Latching relays

[Phil Hobbs]

On 9/24/19 9:30 AM, Michael Terrell wrote:

On Tuesday, September 24, 2019 at 9:00:06 AM UTC-4, pcdh...@gmail.com wrote:
   I am currently designing a replacement 'A23' Attenuator board for
the HP3325A/B function generators. The original relays are long obsolete
so rather than making adapter boards, I am designing a smaller board that
uses modern relays.

If you're getting a bunch made, I'd chip in for three of them. That's the only thing that seems to go bad on those things. So far Deoxit has always fixed it, but I expect it's just a matter of time.

Cheers

Phil Hobbs


I need at least three for myself. I was looking at making ten boards, with a choice of Phono or SMA connectors. If you have a failed board you can reuse the resistors, but I am planning to use SMD for complete boards. I may also make a replacement for the HV output board, and the 10 MHz OCXO as well. I have a pile of them that are missing one or more of those three boards.

It's not a hugely complicated board, so we can stuff them ourselves if
needed.

Thanks

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Michael Terrell]

On Wednesday, September 25, 2019 at 9:30:54 AM UTC-4, Phil Hobbs wrote:

On 9/24/19 9:30 AM, Michael Terrell wrote:
On Tuesday, September 24, 2019 at 9:00:06 AM UTC-4, pcdh...@gmail.com wrote:
   I am currently designing a replacement 'A23' Attenuator board for
the HP3325A/B function generators. The original relays are long obsolete
so rather than making adapter boards, I am designing a smaller board that
uses modern relays.

If you're getting a bunch made, I'd chip in for three of them. That's the only thing that seems to go bad on those things. So far Deoxit has always fixed it, but I expect it's just a matter of time.

Cheers

Phil Hobbs


I need at least three for myself. I was looking at making ten boards, with a choice of Phono or SMA connectors. If you have a failed board you can reuse the resistors, but I am planning to use SMD for complete boards. I may also make a replacement for the HV output board, and the 10 MHz OCXO as well. I have a pile of them that are missing one or more of those three boards.


It's not a hugely complicated board, so we can stuff them ourselves if
needed.

You're correct. It is just a custom DC to 20 MHZ three stage attenuator with latching relays. There is no need for >$500 modules and an interface board. I ran across a deal on 1000 latching Telcom relays in a sealed factory carton. That will allow the construction of up to 250 boards at a reasonable price. The SMA connectors would be edge launched, and only used to add those rear panel connectors, if needed. Otherwise. it will be a drop in replacement with a smaller footprint and possible a shield over the module that HP didn't have.

People who have never used a 3325A/B don't know what they are missing! Both have IEEE-488 interfaces, but the B model also has RS232 which makes it easy to program without expensive interface boards. It's too bad they were designed before Ethernet control was common.

 

[Jeroen Belleman]

Michael Terrell wrote:

On Wednesday, September 25, 2019 at 9:30:54 AM UTC-4, Phil Hobbs
wrote:
On 9/24/19 9:30 AM, Michael Terrell wrote:
On Tuesday, September 24, 2019 at 9:00:06 AM UTC-4,
pcdh...@gmail.com wrote:
I am currently designing a replacement 'A23' Attenuator
board for the HP3325A/B function generators. The original
relays are long obsolete so rather than making adapter
boards, I am designing a smaller board that uses modern
relays.
[Snip!]

People who have never used a 3325A/B don't know what they are
missing! Both have IEEE-488 interfaces, but the B model also has
RS232 which makes it easy to program without expensive interface
boards. It's too bad they were designed before Ethernet control
was common.

If you really want that, you could add a Prologix GPIB to
Ethernet Controller. I'm very happy with that little gadget.

Jeroen Belleman

 

[Rick C]

On Wednesday, September 25, 2019 at 6:13:05 AM UTC-4, pozz wrote:

Il 25/09/2019 05:20, jjhudak4@gmail.com ha scritto:
On Tuesday, September 24, 2019 at 9:40:30 AM UTC-4, pozz wrote:
Il 24/09/2019 14:51, Michael Terrell ha scritto:
On Tuesday, September 24, 2019 at 8:06:18 AM UTC-4, pozz wrote:
Il 24/09/2019 13:03, Michael Terrell ha scritto:
On Tuesday, September 24, 2019 at 3:04:22 AM UTC-4, pozz wrote:
I have to use a latching relay with two coils, exactly TQ2-L2-24V[1].

Is it possible to give power to only one coil forever, as in a "normal"
relay, or is it better to remove power from coils after a short time?

I don't have critical restrictions on power consumption, so I'm going to
maintain one coil active forever (until power is cut off).
Are there any drawback on this approach?


[1] https://www.mouser.it/datasheet/2/315/mech_eng_tq-1299280.pdf

Why use a latching relay, if you plan on keeping it energized?


Thanks for the answer, even if it doesn't answer my question.

It's very simple, because I want a stable condition when the power is
cut off. Latching relays guarantees that the contacts don't change their
position when the mains power goes down.

