later.

There was obviously something wrong with the circuit, or wrong with the environment. Figuring it was hot out anyway, [psgarcha] replaced those caps again and added a fan and a small heatsink to the largest chip on the board. This should solve any overheating problems, but the real testing must be done in summer (or putting the router in a well-insulated enclosure). It’s an easy fix, a good reminder of exactly how often caps fail, and a great example of reducing the electronic cruft building up in landfills.

59 thoughts on “Repairing A Router Plagued By Capacitors

  1. Electrolytic capacitor life is mostly a function of the ratio between rated temperature and internal temperature, which is mostly a function of ripple current and ESR. If you want caps to last much longer and can’t redesign the circuit to reduce ripple current, going to 105C-rated instead of 85C-rated will make a big difference, as will the use of low-ESR caps. The main contributor to cap internal temp is ohmic dissipation from ripple current, so installing low-ESR caps which have thicker foil in them will reduce that dissipation greatly, and therefore the temperature too.

    Putting several smaller ones in parallel (requires that you have plenty of room available) will also help with the heat dissipation. Assuming that a higher capacitance is acceptable, several larger ones can be even better.

    1. One of the things to note that capacitors have multiple parameters besides just voltages and values. There are temperature ratings, lifetime, ripple current rating & ESR. There are multiple capacitor types with different tweaks optimized for different applications. So in this case, you want low ESR caps for switchers. If the caps or the place that you are buying from do not have these parameters available or cap types to choose from, you need to shop elsewhere.

      Larger physical sizes caps and higher voltage rating ones by their nature have larger surface areas, so better heat dissipation, less current density/voltage stress and they tend to have a longer life (even from the same series).

      1. I agree with the surface area issue. It seems that this is always overlooked. As capacitors get made smaller and smaller this becomes even more of an issues. Even in the article picture above you can see that the caps are much smaller than the marking on the PCB.

        When replacing them don’t even bother buying 85°C (RB) capacitors, you need at least 105°C (TKR) or better still the 105°C low series resistance (ESR). ESR’s are generally only available in low capacity values.

        Another thing, if you have bulging caps then any that are close by or have been exposed to the same temperatures that are under 2.2uF (electro’s) are also dead even though they may look OK.

        Often in Switch Mode Power Supply Units you will find the larger secondary filter caps bulging. Often it is because of a faulty smaller cap in the regulation path. It will always look ok. Most often it will be between 0.68uF and 2.2uF.

        Cap meters are not much good for testing because caps have a wide manufacture tolerance. An ESR meter will very quickly tell you when a cap is dried out.

        You would be surprised how many things you can fix and quite quickly with an ESR meter. LCD monitors are a classic. $10 worth of caps generally gets them going again.

        Back to the surface area issue. To get a larger surface area just use a cap with a higher voltage rating.

    2. I always wondered about the placing of capacitors on motherboards right next or even under the heatsink of the hottest parts, and assumed that was the cause of many issues.
      But the manufactures do not seem to consider that issue at all.

      And not just motherboards of course, they often put them next to heat sources even when they could put them elsewhere without impacting the circuit..

  2. Chingie chingie ching ching…ching CHANG chung. Usually it’s one of the three letters brands. Then there is TEAPO made out of the brown dregs of a teapot.
    Double the voltage rating for decent service life.

  3. As much as I love seeing repairs bring back electronics to life, I think its worth considering whether some repairs are actually “worth” the effort/time/resources required. Obviously it wasn’t particularly expensive or time-consuming in this case… But those all-in-one no-name wireless routers are such horrendous pieces of garbage that I would never have bothered.
    Their problems lie beyond the hardware level. The firmware, and particularly the wifi drivers used in the firmware written by companies like ralink/mediatek, in my experience are so unbelievably buggy and outdated.
    This also isn’t the first I have seen of tacking a fan onto a router to try and improve stability. Read back on some forums during the wrt54g era. People really used to go nuts at it.. Attaching massive CPU radiators sticking out the top “cooling” some tiny SOC . All in a hopeless belief that this would fix their wifi from dropping out once a day or “improve routing performance”.
    Another community that likes replacing capacitors thinks they can improve sound quality of audio gear by slapping on the largest most obnoxious caps in place of what the engineers designed it to work with.
    But I digress.
    I would have chucked that thing so fast in the trash and bought a better ADSL modem and a separate router. But good for not doing that.

