HIS PAUL READ THIS TOO, MAYBE A BETTER SOLUTION FOR BUS MASTER SPECIFIC :D i found this too with ds2480b! maybe a single workaround (ugly, but works) is a counter at device for example after 10 reads, close and open device (/dev/ttyUSB0 or other device) this works like cron restart, but it's faster and don't need user to configure a cron server (in ARM this must be compiled on some distros or in android devices) maybe a command line could help a lot (in another post i talked to paul to add a force bus master type at command line, maybe this could work too, but this one is another option) for example: --device-reconnect-counter=0 (don't use reconnect) --device-reconnect-counter=100 (after 100 complete(send/receive, or send/timeout) commands reconnect) this could be good for rfc2217 devices with a noise internet/ethernet network too =D i don't know if we should execute a bus master auto identification after reconnection (check if we are using a ds2480 or a bit bang serial por for example...) maybe the other command (force a bus master) could help here
2012/1/27 ekgnkb3d <achi...@gmx.de>: > > Hi all, > probably I've found a funny bug in the owserver in relation to USB > Bus-Master and longer cables. This was mentioned before in > http://old.nabble.com/Re%3A-How-to-enable-statistics--p33202282.html this > posting very fuzzy, but now I was able to reproduce it in my lab: > > 1. Install OWFS 2.8p13 or p14 on an ARM9 400MHz > 2. Connect one DS9490R USB Bus-Master > 3. To provoke the error to appear fast, use an extra long cable of about > 200m > (28nF, e.g. SFTP, see simulation tool below) > 4. Connect one DS18B20 to the cable, with Vcc powered to +5V > 5. Start owserver with "-uall" > 6. Start a client polling for the temp 2 times a second, eg owread or an > owcapi application > > This works fine for about 2 to 20 minutes. If cable is shorter, 2.8p14 is > used or other quirks are done (see remarks below) it can last for days and > weeks. > But suddenly it don't return the temp and don't show an error, even the > device is not shown in owdir anymore > > One workaround: > Restart the owserver, than it works again for 2 - 20 minutes until it > suddenly don't return any temp. > > Another workaround by keeping the owserver-process alive: > Truncate the cable to about 170m (25nF), than it works again for about 2 to > 20 minutes. > But suddenly it return an errno=22, the device disappears again in owdir. > Truncate the cable again to 50m (8nF), this now works forever, I've checked > at least for several month :-) > > Some remarks: > - To simulate the long cables and the different length I made a > capacity-cascade of 8 parallel > capacities (0.1nF, 0.22nF, 0.47nF, 0.82nF, 1.5nF, 3.3nF, 6.8nF, 15nF and > an 8-pol DIP switch > between pins OWDATA and OWRETURN) this allows simulating cable length of > 1m to 200m. > - Beside the simulation, exactly the same behavior appears every week, at a > real installation > with a real cable of about 100m. > - It have nothing to do with star- or other bad network topology, even > a single cable with a single device at the end is enough to reproduce it > - Disconnecting the cable and wait for minutes, or shorting the device don't > solve the problem, > only restarting the owserver or truncate the cable. > - This is not dependent from dedicated pieces of hardware, it is > reproducible with several DS9490R, > several DS18B20, several ARMs and several types of cable. > - In case several devices are connected to the bus, they all get the problem > exactly at the same time, > or came back when reducing the capacity/restart owserver together > - I've tested also networks via I2C DS2482-100, this works fine since month, > no problem here. > I've not tested serial bus masters > - It looks like the communication capabilities of the USB bus-master is > reduced in stages: > 0. After owserver restart it is able to drive 28nF at extreme. > The statistics counter grow slowly, except the 3 search_errors which > stay at 0 > 1. Error stage: Now it can only drive up to 25nF, but the change is not > smooth, it is an abrupt state > change. This is stable for a while, no drifting, I've tested this with > my cascade, reproducible for > meanwhile dozens of tests: NOK >= 25.2nF, OK <= 25nF. > Now the statistics counter stays at the last value, they don't grow > further, even if the > capacity is above 25nF. > 2. Error stage: Now it can only drive 8nF, again no smooth transition, it > is an abrupt state change. > This seems to be stable forever, NOK >= 8.2nF, OK <= 8nF. > Now the statistics counter stays, except the 3 search_errors they grow > fast if the capacity > is above 8nF. > - The steps to reproduce, mentioned above, are aimed to see the error as > fast as possible. There a > several methods to delay the problem appearance, which is helpful when > needing the OW-network, > but none of them really solve it: > - Use shorter cables from beginning onwards, start with 100m cable will > cause the problem to > appear after 3 to 14 days. Maybe below an dedicated length the problem > completely disappear, > but I need this length > - Using 100 Ohm serial resistors after the bus-master, before the cable > will shift the problem > further by about 50% > - Use parasitic power of the DS18B20 with Vcc to GND and remove the +5V > from the cable. > This will reduce the capacity and therefore allows longer cable or > delays the error appearance. > - Daily restart the owserver via cron-job :-( > -- > View this message in context: > http://old.nabble.com/Problems-in-owserver-with-USB-and-longer-cables-tp33213551p33213551.html > Sent from the OWFS - Dev mailing list archive at Nabble.com. > > > ------------------------------------------------------------------------------ > Try before you buy = See our experts in action! > The most comprehensive online learning library for Microsoft developers > is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3, > Metro Style Apps, more. Free future releases when you subscribe now! > http://p.sf.net/sfu/learndevnow-dev2 > _______________________________________________ > Owfs-developers mailing list > Owfs-developers@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/owfs-developers > -- Roberto Spadim Spadim Technology / SPAEmpresarial ------------------------------------------------------------------------------ Try before you buy = See our experts in action! The most comprehensive online learning library for Microsoft developers is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3, Metro Style Apps, more. Free future releases when you subscribe now! http://p.sf.net/sfu/learndevnow-dev2 _______________________________________________ Owfs-developers mailing list Owfs-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/owfs-developers
