[wiki:Internal Internal] -> Calibration

= !WiFi Card Calibration =

As the first important discovery in ORBIT testbed, some (if not most) !WiFi cards behaved strangely. Therefore, it is really important to identify those '''bad''' cards. The calibration procedure tries to find if the Tx power and RSSI measurement of the card in each 802.11 channel is ''regular''  or not.

For each card, the calibration has two experiments:
 1. Card-RX exp: A standard signaling generator will generator 50 802.11 beacons in 5 seconds, the card would report its measured RSSIs in a database.
 1. Card-TX exp: A VSA equipment will measure the signal strength of 50 beacons generated by the card

Those experiments are going to repeat at each 802.11 a/b/g channel and with 4 different Tx power levels: 1, 10, 20, 100mW.

== Calibration Set-up ==

[[Image(caliber_setup.PNG)]]
== Calibration Procedure ==
 1. Pick up one node from the grid position (x,y) and take it down from the shelf
 1. load '''calibration-zhibin.ndz''' image to the node
 1. Connecting RF cables to both cards
 1. ssh to console.grid
 1. cd /usr/sbin
 1. Run Card experiment '''./cardcalibrate.pl [x] [y] 0'''
 1. Run Card experiment '''./cardcalibrate.pl [x] [y] 1'''
 1. check the two tables in '''calibration_central''' database in ''idb1''. 

== Database Format ==

The database is named  “calibration_central”. It is now in idb1.
It has two tables:
 *  Rx_caliber: has all measurements by  Card-Tx experiment.
 *  Tx_caliber: has all measurements by  Card-Rx experiment.

=== Example table records ===
Rx_caliber

|| Mac address || Date ||Time  || ESGTxPr || channel||RSSI ||
||3F:65:89:4E:2D:0C||	2006-07-07||	12:29:15||	10||	6||	45||
					
Tx_caliber

||Mac-address||	Date ||	Time|| !CardTxPr	|| channel||	VSAreading||
||3F:65:89:4E:2D:0C||	2006-07-07	||12:30:21 ||	100 ||	11 ||	65 ||

=== Estimation of Calibration Workload ===

Each card shall be placed in TX and RX experiment for both 5 seconds per channel per power setting. Suppose there are 24 [11(b/g)+13(a)] channels, 5 power settings. Each card takes 5*2*24*5 seconds = 20 minutes to calibrate.

Some other overheads:
 * There are some 5~8 seconds overhead to start the experiments.
 * Powering on the node needs to take around 2 minutes.
 * Loading a image on the node takes 3 minutes.
Therefore, calibration a node will generally takes 45~46 minutes.

== For developers ==
 * [wiki:Internal/Mapping/Calibration/DesignSpec Design Specification for Calibration Experiment]
 * [wiki:Internal/Mapping/Calibration/MonitorOmlCode Design of "moitor_oml" program]