Well, that has much to do with physics than with software. If you are far enough to get -70, then anybody passing right in front of the AP will reduce the signal much more than if you sat near, and benefited from reflected signal. You patch in combination with what Intel have submitted still improves things a lot, and they should be merged without hesitation. On Thu, Dec 19, 2019 at 11:58 PM Ben Greear <greearb@xxxxxxxxxxxxxxx> wrote: > > On 12/19/2019 06:20 AM, Pavel Nikulin wrote: > > I already tried it with good results. At least, now you don't have > > roaming when you are sitting 5 metres away from the AP. Clearly it > > works. > > > > Again, roaming with moderately low signal levels (-67-70) is still very jumpy > > Can you enable some debugging around where I made changes to see if you can figure out > why it is roaming at the modest signal difference levels? Maybe my patch can be > improved in that case. > > Thanks, > Ben > > > > > On Wed, Dec 18, 2019 at 8:42 PM Ben Greear <greearb@xxxxxxxxxxxxxxx> wrote: > >> > >> On 12/18/19 5:25 AM, Pavel Nikulin wrote: > >>> Hello, > >>> > >>> I want to bring to people's attention that there is still an > >>> outstanding issue with roaming logic defaults. > >>> > >>> At around -60 to -70 dbi roaming decisions are still taken due to > >>> minute disturbances, and at above -50 dbi they are being taken when > >>> there is no real need to roam. > >>> > >>> Current defaults feel to me to be made for non-realistic scenarios. > >>> You may have a lot of APs in range with -60 to -70 signal levels, and > >>> none of them be particularly better than each other. You find this a > >>> lot in office parks. > >>> > >>> All of them can have some packet loss, but you will not improve user > >>> experience by having the roaming happen every few minutes in such > >>> cases. > >>> > >>> Second to that, a minute disturbance like a person passing in front of > >>> an AP, or something electrically conductive being put near the antenna > >>> may also easily add or subtract few decibels from the signal level > >>> quickly. I think there should be at least some moving average used for > >>> the roaming decision, or better a more advanced filtering method. > >>> > >>> Past few patches from Mathew did improve the situation somehow, but I > >>> myself is still having issues with overly aggressive roaming every > >>> day. > >>> > >>> Have anybody took a look at this patch > >>> https://www.spinics.net/lists/hostap/msg05792.html ? > >> > >> Maybe if you test this patch and see good results, it will help it get > >> upstream? > >> > >> Thanks, > >> Ben > >> > >> > >> -- > >> Ben Greear <greearb@xxxxxxxxxxxxxxx> > >> Candela Technologies Inc http://www.candelatech.com > >> > > > > > -- > Ben Greear <greearb@xxxxxxxxxxxxxxx> > Candela Technologies Inc http://www.candelatech.com > _______________________________________________ Hostap mailing list Hostap@xxxxxxxxxxxxxxxxxxx http://lists.infradead.org/mailman/listinfo/hostap