My point was, there is no reason to continue to power a latching relay.

I have two reasons.

The first is the software that must be more complex to generate an
impulse instead of setting/resetting an output pin level forever.

I know, it's simple to create a pulse too with a blocking delay:

activate_coil1();
delay_ms(10);
deactivate_coil1();

however in my software, that is cooperative multitasking, I can't block
for more than 100usec. I need to implement a small state-machine to
generate a pulse.

The other reason is: how long should be the impulse? Datasheet says
maximum 3ms, just to be sure I would generate a 10ms pulse. Could I be
sure the relay has switching after my pulse? I don't have any feedback
from relay.

In my application it's much more important to be sure that the relay has
switched instead of saving some power.


Look at the current draw, vs a non latching version. It takes more power to set or reset a mechanical latch, It doesn't matter if the relay is single or double coil.

8.3mA vs 12.5mA

Why waste power and heat > up the relay? That will shorten its useful life.

It's a pity the datasheet says nothing about expected life vs coil
activation time.


I am currently designing a replacement 'A23' Attenuator board for the HP3325A/B function generators. The original relays are long obsolete so rather than making adapter boards, I am designing a smaller board that uses modern relays. It will also allow you to select the supply voltage for the relays, to allow the use of more common voltages.


Ummmm, this operation is very strange. What CPU are you using? Almost all of them have a timer capability. Write a '1' to the DO port associated with the relay and at the same time, set time to count down x ms. The ISR associated with the timer can write a '0' to the DO port. Doesn't matter if you have an RTOS, a roll your own task scheduler, or a cyclic exec or a big loop. or, make the entire system event driven. No hw timer on the cpu? use the timer or RTC function associated with the RTOS. Busy-waits are inefficient - there is almost no reason to use them. Also, a loop with a gazillion no-ops to implement time related events makes for non-portable code..
Sorry, but your reasons are invalid.

I didn't say I can't implement a software that generates a pulse, I said
that this implementation is more complex compared to set/reset a pin.
I know how to code a software that generates a pulse, however I like to
keep the code simple if possible (bugs are everywhere).

If I understand what you want to do is to run the latching relay as if it were a non-latching relay. Whatever state you want it to be in is set in the software and held rather than pulsed. I think at least one poster thinks you want to energize one coil all the time and pulse the other coil.

If the relay won't overheat this should work. I do get your point, but I would vote for the pulsed operation myself.

--

Rick C.

- Get 2,000 miles of free Supercharging
- Tesla referral code - https://ts.la/richard11209

 

[pozz]

Il 25/09/2019 17:16, Rick C ha scritto:

On Wednesday, September 25, 2019 at 6:13:05 AM UTC-4, pozz wrote:
Il 25/09/2019 05:20, jjhudak4@gmail.com ha scritto:
On Tuesday, September 24, 2019 at 9:40:30 AM UTC-4, pozz wrote:
Il 24/09/2019 14:51, Michael Terrell ha scritto:
On Tuesday, September 24, 2019 at 8:06:18 AM UTC-4, pozz wrote:
Il 24/09/2019 13:03, Michael Terrell ha scritto:
On Tuesday, September 24, 2019 at 3:04:22 AM UTC-4, pozz wrote:
I have to use a latching relay with two coils, exactly TQ2-L2-24V[1].

Is it possible to give power to only one coil forever, as in a "normal"
relay, or is it better to remove power from coils after a short time?

I don't have critical restrictions on power consumption, so I'm going to
maintain one coil active forever (until power is cut off).
Are there any drawback on this approach?


[1] https://www.mouser.it/datasheet/2/315/mech_eng_tq-1299280.pdf

Why use a latching relay, if you plan on keeping it energized?


Thanks for the answer, even if it doesn't answer my question.

It's very simple, because I want a stable condition when the power is
cut off. Latching relays guarantees that the contacts don't change their
position when the mains power goes down.

My point was, there is no reason to continue to power a latching relay.

I have two reasons.

The first is the software that must be more complex to generate an
impulse instead of setting/resetting an output pin level forever.

I know, it's simple to create a pulse too with a blocking delay:

activate_coil1();
delay_ms(10);
deactivate_coil1();

however in my software, that is cooperative multitasking, I can't block
for more than 100usec. I need to implement a small state-machine to
generate a pulse.

The other reason is: how long should be the impulse? Datasheet says
maximum 3ms, just to be sure I would generate a 10ms pulse. Could I be
sure the relay has switching after my pulse? I don't have any feedback
from relay.

In my application it's much more important to be sure that the relay has
switched instead of saving some power.


Look at the current draw, vs a non latching version. It takes more power to set or reset a mechanical latch, It doesn't matter if the relay is single or double coil.

8.3mA vs 12.5mA

Why waste power and heat > up the relay? That will shorten its useful life.

It's a pity the datasheet says nothing about expected life vs coil
activation time.