      1. Exactly. And that’s what many people were doing with the WRT54g. It was a nice hackable piece of hardware, relatively inexpensive, and ran linux. Several GPIOs available to be repurposed. In some ways, I’d be willing to say it was the commercially available precursor to products such as the Beagle Bone and RasPI

    1. Well, as always, we get what we pay for. If some company were to build a modem/router correctly designed, with quality parts, for the adequate price, just a few people would buy it, because of the much cheaper alternatives that “just work the same”.

      I don´t believe in the audio fanatics ideas, but if one buys audio gear from these cheap chinese brands, some improvement can be had swapping the caps and other components for quality ones, just because the “engineers” didn´t *designed* them to work correctly, they determined the cheapest part fo make it *barely* work.

      1. This sounds like the Keeley mod for a Line6 pedal which basically involves shelling out ~100 bux for him to switch the smd garbage out with real components. It still doesn’t fix the ZIF socket for the cpu which gets stomped out of the stomp pedal lol. And the tiny clicky buttons that are not meant to be stepped on and the tiny plastic nubs inside them and I could go on lol. It is funny but folks get really used to their gear and their particular sound color. Whether or not it is rational, is a personal taste I guess, or the salesperson’s ability “dude you can just hear those highs singing on that riff” lol.

    2. I have to agree, I recently repaired a DVD player.. I was ready to bin the thing and get a new player except we needed our DVD back.

      So I cracked it open to get the DVD out, and saw the blown cap… sighed, collected the DVD, pulled out the soldering iron, and replaced the cap with one I had laying around.

      3 months later I still bought a far superior device (one from this decade) and the DVD player is in a bag to go to the thrift store… The problem with the DVD player that finally made me junk it, was the remote not working (which had never really worked properly, and wouldn’t have been an issue if the DVD player was any kind of name brand and would work with any kind of universal remove)

      Some times it’s easier to just junk things and move on.

    3. When the Linksys models of that generation would fail, it was nearly always the wall-wart that was the problem, not the caps on-board or the heat issues (though they ran so close to the edge, extra cooling certainly didn’t hurt anything, it just didn’t really help anything either). It’s like Linksys designed the wallwarts to fail at the 12 month point.

  4. Most modern day SMT parts relies on the PCB to act as heat spreader. Guess what happen when the modem is lying flat? You have effectively reduced the surface area by a factor of 2 as the hot air on the bottom side of the pcb is trapped.

    BTW modems should really be mounted vertically so that you get air flow due to convection. That case with slots across the sides of the case means it is designed to have air flow across the width of the modem. You want the hotter side (if there is one) towards the top so that the hot air escape and creates a low pressure to draw cooler air in from the bottom. That’ll makes a big difference.

    Without air flow, you can stick as many heat sinks on chips/ram etc and it won’t make a difference once it reaches steady state.

    1. Which is the thing so dumb about making PCI cards backwards from ISA slots. PCI came along quite some time after the first tower computer cases, which tipped the power supply end up and positioned the expansion cards with the heat producing components on the top side.

      The *logical* design would have been to make PCI oriented to match ISA, VL-Bus, AGP, and MCA boards so that in tower applications the hot components would be able to convect and radiate heat UPWARDS.

      But since PCI and PCI-X and PCIe are all backwards and the computer industry has resisted all attempts to stand PC cases on the other end*, those cards continue to trap hot air around their components.

      *A few proprietary designs were around in the 90’s. BTX was a good idea, adopted by several large OEMs, but even those continued to produce ATX systems with all the cards hot side down. Once the ISA bus was no more, the entire industry should have said “OK, it’s time to fix this design flaw and flip all tower case designs the other way!” but they stubbornly continue to build systems designed for the heat flow of pre-PCI expansion cards.

