transformer coupled logic isolator...

[server]

On Sun, 02 Jan 2022 17:54:15 +0100, Klaus Kragelund
<klauskvik@hotmail.com> wrote:

02.01.22 16:29, jlarkin@highlandsniptechnology.com wrote:
On Sun, 02 Jan 2022 10:09:25 +0100, Klaus Kragelund
klauskvik@hotmail.com> wrote:

02.01.22 02:38, Anthony William Sloman wrote:
On Sunday, January 2, 2022 at 7:45:52 AM UTC+11, jla...@highlandsniptechnology.com wrote:
On Sat, 01 Jan 2022 13:45:00 -0500, legg <le...@nospam.magma.ca> wrote:

On Sat, 1 Jan 2022 04:01:57 -0800 (PST), Anthony William Sloman
bill....@ieee.org> wrote:

On Saturday, January 1, 2022 at 12:43:32 PM UTC+11, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 20:21:16 -0500, legg <le...@nospam.magma.ca> wrote:
On Fri, 31 Dec 2021 09:23:48 -0800, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 11:38:25 -0500, legg <le...@nospam.magma.ca> wrote:
On Thu, 30 Dec 2021 19:05:06 -0800 (PST), Anthony William Sloman <bill....@ieee.org> wrote:
On Friday, December 31, 2021 at 5:21:07 AM UTC+11, John Larkin wrote:
On Thu, 30 Dec 2021 09:08:03 -0500, legg <le...@nospam.magma.ca> wrote:

On Thu, 30 Dec 2021 19:04:22 +1100, Sylvia Else <syl...@email.invalid
wrote:

On 30-Dec-21 4:11 pm, jla...@highlandsniptechnology.com wrote:
Version 4

snip

What is the use-case for this that a conventional digital isolator
wouldn\'t be suitable for?

Sometimes used in lower frequency isolated gate drive, when minimal
magnetics cost is the aim.

It\'s faster than most isolators, and is DC-coupled, after a power-up
priming shot.

Not a claim that\'s worth making for a purely theoretical transformer driving an LT Spice generic Schmitt trigger.

No parallel capacitance across either inductor, and no current induced in the transformer core - it\'s a little too theoretical too swank about.

It worked fine when I did it in 1979, but I wasn\'t around to see it go into production (if it did).

Getting the model to act like the real thing takes time and effort.

Getting the real thing to act like the model is probably delusional.

Right, it\'s best to avoid designing any electronics. It\'s too hard and
too risky.

Hey! The model works! What\'s HIS problem . . . ?

Do you mean Sloman?

Legg was responding to one of your posts, not mine.

He\'s the group leader on never actually doing anything.

I\'d got what you posted working with real parts back in 1979 - I\'d already done it, so why would I need to do it again?

So naturally he finds reasons why nothing will work.

I didn\'t say it wouldn\'t work - I just pointed out that the transformer model wasn\'t all that realistic, and neither was the Schmitt trigger.

You could have done quite a bit better, and telling us what you had in mind to use for your transformer would have been a good start.

Simulationss are useful in that they suggest what should or
could work.

If you limit it to a specific application, you can introduce
realistic strays and likely operating conditions with increasingly
more accurate models.

The \'party trick\' aspect of this circuit was the miniscule magnetic
component that was possible - though reduction in actual cost shows
diminishing and even reversing returns as you get carried away.

When I cam up with my version of the circuit in 1979 this did strike me as the useful feature. I wasn\'t tempted to try and get it patented.

An integrated magnetic component has been used in some places, though the isolation tended to be compromised.

Integrating anything means realising it within very limited dimensions, and high voltage isolation needs big gaps.

Not really. Just needs to be solid. 0.4mm FR4 is approved reinforced. So spiral coils on either side of the PCB could work. In practical size, leakage inductance is high

There are down-firing surface-mount leds and photodiodes. One could
couple light through a board.

Or e-fields, an FR4 capacitor based signal or power coupler.



I have done both the AM modulated transformer and the E field. Both tend to emit a lot of radiated noise, so you need to keep it to only transitions transmission

Also, they can be subceptible to EMC burst

LED method is clean, but slow

With blind vias one could make a shielded capacitor!

I should try some stacked stripline couplers on a future test board.
Parts made of copper inside PCBs are free!



--

I yam what I yam - Popeye

 

[Phil Hobbs]

jlarkin@highlandsniptechnology.com wrote:

On Sun, 2 Jan 2022 09:39:30 -0500, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

jlarkin@highlandsniptechnology.com wrote:
On Sat, 01 Jan 2022 13:45:00 -0500, legg <legg@nospam.magma.ca> wrote:

On Sat, 1 Jan 2022 04:01:57 -0800 (PST), Anthony William Sloman
bill.sloman@ieee.org> wrote:

On Saturday, January 1, 2022 at 12:43:32 PM UTC+11, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 20:21:16 -0500, legg <le...@nospam.magma.ca> wrote:
On Fri, 31 Dec 2021 09:23:48 -0800, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 11:38:25 -0500, legg <le...@nospam.magma.ca> wrote:
On Thu, 30 Dec 2021 19:05:06 -0800 (PST), Anthony William Sloman <bill....@ieee.org> wrote:
On Friday, December 31, 2021 at 5:21:07 AM UTC+11, John Larkin wrote:
On Thu, 30 Dec 2021 09:08:03 -0500, legg <le...@nospam.magma.ca> wrote:

On Thu, 30 Dec 2021 19:04:22 +1100, Sylvia Else <syl...@email.invalid
wrote:

On 30-Dec-21 4:11 pm, jla...@highlandsniptechnology.com wrote:
Version 4

snip

What is the use-case for this that a conventional digital isolator
wouldn\'t be suitable for?

Sometimes used in lower frequency isolated gate drive, when minimal
magnetics cost is the aim.

It\'s faster than most isolators, and is DC-coupled, after a power-up
priming shot.

Not a claim that\'s worth making for a purely theoretical transformer driving an LT Spice generic Schmitt trigger.

No parallel capacitance across either inductor, and no current induced in the transformer core - it\'s a little too theoretical too swank about.

It worked fine when I did it in 1979, but I wasn\'t around to see it go into production (if it did).

Getting the model to act like the real thing takes time and effort.

Getting the real thing to act like the model is probably delusional.

Right, it\'s best to avoid designing any electronics. It\'s too hard and
too risky.

Hey! The model works! What\'s HIS problem . . . ?

Do you mean Sloman?

Legg was responding to one of your posts, not mine.

He\'s the group leader on never actually doing anything.

I\'d got what you posted working with real parts back in 1979 - I\'d already done it, so why would I need to do it again?

So naturally he finds reasons why nothing will work.

I didn\'t say it wouldn\'t work - I just pointed out that the transformer model wasn\'t all that realistic, and neither was the Schmitt trigger.

You could have done quite a bit better, and telling us what you had in mind to use for your transformer would have been a good start.

Simulationss are useful in that they suggest what should or
could work.

If you limit it to a specific application, you can introduce
realistic strays and likely operating conditions with increasingly
more accurate models.

The \'party trick\' aspect of this circuit was the miniscule magnetic
component that was possible - though reduction in actual cost shows
diminishing and even reversing returns as you get carried away.
An integrated magnetic component has been used in some places,
though the isolation tended to be compromised.

Semiconductor \'pulse stretchers\' go back to the mid 70\'s. Physical
iteration may still be the fastest way to implimentation for a
practical app, though a pencil and paper can cut this work down.
The simulation just eats man-hours.

RL

Spice is great. It lets a person play with ideas quickly, explore
hunches, get quantitative with the things that look promising.
Sometimes I design a circuit and understand it later, if ever.

https://www.youtube.com/watch?v=x6TrbD7-IwU

\"Intuition is the most important part of engineering.\"

It\'s a _sine qua non_ of building new circuit forms, for sure. For
doing bridge abutments and oil wells, not so much.

It\'s good that our designs seldom kill people if they fail.



\"The function of a simulator is to train your instincts.\"

At board level, I mostly agree. Due to crappy models, board-level
simulations generally bear only an impressionistic resemblance to actual
circuit performance, so intuition is about all it\'s good for, apart from
catching gross blunders such as biasing errors or running out of GBW.


I disagree with Mike about the utility of Spice. We do go directly
from subcircuit sims to PCB layout of final products, and it usually
works. Why build a breadboard when you can build a sellable thing? We
only breadboard enough to understand new components that don\'t have
trustable models.


Foundry models are usually much better, because otherwise they wouldn\'t
turn out many functioning chips.

Interestingly I\'m currently dealing with an exception: the Xfab 180 nm
process with APDs and SPADs and such. One gathers that they haven\'t
actually fabbed many APD-based chips lately. (This is the first time
I\'ve actually collaborated closely on an IC design, apart from my
silicon photonics work.)

Mike is in the IC business, so his perspective on simulation is a
little differerent from people who solder parts to boards. But then,
Spice won\'t help much with semiconductor physics problems.

Can you make test chips? Maybe in the corners of a production wafer?

It\'s a multiproject wafer, a Phase 2 SBIR for a bathymetric lidar for
the Navy. Xfab had no idea how large an oxide isolation barrier you
need for a 25V APD. They told us to us _eight_microns_, which doubles
the die area of our 35-um APDs. Of course we were two weeks from our
planned tapeout when they sprung that one on us.

In real life, at the usually-quoted number of 6 MV/cm, that\'s enough for
a 5 kV barrier. Of course, the APD\'s anode is the back contact of the
die, which may make it a bit more difficult, but surely not 200x.

It\'s a follow-on for a time stretcher board (_not_ a pulse stretcher) we
did last spring. The proto used pHEMT samplers so that we could get 24
~100 ps-class T/Hs on a single APD output.