I am currently designing a replacement 'A23' Attenuator board for the HP3325A/B function generators. The original relays are long obsolete so rather than making adapter boards, I am designing a smaller board that uses modern relays. It will also allow you to select the supply voltage for the relays, to allow the use of more common voltages.


Ummmm, this operation is very strange. What CPU are you using? Almost all of them have a timer capability. Write a '1' to the DO port associated with the relay and at the same time, set time to count down x ms. The ISR associated with the timer can write a '0' to the DO port. Doesn't matter if you have an RTOS, a roll your own task scheduler, or a cyclic exec or a big loop. or, make the entire system event driven. No hw timer on the cpu? use the timer or RTC function associated with the RTOS. Busy-waits are inefficient - there is almost no reason to use them. Also, a loop with a gazillion no-ops to implement time related events makes for non-portable code.
Sorry, but your reasons are invalid.

I didn't say I can't implement a software that generates a pulse, I said
that this implementation is more complex compared to set/reset a pin.
I know how to code a software that generates a pulse, however I like to
keep the code simple if possible (bugs are everywhere).

If I understand what you want to do is to run the latching relay as if it were a non-latching relay. Whatever state you want it to be in is set in the software and held rather than pulsed.

Yes.

> I think at least one poster thinks you want to energize one coil all the time and pulse the other coil.

Absolutely no.

If the relay won't overheat this should work. I do get your point, but I would vote for the pulsed operation myself.

Indeed the question is exactly this: is there any drawback to keep one
coil of a latching relay energized?
We all don't care about this with non latching relay, we keep energized
the single coil to maintain the NO contact closed to COM.

 

[Dimitrij Klingbeil]

On 2019-09-25 17:36, pozz wrote:

Il 25/09/2019 17:16, Rick C ha scritto:
On Wednesday, September 25, 2019 at 6:13:05 AM UTC-4, pozz wrote:
Il 25/09/2019 05:20, jjhudak4@gmail.com ha scritto:
On Tuesday, September 24, 2019 at 9:40:30 AM UTC-4, pozz
wrote:
Il 24/09/2019 14:51, Michael Terrell ha scritto:
On Tuesday, September 24, 2019 at 8:06:18 AM UTC-4, pozz
wrote:
Il 24/09/2019 13:03, Michael Terrell ha scritto:
On Tuesday, September 24, 2019 at 3:04:22 AM UTC-4,
pozz wrote:
I have to use a latching relay with two coils,
exactly TQ2-L2-24V[1].

Is it possible to give power to only one coil
forever, as in a "normal" relay, or is it better to
remove power from coils after a short time?

I don't have critical restrictions on power
consumption, so I'm going to maintain one coil
active forever (until power is cut off). Are there
any drawback on this approach?


[1]
https://www.mouser.it/datasheet/2/315/mech_eng_tq-1299280.pdf




Why use a latching relay, if you plan on keeping it energized?


Thanks for the answer, even if it doesn't answer my
question.

It's very simple, because I want a stable condition when
the power is cut off. Latching relays guarantees that
the contacts don't change their position when the mains
power goes down.

My point was, there is no reason to continue to power a
latching relay.

I have two reasons.

The first is the software that must be more complex to
generate an impulse instead of setting/resetting an output
pin level forever.

I know, it's simple to create a pulse too with a blocking
delay:

activate_coil1(); delay_ms(10); deactivate_coil1();

however in my software, that is cooperative multitasking, I
can't block for more than 100usec. I need to implement a
small state-machine to generate a pulse.

The other reason is: how long should be the impulse?
Datasheet says maximum 3ms, just to be sure I would generate
a 10ms pulse. Could I be sure the relay has switching after
my pulse? I don't have any feedback from relay.

In my application it's much more important to be sure that
the relay has switched instead of saving some power.


Look at the current draw, vs a non latching version. It
takes more power to set or reset a mechanical latch, It
doesn't matter if the relay is single or double coil.

8.3mA vs 12.5mA

Why waste power and heat > up the relay? That will
shorten its useful life.

It's a pity the datasheet says nothing about expected life
vs coil activation time.


I am currently designing a replacement 'A23' Attenuator
board for the HP3325A/B function generators. The original
relays are long obsolete so rather than making adapter
boards, I am designing a smaller board that uses modern
relays. It will also allow you to select the supply
voltage for the relays, to allow the use of more common
voltages.


Ummmm, this operation is very strange. What CPU are you using?
Almost all of them have a timer capability. Write a '1' to the
DO port associated with the relay and at the same time, set
time to count down x ms. The ISR associated with the timer
can write a '0' to the DO port. Doesn't matter if you have an
RTOS, a roll your own task scheduler, or a cyclic exec or a
big loop. or, make the entire system event driven. No hw
timer on the cpu? use the timer or RTC function associated with
the RTOS. Busy-waits are inefficient - there is almost no
reason to use them. Also, a loop with a gazillion no-ops to
implement time related events makes for non-portable code.
Sorry, but your reasons are invalid.