      1. It would need confirmation but it wasn´t adopted mainly because it was licensed ( manufacturers would have to pay royalties to someone to use the format ) and also because they didn´t consider important to make the holes and etc compatible with old motherboards. So case makers and board makers would need to have two lines of the same products, cases and motherboards. That was not reasonable cost-wise.

      2. To improve the air flow for the PC slots, one would want to cooler air in from the front of the case to flow toward the expansion card, and warm air sucked out horizontally to exit the case from the side.

        In a servers/frame/shelf, they have forced air cooling across the width of the board to address that and not having individual fans.

        They probably flipped the PCI board originally to take advantage of the existing cutouts in a PC case (AT era?). So you get to choose 3 PCI + 2 ISA + 1 of either kind. Back then, the slots are filled as nothing was integrated. Not fixing it when they had the PCIe was silly.

  5. Do I believe in coincidence? Only this weekend I replaced 11+1 capacitors in a dead “Teledat 300 LAN” DSL modem. The single one was on the primary side of what appeared to be a switched power supply, but that didn’t help, so I took out the 11 caps on the secondary side (note to followers: do make some kind of documentation of the polarity of the caps, e.g. make a photo or put it onto a flatbed scanner) and replaced those. That did the trick and the lights are back on. Testing is due …

    1. The primary side high voltage cap almost never dies because there is no significant ripple current flowing through it There is only 100Hz ripple on it from the bridge rectifier at the mains entry, which is “almost DC” from the point of view of the cap.

      On the other hand, the caps on the secondary that have to conduct the 100kHz-1MHz ripple from the switching supply, this is where a high capacity cap will have a low impedance, thus you will get fairly high currents. Unless they are low ESR, high temp rating they will die rapidly. Even good quality low ESR electrolytic caps will eventually die in this application, it just takes longer – hopefully longer than the practical lifetime of the gadget using them.

    2. In a switch mode converter, the input filter cap do see significant amount of high frequency ripple current. Most likely you are dealing with a buck (step down) converter where the input cap sees discontinuous current drawn from the inductor.

      http://www.maximintegrated.com/en/app-notes/index.mvp/id/842
      >This application note helps a system designer to set-up Mathcad and compute CIN for a particular step-down DC-DC regulator design.

      The trick to deal with these ripple currents is to use ceramic caps for the high frequency high current ripples with parallel electrolytic for bulk capacitance.

      I think these modem manufactures know about trapped heat problems. By throwing excess caps at the symptom, they can get away *longer* with the increase in ESR overtime (due to heating from improper ventilation).

      The cramped Speed Touch ST516V6 is a good example of this practice. The extra caps are actually input filter caps and the heating is caused by them using a linear regulator to drop from the 20 odd volts down to 5 for some analog supply. Having the caps right next to the hot regulator makes it worse. Typical life for these modem is about 2 years. If you mount the modem vertically with the left side at the top, then the modems get some chimney effects cooling and helps with extending its life.

  6. The time spend (wasted) on the fix, even considering common minimal wage levels, costs more than the sodding device. you’ll find better replacements for tuppence in a thrift store.
    Bin it, don’t even bother sourcing parts, just bin it.

      1. Well, where I live the replacement caps would likely cost more than the stupid crappy router. I was doing this sort of capacitor replacement recently on an 8 years old Samsung LCD and an LG LCD TV – each time it was about 50 euro in caps. Factor in the work, the time you spend fixing it and sometime it really doesn’t make any sense to repair it.

        I do hate throwing stuff away, but I don’t like throwing good money after bad neither – that router would likely fail for some other reason later again, it such piss-poor cheap design.

      2. The problem feeds itself. People keep buying cheap electronics which use cheap parts that fails early. When it fails, it is cheaper to throw it away than to get quality replacement parts. What is worse is that even expensive brands are made like this to increase their profits.

        Thankfully they mostly cut corners at the power supplies and not the more complex chips (so far), so some levels of repair is possible. I sometimes do preventive replacement work and swap out caps with power supply grade tantalums for low voltage circuits in some cases or high quality electrolytics. I also replaced the wall warts with much better built ones (from the old days with real Japanese caps.).