The IC has far lower strays, so the hold capacitors can be much smaller,
so it\'s possible to integrate. We\'re doing a single-APD test chip
(really four or five of them for different purposes) and a 64-APD
miniature array. The idea is to get up to 4096 APDs by the end of this
summer.

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

 

[server]

On Sun, 2 Jan 2022 13:25:04 -0500, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

jlarkin@highlandsniptechnology.com wrote:
On Sun, 2 Jan 2022 09:39:30 -0500, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

jlarkin@highlandsniptechnology.com wrote:
On Sat, 01 Jan 2022 13:45:00 -0500, legg <legg@nospam.magma.ca> wrote:

On Sat, 1 Jan 2022 04:01:57 -0800 (PST), Anthony William Sloman
bill.sloman@ieee.org> wrote:

On Saturday, January 1, 2022 at 12:43:32 PM UTC+11, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 20:21:16 -0500, legg <le...@nospam.magma.ca> wrote:
On Fri, 31 Dec 2021 09:23:48 -0800, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 11:38:25 -0500, legg <le...@nospam.magma.ca> wrote:
On Thu, 30 Dec 2021 19:05:06 -0800 (PST), Anthony William Sloman <bill....@ieee.org> wrote:
On Friday, December 31, 2021 at 5:21:07 AM UTC+11, John Larkin wrote:
On Thu, 30 Dec 2021 09:08:03 -0500, legg <le...@nospam.magma.ca> wrote:

On Thu, 30 Dec 2021 19:04:22 +1100, Sylvia Else <syl...@email.invalid
wrote:

On 30-Dec-21 4:11 pm, jla...@highlandsniptechnology.com wrote:
Version 4

snip

What is the use-case for this that a conventional digital isolator
wouldn\'t be suitable for?

Sometimes used in lower frequency isolated gate drive, when minimal
magnetics cost is the aim.

It\'s faster than most isolators, and is DC-coupled, after a power-up
priming shot.

Not a claim that\'s worth making for a purely theoretical transformer driving an LT Spice generic Schmitt trigger.

No parallel capacitance across either inductor, and no current induced in the transformer core - it\'s a little too theoretical too swank about.

It worked fine when I did it in 1979, but I wasn\'t around to see it go into production (if it did).

Getting the model to act like the real thing takes time and effort.

Getting the real thing to act like the model is probably delusional.

Right, it\'s best to avoid designing any electronics. It\'s too hard and
too risky.

Hey! The model works! What\'s HIS problem . . . ?

Do you mean Sloman?

Legg was responding to one of your posts, not mine.

He\'s the group leader on never actually doing anything.

I\'d got what you posted working with real parts back in 1979 - I\'d already done it, so why would I need to do it again?

So naturally he finds reasons why nothing will work.

I didn\'t say it wouldn\'t work - I just pointed out that the transformer model wasn\'t all that realistic, and neither was the Schmitt trigger.

You could have done quite a bit better, and telling us what you had in mind to use for your transformer would have been a good start.

Simulationss are useful in that they suggest what should or
could work.

If you limit it to a specific application, you can introduce
realistic strays and likely operating conditions with increasingly
more accurate models.

The \'party trick\' aspect of this circuit was the miniscule magnetic
component that was possible - though reduction in actual cost shows
diminishing and even reversing returns as you get carried away.
An integrated magnetic component has been used in some places,
though the isolation tended to be compromised.

Semiconductor \'pulse stretchers\' go back to the mid 70\'s. Physical
iteration may still be the fastest way to implimentation for a
practical app, though a pencil and paper can cut this work down.
The simulation just eats man-hours.

RL

Spice is great. It lets a person play with ideas quickly, explore
hunches, get quantitative with the things that look promising.
Sometimes I design a circuit and understand it later, if ever.

https://www.youtube.com/watch?v=x6TrbD7-IwU

\"Intuition is the most important part of engineering.\"

It\'s a _sine qua non_ of building new circuit forms, for sure. For
doing bridge abutments and oil wells, not so much.

It\'s good that our designs seldom kill people if they fail.



\"The function of a simulator is to train your instincts.\"

At board level, I mostly agree. Due to crappy models, board-level
simulations generally bear only an impressionistic resemblance to actual
circuit performance, so intuition is about all it\'s good for, apart from
catching gross blunders such as biasing errors or running out of GBW.


I disagree with Mike about the utility of Spice. We do go directly
from subcircuit sims to PCB layout of final products, and it usually
works. Why build a breadboard when you can build a sellable thing? We
only breadboard enough to understand new components that don\'t have
trustable models.


Foundry models are usually much better, because otherwise they wouldn\'t
turn out many functioning chips.

Interestingly I\'m currently dealing with an exception: the Xfab 180 nm
process with APDs and SPADs and such. One gathers that they haven\'t
actually fabbed many APD-based chips lately. (This is the first time
I\'ve actually collaborated closely on an IC design, apart from my
silicon photonics work.)

Mike is in the IC business, so his perspective on simulation is a
little differerent from people who solder parts to boards. But then,
Spice won\'t help much with semiconductor physics problems.

Can you make test chips? Maybe in the corners of a production wafer?

It\'s a multiproject wafer, a Phase 2 SBIR for a bathymetric lidar for
the Navy. Xfab had no idea how large an oxide isolation barrier you
need for a 25V APD. They told us to us _eight_microns_, which doubles
the die area of our 35-um APDs. Of course we were two weeks from our
planned tapeout when they sprung that one on us.

In real life, at the usually-quoted number of 6 MV/cm, that\'s enough for
a 5 kV barrier. Of course, the APD\'s anode is the back contact of the
die, which may make it a bit more difficult, but surely not 200x.

It\'s a follow-on for a time stretcher board (_not_ a pulse stretcher) we
did last spring. The proto used pHEMT samplers so that we could get 24
~100 ps-class T/Hs on a single APD output.

The IC has far lower strays, so the hold capacitors can be much smaller,
so it\'s possible to integrate. We\'re doing a single-APD test chip
(really four or five of them for different purposes) and a 64-APD
miniature array. The idea is to get up to 4096 APDs by the end of this
summer.

Cheers

Phil Hobbs

Who else could use that? Astronomers?

I know some flourescent-tag molecule imagers who might.



--

I yam what I yam - Popeye

 

[Phil Hobbs]

jlarkin@highlandsniptechnology.com wrote:

On Sun, 2 Jan 2022 13:25:04 -0500, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

jlarkin@highlandsniptechnology.com wrote:
On Sun, 2 Jan 2022 09:39:30 -0500, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

jlarkin@highlandsniptechnology.com wrote:
On Sat, 01 Jan 2022 13:45:00 -0500, legg <legg@nospam.magma.ca> wrote:

On Sat, 1 Jan 2022 04:01:57 -0800 (PST), Anthony William Sloman
bill.sloman@ieee.org> wrote:

On Saturday, January 1, 2022 at 12:43:32 PM UTC+11, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 20:21:16 -0500, legg <le...@nospam.magma.ca> wrote:
On Fri, 31 Dec 2021 09:23:48 -0800, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 11:38:25 -0500, legg <le...@nospam.magma.ca> wrote:
On Thu, 30 Dec 2021 19:05:06 -0800 (PST), Anthony William Sloman <bill....@ieee.org> wrote:
On Friday, December 31, 2021 at 5:21:07 AM UTC+11, John Larkin wrote:
On Thu, 30 Dec 2021 09:08:03 -0500, legg <le...@nospam.magma.ca> wrote:

On Thu, 30 Dec 2021 19:04:22 +1100, Sylvia Else <syl...@email.invalid
wrote:

On 30-Dec-21 4:11 pm, jla...@highlandsniptechnology.com wrote:
Version 4

snip

What is the use-case for this that a conventional digital isolator
wouldn\'t be suitable for?

Sometimes used in lower frequency isolated gate drive, when minimal
magnetics cost is the aim.

It\'s faster than most isolators, and is DC-coupled, after a power-up
priming shot.

Not a claim that\'s worth making for a purely theoretical transformer driving an LT Spice generic Schmitt trigger.

No parallel capacitance across either inductor, and no current induced in the transformer core - it\'s a little too theoretical too swank about.

It worked fine when I did it in 1979, but I wasn\'t around to see it go into production (if it did).

Getting the model to act like the real thing takes time and effort.

Getting the real thing to act like the model is probably delusional.

Right, it\'s best to avoid designing any electronics. It\'s too hard and
too risky.

Hey! The model works! What\'s HIS problem . . . ?

Do you mean Sloman?

Legg was responding to one of your posts, not mine.

He\'s the group leader on never actually doing anything.

I\'d got what you posted working with real parts back in 1979 - I\'d already done it, so why would I need to do it again?

So naturally he finds reasons why nothing will work.

I didn\'t say it wouldn\'t work - I just pointed out that the transformer model wasn\'t all that realistic, and neither was the Schmitt trigger.

You could have done quite a bit better, and telling us what you had in mind to use for your transformer would have been a good start.

Simulationss are useful in that they suggest what should or
could work.

If you limit it to a specific application, you can introduce
realistic strays and likely operating conditions with increasingly
more accurate models.

The \'party trick\' aspect of this circuit was the miniscule magnetic
component that was possible - though reduction in actual cost shows
diminishing and even reversing returns as you get carried away.
An integrated magnetic component has been used in some places,
though the isolation tended to be compromised.

Semiconductor \'pulse stretchers\' go back to the mid 70\'s. Physical
iteration may still be the fastest way to implimentation for a
practical app, though a pencil and paper can cut this work down.
The simulation just eats man-hours.

RL

Spice is great. It lets a person play with ideas quickly, explore
hunches, get quantitative with the things that look promising.
Sometimes I design a circuit and understand it later, if ever.

https://www.youtube.com/watch?v=x6TrbD7-IwU

\"Intuition is the most important part of engineering.\"

It\'s a _sine qua non_ of building new circuit forms, for sure. For
doing bridge abutments and oil wells, not so much.

