> If i were to pick a root cause: inadequate testing. If you are saying that we didn't throw enough things out of airplanes then I couldn't agree more. :)
What do you suggest would count as proper testing? Someone (alright, me) can work on building tests before we even make the thing so that we can do a really good round of testing as soon as we have a prototype. On Wed, Jul 8, 2009 at 3:45 PM, <[email protected]> wrote: > (2009.07.08) [email protected]: >> Quoting [email protected]: >>> If i were doing it i'd use 4 mechanically independent servos. I estimate >>> the reliability of doing this is about 1/2 of the single servo system >>> being proposed. (Can we put the cool linkage drawing on the wiki page?) >>> >>> Despite the reduced reliability, i still think the 4 independent scheme >>> is a winner. >>> >>> * Reliability is still very high >>> >>> * Mechanically ready for full flight control >>> >>> * Mechanically simpler >>> >>> * Same system that must be developed down the road anyway >> >> Add higher cost to the list. > > Certainly it might cost more. Probably no more than 400 $ extra. > > Keep in mind that the per-servo torque requirements might be > considerably reduced in the 4 independent scheme. > > >> The reason I was pushing for a linked system is because there were >> reservations on the team about what could go wrong if the micro messes >> up and the servos go out of sync. Keep in mind we have NEVER >> successfully flown an ARM micro in 3 airborne tests. > > Yeah, what's up with all the ARM failures? (See Doug's question below.) > > The linked system is safer with respect to controller failure. A > controller failure in the independent system has the potential to make > the flight path into a large radius circle, possibly at high spin rate, > not good. > > >> I will be happy to machine both the linked single servo system and the >> four servo system, so multiplying the work required is not a big >> problem. I really want to take baby steps on this, and I think the >> additional insurance is worth the work. We'll do the independent >> version, but I don't think we should do that one first. > > I appreciate what you're saying. > > Since you're doing the work, you should do it the way that seems best to > you. > > The linkage will be hard to get working well, but if it does work well > it will be a thing of beauty. > > >>> In the 4 independent scheme, the servos must be individually trimmed. To >>> do this i would consider an absolute magnetic shaft encoder: >> >> Why add a shaft encoder when the existing servo positioning system gives >> minute of angle precision? I think the additional sensor violates the >> KISS principle. Modern digital servos are strong, fast, and accurate. I >> plan to align the fins the same way RC helicopter blades are aligned; >> using an inclinometer style pitch gauge. > > My experience with off the shelf servos is they won't re-point with single > degree accuracy but i haven't tried the more expensive digital servos. > > -- > I just checked what i think is the Futaba site > > http://www.futaba-rc.com > > I can't find a specification for pointing accuracy, which does not > inspire confidence. > > Ditto for > > http://www.rc.futaba.co.jp > > Every servo i've ever taken apart had a sub-50 cent potentiometer in it. > Will that A) work. B) work under vibration, C) work reliably, D) work to > sub-degree precision? Honestly i think the answers are: Sometimes. No, > no, & no. > > Clearly doing anything but buying and using an off the shelf servo is > way more work than desirable. I'd like someone to prove me wrong and > show that off the shelf servos are just fine. Please. > > I did notice these guys, who are interesting > > http://www.openservo.com/ > > > (2009.07.08) [email protected]: >> What has been the root-cause of each of the three failed ARM flights? A >> robust controller solution would seem to be a primary factor for the >> roll-control project, with either a single- or a multi- servo approach. > > If i recall, first was either a blown GPIO pin or an odd reset due to a > wiring problem. Second was a blown GPIO. Third was a firmware bug. > > If i were to pick a root cause: inadequate testing. > > > _______________________________________________ > psas-airframe mailing list > [email protected] > http://lists.psas.pdx.edu/mailman/listinfo/psas-airframe > _______________________________________________ psas-airframe mailing list [email protected] http://lists.psas.pdx.edu/mailman/listinfo/psas-airframe