    1. I did and still do some similar projects, wasted more hours repairing some crap during which, had I been working, I would have made more than the final value of the repaired object. But …… if you look at the time you spent doing this as “fun time”, time where you did this instead of watching TV or wasting time on Facebook, then it’s a win-win – you had some fun repairing it and you spared the cash you would have used to buy a new one. Yeah, some times you get frustrated when things go awry and the fun part is gone, but I could say the same about a bad movie you paid for at the cinema and walked out half way through. And yeah, I guess only a geek would look at this as “fun time”, but then again, this is Hack-A-Day, right?

      1. Hack-a-day, not Bodge-all-week.

        I agree with some of the post above, that the problem lies in us, as a society, continue to accept that this crap is made because we either choose to buy it or accept we get shafted with it.

        Trying to get this rubbish to work actually makes the problem worst in the same way that the tech-savy neighbors kid from the 90’s facilitated the acceptance of poorly designed slow computers and software for families in that same decade.

    2. If you include the time spend to fix this one, why don’t you also include the time spent looking for another device and setting it up ?
      Also, Ghana is one of the country who thank you for all of your electronic waste.

    3. Or: Dont bin it, save the world from one more piece of electric garbage and just repair it.
      I too have a price far above minimal wage, so in your view my time would be too valueable to fix those things. But I think this is a very shabby argument to trash the world, only because we have a economical system that is so much more significant then the ecological one…
      In the same Argument I have to assume that you run over every red light and just punch the food into your pie-hole, because you know: Your time is so precious.
      Your not wrong, youre just an ass for throwing away every piece of electric only because someone let you know between 1980 and 2030 that your time is worth X and therefore you can not afford to sink an hour into it. Get over it and fix that stuff.

      1. The extra earning would be taxed at your marginal rate, so you really aren’t getting paid at your regular rates. You are fixing stuff vs buying something new with after tax income. Also assuming that your extra work is only for a one time event of fixing something instead of a longer term commitments like extra shift or part time work. :P

        What I see is to fix stuff preemptively to prevent the cost of the down time on a piece of networking equipment. The down time is sometimes worse than just the cost of the equipment especially when you need your interweb or VoIP or not die in a (gaming) raid at that moment. :)

      2. No, just make sure you don’t buy it. If you buy shit, it will be a brick sooner or later anyway. Do you want it to be an inefficient piece of crap risk? You bought it and with that all the environmental impacts that low cost manufacturing brings with it. The damage has been done the moment you deemed it an acceptable product to spend money on in some shape, way or form.

        When I say that the time/money equation isn’t worth it I mean to say that it is both economically and ecologically more sound to simply spend some monetary units on proper gear that will have a measurably lower impact on resources, even if you have been an idiot stuck with crappy hardware.

        Just say no !

    4. True. I work for the ISP, and those home routers live about 4 years. When they fail, capacitors are most often the problem. I suggested that capacitor replacement would be the solution for most of the dead devices, but my bosses told me that those devices costs us about 15€ a piece, and my working hour costs more than that so attempt to repair it is just wasting money/time.
      I took apart few devices, and inside they all look like someone took reference design schematic, transfered it to autoplace/autoroute software and send it directly to the production line.

    5. no, do NOT throw it out!

      all you have done is added to landfill and NOT solved a single problem. you are encouraging chinese vendors to keep using fake caps by paying them again and again for the same box.

      best is to replace the caps with low esr brand-name ones (pany and nichicon are my goto brands) and then you will have a box that lasts years, even decades.

      re-buying won’t ever solve the problem is you are buying the same crap china junk that just blew out!

      1. Nonsense, one encourages them by actually accepting the crap. I do not have to throw it out because I see to it that I do not have to deal with it in the first place. That being said, the idea that binning electronics that have already have been produced only added to landfill is not true. You will in fact have removed an inefficient and statistically more dangerous device from operation.
        But primarily, you have to make sure you do not pay for one.

      2. Stop being an environmentalist. It’s partly your damn fault for encouraging people to discard their single power hungry, lead laced, repairable, CRT that lasts more than twenty-thirty years for four or five power sipping WalMart Black Friday flat screens that last only a year or two each.

