Tuesday, April 06, 2010

Sorry about the move

Google/Blogger decided to change the way they interact with externally hosted blogs - probably a reaction to some search-engine abuse since blogger blog 'index' very well under Google.

(This post is a test, plus to move down the redirect message)

This blog has moved


This blog is now located at http://iatip.blogspot.com/.
You will be automatically redirected in 30 seconds, or you may click here.

For feed subscribers, please update your feed subscriptions to
http://iatip.blogspot.com/feeds/posts/default.

How Long to Obtain Cellular IP?

Well, I finally have a SIM, so I can actually do cellular tests again. It is a GSM 'Capped Plan' with 20/MB of data per month for $30. The 'capped' means that even if I go over the 20MB by dozens of GB, the maximum charge will be $130. It is a truly 'unlimited plan' for that $130. What's the catch? If I hit the 'Cap' two months in a row, then the carrier either cancels the account or forces me to start paying full-rate for a normal 1GB or 5GB/mo plan.

The 20MB is fine for me since I am testing methods to keep data small.

My current tests have a sleep devices waking, then obtaining a cell-link. For example, I am working on a Digi gateway called an X3 (not on Digi's web site yet). The X3 can power itself down completely, using a real-time clock chip to wake it at a future time.

I ran a test for 5 days, waking the X3 device every 10 minutes and clocked how long it too for the GSM carrier to assign an IP to the X3:
  • The fastest time was 23.7 seconds (half a minute)
  • The slowest time was 97.0 seconds (1 and a half minutes)
  • The average time was 28.6 seconds (so most near 30 seconds, with a few long outliers)
  • Note that there will be several seconds of CPU start-up time in there - I can only know the RTC wake-up time (which I loaded into the RTC last cycle) compared to the RTC time when the IP address is assigned.
Why is this important? Because the cellular modem consumes a LOT of power for the duration of this variable time. It is one of the variable which makes estimating battery life difficult.

My next test will include actually doing work - probably sending a Modbus/TCP transaction to a DSL-based server.