Bimetallic switches reliability

[Michael]

I am looking for a thermostat (e.g.: http://www.ngtcontrols.com/gen-automaticreset.html) for basic on-off temperature control. I noticed that specified service life never exceeds 100k cycles. I wonder whether it is due to contacts burnout or bimetallic strip fatigue. Does anybody have any insight?
Thanks.
Michael

[whit3rd]

On Monday, August 10, 2015 at 2:26:31 PM UTC-7, Michael wrote:
> I am looking for a thermostat (e.g.: http://www.ngtcontrols.com/gen-automaticreset.html) for basic on-off temperature control. I noticed that specified service life never exceeds 100k cycles. I wonder whether it is due to contacts burnout or bimetallic strip fatigue. Does anybody have any insight?

Furnaces are best switched infrequently, are rather inefficient if you only operate them
for a few minutes. So, a thermostat with one cycle per hour will take 11 years (or
if heating season is only three months per year, 45 years) to reach that life.

It may be that no one cares to wait for longer testing.

[DecadentLinuxUserNumeroUno]

On Mon, 10 Aug 2015 14:26:24 -0700 (PDT), Michael <mk61...@gmail.com>
Gave us:

>I am looking for a thermostat (e.g.: http://www.ngtcontrols.com/gen-automaticreset.html)
> for basic on-off temperature control. I noticed that specified service

>. life never exceeds 100k cycles. I wonder whether it is due to contacts

> burnout or bimetallic strip fatigue. Does anybody have any insight?
>Thanks.
>Michael

Usually based on contact degradation. Bimetal works forever unless it
rusts away.

[DecadentLinuxUserNumeroUno]

On Mon, 10 Aug 2015 16:07:12 -0700 (PDT), whit3rd <whi...@gmail.com>
Gave us:

>It may be that no one cares to wait for longer testing.

It is based on already known or expected service life of the contacts.
No need to perform some lifelong test, which would not provide accurate
MTBF numbers anyway.

[Jeff Liebermann]

On Mon, 10 Aug 2015 14:26:24 -0700 (PDT), Michael <mk61...@gmail.com>

wrote:

>I am looking for a thermostat (e.g.: http://www.ngtcontrols.com/gen-automaticreset.html) for basic on-off temperature control. I noticed that specified service life never exceeds 100k cycles. I wonder whether it is due to contacts burnout or bimetallic strip fatigue. Does anybody have any insight?
>Thanks.
>Michael

Thank you for not bothering to specify what you are controlling with
the thermostat. That makes answering your question much easier since
I can more easily produce vague generalities instead of something you
might find useful.

The snap disc thermostats in your URL are not designed for a large
number of cycles.
<https://www.google.com/search?q=disk+thermostat>
They are over-temperature thermostats, designed to cycle when
something goes awry, gets too hot, and requires a shut down before the
heat producer burns down the building. Hopefully, such events are
rare.

Also note that the loads are AC and resistive only. That means you're
not going to successfully switch a relay or DC. I found that out the
hard way when I designed a similar thermostat into an RF power
amplifier and found that the thermostat would last about 10 cycles
before welding the contacts together. Of course, nobody noticed until
the PA almost melted. Oops(tm).

If your unspecified application really needs more than 100K cycles,
then I suggest you look at what the typical HVAC thermostat is doing
for temperature control using no moving parts:
<https://www.google.com/search?q=hvac+thermostat&tbm=isch>
and avoid the mechanical flavor.

--
Jeff Liebermann je...@cruzio.com
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558

[Michael]

On Monday, August 10, 2015 at 7:34:36 PM UTC-4, Jeff Liebermann wrote:
> On Mon, 10 Aug 2015 14:26:24 -0700 (PDT), Michael <mkx...@gmail.com>

>>You wrote: Thank you for not bothering to specify what you are controlling ...

I have some time to elaborate.
The goal is to keep temperature in an enclosure within "narrow" range. 20 and 30 degC is considered barely acceptable. The power dissipated inside is... er... probably between 40 and 60W, the box is used outdoors. The box size - ?? My guess that surface area is ~1.5m^2. Outside temperature is -10...+50degC.
I wish I knew whether there will be any wind or sun load.
It is acceptable to thermally insulate the box, I do not think I can ask for more than, say, 10mm.
I have ~110V and =24V as a power source.
Can anybody recommend turnkey solution vendor? I would be the best. I contacted ... forgot the name ... few days ago, haven't heard from them yet.

The plan is to have one (demo) system running in one month ("Nothing gets built on time or within budget"). NO TIME for anything elegant, hence seemingly dumb solution (thermal switch for heating, another one for cooling). Overlapping should be OK, I think.
Heating will be needed more often, but it easier (<=0.5A @ ~110V).
Cooling (TEC, >7A,=24V), requires a relay.
If thermal switch contacts define longevity, I can use them. If it is bimetal disk/strip, I need to keep looking.
Thank you
Michael

[Jeff Liebermann]

On Mon, 10 Aug 2015 17:58:44 -0700 (PDT), Michael <mk61...@gmail.com>
wrote:

>I have some time to elaborate.

I'm waiting for Windoze to update, so I also have some time.

>The goal is to keep temperature in an enclosure within "narrow" range.
>20 and 30 degC is considered barely acceptable.

Sounds like a chicken egg incubator.

>The power dissipated inside is... er... probably between 40 and 60W,
>the box is used outdoors. The box size - ?? My guess that surface
>area is ~1.5m^2. Outside temperature is -10...+50degC.

You'll need a refrigerator. If the outside temp is 50C, you won't be
able to lower it to 20 - 30C with a heater.

Any chance you can supply dimensions (LWH) instead of surface area?
With those, one can calculate volume and surface area. Going the
other way doesn't work.

>I wish I knew whether there will be any wind or sun load.

All that can be calculated if could also produce the mass of the
contents and the box. Also what's the nature of the box or contents:
<http://hyperphysics.phy-astr.gsu.edu/hbase/thermo/spht.html>
Solar head load and wind cooling will complicate matters, but it's
still possible. Basically, what you're trying to calculate is whether
40 to 60 watts is sufficient to stabilize the temperature. I've
already demonstrated that you'll need a refrigerator when the box gets
hot, so the short answer is no, it won't work.

