Introduction

The purpose of this note is to discuss the conversion of signal level in dBm to percent for the ipw2200 driver. The primary reference for this article is

  http://www.wildpackets.com/elements/whitepapers/
      Converting_Signal_Strength.pdf

An associated access point typically sends 10 beacons per second. The ipw2200 driver produces averages (over multiple beacons) for signal level and noise in dBm. The ipw2200 driver converts the associated AP average signal level into a signal quality (0-100%) and also provides a link quality (0-100%) which is the minimum of five qualities: signal quality, beacon quality, transmitted quality, received quality, and rate quality. The iwconfig utility reports link quality along with average signal level and noise.

For each AP found in a scan, the ipw2200 driver produces raw data for signal level in dBm. The iwlist utility reports the signal level in dBm for each scanned AP. This signal level is not an average. It is a one shot measurement. Consequently, the scanned signal levels and associated average signal levels do not always match. As of version 1.0.3, the ipw2200 driver provides signal quality for scanned APs using the same conversion that is applied to associated APs so that applications like NetworkManager do not have to guess at how to do the conversion.

Measures of Wireless Signals

Here is what the IEEE 802.11 standard says about the RSSI metric.

14.2.3.2 RXVECTOR RSSI

The receive signal strength indicator (RSSI) is an optional parameter that has a value of 0 through RSSI Max. This parameter is a measure by the PHY sublayer of the energy observed at the antenna used to receive the current PPDU. RSSI shall be measured between the beginning of the start frame delimiter (SFD) and the end of the PLCP header error check (HEC). RSSI is intended to be used in a relative manner. Absolute accuracy of the RSSI reading is not specified.

There is no specified accuracy to the RSSI reading. That is, there is nothing in the 802.11 standard that stipulates a relationship between RSSI value and any particular energy level as would be measured in mW or dBm. Individual vendors have chosen to provide their own levels of accuracy, granularity, and range for the actual power (measured as mW or dBm) and their range of RSSI values (from 0 to RSSI_Max). Some vendors do not provide an RSSI value and instead convert directly from dBm to percent. This is the case with the ipw2200 driver.

The formulas that Linux wireless drivers use to convert RSSI to percent have not been standardized so it is impossible to compare signal qualities across vendors. There is room for improvement here. This article will offer two methods of conversion.

The IPW2200 Driver

The tables and formulas appearing in the reference above suggest using a simple straight line model for percent as a function of signal level in dBm. What is needed in the Linux wireless driver community is agreement on a standard approach for setting up this model. Here is a methodology for setting up such a model. For the ipw2200 driver, define

  max_signal = -20 (dBm)
  disassociation_signal = -85 (dBm)

where max_signal is roughly the maximum signal you see when holding a laptop up to an access point and disassociation_signal is roughly the signal level at which disassociation occurs. According to the reference above the 'Roaming Threshold' is typically about 20% and the 'Clear Signal Threshold' is about 10%. Also, normal operation should be expected above 30%. Let's assume that the disassociation_signal level and the 'Roaming Threshold' are about the same. Then, we have the linear model

  percent = 100 - 80*(max_signal - signal)/
            (max_signal - disassociation_signal)

What is the Best Model

Version 1.0.0 of the ipw2200 driver converts associated signal level in dBm to associated signal quality (0-100%) using a linear or straight line conversion formula that interpolated between -50 dBm (100%) and -85 dBm (0%). Development version 1.0.1 converts associated signal level in dBm to associated signal quality (0-100%) using a quadratic conversion formula. The coefficients in the quadratic were obtained by a least squares fit of data obtained using the Windows XP version of the driver along with the IBM Access Connections Utility (see NetworkManager, IPW2200). The development version 1.0.3 also converts scanned signal level in dBm to scanned signal quality (0-100%) using the same quadratic conversion formula. Figure 1 contains plots of this quadratic model along with the linear model discussed here.

Figure 1: Quadratic and Linear Models

Which model is best? There is currently no basis for answering this question. Matching Windows XP values for signal quality was a first quess. The alternative linear model was set up so that disassociation occurs at about 20%.