Mechanical Relay Quality

[evilkitty]

Re: Mechanical Relay Quality

@stj, are double throw SSRs a thing?

[evilkitty]

Re: Mechanical Relay Quality

600 seems quite low before failure. So it did only ~10 states per day, i.e. 5 on and 5 off per day, for 2 months?
Hmm...
how are you driving your coil? DC? Full rated current?

the contact failed on the coil powered off state 1st, so i am not sure that the could voltage it the issue, but it is running with 5v -transistor vdrop, but you can see how it was done in this image


"600 seems quite low before failure. So it did only ~10 states per day, i.e. 5 on and 5 off per day, for 2 months?"
yes, though we need to extrapolate time cause it was close to 4 or 4.5 years IRC, but that would be under 15k changes or 7500 full cycles and that includes cycles for both relays

also worth noting that this relay did not have the snubber for probably a week while i wanted on parts

the bottom relay has around 6 years on it, 2 years without a snubber and is still working, but only has 1 side of the relay (coil powered on) that does anything and it gets far less yearly cycles as it does not get used with the AC is active

[stj]

Re: Mechanical Relay Quality

maybe it's time to think about cheap SSR's or triacs, but i dont know if triacs will work at that low voltage.

[eccerr0r]

Re: Mechanical Relay Quality

600 seems quite low before failure. So it did only ~10 states per day, i.e. 5 on and 5 off per day, for 2 months?
Hmm...
how are you driving your coil? DC? Full rated current?

[evilkitty]

Re: Mechanical Relay Quality

I have ~600 state changes in my logs 11/26/21 - 1/31/22
this include both relays, i have sanity checks to detect bad data to prevent random state changes, eg if the temp sensor goes from reporting say 20c to 80C in 2 seconds i call bs and read the senor again

edit: just updated the git repo with my current script reversion

[eccerr0r]

Re: Mechanical Relay Quality

I found my furnace/CAC actually sort of low pass filters switching events so bouncing by the mercury switch actually does not affect it. However, not sure about cooling/heating cycles, though I suspect it will do so too.

Maybe the code is changing state too often. My current microcontroller thermostat also uses a (sealed) relay and so far so good, though I don't recall what snub network is in it. It's also a tiny relay like the size of a very tall DIP14 IIRC... It doesn't seem to switch that often maybe leading to longer life.

[redwire]

Re: Mechanical Relay Quality

Just looking at reasons for the relay to have a short life, even with the snubbers.
I am familiar with separate heating and cooling systems, which use a natural gas furnace or boiler, and separate A/C unit. So that just needs a heat and a cooling relay. What model #'s are the furnace or A/C, to see if the current or spike is much larger than we think, for the hardware.

The Honeywell mercury thermostats have their own issues. If anything touches the coils they can jam a bit, if anything nearby makes heat they will also get fooled into thinking the room is warmer than it is. The mercury bounces, even as people walk by or drafts from doors etc. There is a need to sample then and average over many seconds so the sloshing isn't a problem.
The amount of hysteresis (there's two types) is fixed and is just due to the weight of the mercury blob moving over. So at 70F the blob rolled to the side so it takes more force to make it roll back, say 68F.
Your RPi and the relay coils make enough heat to make the mercury switches far off. Your heat sensor is away, so that is accurate.

In the software, I don't see any averaging of temperature readings, or debouncing of the mercury switches. This is a problem because temperature moves up and down from people walking by, doors opening, cooking etc. I would be averaging over a long time like a minute or so.
As well, the mercury switches don't just switch, they bounce and slosh so their readings need to settle for many seconds.

The software seems to read the mercury switches many times in a subroutine which is a no-no because the reading can change while you are in that routine. This might cause oddball errors. A better way is to read tubes once, atomic read into a temporary variable and only use that var for all the consequent compares and conditionals because you know it won't change until next time you read the switches.
The code to reconcile the tubes with the relays state might not work so great, it could cause chatter of the relays if they get turned on and off many times. But it still shouldn't cause that short a life that we are seeing.

[eccerr0r]

Re: Mechanical Relay Quality

Tubes? or just mercury switches?
I still have the old mechanical heat/cool switch which likewise require to be forced in heat or cool mode but not automatically switch between the two. While it does have a mercury switch in it, I thought it was more of the bimetallic strip that senses temperature and the mercury switch was solely for electrical contacting. Having the switch in a hermetically sealed environment with liquid wetting gives long life... alas a TRIAC should also have a long life.

[evilkitty]

Re: Mechanical Relay Quality

"Do "smart" furnaces now automatically choose cool when it's hot and heat when it's cold?"
i have no idea, my dad had bought a $50 thermostat (not even a smart one) and it did not support aux heat, i got him to return it and made made a hybrid thermostat using the ancient thermostat and a raspberry pi, he was saying the tubes act like they are getting 'stuck' (maybe a metal fatigue issue) i decided well if the only issue is the tubes why re-designed everything when i can just use a relays to in-place of tubes, they are just a STSP and STSP switch

[eccerr0r]

Re: Mechanical Relay Quality

PWM your furnace instead of bang/banging it

(and replace it every month...)

You know what, the computerized thermostat I have now does not automatically go run the refrigerant compressor when it's in natural gas mode, and never uses the natural gas when it's in compressor mode. It's interlocked - you have to specifically choose cooling or heating and won't automatically do either. Do "smart" furnaces now automatically choose cool when it's hot and heat when it's cold? Seems easy enough to do but this old microcontrolled thermostat should be able to do it without an issue yet it's explicitly disabled...as I'd expect it to be...