>It is acceptable to thermally insulate the box, I do not think I
>can ask for more than, say, 10mm.

10 mm of what? Vacuum? Fiberglass? Foam board? 10 mm is not very
much in the way of insulation.

>I have ~110V and =24V as a power source.

At how many amps or watts? Electric heating is 100% efficient so your
sources should be able to produce 40 to 60 watts.

>Can anybody recommend turnkey solution vendor?

Chicken egg incubator. Plenty of kits and instructions on the web.

>I would be the best. I contacted ... forgot the name ...
>few days ago, haven't heard from them yet.

Alien contact is highly over rated.

>The plan is to have one (demo) system running in one month ("Nothing
>gets built on time or within budget").

I never make plans. That way, I'm never disappointed. Unfortunately,
other people tend to make plans for my time, so that doesn't really
work.

>NO TIME for anything elegant, hence seemingly dumb solution (thermal
>switch for heating, another one for cooling). Overlapping should be
>OK, I think.

The box seems small enough that CO2 cooling should work. I used to
work with various environmental test chambers for temperature testing
radios. -40C to +75C was typical. It think they were made by Tenny.
<https://www.google.com/search?q=environmental+test+chamber>
CO2 was a cheap way to avoid refrigeration. This looks close:
<https://www.thermalproductsolutions.com/product/tenney-junior-compact-temperature-test-chamber>

>Heating will be needed more often, but it easier (<=0.5A @ ~110V).
>Cooling (TEC, >7A,=24V), requires a relay.
>If thermal switch contacts define longevity, I can use them. If it is bimetal disk/strip, I need to keep looking.

Controlling the temperature is one of the easiest parts of your
project. There are off the shelf controllers which will probably do a
better job than a bimetalic thermostat, which incidentally has lots of
hysteresis. How does $3.00 sound?
<http://www.ebay.com/itm/331622950879>

Incidentally, if you decide to buy an environmental oven, be sure that
it fits a pizza platter.

[Michael]

On Monday, August 10, 2015 at 10:26:05 PM UTC-4, Jeff Liebermann wrote:
> On Mon, 10 Aug 2015 17:58:44 -0700 (PDT), Michael <xxxx...@gmail.com>

You wrote: > Sounds like a chicken egg incubator.
Same idea, different contents.
You wrote: You'll need a refrigerator. If the outside temp is 50C, you won't be able to lower it to 20 - 30C with a heater.
Very insightful
You wrote: > 10 mm of what? Vacuum? Fiberglass? Foam board?
Probably some kind of foam. Ideally need ~1.5-2" of foam for the internal power to be enough to keep the box warm in winter, much more if it is windy. The box needs to be reasonably sized - proper insulation is out of question.
You wrote: I never make plans. That way, I'm never disappointed.
I am with you 100%. You also said something about other people making plans for you... c'est la vie

$3 controller from ebay sounds very tempting. I will keep it in mind when/if I decide to build incubator at home. For this application something with manufacturer name/part# and datasheet is required ...
I need to do some "googling".

[John S]

Your local Home Depot might have a bimetallic thermostat as was once
used in every household. Have you tried that?

[George Herold]

I've got a few thermostat's with a Mercury switch attached to a bimetalic coil.
Maybe 20 years old... I think they should be good for another 20 years?

George H.

[DecadentLinuxUserNumeroUno]

On Tue, 11 Aug 2015 07:01:36 -0700 (PDT), George Herold
<ghe...@teachspin.com> Gave us:

snip

>I've got a few thermostat's with a Mercury switch attached to a bimetalic coil.
>Maybe 20 years old... I think they should be good for another 20 years?
>
>George H.

More proof that the MTBF determining factor is the contacts.

[tabb...@gmail.com]

On Tuesday, 11 August 2015 03:26:05 UTC+1, Jeff Liebermann wrote:

> Controlling the temperature is one of the easiest parts of your
> project. There are off the shelf controllers which will probably do a
> better job than a bimetalic thermostat, which incidentally has lots of
> hysteresis.

Hysteresis is almost eliminated by compensation. Uncompensated bimetal stats are a curse.

You can increase mtbf with a snubber, and increase it again with a series parallel set of 4 stats, but how far those will get you I don't know. Its also increased by keeping switching current low and avoiding inductance.


NT

[Jeff Liebermann]

On Thu, 13 Aug 2015 15:59:15 -0700 (PDT), tabb...@gmail.com wrote:

>On Tuesday, 11 August 2015 03:26:05 UTC+1, Jeff Liebermann wrote:
>
>> Controlling the temperature is one of the easiest parts of your
>> project. There are off the shelf controllers which will probably do a
>> better job than a bimetalic thermostat, which incidentally has lots of
>> hysteresis.

>Hysteresis is almost eliminated by compensation.

How does one compensate for mechanical hysteresis? To the best of my
limited understanding, this thermostat requires the snap action of the
bimetallic strip, which needs intentional hysteresis in order to
obtain the snap action. No hysteresis -> No snap action. Where the
contacts to slowly drift together or apart, the result would be
arcing, a noisy connection, and generally lousy reliability. The best
my imagination can conjure for a switch where the snap action is
eliminated, is a bimetallic strip moving a variable resistor, which
eventually trips a comparator (presumably without the benefits of
hystersis. At that level of complexity, one might as well use a
thermistor or thermocouple and do it all electronically and eliminate
the bimetallic question. (Apologies to Sherlock Holmes).

>Uncompensated bimetal stats are a curse.

Do you have an example of a compensated thermostat without any or very
little hysteresis?

>You can increase mtbf with a snubber, and increase it again with a series
>parallel set of 4 stats, but how far those will get you I don't know.
>Its also increased by keeping switching current low and avoiding inductance.

With a snap action thermostat, one can reduce turn on and turn off
times sufficiently that any arcing is minimized. Much depends on the
contact material, if there's a DC component, current level, whether
it's switched under load, and environmental issues. In other words,
you can't optimize the life of relay, switch, or thermostat contacts
unless you know the operating conditions. RC snubbers can help, but
only if they're tuned to the specific operating conditions. Done
wrong, they make electrical contact metal migration worse.
<http://electronics.stackexchange.com/questions/42131/how-to-design-an-rc-snubber-for-a-solenoid-relay-driving-an-inductive-load>

Incidentally, bimetallic switches also tend to drift as the two metal
strips tend to delaminate after many thousands of operations. It's
part of "If it moves, it breaks" which means that if you want it to be
reliable, don't design in any moving parts.