I didn't say I can't implement a software that generates a
pulse, I said that this implementation is more complex compared
to set/reset a pin. I know how to code a software that generates
a pulse, however I like to keep the code simple if possible
(bugs are everywhere).

If I understand what you want to do is to run the latching relay
as if it were a non-latching relay. Whatever state you want it to
be in is set in the software and held rather than pulsed.

Yes.

I think at least one poster thinks you want to energize one coil
all the time and pulse the other coil.

Absolutely no.


If the relay won't overheat this should work. I do get your
point, but I would vote for the pulsed operation myself.


Indeed the question is exactly this: is there any drawback to keep
one coil of a latching relay energized? We all don't care about this
with non latching relay, we keep energized the single coil to
maintain the NO contact closed to COM.

There may be an issue concerning long term reliability that would only
manifest itself under unpowered conditions because the magnets degrade.

A latching relay typically has 2 magnetic circuits, each one backed by
its own permanent magnet. Depending on the position of the contact arm,
one of the magnetic circuits is open while the other is closed. None of
them is permanently magnetically reverse-biased however in normal use.

Even when one magnetic circuit is closed, the other (open) one is still
only seeing an air gap, but no external reverse-biasing magnetic field
because the magnets are mounted in a "series like" and not in an
"anti-series like" configuration when considering both circuits
together. Furthermore, even the "open" magnetic circuit still sees a
significant flux through the small air gap that amounts to a "forward" bias.

However, when the coil is energized, it will create a magnetic field
that is stronger than that of either permanent magnet, because it needs
to actively "unlatch" the contact arm from a magnetic circuit that was
closed, before it can set it moving towards the opposite position.

With a higher strength field from the coil applied, one magnetic circuit
will be reinforced, but the other (now open) one will be reverse biased,
even if this reverse bias would have to pass through the open air gap.

Now, commonly available "permanent" magnets are not really permanent
when considered over a lifespan of decades. While they will always
slowly degrade by themselves, the state of the magnetic circuit does
make a difference: a permanent magnet in closed magnetic circuit will
hold up very well and degrade only slowly, while in an open magnetic
circuit it will degrade (demagnetize) faster. However, when reverse-
biased externally, the degradation may speed up rather significantly,
depending on how strong the reverse biasing external field is.

An additional effect can result from increased self-heating of the coil
with higher temperatures accelerating the permanent magnet degradation,
especially for a magnet that is under reverse biasing conditions.

In a normally operated latching relay this state is of no long term
concern because in the passive state neither of the 2 magnets is ever
reverse biased and the coil only operates in short pulses, typically
alternating polarities each time, so it would not contribute any long
term reverse bias in any direction either.

If driven all the time however, the coil can have an influence on the
long term reliability of the permanent magnets. Unbalanced, one magnet
would likely age faster than the other, slowly creating a relay that
does not "hold up" well in the opposite position. Unless very strongly
degraded, the relay may not fail outright, but it can become sensitive
to shock and vibration, preferring to easily flip into one direction.

To avoid the relay getting "strange" (sometimes liking to flip for no
good reason) when turned off, it's better to not drive it permanently.

At least, consider reducing the drive current after a short on time.

 

[Phil Hobbs]

On 9/24/19 12:17 PM, Michael Terrell wrote:

On Tuesday, September 24, 2019 at 9:40:30 AM UTC-4, pozz wrote:
Il 24/09/2019 14:51, Michael Terrell ha scritto:
On Tuesday, September 24, 2019 at 8:06:18 AM UTC-4, pozz wrote:
Il 24/09/2019 13:03, Michael Terrell ha scritto:
On Tuesday, September 24, 2019 at 3:04:22 AM UTC-4, pozz wrote:
I have to use a latching relay with two coils, exactly TQ2-L2-24V[1].

Is it possible to give power to only one coil forever, as in a "normal"
relay, or is it better to remove power from coils after a short time?

I don't have critical restrictions on power consumption, so I'm going to
maintain one coil active forever (until power is cut off).
Are there any drawback on this approach?


[1] https://www.mouser.it/datasheet/2/315/mech_eng_tq-1299280.pdf

Why use a latching relay, if you plan on keeping it energized?


Thanks for the answer, even if it doesn't answer my question.

It's very simple, because I want a stable condition when the power is
cut off. Latching relays guarantees that the contacts don't change their
position when the mains power goes down.

My point was, there is no reason to continue to power a latching relay.

I have two reasons.

The first is the software that must be more complex to generate an
impulse instead of setting/resetting an output pin level forever.

I know, it's simple to create a pulse too with a blocking delay:

activate_coil1();
delay_ms(10);
deactivate_coil1();

however in my software, that is cooperative multitasking, I can't block
for more than 100usec. I need to implement a small state-machine to
generate a pulse.


Look at 'Pulse stretcher' circuits.


