Re: signal strength problem for multiple ap's w/ the same essid

[Sven <sr12 duke edu>]

sorry for being imprecise: in my setup with multiple AP's, most of the
time NM reports a "correct" value of the signal strength. by correct i
mean the signal strength of the AP i'm connected to, and not the signal
strength of one of the AP's. but maybe i'm wrong, which i am if NM uses
the iwconfig or /proc/net/wireless signal quality values... i was
concerned that with multiple AP's with the same essid, NM would not
necessarily take the signal strength of the AP the card is connected to,
but accidentally the signal strength of another AP.

i use a little gdesklet applet to show the signal quality, and (not
knowing better) it is calculated from /proc/net/wireless as math.log
(quality+1)/math.log(92)*100 , which i am pretty happy with ( +1 to have
0 when quality=0 in /proc/net/wireless). when right next to my ap at
home, iwconfig reports 68/92, which gives me 92% in my gdesklet.
restarting NM when close to the AP gives 92% as well... so it seems that
NM uses a log formula as well, no? and 75/94 would translate to 95%... 

should this be a concern? when changing locations, NM _does not_ change
the link quality displayed in the applet, while (getting closer and
further away) obviously changes the signal. i assume that will change at
a later point?

Sven  

On Tue, 2005-02-01 at 10:05 -0500, Dan Williams wrote:
> On Mon, 2005-01-31 at 17:04 -0500, Sven wrote:
> > checked out the CVS. sorry to report that: a few restarts of NM produced
> > 2 bars, 43%, and 39%, while everything else says ~80%... the applet
> > seems not to change the %, ever... unless i restart NM, so it seems. i
> > guess it's fair to say that it shows the signal quality correctly "most
> > of the time" ;-) also, locking the card to a certain AP, staying
> > connected etc, does not change the % either...
> 
> I just tried with my 5212 card, and I get only about 75% quality when
> right next to the AP and at 36MB/s.
> 
> This is _completely_ a driver problem, please file a bug with Madwifi, I
> can't do anything about it in NetworkManager.  For example, when right
> next to the access point, 'iwconfig' reports a quality value of "75/94".
> This is current qual / max qual, as directly reported by the driver.
> NetworkManager takes these two values and divides them to obtain a total
> "quality" percentage.  If the driver reports either quality or
> max_quality as 0, then NM tries to use the Signal and Noise levels to
> obtain quality through Signal-to-Noise ratio.
> 
> The Madwifi driver sources hardcode the max quality:
> (net80211/ieee80211_wireless.c):
> 
> range->max_qual.qual = 0xff - 161
> 
> This value is wrong for both our chipsets, and needs to be fixed in the
> driver.  If driver's cannot provide clear & correct values for their
> supported hardware or don't follow the WEXT API, there's not much I can
> do about it in NetworkManager.
> 
> Dan