[DecadentLinuxUserNumeroUno]

On Thu, 13 Aug 2015 18:23:40 -0700, Jeff Liebermann <je...@cruzio.com>
Gave us:

> Where the
>contacts to slowly drift together or apart, the result would be
>arcing, a noisy connection, and generally lousy reliability.

An encapsulated mercury switch on the end of a bimetal coil reaches a
tipping point and the contacts close or open in a vacuum as the mercury
ball falls to one end or the other of the tube.

There are also steel ball switches which impinge on hard contact reeds
in a similar manner within a vacuum 'tube' assembly. They rely on
gravity and balance.

[Jeff Liebermann]

On Fri, 14 Aug 2015 00:32:33 -0400, DecadentLinuxUserNumeroUno
<DL...@DecadentLinuxUser.org> wrote:

>On Thu, 13 Aug 2015 18:23:40 -0700, Jeff Liebermann <je...@cruzio.com>
>Gave us:
>>Where the
>>contacts to slowly drift together or apart, the result would be
>>arcing, a noisy connection, and generally lousy reliability.

>An encapsulated mercury switch on the end of a bimetal coil reaches a
>tipping point and the contacts close or open in a vacuum as the mercury
>ball falls to one end or the other of the tube.

Such mercury thermostats have a huge hysteresis. I was discussing
what happens if you take the hysteresis out of the thermostat.
Therefore, mercury switches don't qualify and are also banned for
ecological reasons. On the other foot:
<http://802.11junk.com/jeffl/crud/mercury.jpg>

>There are also steel ball switches which impinge on hard contact reeds
>in a similar manner within a vacuum 'tube' assembly. They rely on
>gravity and balance.

I haven't seen any of those. However, I suspect that like any similar
type of "tipping" thermostat, it would have a rather large hysteresis.
I could probably get rid of much of the hysteresis by making the tube
assembly circular and roughly concentric with the arc followed by the
bimetallic spring arm. However, as the spring slooooooowly creeps in
the direction need to close or open the contacts, there will be plenty
of time for the gap to be the proper size to produce an impressive
arc. If a low voltage thermostat, plenty of contact noise instead.

[whit3rd]

On Tuesday, August 11, 2015 at 7:27:22 AM UTC-7, DecadentLinuxUserNumeroUno wrote:
> On Tue, 11 Aug 2015 07:01:36 -0700 (PDT), George Herold
> <ghe...@teachspin.com> Gave us:

> >I've got a few thermostat's with a Mercury switch attached to a bimetalic coil.

> More proof that the MTBF determining factor is the contacts.

Not necessarily the ONLY factor, though; peak current, arcing, dirt sensitivity,
contamination, oxidation... there's a lot of ways to kill a switch.
Mercury switches don't always have a MTBF rating, but mercury relays do.

[tabb...@gmail.com]

On Friday, 14 August 2015 02:23:48 UTC+1, Jeff Liebermann wrote:

> On Thu, 13 Aug 2015 15:59:15 -0700 (PDT), nt wrote:
> >On Tuesday, 11 August 2015 03:26:05 UTC+1, Jeff Liebermann wrote:
> >
> >> Controlling the temperature is one of the easiest parts of your
> >> project. There are off the shelf controllers which will probably do a
> >> better job than a bimetalic thermostat, which incidentally has lots of
> >> hysteresis.
>
> >Hysteresis is almost eliminated by compensation.
>
> How does one compensate for mechanical hysteresis?

A little heater resistor comes on, raising the bimetal by about the same temp rise as the hysteresis. The result is a fraction of a degree hysteresis.

> To the best of my
> limited understanding, this thermostat requires the snap action of the
> bimetallic strip, which needs intentional hysteresis in order to
> obtain the snap action. No hysteresis -> No snap action. Where the
> contacts to slowly drift together or apart, the result would be
> arcing, a noisy connection, and generally lousy reliability. The best
> my imagination can conjure for a switch where the snap action is
> eliminated, is a bimetallic strip moving a variable resistor, which
> eventually trips a comparator (presumably without the benefits of
> hystersis. At that level of complexity, one might as well use a
> thermistor or thermocouple and do it all electronically and eliminate
> the bimetallic question. (Apologies to Sherlock Holmes).
>
> >Uncompensated bimetal stats are a curse.
>
> Do you have an example of a compensated thermostat without any or very
> little hysteresis?

every British central heating stat that hasn't been replaced with an electronic one, eg
http://www.screwfix.com/c/heating-plumbing/central-heating-controls/cat831444?cm_sp=managedredirect-_-plumbing-_-HeatControls
includes three

> >You can increase mtbf with a snubber, and increase it again with a series
> >parallel set of 4 stats, but how far those will get you I don't know.
> >Its also increased by keeping switching current low and avoiding inductance.
>
> With a snap action thermostat, one can reduce turn on and turn off
> times sufficiently that any arcing is minimized.

how? if you want low hysteresis you have to live with low contact pressure, and that means slow switching. Arcing goes with the territory.
You could add a relay to relieve current, but the coil is inductive. You could add electronics and sense either the short or capacitance to enable even a well oxidised contact to work, but then you've got the unreliability of the electronics.

> Much depends on the
> contact material, if there's a DC component, current level, whether
> it's switched under load, and environmental issues. In other words,
> you can't optimize the life of relay, switch, or thermostat contacts
> unless you know the operating conditions. RC snubbers can help, but
> only if they're tuned to the specific operating conditions. Done
> wrong, they make electrical contact metal migration worse.
> <http://electronics.stackexchange.com/questions/42131/how-to-design-an-rc-snubber-for-a-solenoid-relay-driving-an-inductive-load>
>
> Incidentally, bimetallic switches also tend to drift as the two metal
> strips tend to delaminate after many thousands of operations. It's
> part of "If it moves, it breaks" which means that if you want it to be
> reliable, don't design in any moving parts.

