My assumption is that either 1) the sensor self-tests and EOLs when the test fails (I think this is unlikely), or 2) the sensor has a timer running internally, since the time of manufacturing, that runs to the actual lifespan (not the warranted life, the longer design life) before alerting with EOL.
Best-F > On Apr 17, 2017, at 12:33 PM, Tom Metro <tmetro+hhack...@gmail.com> wrote: > > Federico Lucifredi wrote: >> This excess life span I detected in use leads me to think the sensor >> chemistry decays even when warehoused and not in powered use. > > Sure, that makes perfect sense given what we know about the sensors. The > author of the teardown you referenced speculated that the limited > lifespan was due to evaporation of the liquid and accumulation of > contaminants in the activated carbon filter. > > >> I think the lifespan runs from the date of manufacturing, and exceed >> the warranty by an unknown amount... > > Yes, I assumed as much. As with any engineering parameter, the system > should be designed with sufficient margin in excess of what is needed. > And with a life safety device, even greater margins. > > But the issue at hand - with respect to having all your alarms die at > once - is not the actual life of the sensor, but the duration of the EOL > timer. > > >> ...it is at least an extra year... > > We're talking an analog device whose lifespan is going to be partly > dependent on environmental conditions (higher temperature causing more > rapid evaporation; higher concentration of contaminants clogging the > carbon faster), so it is going to be a statistical curve. The > manufacturer might be aiming for 99.9% sensor survival at rated lifespan > + N years. N probably varies by manufacturer and the risk level they're > willing to take. > > Quite possible a CO sensor might work adequately for another 5 to 10 > years after the programmed EOL, if the manufacturers engineering margins > were high and environmental conditions favorable. > > >> This leads to unit warehousing, and the sensor must still deliver its >> full warranty life to the customer once deployed. > > Agreed. You'd think then that they'd have "best sold/used by" dates on > the product packaging, otherwise a unit kept in storage for 5 years runs > a high risk of sensor failure, and if the micro can't detect that, the > detector could fail to alarm in a life threatening condition and be a > liability for the manufacturer. > > Maybe alarm manufacturers have special contracts with their distributors > and retailers that prevent the sale of old units? > > >> For example, the first set I had in my home were warranted 7 years, >> but lasted 8 before the annoy-a-tron triggered and forced me to >> replace all units. > > This part I don't really understand. The EOL alarm should be purely time > based, so ether the EOL alarm is not set to 7 years, or the unit wasn't > powered that entire time, or the unit suffers from significant clock drift. > > Given these devices use cheap micro controllers without a real time > clock, and possibly don't even use a crystal oscillator for the MCU > clock, clock drift is a real possibility. An RC oscillator could easily > be off by 10%, which is getting in the neighborhood of adding a year to > a 7 year timer. > > That could explain why my Kidde alarms haven't EOLed. > > >>> I suppose I could always delay activating subsequent units by a month. >> >> I do not believe this will work. > > The First Alert unit with the built-in battery comes powered off. It has > a switch on the back to "activate it." The switch has a mechanical > interlock, such that you have to snap off a bit of plastic to deactivate > it, and once done, it can't be activated again. > > Presumably they do this to start the EOL clock running and to minimize > cutting in to the 10-year battery life. Given the mechanical interlock, > they've clearly gone to some expense to accommodate this, so it must be > advantageous. > > I'm guessing the mechanical interlock exists so 1. a user can't > intentionally or inadvertently silence an alarm by turning the unit off, > and 2. you don't end up running across a product that might have been > powered for a few years, shelved for several, and then deployed again, > such that the EOL timer trips years later than it should. But you could > argue #2 is no worse than having a product sitting in a warehouse for years. > > Might it have been cheaper for the manufacturer to skip the switch and > interlock, sell the unit already activated, add a "use by" date, and > just derate the lifespan by a few years? > > I'm kind of surprised they didn't use the technique of wedging a bit of > plastic film between a set of spring contacts, which the user pulls out > on activation, as commonly used by products sold with batteries. Much > cheaper. Can't be "unactivated." That they didn't suggests they had a > strong requirement to be able to deactivate units. > > Not clear to me why you would ever want to deactivate one of these. > Perhaps to silence a defective unit, given you can't remove the battery > without disassembly. > > -Tom > _______________________________________________ Hardwarehacking mailing list Hardwarehacking@blu.org http://lists.blu.org/mailman/listinfo/hardwarehacking