The time is done by counting AC cycles; in this case likely through D7 and the R24/R25 divider. 10 years comes to 18,921,600,000 counts on that input pin so you only need a 35 bit counter. My guess is a RAM based counter counts off big chunk of time and then updates an EEPROM value periodically to make the timer non-volatile.
@@crispy_otter If the prescaler divides by 5,184,000 which is the number of AC cycles per day, the count in EEPROM would only need to be updated 3,650 times and the EEPROM in those PICs is good for a million writes. If they prescaled by hours rather than days, then it's 216,000 AC cycles per hour and then 87,600 EEPROM writes. Piece of cake.
I think all of you guys are wrong 😮 these devices usually have some gel meterials or some radioactive elements like americium which decay over time. If these materials becoms too weak like a battery (but it has an ohmic resistance) then the cpu creates a warning. It could be 10 years or 11 maybe 15. Very similar to the low battery signal.
I had one of these in a house I rented. It went off overnight and scared the hell out of me, because there really WAS a CO risk there. .I took it out to the unheated garage and it didn't stop beeping so I knew it wasn't CO. Then I read the fine print. I wasn't yet schooled by IMSAI or EEVblog then, so I never looked inside....
I got a carbon monoxide detector for the house here and for the first few years I had placed about a metre from the ground... (I know I'm an absolute idiot) When I realised it should be up high I placed it to the proper height. I kept it beside the fire place in our sitting room as I didn't trust the chimney. Anyways 2 days after placing it up to the right height it started going crazy. Like they tell you if you're in a house fire get as low to the ground as you can and get out. HOW MANY TIMES WAS OUR SITTING ROOM FULL OF POISONOUS GAS AND WE NEVER KNEW BECAUSE I HAD THE DETECTOR PLACED JUST A LITTLE HIGHER THAN A METER ON THE WALL... WE SORTED THE CHIMNEY AS BEST WE COULD & THANK GOD THE ALARM NEVER WENT OFF AGAIN AFTER THAT EVENING.. 😳 These should be in every home, they're so important. 👍🇮🇪🙏
There's a very early PIC technote from the days of the PIC165x which says the pin protection diodes are plenty strong enough to handle connection to 120V mains through a greater than 2 meg resistor. I think it might have suggested two 2.2 meg 1/4 watt resistors in series to get the voltage rating.
You don't need one. A high value resistor tied to an input will do. The protection diode will clamp the voltage and a schmitt trigger input will avoid noisy transitions.
the CO sensor is based on a catalyst that deteriorates at a specific rate. in your device it around 10 years. if you have other installed at the same time failure is soon
Almost all PIC controllers have an internal oscillator and flash memory, so no need for a crystal, and the flash can remember counts of whatever to get to 10 years.
The PIC16F883 has some nonvolatile memory in it. They probably increment a counter in that memory area every hour or day or so. Nothing fancy, and it doesn't need to be precise.
My guess is that it divides the power grid frequency by 60 to get a second, and then counts a day count to 86 400 seconds before adding one to the year count, resetting the day count to 0 at the end. This keeps going to the point where the year count reaches 3653, then breaks itself and forces you to buy a new one. Chances are this is going to be plugged in all the time, the battery is just a backup if the power goes out. So the counter should be accurate. Additionally it could use it's internal oscillator as a backup clock to keep counting while battery powered.
everything has a use by date nowadays. i expect that as there is no testing or calibration required every year, they have taken sensor life and the drift of the component values and decided that it won't detect Co correctly. Folk forget that if you have a resistive voltage divider supplied from a very good and stable supply to define a trigger point, after a few years it will be drifting away from it set point. So with very high stab components any analogue references will have drifted in that time. Having had to work on designs with with a 25 year service life component selection is a real pain in the butt, even more so when they stopped making mill-spec components.
Smoke detectors which are supposed to last for 10 years only just put a lithium ion battery in there rated for 10 years. Don't think I've ever seen a mains backup carbon monoxide device like this before. Wild of them to do a 10 year timer in only software.
Does not have to be too precise at all, so using internal oscillator will still work well enough. Just has to time out sometime between 10 and 11 years, so internal oscillator when mains is off, and low battery detection, or power fail, makes it do a write to EEPROM of the count. Yes the sensor is an actual battery, with a small air entry that allows sensing, and the voltage is proportional to the oxygen content, and any CO will cause the voltage to change as well, though you also need to correct for ambient temperature as that varies voltage as well.
Both of our CO alarms are mains power with battery backup, and one is a plug-in very much like this. The other is a round ceiling-mount smoke/CO combo alarm.
Ten years is easy, even with the PIC's internal oscillator. It certainly doesn't have to be accurate, and the datasheet says it's accurate to 1%...plenty good enough for this application.