It\'s good that our designs seldom kill people if they fail.



\"The function of a simulator is to train your instincts.\"

At board level, I mostly agree. Due to crappy models, board-level
simulations generally bear only an impressionistic resemblance to actual
circuit performance, so intuition is about all it\'s good for, apart from
catching gross blunders such as biasing errors or running out of GBW.


I disagree with Mike about the utility of Spice. We do go directly
from subcircuit sims to PCB layout of final products, and it usually
works. Why build a breadboard when you can build a sellable thing? We
only breadboard enough to understand new components that don\'t have
trustable models.


Foundry models are usually much better, because otherwise they wouldn\'t
turn out many functioning chips.

Interestingly I\'m currently dealing with an exception: the Xfab 180 nm
process with APDs and SPADs and such. One gathers that they haven\'t
actually fabbed many APD-based chips lately. (This is the first time
I\'ve actually collaborated closely on an IC design, apart from my
silicon photonics work.)

Mike is in the IC business, so his perspective on simulation is a
little differerent from people who solder parts to boards. But then,
Spice won\'t help much with semiconductor physics problems.

Can you make test chips? Maybe in the corners of a production wafer?

It\'s a multiproject wafer, a Phase 2 SBIR for a bathymetric lidar for
the Navy. Xfab had no idea how large an oxide isolation barrier you
need for a 25V APD. They told us to us _eight_microns_, which doubles
the die area of our 35-um APDs. Of course we were two weeks from our
planned tapeout when they sprung that one on us.

In real life, at the usually-quoted number of 6 MV/cm, that\'s enough for
a 5 kV barrier. Of course, the APD\'s anode is the back contact of the
die, which may make it a bit more difficult, but surely not 200x.

It\'s a follow-on for a time stretcher board (_not_ a pulse stretcher) we
did last spring. The proto used pHEMT samplers so that we could get 24
~100 ps-class T/Hs on a single APD output.

The IC has far lower strays, so the hold capacitors can be much smaller,
so it\'s possible to integrate. We\'re doing a single-APD test chip
(really four or five of them for different purposes) and a 64-APD
miniature array. The idea is to get up to 4096 APDs by the end of this
summer.

Who else could use that? Astronomers?

Dunno. It\'s good for time-resolved things using fast pulses with low
rep rates, such as Q-switched lasers.

I know some fluorescent-tag molecule imagers who might.

That might well be a good application, especially in a dose-limited
situation where a stroboscopic approach won\'t work.

If the application looks real, maybe we and our respective customers
should talk. (It\'s a fun technology, for sure.)

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

 

[server]

On Sun, 2 Jan 2022 15:17:10 -0500, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

jlarkin@highlandsniptechnology.com wrote:
On Sun, 2 Jan 2022 13:25:04 -0500, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

jlarkin@highlandsniptechnology.com wrote:
On Sun, 2 Jan 2022 09:39:30 -0500, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

jlarkin@highlandsniptechnology.com wrote:
On Sat, 01 Jan 2022 13:45:00 -0500, legg <legg@nospam.magma.ca> wrote:

On Sat, 1 Jan 2022 04:01:57 -0800 (PST), Anthony William Sloman
bill.sloman@ieee.org> wrote:

On Saturday, January 1, 2022 at 12:43:32 PM UTC+11, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 20:21:16 -0500, legg <le...@nospam.magma.ca> wrote:
On Fri, 31 Dec 2021 09:23:48 -0800, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 11:38:25 -0500, legg <le...@nospam.magma.ca> wrote:
On Thu, 30 Dec 2021 19:05:06 -0800 (PST), Anthony William Sloman <bill....@ieee.org> wrote:
On Friday, December 31, 2021 at 5:21:07 AM UTC+11, John Larkin wrote:
On Thu, 30 Dec 2021 09:08:03 -0500, legg <le...@nospam.magma.ca> wrote:

On Thu, 30 Dec 2021 19:04:22 +1100, Sylvia Else <syl...@email.invalid
wrote:

On 30-Dec-21 4:11 pm, jla...@highlandsniptechnology.com wrote:
Version 4

snip

What is the use-case for this that a conventional digital isolator
wouldn\'t be suitable for?

Sometimes used in lower frequency isolated gate drive, when minimal
magnetics cost is the aim.

It\'s faster than most isolators, and is DC-coupled, after a power-up
priming shot.

Not a claim that\'s worth making for a purely theoretical transformer driving an LT Spice generic Schmitt trigger.

No parallel capacitance across either inductor, and no current induced in the transformer core - it\'s a little too theoretical too swank about.

It worked fine when I did it in 1979, but I wasn\'t around to see it go into production (if it did).

Getting the model to act like the real thing takes time and effort.

Getting the real thing to act like the model is probably delusional.

Right, it\'s best to avoid designing any electronics. It\'s too hard and
too risky.

Hey! The model works! What\'s HIS problem . . . ?

Do you mean Sloman?

Legg was responding to one of your posts, not mine.

He\'s the group leader on never actually doing anything.

I\'d got what you posted working with real parts back in 1979 - I\'d already done it, so why would I need to do it again?

So naturally he finds reasons why nothing will work.

I didn\'t say it wouldn\'t work - I just pointed out that the transformer model wasn\'t all that realistic, and neither was the Schmitt trigger.

You could have done quite a bit better, and telling us what you had in mind to use for your transformer would have been a good start.

Simulationss are useful in that they suggest what should or
could work.

If you limit it to a specific application, you can introduce
realistic strays and likely operating conditions with increasingly
more accurate models.

The \'party trick\' aspect of this circuit was the miniscule magnetic
component that was possible - though reduction in actual cost shows
diminishing and even reversing returns as you get carried away.
An integrated magnetic component has been used in some places,
though the isolation tended to be compromised.

Semiconductor \'pulse stretchers\' go back to the mid 70\'s. Physical
iteration may still be the fastest way to implimentation for a
practical app, though a pencil and paper can cut this work down.
The simulation just eats man-hours.

RL

Spice is great. It lets a person play with ideas quickly, explore
hunches, get quantitative with the things that look promising.
Sometimes I design a circuit and understand it later, if ever.

https://www.youtube.com/watch?v=x6TrbD7-IwU

\"Intuition is the most important part of engineering.\"

It\'s a _sine qua non_ of building new circuit forms, for sure. For
doing bridge abutments and oil wells, not so much.

It\'s good that our designs seldom kill people if they fail.



\"The function of a simulator is to train your instincts.\"

At board level, I mostly agree. Due to crappy models, board-level
simulations generally bear only an impressionistic resemblance to actual
circuit performance, so intuition is about all it\'s good for, apart from
catching gross blunders such as biasing errors or running out of GBW.


I disagree with Mike about the utility of Spice. We do go directly
from subcircuit sims to PCB layout of final products, and it usually
works. Why build a breadboard when you can build a sellable thing? We
only breadboard enough to understand new components that don\'t have
trustable models.


Foundry models are usually much better, because otherwise they wouldn\'t
turn out many functioning chips.

Interestingly I\'m currently dealing with an exception: the Xfab 180 nm
process with APDs and SPADs and such. One gathers that they haven\'t
actually fabbed many APD-based chips lately. (This is the first time
I\'ve actually collaborated closely on an IC design, apart from my
silicon photonics work.)

Mike is in the IC business, so his perspective on simulation is a
little differerent from people who solder parts to boards. But then,
Spice won\'t help much with semiconductor physics problems.

Can you make test chips? Maybe in the corners of a production wafer?

It\'s a multiproject wafer, a Phase 2 SBIR for a bathymetric lidar for
the Navy. Xfab had no idea how large an oxide isolation barrier you
need for a 25V APD. They told us to us _eight_microns_, which doubles
the die area of our 35-um APDs. Of course we were two weeks from our
planned tapeout when they sprung that one on us.

In real life, at the usually-quoted number of 6 MV/cm, that\'s enough for
a 5 kV barrier. Of course, the APD\'s anode is the back contact of the
die, which may make it a bit more difficult, but surely not 200x.

It\'s a follow-on for a time stretcher board (_not_ a pulse stretcher) we
did last spring. The proto used pHEMT samplers so that we could get 24
~100 ps-class T/Hs on a single APD output.

The IC has far lower strays, so the hold capacitors can be much smaller,
so it\'s possible to integrate. We\'re doing a single-APD test chip
(really four or five of them for different purposes) and a 64-APD
miniature array. The idea is to get up to 4096 APDs by the end of this
summer.

Who else could use that? Astronomers?

Dunno. It\'s good for time-resolved things using fast pulses with low
rep rates, such as Q-switched lasers.


I know some fluorescent-tag molecule imagers who might.

That might well be a good application, especially in a dose-limited
situation where a stroboscopic approach won\'t work.

If the application looks real, maybe we and our respective customers
should talk. (It\'s a fun technology, for sure.)

Cheers

Phil Hobbs

Time resolution, fluorescent decay timing, could add another dimension
to the tagged molecule thing. I\'ll send you my contact.