Kind of. In practice electronic goods seem less reliable than electromechanical.


NT

[DecadentLinuxUserNumeroUno]

On Fri, 14 Aug 2015 11:42:06 -0700 (PDT), whit3rd <whi...@gmail.com>
Gave us:

>On Tuesday, August 11, 2015 at 7:27:22 AM UTC-7, DecadentLinuxUserNumeroUno wrote:
>> On Tue, 11 Aug 2015 07:01:36 -0700 (PDT), George Herold
>> <ghe...@teachspin.com> Gave us:
>
>> >I've got a few thermostat's with a Mercury switch attached to a bimetalic coil.
>
>> More proof that the MTBF determining factor is the contacts.
>
>Not necessarily the ONLY factor, though; peak current, arcing, dirt sensitivity,
>contamination, oxidation... there's a lot of ways to kill a switch.

Silly child. Those are all CONTACT failure modes.

>Mercury switches don't always have a MTBF rating, but mercury relays do.

Everything has an MTBF rating otherwise nothing would. It's the
weakest link thing.

That is why America would be absolutely stupid to put an idiot like
Trump in office, much less Clinton.

[Jeff Liebermann]

On Fri, 14 Aug 2015 11:59:50 -0700 (PDT), tabb...@gmail.com wrote:

>> How does one compensate for mechanical hysteresis?

>A little heater resistor comes on, raising the bimetal by about
>the same temp rise as the hysteresis. The result is a fraction
>of a degree hysteresis.

On this side of the planet, we don't have those. Mostly, we have
setback thermostats like these:
<https://www.google.com/search?q=setback+thermostat&tbm=isch>

I don't see how the little heater resistor trick would work. What
turns on the little heater before the main contacts close? When going
from off to on, the contacts haven't closed yet, so there's nothing to
trigger the little heater. When going from on -> off the little
heater is presumably on, so the little heater resistor would need some
way to predict that the contacts are about to open and turn off early.

>> Do you have an example of a compensated thermostat without any or very
>> little hysteresis?
>
>every British central heating stat that hasn't been replaced with an electronic one, eg
>http://www.screwfix.com/c/heating-plumbing/central-heating-controls/cat831444?cm_sp=managedredirect-_-plumbing-_-HeatControls
>includes three

Those are all electronic thermostats that use thermistors or
thermocouples. What I would like to see is the guts of a mechanical
bimetallic thermostat with the little heater resistor trick. I tried
Googling for such a device, but couldn't find anything, probably
because I'm missing a UK specific buzzword. What are they called?

>> With a snap action thermostat, one can reduce turn on and turn off
>> times sufficiently that any arcing is minimized.
>
>how?

Easy. A strong spring will close quicker than a weak spring. If you
want to reduce arcing duration, just switch faster. That's one way
big electrical contactors prevent contact meltdown from arcing. Lots
more under arc suppression, but none which really applies to a fairly
low current thermostat. My only point is that if the thermostat
contacts open slowly, the arc duration is longer, and therefore the
damage more extensive. It may be a tiny arc, but after 100,000
cycles, the damage can accumulate.

>if you want low hysteresis you have to live with low contact pressure,
>and that means slow switching. Arcing goes with the territory.

Yep. Zero contact pressure will give you nearly zero hysteresis.
However, if you want to minimize the damage caused by even minimal
arcing, faster switching will be needed, which means stronger spring
tension, which implies more hysteresis.

I guess I should mention that even the electronic thermostats have
designed in hystersis. It would not do to have the HVAC system
cycling ever few seconds or every time the room door is opened or
closed. Hysteresis keeps the machinery from getting pounded to death
by too many starts and stops.

>You could add a relay to relieve current, but the coil is inductive.
>You could add electronics and sense either the short or capacitance
>to enable even a well oxidised contact to work, but then you've got
>the unreliability of the electronics.

If you're going to add the level of complexity, you might as well go
all electronic.

>> Incidentally, bimetallic switches also tend to drift as the two metal
>> strips tend to delaminate after many thousands of operations. It's
>> part of "If it moves, it breaks" which means that if you want it to be
>> reliable, don't design in any moving parts.

>Kind of. In practice electronic goods seem less reliable than electromechanical.

Not exactly. The components used in electronic solutions are
generally more reliable than a mechanical solution. The problem is
that electronic solutions tend to have far more components than
electronic solutions. In other words, electronic solutions are less
reliable simply because there are more things that can fail. There is
also a tendency to design electronics so that it only lasts 1 day more
than the warranty period, which offers additional opportunities for
failure. While this can also be done with mechanical solutions, I
haven't seen it done as often. However, it's not the reliability of
bimetal thermostats that make them attractive. It's the low price.

[Jeff Liebermann]

On Fri, 14 Aug 2015 11:42:06 -0700 (PDT), whit3rd <whi...@gmail.com>

wrote:

Mercury wetted relays are rated in cycles. I learned this the hard
way while helping troubleshoot a fixture to test 16 bit DAC's for
linearity and missing codes. We used relays that were rated for 2
million cycles. The LSB relay would thrash 65,000 times during one
test cycle. Therefore, the relay was good for:
2*10^6 / 65*10^3 = 31 DACs
Every 31 DACs, we had to replace the LSB relay, and every 62 DAC's, we
also had to replace the next bit, and so on. That rapidly became very
tedious but at the time (mid 1970's) we didn't have a better way.

Incidentally, I came up with a method of rejuvenate the relays. I
would ultrasonically shake the relays which caused the breakup of
whatever slag was covering the contacts. It wasn't reliable, but
lasted long enough until the next batch of replacement relays arrived.

[Phil Hobbs]

Why didn't you swap the relays down 1 bit every few ADCs?

Cheers

Phil Hobbs

[Jasen Betts]

On 2015-08-14, Jeff Liebermann <je...@cruzio.com> wrote:
> On Fri, 14 Aug 2015 11:59:50 -0700 (PDT), tabb...@gmail.com wrote:
>
>>> How does one compensate for mechanical hysteresis?
>
>>A little heater resistor comes on, raising the bimetal by about
>>the same temp rise as the hysteresis. The result is a fraction
>>of a degree hysteresis.
>
> On this side of the planet, we don't have those. Mostly, we have
> setback thermostats like these:
><https://www.google.com/search?q=setback+thermostat&tbm=isch>
>
> I don't see how the little heater resistor trick would work. What
> turns on the little heater before the main contacts close?