The other reason is: how long should be the impulse? Datasheet says
maximum 3ms, just to be sure I would generate a 10ms pulse. Could I be
sure the relay has switching after my pulse? I don't have any feedback
from relay.


Either you trust the relay, or you find another way to design the item. If the datasheet says the maximum is 3ms, you don't need 10ms.


In my application it's much more important to be sure that the relay has
switched instead of saving some power.


Look at the current draw, vs a non latching version. It takes more power to set or reset a mechanical latch, It doesn't matter if the relay is single or double coil.

8.3mA vs 12.5mA

Why waste power and heat > up the relay? That will shorten its useful life.


Implement the timing in the driver circuit. How cay you trigger the second coil. if you are still powering the first coil?

Connect the coils in series between the supply and ground, and connect a
totem-pole driver to the midpoint.

It's a pity the datasheet says nothing about expected life vs coil
activation time.

A parallel RC network in series with each coil is the usual way to
reduce dissipation while still having reliable pull-in.

One caution: in reading a relay datasheet it often seems as though
there's a whole lot of slop between the guaranteed pull-in conditions
and the rated coil voltage. Resist the temptation to reduce the
voltage--the RC gets you the full voltage during pull-in and reduced
dissipation in the hold condition. With a one-coil latching relay you
can in principle use just the capacitor--I did that in the bootstrapped
1G/50G TIA I posted in the driven-switch-body thread:

<https://electrooptical.net/www/sed/QceptTransimpedanceAmpRev1.pdf>

Cheers

Phil Hobbs



--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[pozz]

Il 25/09/2019 21:02, Dimitrij Klingbeil ha scritto:

On 2019-09-25 17:36, pozz wrote:
Il 25/09/2019 17:16, Rick C ha scritto:
On Wednesday, September 25, 2019 at 6:13:05 AM UTC-4, pozz wrote:
Il 25/09/2019 05:20, jjhudak4@gmail.com ha scritto:
On Tuesday, September 24, 2019 at 9:40:30 AM UTC-4, pozz
wrote:
Il 24/09/2019 14:51, Michael Terrell ha scritto:
On Tuesday, September 24, 2019 at 8:06:18 AM UTC-4, pozz
wrote:
Il 24/09/2019 13:03, Michael Terrell ha scritto:
On Tuesday, September 24, 2019 at 3:04:22 AM UTC-4,
pozz wrote:
I have to use a latching relay with two coils,
exactly TQ2-L2-24V[1].

Is it possible to give power to only one coil
forever, as in a "normal" relay, or is it better to
remove power from coils after a short time?

I don't have critical restrictions on power
consumption, so I'm going to maintain one coil
active forever (until power is cut off). Are there
any drawback on this approach?


[1]
https://www.mouser.it/datasheet/2/315/mech_eng_tq-1299280.pdf




Why use a latching relay, if you plan on keeping it energized?

Because I need to maintain the last position when the power is cut off.

Thanks for the answer, even if it doesn't answer my
question.

It's very simple, because I want a stable condition when
the power is cut off. Latching relays guarantees that
the contacts don't change their position when the mains
power goes down.

My point was, there is no reason to continue to power a
latching relay.

I have two reasons.

The first is the software that must be more complex to
generate an impulse instead of setting/resetting an output
pin level forever.

I know, it's simple to create a pulse too with a blocking
delay:

activate_coil1(); delay_ms(10); deactivate_coil1();

however in my software, that is cooperative multitasking, I
can't block for more than 100usec. I need to implement a
small state-machine to generate a pulse.

The other reason is: how long should be the impulse?
Datasheet says maximum 3ms, just to be sure I would generate
a 10ms pulse. Could I be sure the relay has switching after
my pulse? I don't have any feedback from relay.

In my application it's much more important to be sure that
the relay has switched instead of saving some power.


Look at the current draw, vs a non latching version. It
takes more power to set or reset a mechanical latch, It
doesn't matter if the relay is single or double coil.

8.3mA vs 12.5mA

Why waste power and heat  > up the relay? That will
shorten its useful life.

It's a pity the datasheet says nothing about expected life
vs coil activation time.


I am currently designing a replacement 'A23' Attenuator
board for the HP3325A/B function generators. The original
relays are long obsolete so rather than making adapter
boards, I am designing a smaller board that uses modern
relays. It will also allow you to select the supply
voltage for the  relays, to allow the use of more common
voltages.


Ummmm, this operation is very strange. What CPU are you using?
Almost all of them have a timer capability.  Write a '1' to the
DO port associated with the relay and at the same time, set
time to count down  x ms.  The ISR associated with the timer
can write a '0' to the DO port.  Doesn't matter if you have an
RTOS, a roll your own task scheduler, or a cyclic exec or a
big loop.  or, make the entire system event driven.  No hw
timer on the cpu? use the timer or RTC function associated with
the RTOS.  Busy-waits are inefficient - there is almost no
reason to use them. Also, a loop with a gazillion no-ops to
implement time related events makes for non-portable code.
Sorry, but your reasons are invalid.

