Re: [Ipw2100-devel] link quality patch

[Bill Moss <bmoss clemson edu>]

This patch seems to have changed the behaviour of reported link quality 
for the better. I do not see inaccurate values at boot as I did before 
and the values do not seem to jump around as much. The model is most 
accurate for the high percentages say 50-100 but when you plot the model 
it has the roll off that you would expect. See for example,

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

James Ketrenos wrote:

> Bill Moss wrote:
>
> > The link quality as displayed by iwconfig has gone through about 
> > three patches so far if I am counting right. This number is computed 
> > as the minimum of five values, one of which is signal quality. Under 
> > good conditions, the link quality matches the signal quality. Vendors 
> > generally publish a percent to dBm look up table along with RSSI_MAX 
> > to provide a basis for the link quality calculation but I have not 
> > found this data anywhere. The linear conversion between dBm and 
> > percentage does not model this nonlinear process well. I built a 
> > quadratic model which can be found in the patch below. This model 
> > does a much better job. To build the model, I booted by laptop in 
> > Windows XP and obtained four stable percentage values at four 
> > locations in my house. I then booted to FC3 and obtained four stable 
> > dBm values for those locations. I fit this data with a quadratic 
> > which is in the patch. Here is the data.
> >
> > percentage                      dBm
> > 100                                  -20
> > 80                                   -50
> > 57                                   -67
> > 45                                   -74
>
>
> Cool!  I ran through the math (solving qual = (100 * a^2 - b * (15 * a 
> + 62 * b)) / a^2) for the min/max thresholds just to make sure I 
> understood what you were talking about :)
>
> I'll incorporate your patch for the next drop.
>
> James
>
> > The result is my iwconfig link quality values now track my Windows XP 
> > link quality values. A useful reference is 
> > http://www.wildpackets.com/elements/whitepapers/Converting_Signal_Strength.pdf 
> >
> > patch
> > ___________________
> >
> > --- ipw2200.c_orig      2005-02-02 19:31:47.742210264 -0500
> > +++ ipw2200.c   2005-02-02 19:47:19.107621200 -0500
> > @@ -3101,7 +3101,7 @@ static u32 ipw_get_current_rate(struct i
> >       return 0;
> > }
> >
> > -#define PERFECT_RSSI (-50)
> > +#define PERFECT_RSSI (-20)
> > #define WORST_RSSI   (-85)
> > #define IPW_STATS_INTERVAL (2 * HZ)
> > static void ipw_gather_stats(struct ipw_priv *priv)
> > @@ -3114,6 +3114,7 @@ static void ipw_gather_stats(struct ipw_
> >       s16 rssi;
> >       u32 beacon_quality, signal_quality, tx_quality, rx_quality,
> >               rate_quality;
> > +       u32 max_rate;
> >
> >       if (!(priv->status & STATUS_ASSOCIATED)) {
> >               priv->quality = 0;
> > @@ -3171,7 +3172,8 @@ static void ipw_gather_stats(struct ipw_
> >                       beacon_quality, missed_beacons_percent);
> >
> >       priv->last_rate = ipw_get_current_rate(priv);
> > -       rate_quality =  priv->last_rate * 40 / priv->last_rate + 60;
> > +       max_rate = ipw_get_max_rate(priv);
> > +       rate_quality =  priv->last_rate * 40 / max_rate + 60;
> >       IPW_DEBUG_STATS("Rate quality : %3d%% (%dMbs)\n",
> >                       rate_quality, priv->last_rate / 1000000);
> >
> > @@ -3199,8 +3201,11 @@ static void ipw_gather_stats(struct ipw_
> >       else if (rssi < WORST_RSSI)
> >               signal_quality = 0;
> >       else
> > -               signal_quality = (rssi - WORST_RSSI) * 100 /
> > -                       (PERFECT_RSSI - WORST_RSSI);
> > +               signal_quality = ( 100 *
> > +               (PERFECT_RSSI - WORST_RSSI) * (PERFECT_RSSI - 
> > WORST_RSSI)
> > +               - (PERFECT_RSSI - rssi) * (15 * (PERFECT_RSSI - 
> > WORST_RSSI)
> > +               + 62 * (PERFECT_RSSI - rssi)) ) /
> > +               ( (PERFECT_RSSI - WORST_RSSI) * (PERFECT_RSSI - 
> > WORST_RSSI) );
> >       IPW_DEBUG_STATS("Signal level : %3d%% (%d dBm)\n",
> >                       signal_quality, rssi);

--
Bill Moss
Professor, Mathematical Sciences
Clemson University