Assume it's wired parallel with the contacts


--
\_(ツ)_

[Jeff Liebermann]

Thanks, that should work. With the contacts open, there's some
current going through the heater resistor, which increases the bimetal
strip temperature to something less than the hysteresis. When the
contacts close, the resistor is shorted, eliminating that source of
heat. Assuming the ambient temperature doesn't change, the bimetal
strip now drops in temperature to again something less than the
hysteresis. It sits there forever, or until the device temperature
drops the remaining difference. The bimetal strip then trips, opens
the contacts, and the cycle repeats.

Looks like it needs more than just a resistor:
<http://www.google.com/patents/US3940727>
<http://www.google.com/patents/US3629607>
<http://www.google.com/patents/CA1276214C>
Looks like the missing buzzword might be "heat anticipation
thermostat".

I can see some problems adjusting such an arrangement, but with a
consistent enough bimetal strip, it should work.

However, I still question if it's necessary or desirable. A zero
hysteresis thermostat will be constantly cycling with every passing
breeze. That would beat up the HVAC machinery or whatever it
controls. A time delay could be added to either the thermostat or the
machinery, but that puts us back to an all electronic solution. Do
the heaters in the UK have built in on/off delays?

[Jeff Liebermann]

On Fri, 14 Aug 2015 15:31:33 -0700 (PDT), Phil Hobbs
<pcdh...@gmail.com> wrote:

>Why didn't you swap the relays down 1 bit every few ADCs?

>Phil Hobbs

Bad planning and a lazy engineer (me). Rotating the relays is what
everyone wanted and suggested. The hourly labor hired to run the test
turned out to be incapable of replacing the relays because the fixture
had not been designed with relay sockets[1] and ease of access. Since
it was a one of kind, no effort was spent making it easily repairable.
At about 15 mins per full test per device (or 15 msec/bit), 31 tests
would be about 8 hrs before one or more relays required replacement.
There were two shifts, so I was sentenced to replacing the necessary
relays when I arrived and again when I left work. Since the relays
had to be desoldered, I was not particularly thrilled with spending an
extra hour or so moving all the relays towards the LSB. So, I made a
chart of which relays had to be replaced, and performed only the
minimum replacement, which usually took about 10 mins. There was also
the problem of the nearby precision resistors, which tended to drift
when exposed to soldering iron heat. Fortunately, the typical lot run
was about 100 DAC's quarterly, so the most I had to deal with this
was:
100 / 31 = 3.2 days
of swearing and snarling every 90 days. Manufacturing declared that
it was easier to tolerate my complaining than to rebuild the
fixture[2]. I vaguely recall performing this ordeal for about 3 years
until the product finally died a merciful and well deserved death. The
worst part of this was that it wasn't my project or even my division.
The company had two divisions (modular products and marine radios) and
I worked for marine radios. I was "helping out".

Lesson learned: Never make a suggestion where I'm the only person
available or qualified to implement it.


[1] Actually, some early fixture used sockets. It was discovered
that the sockets were less reliable than the relays and were rapidly
discarded in favor of soldered relays. There was also an aborted
attempt to use FET switches, but I don't recall why that failed.

[2] I signed all the associated documents as Sisyphus.

[tabb...@gmail.com]

On Friday, 14 August 2015 22:14:09 UTC+1, Jeff Liebermann wrote:

> On Fri, 14 Aug 2015 11:59:50 -0700 (PDT), nt wrote:
>
> >> How does one compensate for mechanical hysteresis?
>
> >A little heater resistor comes on, raising the bimetal by about
> >the same temp rise as the hysteresis. The result is a fraction
> >of a degree hysteresis.
>
> On this side of the planet, we don't have those. Mostly, we have
> setback thermostats like these:
> <https://www.google.com/search?q=setback+thermostat&tbm=isch>

With bimetals the heating is just switched off at night. Houses only drop a couple of degrees overnight on a very cold day.

> I don't see how the little heater resistor trick would work. What
> turns on the little heater before the main contacts close? When going
> from off to on, the contacts haven't closed yet, so there's nothing to
> trigger the little heater. When going from on -> off the little
> heater is presumably on, so the little heater resistor would need some
> way to predict that the contacts are about to open and turn off early.

http://wiki.diyfaq.org.uk/index.php?title=Thermostat#Types_of_thermostat
explains a little more

> >> Do you have an example of a compensated thermostat without any or very
> >> little hysteresis?
> >
> >every British central heating stat that hasn't been replaced with an electronic one, eg
> >http://www.screwfix.com/c/heating-plumbing/central-heating-controls/cat831444?cm_sp=managedredirect-_-plumbing-_-HeatControls
> >includes three
>
> Those are all electronic thermostats that use thermistors or
> thermocouples. What I would like to see is the guts of a mechanical
> bimetallic thermostat with the little heater resistor trick. I tried
> Googling for such a device, but couldn't find anything, probably
> because I'm missing a UK specific buzzword. What are they called?

Google isn't showing compensated ones. I'll see if I can get a pic.

> >> With a snap action thermostat, one can reduce turn on and turn off
> >> times sufficiently that any arcing is minimized.
> >
> >how?
>
> Easy. A strong spring will close quicker than a weak spring. If you
> want to reduce arcing duration, just switch faster. That's one way
> big electrical contactors prevent contact meltdown from arcing. Lots
> more under arc suppression, but none which really applies to a fairly
> low current thermostat. My only point is that if the thermostat
> contacts open slowly, the arc duration is longer, and therefore the
> damage more extensive. It may be a tiny arc, but after 100,000
> cycles, the damage can accumulate.

but that's not much use when you want low hysteresis

> >if you want low hysteresis you have to live with low contact pressure,
> >and that means slow switching. Arcing goes with the territory.
>
> Yep. Zero contact pressure will give you nearly zero hysteresis.
> However, if you want to minimize the damage caused by even minimal
> arcing, faster switching will be needed, which means stronger spring
> tension, which implies more hysteresis.

there are other options

another reason is simply that typically with (inevitably complex) electronics, some failure mode was not predicted. Goods can be very well designed in almost every respect, it only takes one little oversight to render them dead. And its tough to avoid that.