It sure has a bunch of zero-ohm resistors as jumpers to avoid a double-sided PCB and vias... Interesting view on the economic trade-offs at work for volume manufacturing.
the sensor is a chemical cell, so it has a finite lifetime. there's liquid in there and eventually it dries out. 10 years is probably a worst case scenario with a bit of safety margin.
First time I've seen a single-sided PCB with dozens of 0 Ohm, SMT resistors used as jumpers. The look like 1206, long enough to get a couple of traces under. I wonder if they are on the schematic, or put there at the PCB designers discretion. Most CAD tools would complain.... They would need to be on the BOM, Pick and Place, etc..
I know if you shake the sensor there is some kind of liquid inside and some little pillow pack on the end under the yellow heat shrink. Have no idea how it senses CO.
probably the sensor has a lifetime of 10 years and after ten years it goes out of whatever calibration standard they are held to for carbon monoxide detection
They can last quite a bit longer but the electrochemical cells are affected by various factors like temperature, exposure to certain other gases and the most important factor that limits the age to 10 years, lawyers and liabilities...
I yanked all the battery eating smoke detectors and carbon monoxide detectors out of our house and replaced them with combination smoke and carbon monoxide detectors with a ten year battery.
When you have the time of production and the time of failing you can compare it with the ten years it is specified for. Then from the precision of the time measurement you might exclude some means of time measurement like radioactive decay, battery lifetime, RC clock count or mains cycles. If it is very precise there might be a hidden Rubidium vapor frequency standard below the transformer.
"Just the right size to poke myself and cause great damage..." We all know that one well.
The time is done by counting AC cycles; in this case likely through D7 and the R24/R25 divider. 10 years comes to 18,921,600,000 counts on that input pin so you only need a 35 bit counter. My guess is a RAM based counter counts off big chunk of time and then updates an EEPROM value periodically to make the timer non-volatile.
That was my first thought - I was thinking there may be a divide by x before storage to lower the count to a managable level
@@crispy_otter If the prescaler divides by 5,184,000 which is the number of AC cycles per day, the count in EEPROM would only need to be updated 3,650 times and the EEPROM in those PICs is good for a million writes. If they prescaled by hours rather than days, then it's 216,000 AC cycles per hour and then 87,600 EEPROM writes. Piece of cake.
It could also count CPU clocks. The RC clock on a PIC is close enough that most people won't notice a few days of error.
@@misterhat5823 If that's the case then you have you come up with a justification for D7, R24, R25, and C10 which give a 60 Hz timebase to pin 27.
I think all of you guys are wrong 😮 these devices usually have some gel meterials or some radioactive elements like americium which decay over time. If these materials becoms too weak like a battery (but it has an ohmic resistance) then the cpu creates a warning. It could be 10 years or 11 maybe 15. Very similar to the low battery signal.
Put the CO sensor element in the curve tracer (then feed it some CO and see what happens)!
That's a good idea.
Then sounds like you've never used a curve tracer. All he needs is a variable supply, a voltmeter and a resistor. Of course a candle is a must!
I had one of these in a house I rented. It went off overnight and scared the hell out of me, because there really WAS a CO risk there. .I took it out to the unheated garage and it didn't stop beeping so I knew it wasn't CO. Then I read the fine print. I wasn't yet schooled by IMSAI or EEVblog then, so I never looked inside....
A real transformer! Well, salvage the LEDs anyway.
I had two identical non-replaceable battery detectors which failed within a day of each other. They used a cheapo 32KHz watch crystal.
I got a carbon monoxide detector for the house here and for the first few years I had placed about a metre from the ground... (I know I'm an absolute idiot) When I realised it should be up high I placed it to the proper height. I kept it beside the fire place in our sitting room as I didn't trust the chimney. Anyways 2 days after placing it up to the right height it started going crazy.
Like they tell you if you're in a house fire get as low to the ground as you can and get out.
HOW MANY TIMES WAS OUR SITTING ROOM FULL OF POISONOUS GAS AND WE NEVER KNEW BECAUSE I HAD THE DETECTOR PLACED JUST A LITTLE HIGHER THAN A METER ON THE WALL... WE SORTED THE CHIMNEY AS BEST WE COULD & THANK GOD THE ALARM NEVER WENT OFF AGAIN AFTER THAT EVENING.. 😳
These should be in every home, they're so important. 👍🇮🇪🙏
Some PIC microcontrollers have a Zero-crossing detector also used for counting AC cycles.
There's a very early PIC technote from the days of the PIC165x which says the pin protection diodes are plenty strong enough to handle connection to 120V mains through a greater than 2 meg resistor. I think it might have suggested two 2.2 meg 1/4 watt resistors in series to get the voltage rating.