I didn't say I can't implement a software that generates a
pulse, I said that this implementation is more complex compared
to set/reset a pin. I know how to code a software that generates
a pulse, however I like to keep the code simple if possible
(bugs are everywhere).

If I understand what you want to do is to run the latching relay
as if it were a non-latching relay.  Whatever state you want it to
be in is set in the software and held rather than pulsed.

Yes.

I think at least one poster thinks you want to energize one coil
all the time and pulse the other coil.

Absolutely no.


If the relay won't overheat this should work.  I do get your
point, but I would vote for the pulsed operation myself.


Indeed the question is exactly this: is there any drawback to keep
one coil of a latching relay energized? We all don't care about this
with non latching relay, we keep energized the single coil to
maintain the NO contact closed to COM.

There may be an issue concerning long term reliability that would only
manifest itself under unpowered conditions because the magnets degrade.

A latching relay typically has 2 magnetic circuits, each one backed by
its own permanent magnet. Depending on the position of the contact arm,
one of the magnetic circuits is open while the other is closed. None of
them is permanently magnetically reverse-biased however in normal use.

Even when one magnetic circuit is closed, the other (open) one is still
only seeing an air gap, but no external reverse-biasing magnetic field
because the magnets are mounted in a "series like" and not in an
"anti-series like" configuration when considering both circuits
together. Furthermore, even the "open" magnetic circuit still sees a
significant flux through the small air gap that amounts to a "forward"
bias.

However, when the coil is energized, it will create a magnetic field
that is stronger than that of either permanent magnet, because it needs
to actively "unlatch" the contact arm from a magnetic circuit that was
closed, before it can set it moving towards the opposite position.

With a higher strength field from the coil applied, one magnetic circuit
will be reinforced, but the other (now open) one will be reverse biased,
even if this reverse bias would have to pass through the open air gap.

Now, commonly available "permanent" magnets are not really permanent
when considered over a lifespan of decades. While they will always
slowly degrade by themselves, the state of the magnetic circuit does
make a difference: a permanent magnet in closed magnetic circuit will
hold up very well and degrade only slowly, while in an open magnetic
circuit it will degrade (demagnetize) faster. However, when reverse-
biased externally, the degradation may speed up rather significantly,
depending on how strong the reverse biasing external field is.

An additional effect can result from increased self-heating of the coil
with higher temperatures accelerating the permanent magnet degradation,
especially for a magnet that is under reverse biasing conditions.

In a normally operated latching relay this state is of no long term
concern because in the passive state neither of the 2 magnets is ever
reverse biased and the coil only operates in short pulses, typically
alternating polarities each time, so it would not contribute any long
term reverse bias in any direction either.

If driven all the time however, the coil can have an influence on the
long term reliability of the permanent magnets. Unbalanced, one magnet
would likely age faster than the other, slowly creating a relay that
does not "hold up" well in the opposite position. Unless very strongly
degraded, the relay may not fail outright, but it can become sensitive
to shock and vibration, preferring to easily flip into one direction.

To avoid the relay getting "strange" (sometimes liking to flip for no
good reason) when turned off, it's better to not drive it permanently.

At least, consider reducing the drive current after a short on time.

Thank you for the explanation

 

[Piotr Wyderski]

Phil Hobbs wrote:

With a one-coil latching relay you can in principle use just the capacitor--I did that in the bootstrapped
1G/50G TIA I posted in the driven-switch-body thread:

Really neat, but why would you do that if the two-coil versions cost
basically the same and the ready-made relay driver chips (SZNUD3124DMT1G
for instance) are so tiny and so cheap?

Best regards, Piotr

 

Really neat, but why would you do that if the two-coil versions cost
basically the same and the ready-made relay driver chips (SZNUD3124DMT1G
for instance) are so tiny and so cheap

Capacitance. I needed to bootstrap the coil-to-contact capacitance (0.4 pF), because it would have trashed the bandwidth.

I also had to short out the 1G resistor when the amp was in 50G mode--otherwise enough of its Johnson noise would have got through the open contacts (0.2 pF) to dominate the noise floor.

The instrument was a scanning surface voltage tool, which used a 100-200 um diameter probe hovering a few tens of microns above a spinning wafer to detect sub-monolayer contamination by the change in the surface Fermi level.

(I didn't invent the technique, just the preamp.)

Cheers

Phil Hobbs

 

pcdhobbs@gmail.com wrote in
news:4c1160b6-740e-4810-b0a1-9c6f2c530ea1@googlegroups.com:

Really neat, but why would you do that if the two-coil versions
cost basically the same and the ready-made relay driver chips
(SZNUD3124DMT1G for instance) are so tiny and so cheap

Capacitance. I needed to bootstrap the coil-to-contact capacitance
(0.4 pF), because it would have trashed the bandwidth.

I also had to short out the 1G resistor when the amp was in 50G
mode--otherwise enough of its Johnson noise would have got through
the open contacts (0.2 pF) to dominate the noise floor.