> However, it's not the reliability of
> bimetal thermostats that make them attractive. It's the low price.

both are attractive. I've replaced electronic with bimetal for reliability. The old dream of replacing everything mechanical with electronics to solve the reliability issues hasn't really worked out. In some cases electronic is better, in some not.


NT

[tabb...@gmail.com]

On Saturday, 15 August 2015 02:43:48 UTC+1, Jeff Liebermann wrote:
> On 15 Aug 2015 00:04:48 GMT, Jasen Betts <ja...@xnet.co.nz> wrote:
> >On 2015-08-14, Jeff Liebermann <je...@cruzio.com> wrote:

> >> On Fri, 14 Aug 2015 11:59:50 -0700 (PDT), nt wrote:
> >>
> >>>> How does one compensate for mechanical hysteresis?
> >>
> >>>A little heater resistor comes on, raising the bimetal by about
> >>>the same temp rise as the hysteresis. The result is a fraction
> >>>of a degree hysteresis.
> >>
> >> On this side of the planet, we don't have those. Mostly, we have
> >> setback thermostats like these:
> >><https://www.google.com/search?q=setback+thermostat&tbm=isch>
> >>
> >> I don't see how the little heater resistor trick would work. What
> >> turns on the little heater before the main contacts close?
>
> >Assume it's wired parallel with the contacts
>
> Thanks, that should work. With the contacts open, there's some
> current going through the heater resistor, which increases the bimetal
> strip temperature to something less than the hysteresis. When the
> contacts close, the resistor is shorted, eliminating that source of
> heat. Assuming the ambient temperature doesn't change, the bimetal
> strip now drops in temperature to again something less than the
> hysteresis. It sits there forever, or until the device temperature
> drops the remaining difference. The bimetal strip then trips, opens
> the contacts, and the cycle repeats.
>
> Looks like it needs more than just a resistor:
> <http://www.google.com/patents/US3940727>
> <http://www.google.com/patents/US3629607>
> <http://www.google.com/patents/CA1276214C>

its just a resistor

> Looks like the missing buzzword might be "heat anticipation
> thermostat".

if you want to find patents maybe.

> I can see some problems adjusting such an arrangement, but with a
> consistent enough bimetal strip, it should work.
>
> However, I still question if it's necessary or desirable. A zero
> hysteresis thermostat will be constantly cycling with every passing
> breeze. That would beat up the HVAC machinery or whatever it
> controls. A time delay could be added to either the thermostat or the
> machinery, but that puts us back to an all electronic solution. Do
> the heaters in the UK have built in on/off delays?

The point is not to zero hysteresis, its to reduce it to the level desired, typically about 0.5C. Electronic room stats tend to shrink this further to trim energy use, down to 0.25C.


NT

[Neon John]

On Fri, 14 Aug 2015 18:43:40 -0700, Jeff Liebermann <je...@cruzio.com>
wrote:

>Looks like the missing buzzword might be "heat anticipation
>thermostat".

That is correct and the resistance wire is called the anticipator.

>
>I can see some problems adjusting such an arrangement, but with a
>consistent enough bimetal strip, it should work.

It work extremely well. The resistance wire is arranged in a 270 deg
loop, similar to a pot with a wiper. The wiper has a calibrated
scale, calibrated in AC mA. One measures the 24 volt loop current and
sets the wiper to that value. That sets the thermostat for 1 deg of
hysteresis.

One can get more or less hysteresis by deviating from this procedure.
I have my Honeywell "Round" (with the mercury switch, installed in
1970) tuned to about 3/4 degree room hysteresis. That's where I'm the
most comfortable.

My heat pump cycles 4-5 times an hour. It uses a scroll type
compressor that has low inrush and starts easily. So my standby
generator can power it, I've installed time delay relays so the
compressor starts first, then the condenser cooling fan and finally
the evaporator fan.

>
>However, I still question if it's necessary or desirable. A zero
>hysteresis thermostat will be constantly cycling with every passing
>breeze. That would beat up the HVAC machinery or whatever it
>controls. A time delay could be added to either the thermostat or the
>machinery, but that puts us back to an all electronic solution. Do
>the heaters in the UK have built in on/off delays?

All this had been worked out eletro-mechanically by at least the 50s.
In the early 60s my family had a new-fangled heat pump installed.
Being the kid with 1000 questions, I asked the installers about every
aspect of the install including the anticipator setup.

Back then, before the government was sucking down the economy, the
HVAC company could afford to do a complete job for a reasonable price
which including leaving a spring wound circular chart recorder in the
house overnight to verify the anticipator setup and the thermostat
calibration.

John
John DeArmond
http://www.neon-john.com
http://www.fluxeon.com
Tellico Plains, Occupied TN
See website for email address

[Jeff Liebermann]

On Sat, 15 Aug 2015 13:02:43 -0400, Neon John <n...@never.com> wrote:

>On Fri, 14 Aug 2015 18:43:40 -0700, Jeff Liebermann <je...@cruzio.com>
>wrote:
>
>
>>Looks like the missing buzzword might be "heat anticipation
>>thermostat".
>
>That is correct and the resistance wire is called the anticipator.

Ok. Got it, maybe:
<http://inspectapedia.com/heat/Heat_Anticipator_Adustment.php>

What surprised me is this line from the above URL:
"The anticipator also creates hysteresis, so the furnace does
not cycle too often."
You suggest that it reduces hysteresis which also what I would expect.
So, is the web page in error?

In addition, it mentions:
"QP means by hysterisis that we introduce a lag or delay between
the change in room temperature and the response of the thermostat.
However the heat anticipator is only operating to turn the
thermostat "off" a little "early". It does not introduce a delay
in turning the thermostat back on. - Ed."
I thought it worked in both directions?

I still have a problem with reducing the hysteresis. You mention
0.75F hysteresis. I would think that even slight changes to air
temperature would cause the thermostat to cycle. I would also think
that a smaller hysteresis means a short run/off time, which might
cause thrashing. Is 0.75F a reasonable value for hysteresis?