--

I yam what I yam - Popeye

 

[Anthony William Sloman]

On Monday, January 3, 2022 at 2:41:35 AM UTC+11, jla...@highlandsniptechnology.com wrote:

On Sun, 2 Jan 2022 09:39:30 -0500, Phil Hobbs <pcdhSpamM...@electrooptical.net> wrote:
jla...@highlandsniptechnology.com wrote:
On Sat, 01 Jan 2022 13:45:00 -0500, legg <le...@nospam.magma.ca> wrote:
On Sat, 1 Jan 2022 04:01:57 -0800 (PST), Anthony William Sloman <bill....@ieee.org> wrote:
On Saturday, January 1, 2022 at 12:43:32 PM UTC+11, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 20:21:16 -0500, legg <le...@nospam.magma.ca> wrote:
On Fri, 31 Dec 2021 09:23:48 -0800, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 11:38:25 -0500, legg <le...@nospam.magma.ca> wrote:
On Thu, 30 Dec 2021 19:05:06 -0800 (PST), Anthony William Sloman <bill....@ieee.org> wrote:
On Friday, December 31, 2021 at 5:21:07 AM UTC+11, John Larkin wrote:
On Thu, 30 Dec 2021 09:08:03 -0500, legg <le...@nospam.magma.ca> wrote:
On Thu, 30 Dec 2021 19:04:22 +1100, Sylvia Else <syl...@email..invalid> wrote:
On 30-Dec-21 4:11 pm, jla...@highlandsniptechnology.com wrote:

<snip>

I disagree with Mike about the utility of Spice. We do go directly
from subcircuit sims to PCB layout of final products, and it usually
works.

John Larkin does product development in very small steps. Each small step is quite likely to work.

> Why build a breadboard when you can build a sellable thing?

It is a good question. If you are only making small steps, the risk of going directly to something you can sell can be quite small.

> We only breadboard enough to understand new components that don\'t have trustable models.

Not that you seem to understand all that much about the components you could model better if you put in more parameters - like the parallel capaciutance of an inductor winding.

> >Foundry models are usually much better, because otherwise they wouldn\'t turn out many functioning chips.

There\'s a lot more money hanging on getting a chip to work on first iteration, and investing a lot of time in getting the models right pays off much more generously for them

Interestingly I\'m currently dealing with an exception: the Xfab 180 nm
process with APDs and SPADs and such. One gathers that they haven\'t
actually fabbed many APD-based chips lately. (This is the first time
I\'ve actually collaborated closely on an IC design, apart from my
silicon photonics work.)

Mike is in the IC business, so his perspective on simulation is a
little different from people who solder parts to boards. But then,
Spice won\'t help much with semiconductor physics problems.

Mike developed LTSpice as a marketing tool for Linear Technology. His perspective is very much that of people who solder parts to boards.

Semiconductor manufacturers use more powerful programs running on much more powerful processors, and can afford to use tools like electron beam testers (and Mike Engelhardt worked for Schlumberger on their - very successful - electron beam tester that ended up with 98% of the market) which can let you test your simulation programs on working integrated circuits.

> Can you make test chips? Maybe in the corners of a production wafer?

Each test chip needs mask - probably quite a few of them. There are direct write options, but they aren\'t cheap.

--
Bill Sloman, Sydney

 

[Klaus Kragelund]

On Sunday, 2 January 2022 at 18:19:00 UTC+1, legg wrote:

On Sat, 1 Jan 2022 17:38:31 -0800 (PST), Anthony William Sloman
bill....@ieee.org> wrote:

snip
Physical iteration may still be the fastest way to implimentation for a practical app, though a pencil and paper can cut this work down.
The simulation just eats man-hours.

A lot less than building the hardware. It\'s good for getting rid of ideas that can\'t work, but it isn\'t a reliable way of working out whether real hardware will work.

Spice is great. It lets a person play with ideas quickly, explore
hunches, get quantitative with the things that look promising.

But only to the extent to which the simulation is realisitic. An inductor winding without any parallel capacitance isn\'t. For a single layer winding 1pF of parallel capacitance is sort of realistic. Multilayer winding need more.
Unless it\'s a saturating core model, miniaturization and
minimalization efforts will be misleading.

As L values reduce, inductive strays also become important,
as do recovery times in the semiconductors of the transmitting
and receiving structures, when impressed with below-ground
or above rail transients or reflected emf.

The physical dimensions of beads and insulated wire create
clusters of performance features that have sometimes been
spelled out quite effectively, in millimeters, in patents.

I expect it\'s much the same for the integrated case, given
the limited types of core matl and dimensional constraints
present.

Even more so, if there\'s no core material.

The digital isolators has the advantage of using oxide layer as isolation, so they can couple well. And all the advantages of ASICs with lots of free transistors to cope with parasitic annoyances

 

[Klaus Kragelund]

On Sunday, 2 January 2022 at 18:48:05 UTC+1, jla...@highlandsniptechnology.com wrote:

On Sun, 02 Jan 2022 17:54:15 +0100, Klaus Kragelund
klau...@hotmail.com> wrote:

02.01.22 16:29, jla...@highlandsniptechnology.com wrote:
On Sun, 02 Jan 2022 10:09:25 +0100, Klaus Kragelund
klau...@hotmail.com> wrote:

02.01.22 02:38, Anthony William Sloman wrote:
On Sunday, January 2, 2022 at 7:45:52 AM UTC+11, jla...@highlandsniptechnology.com wrote:
On Sat, 01 Jan 2022 13:45:00 -0500, legg <le...@nospam.magma.ca> wrote:

On Sat, 1 Jan 2022 04:01:57 -0800 (PST), Anthony William Sloman
bill....@ieee.org> wrote:

On Saturday, January 1, 2022 at 12:43:32 PM UTC+11, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 20:21:16 -0500, legg <le...@nospam.magma.ca> wrote:
On Fri, 31 Dec 2021 09:23:48 -0800, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 11:38:25 -0500, legg <le...@nospam.magma.ca> wrote:
On Thu, 30 Dec 2021 19:05:06 -0800 (PST), Anthony William Sloman <bill....@ieee.org> wrote:
On Friday, December 31, 2021 at 5:21:07 AM UTC+11, John Larkin wrote:
On Thu, 30 Dec 2021 09:08:03 -0500, legg <le...@nospam.magma.ca> wrote:

On Thu, 30 Dec 2021 19:04:22 +1100, Sylvia Else <syl...@email.invalid
wrote:

On 30-Dec-21 4:11 pm, jla...@highlandsniptechnology.com wrote:
Version 4

snip

What is the use-case for this that a conventional digital isolator
wouldn\'t be suitable for?

Sometimes used in lower frequency isolated gate drive, when minimal
magnetics cost is the aim.

It\'s faster than most isolators, and is DC-coupled, after a power-up
priming shot.

Not a claim that\'s worth making for a purely theoretical transformer driving an LT Spice generic Schmitt trigger.

No parallel capacitance across either inductor, and no current induced in the transformer core - it\'s a little too theoretical too swank about.

It worked fine when I did it in 1979, but I wasn\'t around to see it go into production (if it did).

Getting the model to act like the real thing takes time and effort.

Getting the real thing to act like the model is probably delusional.

Right, it\'s best to avoid designing any electronics. It\'s too hard and
too risky.

Hey! The model works! What\'s HIS problem . . . ?

Do you mean Sloman?

Legg was responding to one of your posts, not mine.

He\'s the group leader on never actually doing anything.

I\'d got what you posted working with real parts back in 1979 - I\'d already done it, so why would I need to do it again?

So naturally he finds reasons why nothing will work.

I didn\'t say it wouldn\'t work - I just pointed out that the transformer model wasn\'t all that realistic, and neither was the Schmitt trigger.

You could have done quite a bit better, and telling us what you had in mind to use for your transformer would have been a good start.

Simulationss are useful in that they suggest what should or
could work.

If you limit it to a specific application, you can introduce
realistic strays and likely operating conditions with increasingly
more accurate models.

The \'party trick\' aspect of this circuit was the miniscule magnetic
component that was possible - though reduction in actual cost shows
diminishing and even reversing returns as you get carried away.

When I cam up with my version of the circuit in 1979 this did strike me as the useful feature. I wasn\'t tempted to try and get it patented.

An integrated magnetic component has been used in some places, though the isolation tended to be compromised.

Integrating anything means realising it within very limited dimensions, and high voltage isolation needs big gaps.

Not really. Just needs to be solid. 0.4mm FR4 is approved reinforced. So spiral coils on either side of the PCB could work. In practical size, leakage inductance is high

There are down-firing surface-mount leds and photodiodes. One could
couple light through a board.

Or e-fields, an FR4 capacitor based signal or power coupler.



I have done both the AM modulated transformer and the E field. Both tend to emit a lot of radiated noise, so you need to keep it to only transitions transmission

Also, they can be subceptible to EMC burst

LED method is clean, but slow
With blind vias one could make a shielded capacitor!

I should try some stacked stripline couplers on a future test board.
Parts made of copper inside PCBs are free!

That would be very interesting to try out, for sure

 

[John Walliker]

On Monday, 3 January 2022 at 10:29:23 UTC, klaus.k...@gmail.com wrote:

On Sunday, 2 January 2022 at 18:48:05 UTC+1, jla...@highlandsniptechnology.com wrote:
On Sun, 02 Jan 2022 17:54:15 +0100, Klaus Kragelund
klau...@hotmail.com> wrote:

02.01.22 16:29, jla...@highlandsniptechnology.com wrote:
On Sun, 02 Jan 2022 10:09:25 +0100, Klaus Kragelund
klau...@hotmail.com> wrote:

02.01.22 02:38, Anthony William Sloman wrote:
On Sunday, January 2, 2022 at 7:45:52 AM UTC+11, jla...@highlandsniptechnology.com wrote:
On Sat, 01 Jan 2022 13:45:00 -0500, legg <le...@nospam.magma.ca> wrote:

On Sat, 1 Jan 2022 04:01:57 -0800 (PST), Anthony William Sloman
bill....@ieee.org> wrote:

On Saturday, January 1, 2022 at 12:43:32 PM UTC+11, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 20:21:16 -0500, legg <le...@nospam.magma.ca> wrote:
On Fri, 31 Dec 2021 09:23:48 -0800, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 11:38:25 -0500, legg <le...@nospam.magma.ca> wrote:
On Thu, 30 Dec 2021 19:05:06 -0800 (PST), Anthony William Sloman <bill....@ieee.org> wrote:
On Friday, December 31, 2021 at 5:21:07 AM UTC+11, John Larkin wrote:
On Thu, 30 Dec 2021 09:08:03 -0500, legg <le...@nospam.magma.ca> wrote:

On Thu, 30 Dec 2021 19:04:22 +1100, Sylvia Else <syl...@email.invalid
wrote:

On 30-Dec-21 4:11 pm, jla...@highlandsniptechnology.com wrote:
Version 4

snip

What is the use-case for this that a conventional digital isolator
wouldn\'t be suitable for?

