Ten Times :: Mike Neir's Page

I'm really tired of needing to apologize to Comcast. They're evil. I know this. But I'm really getting annoyed with having what I perceive to be their problems actually be problems on my end.

I've been having a lot of problems with random cable modem disconnects due to lack of signal over the past few weeks. I know this because I'm a dork and wrote a script that plugs into my Cacti installation and logs the signal levels displayed in my cable modem's admin interface. I did this years ago because - you guessed it - Comcast is evil and their stuff cut out on me enough for me to want to be able to monitor it. There has been a steady and constant trend of ever-dropping signal levels that has made itself apparent since the beginning of the year. I've had the same general layout in the apartment in terms of devices and number signal paths, so I could only assume that the problem was upstream on Comcast's side. I was so confident that this was yet another Comcast screw-up that I ordered DSL.

I had a service appointment scheduled with Comcast on Friday. The guy did some tests outside where my line plugs in, and he said he noticed a lot of signal "reflection", and said that my splitters and/or wiring were probably bad. He looked around, and also noticed that I was running my cable through the Monster way-too-much-money power strip, and called that a no-no as well. He replaced three splitters with two nicer ones, and ran custom length cables to each one of my devices. The result?

As you can see, the signal levels improved dramatically. The downstream power, which is what I receive from Comcast, is up by at least 10dBmV (depending on the time of day). 10dB corresponds to a ten-fold signal power increase. Signal-to-noise ratio is also better than it was, but not by as striking of an amount.

What's are the lessons here? Use good splitters. Use good cable. Don't assume Comcast is at fault because they're evil.

EDIT: My dad has corrected me. The 10dB increase probably only corresponds to around a 3.1 fold increase in power. I think I read the wikipedia chart wrong.

Phil :: 08/19/2008, 6:50 am :: Reply

I have fought with cable problems for many years. Cables and connection points are the weakest link in every electronic system. The higher the frequency, the more effect there is. Use good stuff, don't kink it, don't yank it, and keep the connections clean.

CR :: 09/05/2008, 12:58 pm :: Reply

As an NVH guy I'm always dealing with dB but wrt sound. Anyway, 1 lesson learned to double check dB calculations is to not do dB calculations! Instead convert to linear scale - where we're all very comfy adding/subtracting/multiplying/dividing - perform your math and then re-convert back to dB. Saves you lots of head scatching and second guessing. I do it all the time in my macros here at work to make sure I'm getting the math right!:)

KC :: 09/18/2008, 4:12 am :: Reply

What are you using to generate that nice graph? I, also, have a flaky Comcast cable modem setup, and I'd like to start graphing the dB in/out to track as well...

Mike Neir :: 09/24/2008, 1:00 pm :: Reply

It's graphed with cacti, using data collected using a screen-scraper script I wrote. The script is nothing too complex, but getting it to work inside cacti was a little more trying if I remember right. It's been a few years since I set it up.

Ben :: 10/01/2008, 11:25 am :: Reply

For every 3db gain, you double your signal level.

Spankey :: 11/21/2009, 2:06 am :: Reply

To be more precise it is 10^(X dB/10) times. So 10dB increase is a 10x increase in power, i.e. 10^(10/10) = 10^1 = 10. 3dB = 2x is a good approximation, the actual value is a bit over 3dB for 2x gain.