Full disclosure: I don't know much about HVAC and missed the step in
growing up where I was suppose to tear apart the wall thermostat to
see how it works. I've also played manager in my fathers apartment
building and rental landlord, where I've replaced such thermostats and
never needed or bothered to adjust any.

>One can get more or less hysteresis by deviating from this procedure.
>I have my Honeywell "Round" (with the mercury switch, installed in
>1970) tuned to about 3/4 degree room hysteresis. That's where I'm the
>most comfortable.
>
>My heat pump cycles 4-5 times an hour. It uses a scroll type
>compressor that has low inrush and starts easily. So my standby
>generator can power it, I've installed time delay relays so the
>compressor starts first, then the condenser cooling fan and finally
>the evaporator fan.

Ok, so you do have an external timing delay to prevent "thrashing". I
guess 12 to 15 mins cycle time is tolerable. At this point, I'm not
sure which is worse, a fan that cycles every 12 mins, or (for example)
a 5 degree variation in temperature over a longer period.

>All this had been worked out eletro-mechanically by at least the 50s.
>In the early 60s my family had a new-fangled heat pump installed.
>Being the kid with 1000 questions, I asked the installers about every
>aspect of the install including the anticipator setup.

I didn't ask, possibly because my parents considered my messing with
the house thermostat a capital crime. Other than clean out the dust,
I don't think we ever did anything with it for about 40 years.

>Back then, before the government was sucking down the economy, the
>HVAC company could afford to do a complete job for a reasonable price
>which including leaving a spring wound circular chart recorder in the
>house overnight to verify the anticipator setup and the thermostat
>calibration.

Yep. A very sad situation. When the cost of a truck roll exceeds the
value of the equipment being adjusted, the equipment soon becomes
uneconomical to service. We've been there for at least 20 years.

Thanks for the details.

[tabb...@gmail.com]

On Sunday, 16 August 2015 03:10:22 UTC+1, Jeff Liebermann wrote:
> On Sat, 15 Aug 2015 13:02:43 -0400, Neon John <n...@never.com> wrote:
> >On Fri, 14 Aug 2015 18:43:40 -0700, Jeff Liebermann <je...@cruzio.com>
> >wrote:
> >
> >
> >>Looks like the missing buzzword might be "heat anticipation
> >>thermostat".
> >
> >That is correct and the resistance wire is called the anticipator.
>
> Ok. Got it, maybe:
> <http://inspectapedia.com/heat/Heat_Anticipator_Adustment.php>
>
> What surprised me is this line from the above URL:
> "The anticipator also creates hysteresis, so the furnace does
> not cycle too often."
> You suggest that it reduces hysteresis which also what I would expect.
> So, is the web page in error?

You can have it either way. A bimetal with spring or magnet action has mechanical hysteresis, and the resistor reduces the hysteresis. A bimetal without spring or magnet action has no inbuilt hysteresis, but also near zero contact pressure at switching - they are used for some things.

> In addition, it mentions:
> "QP means by hysterisis that we introduce a lag or delay between
> the change in room temperature and the response of the thermostat.
> However the heat anticipator is only operating to turn the
> thermostat "off" a little "early". It does not introduce a delay
> in turning the thermostat back on. - Ed."
> I thought it worked in both directions?
>
> I still have a problem with reducing the hysteresis. You mention
> 0.75F hysteresis. I would think that even slight changes to air
> temperature would cause the thermostat to cycle.

about 0.75F no less. Don't forget in practice the bimetal is in a case attached to metal, so if you breathe on it once it won't respond.

> I would also think
> that a smaller hysteresis means a short run/off time, which might
> cause thrashing. Is 0.75F a reasonable value for hysteresis?

we use about 0.5C here

> Full disclosure: I don't know much about HVAC and missed the step in
> growing up where I was suppose to tear apart the wall thermostat to
> see how it works. I've also played manager in my fathers apartment
> building and rental landlord, where I've replaced such thermostats and
> never needed or bothered to adjust any.
>
> >One can get more or less hysteresis by deviating from this procedure.
> >I have my Honeywell "Round" (with the mercury switch, installed in
> >1970) tuned to about 3/4 degree room hysteresis. That's where I'm the
> >most comfortable.
> >
> >My heat pump cycles 4-5 times an hour. It uses a scroll type
> >compressor that has low inrush and starts easily. So my standby
> >generator can power it, I've installed time delay relays so the
> >compressor starts first, then the condenser cooling fan and finally
> >the evaporator fan.
>
> Ok, so you do have an external timing delay to prevent "thrashing". I

temperature hysteresis plus slow stat response is all you need

> guess 12 to 15 mins cycle time is tolerable. At this point, I'm not
> sure which is worse, a fan that cycles every 12 mins, or (for example)
> a 5 degree variation in temperature over a longer period.

5C or 5F is way too large, it would waste a pile of energy/money.

> >All this had been worked out eletro-mechanically by at least the 50s.
> >In the early 60s my family had a new-fangled heat pump installed.
> >Being the kid with 1000 questions, I asked the installers about every
> >aspect of the install including the anticipator setup.
>
> I didn't ask, possibly because my parents considered my messing with
> the house thermostat a capital crime. Other than clean out the dust,
> I don't think we ever did anything with it for about 40 years.
>
> >Back then, before the government was sucking down the economy, the
> >HVAC company could afford to do a complete job for a reasonable price
> >which including leaving a spring wound circular chart recorder in the
> >house overnight to verify the anticipator setup and the thermostat
> >calibration.
>
> Yep. A very sad situation. When the cost of a truck roll exceeds the
> value of the equipment being adjusted, the equipment soon becomes
> uneconomical to service. We've been there for at least 20 years.
>
> Thanks for the details.

Hang on. More people can afford ac now than in the 60s because the equipment is cheaper, and run cost lower. Same reason kit doesn't get serviced, its price has fallen, meaning we do better today. Its a shame to see so much kit get junked, but it is economic progress.


NT

[Jasen Betts]

On 2015-08-16, Jeff Liebermann <je...@cruzio.com> wrote:

> In addition, it mentions:
> "QP means by hysterisis that we introduce a lag or delay between
> the change in room temperature and the response of the thermostat.
> However the heat anticipator is only operating to turn the
> thermostat "off" a little "early". It does not introduce a delay
> in turning the thermostat back on. - Ed."
> I thought it worked in both directions?

it brings "off" closer to "on", technically that's only working in one
direction, but the net effect ix that at both on and off periods
will be reduced.

