Hi, On 2020-12-21 09:02, Poul-Henning Kamp wrote: > -------- > Bob kb8tq writes: > >> I have seen cases of “goes away until power cycled”. I have not seen any >> cases of >> “goes away forever” other than the obvious ( = feed it an insane almanac >> that prevents >> if from ever locking up ). Even with that said, I have not seen an example >> ot that sort of >> thing living through a hard reset … ( which isn’t quite the same thing as a >> power cycle ). > I have: An corrupt alamanac in NV storage contained something which > made a particular GPS receiver divide by zero, shit its pants and > wedge during startup. > > The post mortem report said that the alamanac passed the "technical > consistency checks", by which I suppose they mean the Hamming code, > but it still caused a divide by zero. > > This incident is the reasons why GPS receivers in some critical > applications are not allowed to have NV storage for "operational > purposes" and get the almanac downloaded from the attached systems > at startup. > With this new language, they would be required to be Level 1 compliant, at which it would always be a way for the user to reset corrupt data in the field. We never even discussed the possibility of prohibiting the NV, rather we focused on the ability to recover in the field. NV has it's upsides to boot quickly etc. but as any cache, it needs to be clearable.
In a similar sense, we also had a good discussion about being able to upgrade the receiver. Several of us was driving hard to ensure that receivers can be upgradeable in the field, so that bugs can be removed continuously throughout the operational lifetime. In fact, I pointed out that the operational lifetime should be limited by how long you can maintain the receiver updated. The whole life cycle aspect is important in that as you loose support on receivers, they should go out of service for critical things. You should also make sure there is contracts on how long they will be maintained. Cheers, Magnus _______________________________________________ time-nuts mailing list -- [email protected] To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com and follow the instructions there.