Sometimes used in lower frequency isolated gate drive, when minimal
magnetics cost is the aim.

It\'s faster than most isolators, and is DC-coupled, after a power-up
priming shot.

Not a claim that\'s worth making for a purely theoretical transformer driving an LT Spice generic Schmitt trigger.

No parallel capacitance across either inductor, and no current induced in the transformer core - it\'s a little too theoretical too swank about.

It worked fine when I did it in 1979, but I wasn\'t around to see it go into production (if it did).

Getting the model to act like the real thing takes time and effort.

Getting the real thing to act like the model is probably delusional.

Right, it\'s best to avoid designing any electronics. It\'s too hard and
too risky.

Hey! The model works! What\'s HIS problem . . . ?

Do you mean Sloman?

Legg was responding to one of your posts, not mine.

He\'s the group leader on never actually doing anything.

I\'d got what you posted working with real parts back in 1979 - I\'d already done it, so why would I need to do it again?

So naturally he finds reasons why nothing will work.

I didn\'t say it wouldn\'t work - I just pointed out that the transformer model wasn\'t all that realistic, and neither was the Schmitt trigger.

You could have done quite a bit better, and telling us what you had in mind to use for your transformer would have been a good start.

Simulationss are useful in that they suggest what should or
could work.

If you limit it to a specific application, you can introduce
realistic strays and likely operating conditions with increasingly
more accurate models.

The \'party trick\' aspect of this circuit was the miniscule magnetic
component that was possible - though reduction in actual cost shows
diminishing and even reversing returns as you get carried away.

When I cam up with my version of the circuit in 1979 this did strike me as the useful feature. I wasn\'t tempted to try and get it patented.

An integrated magnetic component has been used in some places, though the isolation tended to be compromised.

Integrating anything means realising it within very limited dimensions, and high voltage isolation needs big gaps.

Not really. Just needs to be solid. 0.4mm FR4 is approved reinforced. So spiral coils on either side of the PCB could work. In practical size, leakage inductance is high

There are down-firing surface-mount leds and photodiodes. One could
couple light through a board.

Or e-fields, an FR4 capacitor based signal or power coupler.



I have done both the AM modulated transformer and the E field. Both tend to emit a lot of radiated noise, so you need to keep it to only transitions transmission

Also, they can be subceptible to EMC burst

LED method is clean, but slow
With blind vias one could make a shielded capacitor!

I should try some stacked stripline couplers on a future test board.
Parts made of copper inside PCBs are free!
That would be very interesting to try out, for sure

I\'ve done something similar in a high-volume automotive product where
a pulsed infra-red diode (for the MOST bus) was producing excessive rf
emissions. It was not possible to place decoupling capacitors close
enough to the pins of the driver module because of keep-out rules for the
selective wave soldering of the module.
The solution was to build a multi-layer capacitor from alternating power
and ground planes directly under the device. It worked really well.
John

 

[legg]

On Mon, 3 Jan 2022 02:28:39 -0800 (PST), Klaus Kragelund
<klaus.kragelund@gmail.com> wrote:

On Sunday, 2 January 2022 at 18:19:00 UTC+1, legg wrote:
On Sat, 1 Jan 2022 17:38:31 -0800 (PST), Anthony William Sloman
bill....@ieee.org> wrote:

snip
Physical iteration may still be the fastest way to implimentation for a practical app, though a pencil and paper can cut this work down.
The simulation just eats man-hours.

A lot less than building the hardware. It\'s good for getting rid of ideas that can\'t work, but it isn\'t a reliable way of working out whether real hardware will work.

Spice is great. It lets a person play with ideas quickly, explore
hunches, get quantitative with the things that look promising.

But only to the extent to which the simulation is realisitic. An inductor winding without any parallel capacitance isn\'t. For a single layer winding 1pF of parallel capacitance is sort of realistic. Multilayer winding need more.
Unless it\'s a saturating core model, miniaturization and
minimalization efforts will be misleading.

As L values reduce, inductive strays also become important,
as do recovery times in the semiconductors of the transmitting
and receiving structures, when impressed with below-ground
or above rail transients or reflected emf.

The physical dimensions of beads and insulated wire create
clusters of performance features that have sometimes been
spelled out quite effectively, in millimeters, in patents.

I expect it\'s much the same for the integrated case, given
the limited types of core matl and dimensional constraints
present.

Even more so, if there\'s no core material.

The digital isolators has the advantage of using oxide layer as isolation, so they can couple well. And all the advantages of ASICs with lots of free transistors to cope with parasitic annoyances

I think the bulk of the commercially viable ones use differential
capacitive coupling for data, but there\'s all kinds, now. Many
ways to skin a cat.

RL

 

[server]

On Mon, 3 Jan 2022 04:09:09 -0800 (PST), John Walliker
<jrwalliker@gmail.com> wrote:

On Monday, 3 January 2022 at 10:29:23 UTC, klaus.k...@gmail.com wrote:
On Sunday, 2 January 2022 at 18:48:05 UTC+1, jla...@highlandsniptechnology.com wrote:
On Sun, 02 Jan 2022 17:54:15 +0100, Klaus Kragelund
klau...@hotmail.com> wrote:

02.01.22 16:29, jla...@highlandsniptechnology.com wrote:
On Sun, 02 Jan 2022 10:09:25 +0100, Klaus Kragelund
klau...@hotmail.com> wrote:

02.01.22 02:38, Anthony William Sloman wrote:
On Sunday, January 2, 2022 at 7:45:52 AM UTC+11, jla...@highlandsniptechnology.com wrote:
On Sat, 01 Jan 2022 13:45:00 -0500, legg <le...@nospam.magma.ca> wrote:

On Sat, 1 Jan 2022 04:01:57 -0800 (PST), Anthony William Sloman
bill....@ieee.org> wrote:

On Saturday, January 1, 2022 at 12:43:32 PM UTC+11, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 20:21:16 -0500, legg <le...@nospam.magma.ca> wrote:
On Fri, 31 Dec 2021 09:23:48 -0800, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 11:38:25 -0500, legg <le...@nospam.magma.ca> wrote:
On Thu, 30 Dec 2021 19:05:06 -0800 (PST), Anthony William Sloman <bill....@ieee.org> wrote:
On Friday, December 31, 2021 at 5:21:07 AM UTC+11, John Larkin wrote:
On Thu, 30 Dec 2021 09:08:03 -0500, legg <le...@nospam.magma.ca> wrote:

On Thu, 30 Dec 2021 19:04:22 +1100, Sylvia Else <syl...@email.invalid
wrote:

On 30-Dec-21 4:11 pm, jla...@highlandsniptechnology.com wrote:
Version 4

snip

What is the use-case for this that a conventional digital isolator
wouldn\'t be suitable for?

Sometimes used in lower frequency isolated gate drive, when minimal
magnetics cost is the aim.

It\'s faster than most isolators, and is DC-coupled, after a power-up
priming shot.

Not a claim that\'s worth making for a purely theoretical transformer driving an LT Spice generic Schmitt trigger.

No parallel capacitance across either inductor, and no current induced in the transformer core - it\'s a little too theoretical too swank about.

It worked fine when I did it in 1979, but I wasn\'t around to see it go into production (if it did).

Getting the model to act like the real thing takes time and effort.

Getting the real thing to act like the model is probably delusional.

Right, it\'s best to avoid designing any electronics. It\'s too hard and
too risky.

Hey! The model works! What\'s HIS problem . . . ?

Do you mean Sloman?

Legg was responding to one of your posts, not mine.

He\'s the group leader on never actually doing anything.

I\'d got what you posted working with real parts back in 1979 - I\'d already done it, so why would I need to do it again?

So naturally he finds reasons why nothing will work.

I didn\'t say it wouldn\'t work - I just pointed out that the transformer model wasn\'t all that realistic, and neither was the Schmitt trigger.

You could have done quite a bit better, and telling us what you had in mind to use for your transformer would have been a good start.

Simulationss are useful in that they suggest what should or
could work.

If you limit it to a specific application, you can introduce
realistic strays and likely operating conditions with increasingly
more accurate models.

The \'party trick\' aspect of this circuit was the miniscule magnetic
component that was possible - though reduction in actual cost shows
diminishing and even reversing returns as you get carried away.

When I cam up with my version of the circuit in 1979 this did strike me as the useful feature. I wasn\'t tempted to try and get it patented.

An integrated magnetic component has been used in some places, though the isolation tended to be compromised.

Integrating anything means realising it within very limited dimensions, and high voltage isolation needs big gaps.

Not really. Just needs to be solid. 0.4mm FR4 is approved reinforced. So spiral coils on either side of the PCB could work. In practical size, leakage inductance is high

There are down-firing surface-mount leds and photodiodes. One could
couple light through a board.

Or e-fields, an FR4 capacitor based signal or power coupler.



I have done both the AM modulated transformer and the E field. Both tend to emit a lot of radiated noise, so you need to keep it to only transitions transmission

Also, they can be subceptible to EMC burst

LED method is clean, but slow
With blind vias one could make a shielded capacitor!

I should try some stacked stripline couplers on a future test board.
Parts made of copper inside PCBs are free!
That would be very interesting to try out, for sure

I sometimes want a differentiator to make a fast edge into a pretty
gaussian pulse. That might be done with coupled striplines. Or make
the isolator transformer!

