Re: Re: Differences in iwspy !!!

If I recall correctly, this issue has been around for a while and is 
fairly well known - it even shows up in some widely published Linux/WiFi 
buglists. The basic problem is that _every_ WLAN card manufacturer uses 
a different metric for reporting signal & noise power. Consequently, the 
power numbers you get from the cards are not apples/apples and need to 
be recalibrated depending on who made them. Adding that capability to 
the WLAN drivers is not considered critical since it doesn't have any 
noticable effect on data traffic.

I used to work at a (then) major WLAN silicon mfg and can confirm that 
the firmware/driver guys didn't put a lot of emphasis on the accuracy of 
these measurments, nor did they devote much time to ensuring that the 
measurements were consistent across product lines or across the industry 
in general.

If you've got an application that needs consistent power measurement, 
you've got a couple options:

0) Design your app for one particular type of card.
1) Query the drivers to figure out what mfg made each card you're 
talking to and re-scale all the measurements in your userland app.
2) Read/understand/patch/submit fixes for all the drivers you anticipate 
using.
3) Organize a grass-roots campaign to have all the various driver 
maintainers fix this issue.

I imagine you can guess which one of these options is the least painful.

Eric

Vikram KAUL wrote:

>Folks,
>
> I haven't heard from anyone on this. Resending it to stress the 
>importance, not just for me, but for everyone out there. 
>
> Look at my original email and see why the stats are not correct. Is there 
>a way to validate this ? Can anyone on the list do a simple experiment and 
>cross check/validate/invalidate what I am seeing. 
>
> Thanks in anticipation
>
>Vikram
>
>On Wed, 23 Nov 2005, Vikram KAUL wrote:
>
>vkaul>Date: Wed, 23 Nov 2005 14:54:42 -0500 (Eastern Standard Time)
>vkaul>From: Vikram KAUL <[email protected]>
>vkaul>To: [email protected]
>vkaul>Cc: [email protected]
>vkaul>Subject: Differences in iwspy !!!
>vkaul>
>vkaul>
>vkaul>Hi Folks,
>vkaul>
>vkaul> I have the same laptop running two different cards with two different
>vkaul>drivers:
>vkaul>
>vkaul>eth0:  Lucent Orinoco Gold               -> "orinoco-0.15rc3" driver
>vkaul>wlan0: Linksys WCF12 with pcmcia adapter -> linux-wlan-ng-0.2.2"
>vkaul>
>vkaul>Both belong to the same ad-hoc group and are "iwspy"ing two other nodes
>vkaul>which belong to the same ad-hoc group as well. A broadcast ping with 200ms
>vkaul>interval ensures 'freshness' of the iwspy measurements. Here is what I
>vkaul>see:
>vkaul>
>vkaul>wlan0     Statistics collected:
>vkaul>    00:02:2D:64:5A:55 : Quality=32/92  Signal level=-100 dBm  Noise
>vkaul>level=-132 dBm
>vkaul>    00:02:2D:64:5A:4B : Quality=36/92  Signal level=-96 dBm  Noise
>vkaul>level=-132 dBm
>vkaul>    Link/Cell/AP      : Quality=0/92  Signal level=-100 dBm  Noise
>vkaul>level=-100 dBm
>vkaul>    Typical/Reference : Quality:0  Signal level:0  Noise level:0
>vkaul>
>vkaul>eth1      Statistics collected:
>vkaul>    00:02:2D:64:5A:55 : Quality=49/92  Signal level=-42 dBm  Noise
>vkaul>level=-91 dBm
>vkaul>    00:02:2D:64:5A:4B : Quality=56/92  Signal level=-39 dBm  Noise
>vkaul>level=-95 dBm
>vkaul>    Link/Cell/AP      : Quality:49/92  Signal level:-42 dBm  Noise
>vkaul>level:-91 dBm
>vkaul>    Typical/Reference : Quality:36/92  Signal level:-62 dBm  Noise
>vkaul>level:-98 dBm
>vkaul>
>vkaul>Notice that the signal level and noise level are off by a very large
>vkaul>margin in the two measurements.
>vkaul>
>vkaul>Why is that ? The measurement cards are in two adjacent slots. I am seeing
>vkaul>this very consistently.
>vkaul>
>vkaul>I can provide more derails if anyone wanted so
>vkaul>
>vkaul>Thanks in advance
>vkaul>
>vkaul>Vikram
>vkaul>
>vkaul>
>_______________________________________________
>Linux-wlan-user mailing list
>[email protected]
>http://lists.linux-wlan.com/mailman/listinfo/linux-wlan-user
>
>  
>