    6. My thoughts here are heavily dependent upon whether the cap failure is a parts quality problem, or a design problem. But if the replacement parts are of better quality than the originals, you may well find that it lasts much longer.

      Consider that you probably won’t find a replacement that’s actually better in your thrift store, instead you’ll find something of similar quality (crap) to the original. If these things are lasting two years, but your repaired unit lasts somewhat longer, say 6 years, then you get to skip two cycles of downtime in those six years.

      Especially if you’re running firmware like OpenWRT on your router, finding a replacement that meets all your needs can be a struggle; if you have one that you already like, but it isn’t sold anymore, repairing what you have may be of greater value than the struggle of finding a new one that provides good WiFi coverage in your particular home.

      1. OpenWRT compatible hardware isn’t as exotic as Tomato and variants firmware that needs to be run on Broadcom MIPS chipset with their CFE. At any time, there was just a handful of routers would run it. (ARM port is being worked on).

  7. Repairing cheap crappy stuff is often more rewarding than (attempting to) repair more expensive stuff. Cheap stuff is often poorly designed and quite often the failure is obvious or at least not difficult to find. A quick fix or a quick decision to through it in the bin, that’s what I like.

    On the other hand more expensive stuff is better designed and the reason for failure is less common and harder to find. Also, a lot of the time it is a very complex failure and turns into a bin job anyway because it takes far too long or is simply too expensive to buy all the parts. Expensive equipment often has a catastrophic failure.

    In any case, if you don’t gain experience by fixing cheap crappy stuff then you probably won’t be able to fix the more expensive stuff.

    I worked in the repair industry for a long time and at the end of that I am sure that I knew more about failure than design engineers. They got me back though by designing equipment with the intention for it to fail after a specific period and they did that with absolute perfection.

  8. I’m sure I would not spend the time to repair it if my router goes out. Like [psgarcha], my router comes right from my ISP (Big “C” in my case). Since I have to pay rental on the thing to begin with, all I have to do if it goes out is to trudge it down to the local office and swap it out with a new one. Then their repair techs determine if it’s fixable or not. Had Ii purchased my router, it would be a different story.

    1. Because rented devices are replaced by new ones in case of lightning or overvoltage accident. If you bought it and lightning destroys it you can only throw it in the garbage and buy a new one. Much more expensive.

      1. Put a good, gas-discharge lightning arrester ($15 on amazon) with a solid ground connection in the line, and with 3 extra months before you start saving on your bill, you will save over a much longer period of time.

        Even if you’re renting the modem, you should consider this — unless you like downtime. The arrestor the cable company puts on your house because they’re required to doesn’t work very well, as you’ve heard.

        (Are you really replacing your modem from lightning more frequently than 14 months? Mine was around $70, and the monthly rental fee would have been in the $5 to $7 range if I remember right.)

    2. Really mostly an ISP issue or people just don’t have $100 lying around for a piece of equipment. (It is far cheaper to run a credit card balance and pay back over months with the heavy interest than to rent it.)

      Some of thee Helldesk/”techs” like to blame it on customer’s equipment than their infra structure, so the first thing they would try to fix any issues is to ask you to swap it. For a rental it is a matter of picking up a new one or swap it when the tech shows up.

  9. I once bought a rackmount Netgear 48 port managed Gigabit switch, knowing that it didn’t work, from a surplus store for $30, in the hopes that I could get it working again. When I plugged it in to test it before purchasing it, the led boot up initialization would start and then freeze. I could also hear a high pitched squeal from the power supply. After I got it home and opened it up I saw a dozen or so bulging Teapo caps, which I replaced with 105 deg Panasonics. The open frame power supply still squealed (but worked), so I found a higher wattage one that happened to be the perfect footprint on Mouser and replaced it too for good measure. It worked great for another $50 or so in parts (mostly the power supply).

  10. 1st failure due to cheap capacitors. 2nd failure accelerated due to replacement with capacitors that weren’t low ESR.

    Low ESR capacitors must almost always be used in switching power supplies, or they will overheat and eventually die. Neither [psgarcha] or Brian acknowledged that – although the article was TLDR, a search reveals ESR was not mentioned once. That is important to know about! Lacking that info, I suppose a fan is a clever workaround; but I’m not sure how long it will extend the life of a cap used for a purpose it wasn’t designed for.