I\'ve done something similar in a high-volume automotive product where
a pulsed infra-red diode (for the MOST bus) was producing excessive rf
emissions. It was not possible to place decoupling capacitors close
enough to the pins of the driver module because of keep-out rules for the
selective wave soldering of the module.
The solution was to build a multi-layer capacitor from alternating power
and ground planes directly under the device. It worked really well.
John

Ground+power planes are better hf capacitors than any discrete cap,
assuming the vias up to the parts don\'t wreck things.



--

I yam what I yam - Popeye

 

[Piglet]

On 01/01/2022 17:21, jlarkin@highlandsniptechnology.com wrote:

I like the nice cheap surface-mount dual inductors, the DRQ74 types.
No CT. They have near zero leakage inductance.

Near zero leakage inductance but gobs of capacitance!

piglet

 

[server]

On Mon, 3 Jan 2022 16:45:37 +0000, Piglet <erichpwagner@hotmail.com>
wrote:

On 01/01/2022 17:21, jlarkin@highlandsniptechnology.com wrote:
I like the nice cheap surface-mount dual inductors, the DRQ74 types.
No CT. They have near zero leakage inductance.


Near zero leakage inductance but gobs of capacitance!

piglet

They are very nice parts, DRQ74 to DRQ127, but of course you have to
understand them to use them. A lot of parts are annoying that way.

They make cool HV power supplies. I like using one as a flyboost
(drive the ct in flyback mode and get 2x out) followed by a C-W
multiplier.



--

I yam what I yam - Popeye

 

[Klaus Vestergaard Kragelund]

On 02/01/2022 13.05, Anthony William Sloman wrote:

On Sunday, January 2, 2022 at 8:09:33 PM UTC+11, Klaus Kragelund wrote:
02.01.22 02:38, Anthony William Sloman wrote:
On Sunday, January 2, 2022 at 7:45:52 AM UTC+11, jla...@highlandsniptechnology.com wrote:
On Sat, 01 Jan 2022 13:45:00 -0500, legg <le...@nospam.magma.ca> wrote:

On Sat, 1 Jan 2022 04:01:57 -0800 (PST), Anthony William Sloman
bill....@ieee.org> wrote:

On Saturday, January 1, 2022 at 12:43:32 PM UTC+11, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 20:21:16 -0500, legg <le...@nospam.magma.ca> wrote:
On Fri, 31 Dec 2021 09:23:48 -0800, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 11:38:25 -0500, legg <le...@nospam.magma.ca> wrote:
On Thu, 30 Dec 2021 19:05:06 -0800 (PST), Anthony William Sloman <bill....@ieee.org> wrote:
On Friday, December 31, 2021 at 5:21:07 AM UTC+11, John Larkin wrote:
On Thu, 30 Dec 2021 09:08:03 -0500, legg <le...@nospam.magma.ca> wrote:

On Thu, 30 Dec 2021 19:04:22 +1100, Sylvia Else <syl...@email.invalid
wrote:

On 30-Dec-21 4:11 pm, jla...@highlandsniptechnology.com wrote:
Version 4

snip

What is the use-case for this that a conventional digital isolator
wouldn\'t be suitable for?

Sometimes used in lower frequency isolated gate drive, when minimal
magnetics cost is the aim.

It\'s faster than most isolators, and is DC-coupled, after a power-up
priming shot.

Not a claim that\'s worth making for a purely theoretical transformer driving an LT Spice generic Schmitt trigger.

No parallel capacitance across either inductor, and no current induced in the transformer core - it\'s a little too theoretical too swank about.

It worked fine when I did it in 1979, but I wasn\'t around to see it go into production (if it did).

Getting the model to act like the real thing takes time and effort.

Getting the real thing to act like the model is probably delusional.

Right, it\'s best to avoid designing any electronics. It\'s too hard and
too risky.

Hey! The model works! What\'s HIS problem . . . ?

Do you mean Sloman?

Legg was responding to one of your posts, not mine.

He\'s the group leader on never actually doing anything.

I\'d got what you posted working with real parts back in 1979 - I\'d already done it, so why would I need to do it again?

So naturally he finds reasons why nothing will work.

I didn\'t say it wouldn\'t work - I just pointed out that the transformer model wasn\'t all that realistic, and neither was the Schmitt trigger.

You could have done quite a bit better, and telling us what you had in mind to use for your transformer would have been a good start.

Simulationss are useful in that they suggest what should or
could work.

If you limit it to a specific application, you can introduce
realistic strays and likely operating conditions with increasingly
more accurate models.

The \'party trick\' aspect of this circuit was the miniscule magnetic
component that was possible - though reduction in actual cost shows
diminishing and even reversing returns as you get carried away.

When I cam up with my version of the circuit in 1979 this did strike me as the useful feature. I wasn\'t tempted to try and get it patented.

An integrated magnetic component has been used in some places, though the isolation tended to be compromised.

Integrating anything means realising it within very limited dimensions, and high voltage isolation needs big gaps.

Not really. Just needs to be solid. 0.4mm FR4 is approved reinforced. So spiral coils on either side of the PCB could work. In practical size, leakage inductance is high

Integration is usually taken to be putting stuff together on a silicon substrate (or GaAs or some other kind of semi-conductor). Anything that gets soldered onto FR4 creates an assembled device, not an integrated part.

That\'s your definition. If I integrate a coil into a PCB with no added
components, I would call that.... integrated ;-)

> Alumina would probably be a better high voltage substrate than FR4-epoxy bonded fibre-glass, but arcs have a nasty way of tracking across insulators.

Not practical for a PCB. Mixing alumina and FR4 makes it very expensive

 

[Klaus Vestergaard Kragelund]

On 03/01/2022 13.09, John Walliker wrote:

On Monday, 3 January 2022 at 10:29:23 UTC, klaus.k...@gmail.com wrote:
On Sunday, 2 January 2022 at 18:48:05 UTC+1, jla...@highlandsniptechnology.com wrote:
On Sun, 02 Jan 2022 17:54:15 +0100, Klaus Kragelund
klau...@hotmail.com> wrote:

02.01.22 16:29, jla...@highlandsniptechnology.com wrote:
On Sun, 02 Jan 2022 10:09:25 +0100, Klaus Kragelund
klau...@hotmail.com> wrote:

02.01.22 02:38, Anthony William Sloman wrote:
On Sunday, January 2, 2022 at 7:45:52 AM UTC+11, jla...@highlandsniptechnology.com wrote:
On Sat, 01 Jan 2022 13:45:00 -0500, legg <le...@nospam.magma.ca> wrote:

On Sat, 1 Jan 2022 04:01:57 -0800 (PST), Anthony William Sloman
bill....@ieee.org> wrote:

On Saturday, January 1, 2022 at 12:43:32 PM UTC+11, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 20:21:16 -0500, legg <le...@nospam.magma.ca> wrote:
On Fri, 31 Dec 2021 09:23:48 -0800, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 11:38:25 -0500, legg <le...@nospam.magma.ca> wrote:
On Thu, 30 Dec 2021 19:05:06 -0800 (PST), Anthony William Sloman <bill....@ieee.org> wrote:
On Friday, December 31, 2021 at 5:21:07 AM UTC+11, John Larkin wrote:
On Thu, 30 Dec 2021 09:08:03 -0500, legg <le...@nospam.magma.ca> wrote:

On Thu, 30 Dec 2021 19:04:22 +1100, Sylvia Else <syl...@email.invalid
wrote:

On 30-Dec-21 4:11 pm, jla...@highlandsniptechnology.com wrote:
Version 4

snip

What is the use-case for this that a conventional digital isolator
wouldn\'t be suitable for?

Sometimes used in lower frequency isolated gate drive, when minimal
magnetics cost is the aim.

It\'s faster than most isolators, and is DC-coupled, after a power-up
priming shot.

Not a claim that\'s worth making for a purely theoretical transformer driving an LT Spice generic Schmitt trigger.

No parallel capacitance across either inductor, and no current induced in the transformer core - it\'s a little too theoretical too swank about.

It worked fine when I did it in 1979, but I wasn\'t around to see it go into production (if it did).

Getting the model to act like the real thing takes time and effort.

Getting the real thing to act like the model is probably delusional.

Right, it\'s best to avoid designing any electronics. It\'s too hard and
too risky.

Hey! The model works! What\'s HIS problem . . . ?

Do you mean Sloman?

Legg was responding to one of your posts, not mine.

He\'s the group leader on never actually doing anything.

I\'d got what you posted working with real parts back in 1979 - I\'d already done it, so why would I need to do it again?

So naturally he finds reasons why nothing will work.

I didn\'t say it wouldn\'t work - I just pointed out that the transformer model wasn\'t all that realistic, and neither was the Schmitt trigger.

You could have done quite a bit better, and telling us what you had in mind to use for your transformer would have been a good start.

Simulationss are useful in that they suggest what should or
could work.

If you limit it to a specific application, you can introduce
realistic strays and likely operating conditions with increasingly
more accurate models.

The \'party trick\' aspect of this circuit was the miniscule magnetic
component that was possible - though reduction in actual cost shows
diminishing and even reversing returns as you get carried away.

When I cam up with my version of the circuit in 1979 this did strike me as the useful feature. I wasn\'t tempted to try and get it patented.

An integrated magnetic component has been used in some places, though the isolation tended to be compromised.

Integrating anything means realising it within very limited dimensions, and high voltage isolation needs big gaps.

Not really. Just needs to be solid. 0.4mm FR4 is approved reinforced. So spiral coils on either side of the PCB could work. In practical size, leakage inductance is high

There are down-firing surface-mount leds and photodiodes. One could
couple light through a board.

Or e-fields, an FR4 capacitor based signal or power coupler.