The instrument was a scanning surface voltage tool, which used a
100-200 um diameter probe hovering a few tens of microns above a
spinning wafer to detect sub-monolayer contamination by the change
in the surface Fermi level.

(I didn't invent the technique, just the preamp.)

Cheers

Phil Hobbs

Like a smart, sensitive hall effect sensor?

So instead of checking individual chips, the entire wafer spins
under the detector and it senses contaminants at the micron level.
That is really cool. Does it spiral out like a record album?

If so, it could probably be set up to make an audio response to the
surface differences. The disco DJs would love it. Take an old wafer
and then even fingerprints could register a sound. ten micron hover
gap is pretty tight though.

 

[Piotr Wyderski]

pcdhobbs@gmail.com wrote:

> Capacitance. I needed to bootstrap the coil-to-contact capacitance (0.4 pF), because it would have trashed the bandwidth.

Whoa, THAT is something I have not considered! Appreciated, Phil,
another goodie to know.

Best regards, Piotr

 

[Winfield Hill]

pcdhobbs@gmail.com wrote...

Capacitance. I needed to bootstrap the coil-to-contact
capacitance (0.4 pF), because it would have trashed the bandwidth.

What were the bandwidths for the two gains?

I also had to short out the 1G resistor when the amp was in 50G
mode --otherwise enough of its Johnson noise would have got
through the open contacts (0.2 pF) to dominate the noise floor.

Can you give us the calculation behind that? If the 50G
bandwidth was modest (2Hz for 0.15pF self capacitance?),
how did the 0.2pF noise feed-path figure in?


--
Thanks,
- Win

 

[Winfield Hill]

Piotr Wyderski wrote...

pcdhobbs@gmail.com wrote:

Capacitance. I needed to bootstrap the coil-to-contact
capacitance (0.4 pF), because it would have trashed the bandwidth.

Whoa, THAT is something I have not considered!
Appreciated, Phil, another goodie to know.

We're often struggling with a limit set by the
feedback resistor's self-capacitance, which is
from 0.06 to 0.15pF depending on various things.
And if we're sufficiently determined, we can
force the effect of the capacitance down by
another factor of 10 or so, by using the trick
in AoE3, Figure 8.80.C, page 545.


--
Thanks,
- Win

 

[Phil Hobbs]

On 9/26/19 8:15 AM, DecadentLinuxUserNumeroUno@decadence.org wrote:

pcdhobbs@gmail.com wrote in
news:4c1160b6-740e-4810-b0a1-9c6f2c530ea1@googlegroups.com:

Really neat, but why would you do that if the two-coil versions
cost basically the same and the ready-made relay driver chips
(SZNUD3124DMT1G for instance) are so tiny and so cheap

Capacitance. I needed to bootstrap the coil-to-contact capacitance
(0.4 pF), because it would have trashed the bandwidth.

I also had to short out the 1G resistor when the amp was in 50G
mode--otherwise enough of its Johnson noise would have got through
the open contacts (0.2 pF) to dominate the noise floor.

The instrument was a scanning surface voltage tool, which used a
100-200 um diameter probe hovering a few tens of microns above a
spinning wafer to detect sub-monolayer contamination by the change
in the surface Fermi level.

(I didn't invent the technique, just the preamp.)

Cheers

Phil Hobbs


Like a smart, sensitive hall effect sensor?

So instead of checking individual chips, the entire wafer spins
under the detector and it senses contaminants at the micron level.
That is really cool. Does it spiral out like a record album?

Yes, exactly. (CDs spiral out, records spiral in.)

If so, it could probably be set up to make an audio response to the
surface differences. The disco DJs would love it. Take an old wafer
and then even fingerprints could register a sound. ten micron hover
gap is pretty tight though.

It would be pretty slow for audio--maybe put its output into an audio VCO.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Phil Hobbs]

On 9/26/19 8:47 AM, Winfield Hill wrote:

pcdhobbs@gmail.com wrote...

Capacitance. I needed to bootstrap the coil-to-contact
capacitance (0.4 pF), because it would have trashed the bandwidth.

What were the bandwidths for the two gains?

Just trying to dig this up out of my old emails--it was from 2012-3, so
i forget. Here's a preliminary test result:
<https://electrooptical.net/www/sed/QceptBigTIA_Wringout.pdf>

The bandwidth was around 10 MHz or so, because the coupling between the
input signal and the TIA was capacitive, and it was the voltage we cared
about.

I also had to short out the 1G resistor when the amp was in 50G
mode --otherwise enough of its Johnson noise would have got
through the open contacts (0.2 pF) to dominate the noise floor.

Can you give us the calculation behind that? If the 50G
bandwidth was modest (2Hz for 0.15pF self capacitance?),
how did the 0.2pF noise feed-path figure in?