    I keep a decent selection of quality low-ESR caps on hand for this type of repair. Panasonic is one of the good brands, and my personal favorite. In general, stay away from caps ending with “con” in the brand name, most are cheap brands that will fail; although Nippon Chemicon seems to be an exception.

    As for whether this type of repair is worthwhile… If the item lasts long enough for cheap capacitors to fail, it will usually last darn near forever once quality replacement caps are installed. Which can’t be said about any newer item you might replace it with. Older stuff has an advantage, because the quality of new stuff just keeps getting worse. Increasingly finer pitch components combined with shoddier solder joints. Flakier firmware. Often more artificial restrictions on use too. In my opinion, repair of an item still useful to you is almost always worth it.

    1. Nichicon is another a good capacitor brand. Nichicon, Nippon Chemicon and Panasonic in that order are the ones I pick. Haven’t used Rubycon parts, but they are supposed to be a good brand too.

      Capacitor selection chart: http://www.nichicon.co.jp/english/products/alm_mini/daia2_c.htm
      That page alone was just for low ESR long life caps. It’s like Hallmark cards, a cap series for every occasions.

      The chinese crap ending in *con names are there to con you. :P

      1. Hehe, yes! I’m comfortable with Nichicon too. Rubycon seems to be hit and miss though, having dealt with a few that died prematurely.

        All this reminds me of a fellow who specialized in repairing motherboards with failed caps. He kept all the old caps and posted a memorable picture called “The Green Mile”. Seems he’s not in business anymore, but I found a copy of the picture:

        http://icrontic.com/uploads/7/9/9011.attach

        That’s a lot of fail. And a goldmine of reliability info for various brands too. Out of curiosity I Waybacked his site, and though perhaps dated now, his brand recommendations were Nichicon and Panasonic.

        I sure wish Radio Shack carried low ESR caps. With cap failure being so common and easy to diagnose, it would be a great convenience, that many would gladly pay their typical markups for.

      2. The -con part is short for condenser (aka cap) and Japanese have a habit of stringing or using a single syllable to make a short form.

        Hope those new solid state caps in new motherboards would last forever as there are no longer electrolyte to dry up. Only time would tell.

  11. I replace caps with the 105c and the next voltage capacity up AT LEAST. If it’s in a location where I know it’s getting the occasional spike, I go as high as the space inside will allow me.

    How many LCD televisions do you think I’ve salvaged from the crap heap at work just to fix them $10 and about 30 minutes of labor later? Makes great throwabout monitors and friends are always needing a set to hold them until they can replace a main unit that crapped out.

    You get what you pay for, or you hope you do anyway.

  12. Rather than try to make the economics work by fixing my stuff, I source it already broken; either cheap or free. I’m watching a tv–23″ Toshiba lcd–that I pulled out of the dump (they sometimes let me make a withdrawal when I dump off stuff). I thought it probably needed new caps, like its brother that I sourced the same way, but in fact it needed nothing at all. It was a combo dvd/tv unit and the dvd was iffy. I didn’t care. It works fine as a tv.

    I buy next to nothing new. For a long time I couldn’t afford it. Now I can, but why? That said, in fact I’m considering a new 40″ lcd set, because used prices are not keeping up with the extreme price-drop of new units. I can get one for $300 (40″ Samsung), which is about the same price I’d pay at a pawn shop for older tech.

    Well, there’s the problem. New capabilities and new standards are forcing old gear into the bin. My TV requires an outboard tuner to pick up ATSC signals. My old WR54 didn’t have wireless ‘N’ capability, so it went away (to be replaced by a thrift-store unit that did). My DVD player gets very little use, and we won’t even mention the VCR. My XP laptop works fine, but got replaced because the new one cost less than upgrading the OS.

    I don’t have all the answers, but a start would be to make things repairable and modular. If we could replace the parts that were superseded, that would be a plus. I did upgrade the wireless card and DVD unit in the old XP machine, but it’s not worth fixing further.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.