I have done both the AM modulated transformer and the E field. Both tend to emit a lot of radiated noise, so you need to keep it to only transitions transmission

Also, they can be subceptible to EMC burst

LED method is clean, but slow
With blind vias one could make a shielded capacitor!

I should try some stacked stripline couplers on a future test board.
Parts made of copper inside PCBs are free!
That would be very interesting to try out, for sure

I\'ve done something similar in a high-volume automotive product where
a pulsed infra-red diode (for the MOST bus) was producing excessive rf
emissions. It was not possible to place decoupling capacitors close
enough to the pins of the driver module because of keep-out rules for the
selective wave soldering of the module.
The solution was to build a multi-layer capacitor from alternating power
and ground planes directly under the device. It worked really well.

You could elaborate that to EMC Y caps over a barrier (needed to return
the CM current). A large PCB inner layer plate has very good HF
properties (and free)

 

[Klaus Vestergaard Kragelund]

On 01/01/2022 02.30, Anthony William Sloman wrote:

On Saturday, January 1, 2022 at 4:24:04 AM UTC+11, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 11:38:25 -0500, legg <le...@nospam.magma.ca> wrote:

On Thu, 30 Dec 2021 19:05:06 -0800 (PST), Anthony William Sloman
bill....@ieee.org> wrote:

On Friday, December 31, 2021 at 5:21:07 AM UTC+11, John Larkin wrote:
On Thu, 30 Dec 2021 09:08:03 -0500, legg <le...@nospam.magma.ca> wrote:

On Thu, 30 Dec 2021 19:04:22 +1100, Sylvia Else <syl...@email.invalid
wrote:

On 30-Dec-21 4:11 pm, jla...@highlandsniptechnology.com wrote:
Version 4

snip

What is the use-case for this that a conventional digital isolator
wouldn\'t be suitable for?

Sometimes used in lower frequency isolated gate drive, when minimal
magnetics cost is the aim.

It\'s faster than most isolators, and is DC-coupled, after a power-up
priming shot.

Not a claim that\'s worth making for a purely theoretical transformer driving an LT Spice generic Schmitt trigger.

No parallel capacitance across either inductor, and no current induced in the transformer core - it\'s a little too theoretical too swank about.

It worked fine when I did it in 1979, but I wasn\'t around to see it go into production (if it did).

Getting the model to act like the real thing takes time
and effort.

Getting the real thing to act like the model is probably
delusional.

Right, it\'s best to avoid designing any electronics. It\'s too hard and too risky.

Anybody else would being intentionally satirical, but since John Larkin doesn\'t design electronics but rather evolves his circuits by making lots of small changes and seeing what effect they have, which may be tedious but doesn\'t seem to be difficult or all that expensive (if you can do it by hacking existing board) he may be posting with a straight face.

For the record, designing novel circuits is hard and risky, but it can pay off.

In todays world of offers like JLCPCB it costs next to nothing to get an
assembled board done. Why don\'t you try to build the Bandaxall converter
and tell us about the results?

 

[Anthony William Sloman]

On Tuesday, January 4, 2022 at 7:22:13 AM UTC+11, Klaus Kragelund wrote:

On 02/01/2022 13.05, Anthony William Sloman wrote:
On Sunday, January 2, 2022 at 8:09:33 PM UTC+11, Klaus Kragelund wrote:
02.01.22 02:38, Anthony William Sloman wrote:
On Sunday, January 2, 2022 at 7:45:52 AM UTC+11, jla...@highlandsniptechnology.com wrote:
On Sat, 01 Jan 2022 13:45:00 -0500, legg <le...@nospam.magma.ca> wrote:

On Sat, 1 Jan 2022 04:01:57 -0800 (PST), Anthony William Sloman
bill....@ieee.org> wrote:

On Saturday, January 1, 2022 at 12:43:32 PM UTC+11, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 20:21:16 -0500, legg <le...@nospam.magma.ca> wrote:
On Fri, 31 Dec 2021 09:23:48 -0800, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 11:38:25 -0500, legg <le...@nospam.magma.ca> wrote:
On Thu, 30 Dec 2021 19:05:06 -0800 (PST), Anthony William Sloman <bill....@ieee.org> wrote:
On Friday, December 31, 2021 at 5:21:07 AM UTC+11, John Larkin wrote:
On Thu, 30 Dec 2021 09:08:03 -0500, legg <le...@nospam.magma..ca> wrote:

On Thu, 30 Dec 2021 19:04:22 +1100, Sylvia Else <syl...@email.invalid
wrote:

On 30-Dec-21 4:11 pm, jla...@highlandsniptechnology.com wrote:
Version 4

snip

What is the use-case for this that a conventional digital isolator
wouldn\'t be suitable for?

Sometimes used in lower frequency isolated gate drive, when minimal
magnetics cost is the aim.

It\'s faster than most isolators, and is DC-coupled, after a power-up
priming shot.

Not a claim that\'s worth making for a purely theoretical transformer driving an LT Spice generic Schmitt trigger.

No parallel capacitance across either inductor, and no current induced in the transformer core - it\'s a little too theoretical too swank about.

It worked fine when I did it in 1979, but I wasn\'t around to see it go into production (if it did).

Getting the model to act like the real thing takes time and effort.

Getting the real thing to act like the model is probably delusional.

Right, it\'s best to avoid designing any electronics. It\'s too hard and
too risky.

Hey! The model works! What\'s HIS problem . . . ?

Do you mean Sloman?

Legg was responding to one of your posts, not mine.

He\'s the group leader on never actually doing anything.

I\'d got what you posted working with real parts back in 1979 - I\'d already done it, so why would I need to do it again?

So naturally he finds reasons why nothing will work.

I didn\'t say it wouldn\'t work - I just pointed out that the transformer model wasn\'t all that realistic, and neither was the Schmitt trigger.

You could have done quite a bit better, and telling us what you had in mind to use for your transformer would have been a good start.

Simulationss are useful in that they suggest what should or
could work.

If you limit it to a specific application, you can introduce
realistic strays and likely operating conditions with increasingly
more accurate models.

The \'party trick\' aspect of this circuit was the miniscule magnetic
component that was possible - though reduction in actual cost shows
diminishing and even reversing returns as you get carried away.

When I cam up with my version of the circuit in 1979 this did strike me as the useful feature. I wasn\'t tempted to try and get it patented.

An integrated magnetic component has been used in some places, though the isolation tended to be compromised.

Integrating anything means realising it within very limited dimensions, and high voltage isolation needs big gaps.

Not really. Just needs to be solid. 0.4mm FR4 is approved reinforced. So spiral coils on either side of the PCB could work. In practical size, leakage inductance is high

Integration is usually taken to be putting stuff together on a silicon substrate (or GaAs or some other kind of semi-conductor). Anything that gets soldered onto FR4 creates an assembled device, not an integrated part.

That\'s your definition. If I integrate a coil into a PCB with no added
components, I would call that.... integrated ;-)
Alumina would probably be a better high voltage substrate than FR4-epoxy bonded fibre-glass, but arcs have a nasty way of tracking across insulators.

Not practical for a PCB. Mixing alumina and FR4 makes it very expensive.

You used to be able to get thick film hybrids made on alumina. They might have been described as the rich man\'s printed circuit, through the conductors were silk-screened onto the alumina, along with the resistors and the solder paste. I\'ve no idea whether anybody still does this, but FR4-expoxy-glass substrates are just one of many available substrates.

--
Bill Sloman, Sydney

 

[Anthony William Sloman]

On Tuesday, January 4, 2022 at 7:28:48 AM UTC+11, Klaus Kragelund wrote:

On 01/01/2022 02.30, Anthony William Sloman wrote:
On Saturday, January 1, 2022 at 4:24:04 AM UTC+11, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 11:38:25 -0500, legg <le...@nospam.magma.ca> wrote:

On Thu, 30 Dec 2021 19:05:06 -0800 (PST), Anthony William Sloman
bill....@ieee.org> wrote:

On Friday, December 31, 2021 at 5:21:07 AM UTC+11, John Larkin wrote:
On Thu, 30 Dec 2021 09:08:03 -0500, legg <le...@nospam.magma.ca> wrote:

On Thu, 30 Dec 2021 19:04:22 +1100, Sylvia Else <syl...@email.invalid
wrote:

On 30-Dec-21 4:11 pm, jla...@highlandsniptechnology.com wrote:
Version 4

snip

What is the use-case for this that a conventional digital isolator
wouldn\'t be suitable for?

Sometimes used in lower frequency isolated gate drive, when minimal
magnetics cost is the aim.

It\'s faster than most isolators, and is DC-coupled, after a power-up
priming shot.

Not a claim that\'s worth making for a purely theoretical transformer driving an LT Spice generic Schmitt trigger.

No parallel capacitance across either inductor, and no current induced in the transformer core - it\'s a little too theoretical too swank about.

It worked fine when I did it in 1979, but I wasn\'t around to see it go into production (if it did).

Getting the model to act like the real thing takes time
and effort.

Getting the real thing to act like the model is probably
delusional.

Right, it\'s best to avoid designing any electronics. It\'s too hard and too risky.

Anybody else would being intentionally satirical, but since John Larkin doesn\'t design electronics but rather evolves his circuits by making lots of small changes and seeing what effect they have, which may be tedious but doesn\'t seem to be difficult or all that expensive (if you can do it by hacking existing board) he may be posting with a straight face.

For the record, designing novel circuits is hard and risky, but it can pay off.

In todays world of offers like JLCPCB it costs next to nothing to get an
assembled board done. Why don\'t you try to build the Bandaxall converter
and tell us about the results?

The Baxandall inverter was made very popular by Jim Williams (Linear Technology application notes AN45, AN49, AN51, AN55, AN61, AN65). He didn\'t call it that, because whoever it was in England who told him about the circuit didn\'t tell him where it came from. Why would I bother to duplicate his results?