Since the summing junction impedance is relatively low compared with 1 G
at most frequencies, the Johnson noise current of the 1G resistor splits
itself between the summing junction via 0.2 pF and the parallel
capacitance of the resistor, about 0.05 pF. So above the corner frequency

f_c = 1/(2 pi * 0.25 pF * 1Gohm) = 640 Hz,

about 80% of the 1G resistor's noise goes into the SJ. Since it's 7
times larger than the 50G resistor's noise, that's a noise contribution
well worth going to a Form C relay to eliminate. (The eN*C noise
doesn't start to dominate until about 13 kHz with the 50G resistor, so
the 1G's noise would be a problem up to at least 100 kHz, and almost all
the useful signal info is below there.)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Winfield Hill]

Phil Hobbs wrote...

On 9/26/19 8:47 AM, Winfield Hill wrote:
pcdhobbs@gmail.com wrote...

Capacitance. I needed to bootstrap the coil-to-contact
capacitance (0.4 pF), because it would have trashed the bandwidth.

What were the bandwidths for the two gains?

Just trying to dig this up out of my old emails--it was from 2012-3,
so i forget. Here's a preliminary test result:
https://electrooptical.net/www/sed/QceptBigTIA_Wringout.pdf

The bandwidth was around 10 MHz or so, because the coupling between
the input signal and the TIA was capacitive, and it was the voltage
we cared about.

Ah, a capacitive-coupled-input with a gain of about 10, nice!


--
Thanks,
- Win

 

[Winfield Hill]

George Herold wrote...

On September 26, 2019, Phil Hobbs wrote:

What were the bandwidths for the two gains?

The bandwidth was around 10 MHz or so, because the coupling between the
input signal and the TIA was capacitive, and it was the voltage we cared
about.

10 MHz at 1 G ohm?!! I'm blown away...

No, this was not a TIA. The feedback element was the
0.35pF Cf self-capacitance of the 50G resistor and the
associated wiring, against a 0.035pF Cin input source.
G = Cf/Cin. The 50G merely provides DC zeroing, so
the Cf integrated output voltage won't soar, and it
makes a low-frequency rolloff. A pretty cool circuit!


--
Thanks,
- Win

 

[George Herold]

On Thursday, September 26, 2019 at 12:11:15 PM UTC-4, Phil Hobbs wrote:

On 9/26/19 8:47 AM, Winfield Hill wrote:
pcdhobbs@gmail.com wrote...

Capacitance. I needed to bootstrap the coil-to-contact
capacitance (0.4 pF), because it would have trashed the bandwidth.

What were the bandwidths for the two gains?


Just trying to dig this up out of my old emails--it was from 2012-3, so
i forget. Here's a preliminary test result:
https://electrooptical.net/www/sed/QceptBigTIA_Wringout.pdf

The bandwidth was around 10 MHz or so, because the coupling between the
input signal and the TIA was capacitive, and it was the voltage we cared
about.
10 MHz at 1 G ohm?!! I'm blown away...

George H.

I also had to short out the 1G resistor when the amp was in 50G
mode --otherwise enough of its Johnson noise would have got
through the open contacts (0.2 pF) to dominate the noise floor.

Can you give us the calculation behind that? If the 50G
bandwidth was modest (2Hz for 0.15pF self capacitance?),
how did the 0.2pF noise feed-path figure in?

Since the summing junction impedance is relatively low compared with 1 G
at most frequencies, the Johnson noise current of the 1G resistor splits
itself between the summing junction via 0.2 pF and the parallel
capacitance of the resistor, about 0.05 pF. So above the corner frequency

f_c = 1/(2 pi * 0.25 pF * 1Gohm) = 640 Hz,

about 80% of the 1G resistor's noise goes into the SJ. Since it's 7
times larger than the 50G resistor's noise, that's a noise contribution
well worth going to a Form C relay to eliminate. (The eN*C noise
doesn't start to dominate until about 13 kHz with the 50G resistor, so
the 1G's noise would be a problem up to at least 100 kHz, and almost all
the useful signal info is below there.)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Phil Hobbs]

On 9/26/19 2:50 PM, George Herold wrote:

On Thursday, September 26, 2019 at 12:11:15 PM UTC-4, Phil Hobbs wrote:
On 9/26/19 8:47 AM, Winfield Hill wrote:
pcdhobbs@gmail.com wrote...

Capacitance. I needed to bootstrap the coil-to-contact
capacitance (0.4 pF), because it would have trashed the bandwidth.

What were the bandwidths for the two gains?


Just trying to dig this up out of my old emails--it was from 2012-3, so
i forget. Here's a preliminary test result:
https://electrooptical.net/www/sed/QceptBigTIA_Wringout.pdf

The bandwidth was around 10 MHz or so, because the coupling between the
input signal and the TIA was capacitive, and it was the voltage we cared
about.
10 MHz at 1 G ohm?!! I'm blown away...

George H.

Well, it's a bit of a cheat really--the 50G just sets the bias and low
frequency rolloff, and then the capacitances kick in and do the real work.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com