> I still have a problem with reducing the hysteresis. You mention
> 0.75F hysteresis. I would think that even slight changes to air
> temperature would cause the thermostat to cycle. I would also think
> that a smaller hysteresis means a short run/off time, which might
> cause thrashing. Is 0.75F a reasonable value for hysteresis?

thermostats have themal mass, it'll take a minute or so to
react to a 1 degree change.

--
\_(ツ)_

[Joe Gwinn]

In article <96kssa12jfbl6v3i6...@4ax.com>, Jeff

Liebermann <je...@cruzio.com> wrote:

> On Fri, 14 Aug 2015 11:59:50 -0700 (PDT), tabb...@gmail.com wrote:
>
> >> How does one compensate for mechanical hysteresis?
>
> >A little heater resistor comes on, raising the bimetal by about
> >the same temp rise as the hysteresis. The result is a fraction
> >of a degree hysteresis.
>
> On this side of the planet, we don't have those. Mostly, we have
> setback thermostats like these:
> <https://www.google.com/search?q=setback+thermostat&tbm=isch>
>
> I don't see how the little heater resistor trick would work. What
> turns on the little heater before the main contacts close? When going
> from off to on, the contacts haven't closed yet, so there's nothing to
> trigger the little heater. When going from on -> off the little
> heater is presumably on, so the little heater resistor would need some
> way to predict that the contacts are about to open and turn off early.

There is a hint: Those little heaters are called anticipators.

The purpose is not to eliminate hysteresis, it is to roughly model the
thermal inertia of the room, to reduce overshoot.

I would look into the old patents on the Honeywell round thermostat,
the ones with bimetal spiral plus mercury switch (which are still in
use in my home).


Joe Gwinn

[tabb...@gmail.com]

On Monday, 17 August 2015 15:49:14 UTC+1, Joe Gwinn wrote:
> In article <96kssa12jfbl6v3i6...@4ax.com>, Jeff
> Liebermann <je...@cruzio.com> wrote:

> > On Fri, 14 Aug 2015 11:59:50 -0700 (PDT), nt wrote:
> >
> > >> How does one compensate for mechanical hysteresis?
> >
> > >A little heater resistor comes on, raising the bimetal by about
> > >the same temp rise as the hysteresis. The result is a fraction
> > >of a degree hysteresis.
> >
> > On this side of the planet, we don't have those. Mostly, we have
> > setback thermostats like these:
> > <https://www.google.com/search?q=setback+thermostat&tbm=isch>
> >
> > I don't see how the little heater resistor trick would work. What
> > turns on the little heater before the main contacts close? When going
> > from off to on, the contacts haven't closed yet, so there's nothing to
> > trigger the little heater. When going from on -> off the little
> > heater is presumably on, so the little heater resistor would need some
> > way to predict that the contacts are about to open and turn off early.
>
> There is a hint: Those little heaters are called anticipators.
>
> The purpose is not to eliminate hysteresis, it is to roughly model the
> thermal inertia of the room, to reduce overshoot.

You'd need a lot of thermal inertia inside the stat to get it to warm up from the compensation R as slowly as a room.



NT

[tabb...@gmail.com]

On Saturday, 15 August 2015 11:06:17 UTC+1, nt wrote:
> On Friday, 14 August 2015 22:14:09 UTC+1, Jeff Liebermann wrote:
> > On Fri, 14 Aug 2015 11:59:50 -0700 (PDT), nt wrote:

> > >> Do you have an example of a compensated thermostat without any or very
> > >> little hysteresis?
> > >
> > >every British central heating stat that hasn't been replaced with an electronic one, eg
> > >http://www.screwfix.com/c/heating-plumbing/central-heating-controls/cat831444?cm_sp=managedredirect-_-plumbing-_-HeatControls
> > >includes three
> >
> > Those are all electronic thermostats that use thermistors or
> > thermocouples. What I would like to see is the guts of a mechanical
> > bimetallic thermostat with the little heater resistor trick. I tried
> > Googling for such a device, but couldn't find anything, probably
> > because I'm missing a UK specific buzzword. What are they called?
>
> Google isn't showing compensated ones. I'll see if I can get a pic.

http://i59.tinypic.com/34xi3yv.jpg
http://i59.tinypic.com/jkbs7r.jpg
http://i60.tinypic.com/2vhu5vb.jpg


NT

[Shaun]

"Michael" wrote in message
news:4337e6f4-e9ea-4b20...@googlegroups.com...

I am looking for a thermostat (e.g.:
http://www.ngtcontrols.com/gen-automaticreset.html) for basic on-off
temperature control. I noticed that specified service life never exceeds
100k cycles. I wonder whether it is due to contacts burnout or bimetallic
strip fatigue. Does anybody have any insight?
Thanks.
Michael

MAN! That technology has probably been around since to 30's, and your
asking questions like that, it's not even a controller even though they call
it one. Replace it with a thyristor circuit and a temperature sensor so
that it really is a controller! This is 2015! You could get fancy and skip
the temp sensor and have a microcontroller monitor turn on current of the
heating element, with that you could determine the temperature by the
resistance of the heating element and how fast it reaches full temperature /
minimum current.

Shaun

[DecadentLinuxUserNumeroUno]

On Wed, 19 Aug 2015 23:33:00 -0500, "Shaun" <stereo...@gmail.com>
Gave us:

You're right. $23 is pretty damned cheap...

http://www.amazon.com/AGPtek%C2%AE-Universal-Temperature-Controller-Fahrenheit/dp/B007MMOEWY

[tabb...@gmail.com]

Actually its an 1800s invention, 1830s & 1880s. Well over a century of technological development has given us less reliable electronic stats.


NT

[DecadentLinuxUserNumeroUno]

On Thu, 20 Aug 2015 04:27:44 -0700 (PDT), tabb...@gmail.com Gave us:

Well, that was a boredom tube post.

[krw]

If reliability were the end-all, we'd still be writing on stone
tablets.