You don't need one. A high value resistor tied to an input will do. The protection diode will clamp the voltage and a schmitt trigger input will avoid noisy transitions.
the CO sensor is based on a catalyst that deteriorates at a specific rate. in your device it around 10 years. if you have other installed at the same time failure is soon
Almost all PIC controllers have an internal oscillator and flash memory, so no need for a crystal, and the flash can remember counts of whatever to get to 10 years.
The PIC16F883 has some nonvolatile memory in it. They probably increment a counter in that memory area every hour or day or so. Nothing fancy, and it doesn't need to be precise.
My guess is that it divides the power grid frequency by 60 to get a second, and then counts a day count to 86 400 seconds before adding one to the year count, resetting the day count to 0 at the end. This keeps going to the point where the year count reaches 3653, then breaks itself and forces you to buy a new one. Chances are this is going to be plugged in all the time, the battery is just a backup if the power goes out. So the counter should be accurate. Additionally it could use it's internal oscillator as a backup clock to keep counting while battery powered.
everything has a use by date nowadays. i expect that as there is no testing or calibration required every year, they have taken sensor life and the drift of the component values and decided that it won't detect Co correctly. Folk forget that if you have a resistive voltage divider supplied from a very good and stable supply to define a trigger point, after a few years it will be drifting away from it set point. So with very high stab components any analogue references will have drifted in that time.
Having had to work on designs with with a 25 year service life component selection is a real pain in the butt, even more so when they stopped making mill-spec components.
Smoke detectors which are supposed to last for 10 years only just put a lithium ion battery in there rated for 10 years. Don't think I've ever seen a mains backup carbon monoxide device like this before. Wild of them to do a 10 year timer in only software.
Does not have to be too precise at all, so using internal oscillator will still work well enough. Just has to time out sometime between 10 and 11 years, so internal oscillator when mains is off, and low battery detection, or power fail, makes it do a write to EEPROM of the count. Yes the sensor is an actual battery, with a small air entry that allows sensing, and the voltage is proportional to the oxygen content, and any CO will cause the voltage to change as well, though you also need to correct for ambient temperature as that varies voltage as well.
Both of our CO alarms are mains power with battery backup, and one is a plug-in very much like this. The other is a round ceiling-mount smoke/CO combo alarm.
Ten years is easy, even with the PIC's internal oscillator. It certainly doesn't have to be accurate, and the datasheet says it's accurate to 1%...plenty good enough for this application.
You were nice to it. Big Clive would have put it in the "vice of enlightenment".
Vice of knowledge.
I have a "blunt probe of discovery" rubber mallet that might also work.
It sure has a bunch of zero-ohm resistors as jumpers to avoid a double-sided PCB and vias... Interesting view on the economic trade-offs at work for volume manufacturing.
the sensor is a chemical cell, so it has a finite lifetime. there's liquid in there and eventually it dries out. 10 years is probably a worst case scenario with a bit of safety margin.
First time I've seen a single-sided PCB with dozens of 0 Ohm, SMT resistors used as jumpers. The look like 1206, long enough to get a couple of traces under.
I wonder if they are on the schematic, or put there at the PCB designers discretion. Most CAD tools would complain.... They would need to be on the BOM, Pick and Place, etc..
Probably the CO sensor last ten years and the circuit can measure it.
I wonder what that tamper switch does.
I disabled my CO2 detectors, because the constant beeping was giving me a headache 🤣🤣🤣
Good of you to rotate the board to share your confusion about the ICs. 🙂
are you able to pull that PIC and read it? maybe reprogram it?
So it’s a subscription service. 😊
I know if you shake the sensor there is some kind of liquid inside and some little pillow pack on the end under the yellow heat shrink. Have no idea how it senses CO.
Interesting. No SMPS? I'd certainly salvage that transformer.
Google "Naoyoshi Taguchi" and his historical patents, if you are serious about CO gas sensors. But who knows what Kidde is doing these days.
probably the sensor has a lifetime of 10 years and after ten years it goes out of whatever calibration standard they are held to for carbon monoxide detection
They can last quite a bit longer but the electrochemical cells are affected by various factors like temperature, exposure to certain other gases and the most important factor that limits the age to 10 years, lawyers and liabilities...
IPhone technology to the rescue!
I yanked all the battery eating smoke detectors and carbon monoxide detectors out of our house and replaced them with combination smoke and carbon monoxide detectors with a ten year battery.
kidde FENWAL?
Anyone else notice the 2017 manufacture date? It only lasted 7 years.
I got ripped off
Oct 17 2013 not 2013.
@@IMSAIGuy I guess @timgraham8975 is correct. I can't read. Love your videos!
got my moneys' worth
When you have the time of production and the time of failing you can compare it with the ten years it is specified for. Then from the precision of the time measurement you might exclude some means of time measurement like radioactive decay, battery lifetime, RC clock count or mains cycles. If it is very precise there might be a hidden Rubidium vapor frequency standard below the transformer.
So it’s a subscription service. 😊