[evilkitty]

Re: Mechanical Relay Quality

There is a slide switch on the thermostat to select heat/cool/ emergency heat/off, in my software you have to set what the physical switch is set to, though software has a option to just pass-though the readings from the mercury tubes (this is also a fallback mode if there is a fatal error)

all i did was use relays in place of the tubes, all other stock hardware is still in use

the 'code' if you can call it that i posted, was just a simple oversimplify logic meant to be easily readable

if you want to see the actual script: https://github.com/GM-Script-Writer-...bin/thermostat

is it possible to have the snubber circuit backwards? no idea if common is the high or low side, i just consider it a mechanical switch so it should not matter it may as well be a piece of wire as far as current flow is concerned

[redwire]

Re: Mechanical Relay Quality

Sorry I clicked on the resistor Mouser link and saw "100J" and thought 100 ohms, but it's 10. I'd say the relay should be lasting much longer then.

The top relay seems to be a master, its coil is on whenever there is demand- for either heat or cooling. The bottom relay is for switching on aux. heat. is that right? How does it switch between heating or cooling? I'd expect one relay for demand and another for selecting heat/cooling. The Honeywell thermostats are brutal without a schematic showing the A/C or changeover valve etc.

You might have an issue with the relatively fast transition from heat/cool as the top relay pole moves, as an arc between N.C. and N.O ends through the center pole. But that's kind of grasping at a straw.
Or maybe software glitching the relay because it looks like you first switch the relay then conditionally update flags, it should be the other way around. The decision (relay on/off) should be the last action taken. But I'm not sure what flags 'heating' or 'cooling' are in your program. I might use code like this knowing that some undefined states exist and need error handling. It's good to never assume things and have extra code to catch invalid flags. This isn't correct code but trying to show what I mean.

if (room_temperature < (setpoint - hysteresis)) {
	heat_demand = true;
	cooling_demand = false;
}
elseif (room_temperature > (setpoint + hysteresis)) {
	cooling_demand = true;
	heat_demand = false;
}
else { // we are at target setpoint in the middle
	cooling_demand = false;
	heat_demand = false;
}
//-------------- now switch outputs --------------------
if (heat_demand == true) {
	Top relay coil = on; 
}
elseif (cooling_demand == true) {
	Top relay coil = off; 
}
else { // we are at setpoint or an invalid i.e. both true heat +cool requested
    do nothing?
}

[evilkitty]

Re: Mechanical Relay Quality

"0.1uF 400V and 100R is not doing great"
"Why did you go 100R not 10R?"
uhh the spec page says it is a 10 Ohm resistor, I'm not sure what you are talking about, i do not see 100 R on the spec page for either the cap or the resistor. does R not mean resistance?

as for why there is wear on both NC and NO

Image for reference: https://github.com/GM-Script-Writer-...0.JPG?raw=true

I have 2 relays, one simulates the top mercury tube and the other emulates the bottom tube
the top one acts like a double throw switch and the bottom is a single throw
both failures have been the top relay so far (it gets the most cycles)
the coil powered off state is what failed 1st on this relay
the bottom relay only gets switched when the temp goes down even though heat is on (this uses a lot of electricity) but when it gets really cold we have a wood stove for heat so that cuts the cycles the bottom relay gets alot

this is the 2 bit logic my relays have
Top relay coil = on; if cooling: cool=true; else if heat: heat=false
Top relay coil = off; if cooling: cool=false; else if heat: heat=true
Bottom relay coil = on; if heat: moreheat=true
Bottom relay coil = off; if heat: moreheat=false

[redwire]

Re: Mechanical Relay Quality

The relay contact's plating is gone on both N.O. and N.C. which is a bit strange. A thermostat uses one set only I thought. But no pitting or etching and not black. This is 24VAC HVAC? Looks like your snubber 0.1uF 400V and 100R is not doing great. The resistor is too high value I think.
I find 24VAC HVAC makes some pretty big arcs with zone valves due to their large inductance. A Honeywell V8043 measures 200mH and 21 ohms, but is only about 8VA load. Best I could fit is 0.22uF and 10R after using scope traces to calm down the spike. Snubber values for mains circuits have different values, there you can use a bigger resistor 33-100R and smaller cap 0.01-0.1uF

Before, I recommended 10R resistor and maybe double the cap size. I don't know what the load is, if it's a contactor or furnace gas valve etc. Why did you go 100R not 10R? I'd change the snubber or maybe a MOV is really what's needed.

[evilkitty]

Re: Mechanical Relay Quality

Lasted ~4 years this time
Here is a image of the relay contacts after failure
these are the parts i used:
https://www.mouser.com/ProductDetail...S2G031001K00KO
https://www.mouser.com/ProductDetail...OS1/2CT52R100J

[Curious.George]

Re: Mechanical Relay Quality

a film cap can hadle high current pulses well - that's why you use film.

Metallized film is also self-healing.

The drawback with film is usually size/cost as you look for higher voltages and values of capacitance as well as a much smaller range of options for SMT production.

[stj]

Re: Mechanical Relay Quality

a film cap can hadle high current pulses well - that's why you use film.

[evilkitty]

Re: Mechanical Relay Quality

is that what the film cap does?
does that mean i could use 2 normal caps back to back to do the same thing?

[Curious.George]

Re: Mechanical Relay Quality

When using a snubber with a AC circuit the direction of current flow matter, if so should the cap be on the Line side or the neutral side of the switch/relay?

Polarity wouldn't affect a devices location IN a circuit (line vs neutral) but, rather, the device's orientation as it is inserted. As this circuit will see both polarities (which is something more than just saying "AC"), you need an NPO device (which can be made from polarized devices placed "back-to-back")

[evilkitty]

Re: Mechanical Relay Quality

thanks, hopefully i can start my projects next weekend