Building my pfSense Machine - Reviving my Old Desktop

For quite some time, I’ve wanted to have a second internet connection. We’ve been served by PLDT DSL for 7 years now and I can say that the connection is quite reliable, though, it also has down times. I’ve wanted to have a continuous internet connection even when one connection is down, a failover. I also wanted to distribute the load between two connections, to minimize slowdown when somebody is downloading a large file or watching some video in YouTube.

Just a week ago, we registered for that second internet connection, wi-tribe’s WiMAX network. We opted to try wi-tribe, as their nearest base station to our home is just a block away, or, roughly, just 30 meters away. I cannot yet say whether the connection will be reliable for the days to come, but so far, it’s providing a continuous connection at consistent speeds of 0.7Mbps to 1Mbps.

Now that I have two internet connections, I’ll then only need to purchase a Load Balancing Router, or upload a Dual WAN firmware to my CD-R King CW-5354U Tomato Router. Since the former will cost me around PhP 4,500 and I have other plans for the Tomato Router, I needed more options (aside from buying another Tomato Router). Then I remembered my old desktop computer (that’s doing nothing but collecting dust), and pfSense.

pfSense is an open source firewall and router distribution based on FreeBSD. Aside from being a firewall and a router, there are plenty of other features and packages available for expansion as the user see fits for his application. Among the features and applications are caching proxy Squid, which I’ll use to cache my daughter, Anika’s, mostly viewed YouTube videos, and Captive Portal, which I am currently exploring as needed for one of my projects at work. I now have an excuse to run a more power hungry machine than a simple and small Load Balancing Router. :D

My old desktop computer is an AMD Athlon 64 on Biostar NF61V with 1GB of RAM. Since the motherboard is microATX, I purchased a microATX chassis to keep the overall size of the machine small and can easily be placed where my router sits now.

Unfortunately, while transferring the board I noticed the capacitors in the VCore supply are already bulging, a very obvious sign of a bad capacitor. I tested it and yes, the capacitors are bad, VCore voltage is down to around 100mV, and the BIOS doesn’t enter POST (Power-On Self-Test).

I am wary of buying replacement capacitors locally as they may not satisfy the original design criteria of very low ESR, and high temperature handling. I’ll then be waiting for the next time that we’ll have to purchase parts from Digikey and order some Nichicon or any other high quality capacitors. But to try my luck and for my machine to start progressing, I passed by Raon the next morning and found replacement capacitors of the same size and value at DEECO, the brand, Acon. I don’t have an ESR meter so I just measured the relative difference in the ESR of the bad capacitor and the replacement using a signal generator and an oscilloscope at the office.

Desoldering and replacing these capacitors is actually not a simple task as they are connected to very large ground and power planes at the motherboard PCB. It took me around 10 minutes to replace all of them. After replacing the capacitors, it’s a good idea to test the VCore voltage while the CPU is not plugged in. It’s also a good idea to cover the new capacitors before powering up the board as there may be other problems like shorted MOSFET’s, or a dead controller IC that can make the capacitors go ballistic. I’ll charge it to experience working with switching power supplies for quite some time now.

After making sure that there are no problems with the new capacitors, and the VCore is stable at around 1.2V, I plugged the CPU. The BIOS now enters POST and makes those beeping sounds as there is no RAM plugged. Measured VCore voltage with the CPU plugged in is 1.218V, while the BIOS reports 1.35V.

To verify whether the locally available replacement capacitors have acceptably low ESR, I measured the ripple voltage across the capacitors. The ripple voltage in a capacitor is directly proportional to ESR, hence, a low ripple voltage will mean the capacitor has a low ESR.

The waveform has spikes and oscillations, most probably because I’ve used the ground clip of the probe which has a finite inductance that will resonate with the capacitors. Even so, the recorded ripple voltage is just 23mV Pk-Pk, including the spikes. For me, it’s a remarkably low ripple voltage as I’m used to seeing 100mV ripples with cheap locally available capacitors. I am actually excited to compare this result with that from high quality capacitors. Perhaps I’ll post a follow up once they’ve arrived. J