NetGear FS108 is eight port switch. Switch - this is the box, where you put one network RJ45 cable from router (best to port 1, so things remain organized) and in the rest seven ports you can put cables from the single PCs and suddently everyone is in the network and on the net Anyone who is seriously crazy about computers and got several of them home surely know this. I got too much of them (not to mention my stepbro and my sis), so except for the four ports on my router/firewall/WiFi gate I need another two eightport swiches. So about 3 years ago I bought this switch - http://s4.photobucket.com/albums/y134/trodas/?action=view¤t=networking_by_netgear.jpg - is the blue sturdy metal box (metal is 2mm thick!) bellow with writings NetGear Fast Ethernet Switch FS108 What I did not suspected was, that inside of the switch are plenty of bad caps - G-Luxons! G-Luxons are used both as the small little support caps and as the big voltage filtering caps for main chip. This is pretty big problem already, but wait, it get worser. It is also good to say that it is a very bad idea use those inferior caps for voltage filtering. The regulator was tortured by them quite badly, Guantanamo are nothing compared to this, you will see. Even on the main voltage input filter is sitting the G-Luxon like frog on the fount. Complete look at the insides of the switch - and such nice clean design has been turned into completely useless crap by inferior caps. And now to the before mentioned regulator. Even the voltage filtering for the main chip of the switcher is not a hard task, with G-Luxon "on your back" it become quite a task. It it not visible as well, as with naked eye thanks to the reflection, but you can clearly see that the regulator was overheating that much, that the PCB color is changed from green to brown under him! I can only hope that these images show all the skeptical of the "caps mania" users that this is not a "caps mania" or crazyness, much less deviation. To put it simply, bad caps overstress other components till they break of the cap explode. Or till the failure is catastrophic - it catch on fire. As you can see for yourself, it was not hat far from the worst case - regulator between G-Luxons, note the PCB color changes: So, how that worked out? Well, I bought the switch like 3 years ago and just after two or there months it begin emitting a little hi-pitch noise when on. Not too loud, but using a hot glue on the coil and adding sound-absorbing foam helped only to the point where it was hard to hear the noise from like 6 feet and more, but it did not quiet the switch. After like a year the switch become so hot in operation, that I was forced to cut a big square hole into bottom of it, just bellow the main chip to get it cool down. Of course against G-Luxons that helped only a bit and just for a while - in the last year of usage is the switcher unstable, packets are lost and need to be retransmissed so after it almost killing my net (even that - God thanks for the TCP/IP packets checking and resending!) and getting a HighID (that just means you have public IP and your ports are correctly open and used) on eMule server in eMule on machine connected to net thru this switch required mostly 5 attempts, but sometimes more and in like 10 attempts you get temporally banned for repeated connections flood... In short, this was not going to get like that any longer. Recap cap need: NetGear switcher FS108 rev. B1 ------------------------------ 1x 1000uF 25V d10 x 16 - Samxon GD 1000uF 16V d10 x 16 3x 470uF 16V d8 x 16 - Samxon GC 1000uF 6.3V d8 9x 47uF 16V d5 x 16 - Samxon GK 47uF 25V d5 In these 470uF -> 1000uF caps I can go from 16V to 6.3V because I measured the voltage and just after the input regulator AP34063 ( http://www.diodes.com/datasheets/AP34063.pdf ) it is from like 12 - 13V on input are made just 3.3V for everything else and in the end, from the 3.3V are made 1.8V for the main chip. So using here a 16V caps means cheat itself unbelievably on the caps parameters, that are made for 16V and that lost most of it's quality if it run on only 1.8V instead... Input cap has to be 16V one, tough. Regulator AME 1085 ( http://www.digchip.com/datasheets/download_datasheet.php?id=143806&part-number=AME1085 ) will be now fondled with big and good caps - after all, his operation temp is only up to 70°C, so it is not need that the PCB will go brown on the run, right? Little support caps could be probably better for lower voltage - I'm sure there will be only 3.3V on them top - but what the hell. At least I used high quality (in these small capacitances) Samxon GK caps. And last image - last Samxon GC before main chip that regulate and cleain it's voltage - 1,8V Results? Difference like between night and day on my total network load. Before upload 20k in eMule is killing my browsing and render internet unusable and now I did not even noticed that it is already 1:00 at night and eMule scheduled upload speed jumped into 20k...! eMule is connecting w/o problems on the server, no more KAD network connection drops in eMule when 2000+ connections are used and speed of data transfers between PCs also increased dramatically. Almost to twice the speed that was before, when it is between two PCs with gigabite NICs, then it at peaks almost touch 12Mby/sec I'm very happy with it. Finally. G-Suxxons never more!