On Sunday 04 Sep 2005 15:11, Andrew de Quincey wrote: > On Sunday 04 Sep 2005 14:47, Jon Burgess wrote: > > Andrew de Quincey wrote: > > > Removing the code obviously means that it will always cycle through the > > > inversions when lock is accidentally lost. I don't understand why that > > > is necessary though - perhaps we have a bug in the frontend driver? > > > > I reckon that the FE is indicating lock when it shouldn't be, either it > > happens at random or perhaps something like this happens because we > > don't give the FE enough time to update the lock status. Perhaps the > > "min_delay" parameter needs increasing for this FE? > > > > It looks like most callers set min_delay_ms = 100: > > > > [jburgess@t41 dvb]$ grep -r "\.min_delay_ms" . > > ./ttusb-budget/dvb-ttusb-budget.c: .min_delay_ms = 100, > > ./b2c2/flexcop-fe-tuner.c: .min_delay_ms = 100, > > ./ttpci/budget-av.c: .min_delay_ms = 100, > > ./ttpci/budget-av.c: .min_delay_ms = 100, > > ./ttpci/budget-patch.c: .min_delay_ms = 100, > > ./ttpci/budget-ci.c: .min_delay_ms = 100, > > ./ttpci/budget-ci.c: .min_delay_ms = 50, > > ./ttpci/av7110.c: .min_delay_ms = 100, > > ./ttpci/av7110.c: .min_delay_ms = 100, > > ./ttpci/budget.c: .min_delay_ms = 100, > > > > Interestingly the odd-one-out at 50ms is the philips_su1278_tt_config > > which was the one implicated by Marian in the start of this thread: > > > > kernel: DVB: registering new adapter (TT-Budget/WinTV-NOVA-CI PCI). > > kernel: stv0299: try to attach to TT-Budget/WinTV-NOVA-CI PCI > > kernel: stv0299: setup for tuner SU1278 (TSA5059 synth) on > > TechnoTrend hardware > > > > Perhaps we should try increasing this to 100ms for this one too. Patch > > attached. > > Yup. I'm just getting my test environment setup on the site with the > issue... will let you know if it helps. Yeah, removing the auto-inversion code did seem to help - it now regains the lock even after losing it on that site. Note though, that on other sites, it has no problem with regaining the lock with the code in place (when I wrote that bit of dvb_frontend, it was throughly tested). We're using the same cards - budget-ci (13c2/100f) with stv0299, and the same drivers on all sites. But: this site with the problem had the cards bought much more recently - maybe there has been some small manufacturing change? (I know TT have fixed altered the design slightly several times before on the 13c2/100f cards). I don't think the change to 100ms made much of a difference for me though...
