Ideally, TPC should be a fully automatic mechanism that reduces transmit power between the two points to as low of a level as possible while delivering the same quality of service. The purpose is to reduce the excess headroom in each link. I.e., if you could still link with 65Mb/s towards a given direction using 14dBmW, you should not transmit with 20dBmW. Some only set the AP TX power globally (i.e., same towards all of its connected clients at the moment) to tackle the hidden/exposed node problem, but again must do this adaptively and change this constantly without human intervention. There exist multiple advanced algorithms for this, some proprietary tuned for corporate deployment. Actually, if we accepted automatically retuning tx power with iw based on actual link stats of momentarily connected clients every 60s with cron, this could be added to OpenWrt pretty easily. > Class A devices control their transmit power within ±3 dB and class B devices control their power within ±9 dB. - https://www.litepoint.com/blog/wi-fi-6-ofdma/ - https://www.cisco.com/c/en/us/td/docs/wireless/controller/technotes/8-3/b_RRM_White_Paper/tpc.html On Thu, Jan 4, 2024 at 10:42 AM Johannes Berg <johannes@xxxxxxxxxxxxxxxx> wrote: > > On Thu, 2024-01-04 at 10:07 +0100, b.K.il.h.u+tigbuh@xxxxxxxxx wrote: > > You can find a wording in most such regulations that if TPC is not > > supported, the maximal TX power must be reduced by 3 dBmW. Hence in > > all such cases, the entries in db.txt contain 3 less than the maximum. > > > > If, on the other hand, you know of a country that specifies that > > devices lacking TPC may not use the band at all, all such band entries > > must be omitted (commented out along with a URL). > > Yeah, that's how we (currently) handle things. > > I'm not even sure what the requirements would be for "TPC" to be > implemented, tbh. > > > > As far as I know the kernel doesn't have facilities to support TPC or > > > transmit power control, > > Right. I have, however, heard the interpretation that the fact that we > have - even if nobody uses it - the "iw set txpower" command means that > we *do* have TPC ... Not really sure what to make of that though. > > > > nor does the wireless-regdb database. > > Correct. With the new regdb format we could add something that would > enable these ranges in the kernel only with some additional > requirements, but > (a) we don't implement that now, and > (b) I don't know what the requirements would actually be, e.g. would > it be enough that the driver promises it implements "TPC" in some > way? Or even the manual setting? > > > > And so > > > in the database we would either omit rules that require TPC, or include > > > alternative rules (as specified by local regulations) not requiring TPC. > > > Am I right? > > Right. > > johannes