I am tempted to build an example of my low distortion tunable oscillator - I did start putting it into KiCad - but without a prospective customer I can\'t work up any enthusiasm at all.

--
Bill Sloman, Sydney

 

[Klaus Vestergaard Kragelund]

On 04/01/2022 03.06, Anthony William Sloman wrote:

On Tuesday, January 4, 2022 at 7:22:13 AM UTC+11, Klaus Kragelund wrote:
On 02/01/2022 13.05, Anthony William Sloman wrote:
On Sunday, January 2, 2022 at 8:09:33 PM UTC+11, Klaus Kragelund wrote:
02.01.22 02:38, Anthony William Sloman wrote:
On Sunday, January 2, 2022 at 7:45:52 AM UTC+11, jla...@highlandsniptechnology.com wrote:
On Sat, 01 Jan 2022 13:45:00 -0500, legg <le...@nospam.magma.ca> wrote:

On Sat, 1 Jan 2022 04:01:57 -0800 (PST), Anthony William Sloman
bill....@ieee.org> wrote:

On Saturday, January 1, 2022 at 12:43:32 PM UTC+11, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 20:21:16 -0500, legg <le...@nospam.magma.ca> wrote:
On Fri, 31 Dec 2021 09:23:48 -0800, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 11:38:25 -0500, legg <le...@nospam.magma.ca> wrote:
On Thu, 30 Dec 2021 19:05:06 -0800 (PST), Anthony William Sloman <bill....@ieee.org> wrote:
On Friday, December 31, 2021 at 5:21:07 AM UTC+11, John Larkin wrote:
On Thu, 30 Dec 2021 09:08:03 -0500, legg <le...@nospam.magma.ca> wrote:

On Thu, 30 Dec 2021 19:04:22 +1100, Sylvia Else <syl...@email.invalid
wrote:

On 30-Dec-21 4:11 pm, jla...@highlandsniptechnology.com wrote:
Version 4

snip

What is the use-case for this that a conventional digital isolator
wouldn\'t be suitable for?

Sometimes used in lower frequency isolated gate drive, when minimal
magnetics cost is the aim.

It\'s faster than most isolators, and is DC-coupled, after a power-up
priming shot.

Not a claim that\'s worth making for a purely theoretical transformer driving an LT Spice generic Schmitt trigger.

No parallel capacitance across either inductor, and no current induced in the transformer core - it\'s a little too theoretical too swank about.

It worked fine when I did it in 1979, but I wasn\'t around to see it go into production (if it did).

Getting the model to act like the real thing takes time and effort.

Getting the real thing to act like the model is probably delusional.

Right, it\'s best to avoid designing any electronics. It\'s too hard and
too risky.

Hey! The model works! What\'s HIS problem . . . ?

Do you mean Sloman?

Legg was responding to one of your posts, not mine.

He\'s the group leader on never actually doing anything.

I\'d got what you posted working with real parts back in 1979 - I\'d already done it, so why would I need to do it again?

So naturally he finds reasons why nothing will work.

I didn\'t say it wouldn\'t work - I just pointed out that the transformer model wasn\'t all that realistic, and neither was the Schmitt trigger.

You could have done quite a bit better, and telling us what you had in mind to use for your transformer would have been a good start.

Simulationss are useful in that they suggest what should or
could work.

If you limit it to a specific application, you can introduce
realistic strays and likely operating conditions with increasingly
more accurate models.

The \'party trick\' aspect of this circuit was the miniscule magnetic
component that was possible - though reduction in actual cost shows
diminishing and even reversing returns as you get carried away.

When I cam up with my version of the circuit in 1979 this did strike me as the useful feature. I wasn\'t tempted to try and get it patented.

An integrated magnetic component has been used in some places, though the isolation tended to be compromised.

Integrating anything means realising it within very limited dimensions, and high voltage isolation needs big gaps.

Not really. Just needs to be solid. 0.4mm FR4 is approved reinforced. So spiral coils on either side of the PCB could work. In practical size, leakage inductance is high

Integration is usually taken to be putting stuff together on a silicon substrate (or GaAs or some other kind of semi-conductor). Anything that gets soldered onto FR4 creates an assembled device, not an integrated part.

That\'s your definition. If I integrate a coil into a PCB with no added
components, I would call that.... integrated ;-)
Alumina would probably be a better high voltage substrate than FR4-epoxy bonded fibre-glass, but arcs have a nasty way of tracking across insulators.

Not practical for a PCB. Mixing alumina and FR4 makes it very expensive.

You used to be able to get thick film hybrids made on alumina. They might have been described as the rich man\'s printed circuit, through the conductors were silk-screened onto the alumina, along with the resistors and the solder paste. I\'ve no idea whether anybody still does this, but FR4-expoxy-glass substrates are just one of many available substrates.

You can get cheap alumina PCBs, but then it\'s single layer traces, and
that can only be used for very simple designs

 

[Anthony William Sloman]

On Tuesday, January 4, 2022 at 6:53:42 PM UTC+11, Klaus Kragelund wrote:

On 04/01/2022 03.06, Anthony William Sloman wrote:
On Tuesday, January 4, 2022 at 7:22:13 AM UTC+11, Klaus Kragelund wrote:
On 02/01/2022 13.05, Anthony William Sloman wrote:
On Sunday, January 2, 2022 at 8:09:33 PM UTC+11, Klaus Kragelund wrote:
02.01.22 02:38, Anthony William Sloman wrote:
On Sunday, January 2, 2022 at 7:45:52 AM UTC+11, jla...@highlandsniptechnology.com wrote:
On Sat, 01 Jan 2022 13:45:00 -0500, legg <le...@nospam.magma.ca> wrote:

On Sat, 1 Jan 2022 04:01:57 -0800 (PST), Anthony William Sloman
bill....@ieee.org> wrote:

On Saturday, January 1, 2022 at 12:43:32 PM UTC+11, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 20:21:16 -0500, legg <le...@nospam.magma.ca> wrote:
On Fri, 31 Dec 2021 09:23:48 -0800, jla...@highlandsniptechnology.com wrote:
On Fri, 31 Dec 2021 11:38:25 -0500, legg <le...@nospam.magma.ca> wrote:
On Thu, 30 Dec 2021 19:05:06 -0800 (PST), Anthony William Sloman <bill....@ieee.org> wrote:
On Friday, December 31, 2021 at 5:21:07 AM UTC+11, John Larkin wrote:
On Thu, 30 Dec 2021 09:08:03 -0500, legg <le...@nospam.magma.ca> wrote:

On Thu, 30 Dec 2021 19:04:22 +1100, Sylvia Else <syl...@email.invalid
wrote:

On 30-Dec-21 4:11 pm, jla...@highlandsniptechnology.com wrote:
Version 4

snip

What is the use-case for this that a conventional digital isolator
wouldn\'t be suitable for?

Sometimes used in lower frequency isolated gate drive, when minimal
magnetics cost is the aim.

It\'s faster than most isolators, and is DC-coupled, after a power-up
priming shot.

Not a claim that\'s worth making for a purely theoretical transformer driving an LT Spice generic Schmitt trigger.

No parallel capacitance across either inductor, and no current induced in the transformer core - it\'s a little too theoretical too swank about.

It worked fine when I did it in 1979, but I wasn\'t around to see it go into production (if it did).

Getting the model to act like the real thing takes time and effort.

Getting the real thing to act like the model is probably delusional.

Right, it\'s best to avoid designing any electronics. It\'s too hard and
too risky.

Hey! The model works! What\'s HIS problem . . . ?

Do you mean Sloman?

Legg was responding to one of your posts, not mine.

He\'s the group leader on never actually doing anything.

I\'d got what you posted working with real parts back in 1979 - I\'d already done it, so why would I need to do it again?

So naturally he finds reasons why nothing will work.

I didn\'t say it wouldn\'t work - I just pointed out that the transformer model wasn\'t all that realistic, and neither was the Schmitt trigger.

You could have done quite a bit better, and telling us what you had in mind to use for your transformer would have been a good start.

Simulationss are useful in that they suggest what should or
could work.

If you limit it to a specific application, you can introduce
realistic strays and likely operating conditions with increasingly
more accurate models.

The \'party trick\' aspect of this circuit was the miniscule magnetic
component that was possible - though reduction in actual cost shows
diminishing and even reversing returns as you get carried away.

When I cam up with my version of the circuit in 1979 this did strike me as the useful feature. I wasn\'t tempted to try and get it patented.

An integrated magnetic component has been used in some places, though the isolation tended to be compromised.

Integrating anything means realising it within very limited dimensions, and high voltage isolation needs big gaps.

Not really. Just needs to be solid. 0.4mm FR4 is approved reinforced.. So spiral coils on either side of the PCB could work. In practical size, leakage inductance is high

Integration is usually taken to be putting stuff together on a silicon substrate (or GaAs or some other kind of semi-conductor). Anything that gets soldered onto FR4 creates an assembled device, not an integrated part.

That\'s your definition. If I integrate a coil into a PCB with no added
components, I would call that.... integrated ;-)
Alumina would probably be a better high voltage substrate than FR4-epoxy bonded fibre-glass, but arcs have a nasty way of tracking across insulators.

Not practical for a PCB. Mixing alumina and FR4 makes it very expensive.

You used to be able to get thick film hybrids made on alumina. They might have been described as the rich man\'s printed circuit, through the conductors were silk-screened onto the alumina, along with the resistors and the solder paste. I\'ve no idea whether anybody still does this, but FR4-expoxy-glass substrates are just one of many available substrates.

You can get cheap alumina PCBs, but then it\'s single layer traces, and that can only be used for very simple designs.

\"Cheap\" does exclude some schemes for doing better. Putting a via through an alumina substrate isn\'t cheap, but it can be done.

--
Bill Sloman, Sydney