Constant current to capacifor / rising voltage...

  • Thread starter sonnic...@gmail.com
  • Start date
John Larkin <jlarkin@highlandSNIPMEtechnology.com> wrote:
On Tue, 20 Jun 2023 09:13:14 -0700 (PDT), whit3rd <whit3rd@gmail.com
wrote:

On Tuesday, June 20, 2023 at 7:31:09?AM UTC-7, Phil Hobbs wrote:
On 2023-06-19 23:18, Gerhard Hoffmann wrote:
Am 19.06.23 um 23:30 schrieb Phil Hobbs:
John Larkin <jla...@highlandSNIPMEtechnology.com> wrote:
On Mon, 19 Jun 2023 17:08:58 +0100, Brian Gregory
void-invalid...@email.invalid> wrote:

On 19/06/2023 11:25, John Larkin wrote:
If someone suggests a current mirror, it\'s probably a bad idea.

Yes. AIUI they\'re more something that gets built in to integrated
circuits since they don\'t work well when built from discrete components
because they\'re not exactly matched, and not thermally linked together.

Exact is a relative term; thermally linked is also a relative term, you can glue
two TO92\'s together and they ARE linked, thermally and physically: it
works better with C1841 NPN transistors than with PN2222\'s, because
the Japanese case has its metal tab oriented differently.


Most dual transistors are really two die, not thermally coupled.

That\'s a packaging issue.

Discrete transistor designs have the collector as the substrate, so
you can\'t easily electrically separate collectors of two transistors mounted
to the same metal plate. To make a good-quality dual, epitaxial
silicon grown over oxide can be electrically isolated, OR you need
a seven-pin package, so you can bias the substrate (as an IC would
do) to create a depletion region. Things like LM13700 have the
specified bias on that most-negative pin in order to work,
as did the (hard-to-find) transistor multiples of yesteryear (CA3046, anyone?).

A pure-silicon current mirror runs one transistor at 0.6ish volts and
low dissipation, and the other at whatever Vce and higher dissipation,
so tight thermal coupling is mandatory.

Really good monolithic pairs have many equivalent transistors
interleaved in a tricky pattern.

I guess you could bolt two TO-220s or TO-247s metal-to-metal with a
thin AlN insulator.

Trouble is, thermal conduction is proportional to the temperature gradient,
and there’s no way for a spreader to make the heat flow preferentially into
the cooler die.

They really need to be interleaved on a single chip for good matching
above, say, 50 uW dissipation.

That’s not at all hard to do, but virtually no one does it anymore.

There are various semicustom array offerings, but the devices available
aren’t exactly cutting edge, unfortunately.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC /
Hobbs ElectroOptics Optics, Electro-optics, Photonics, Analog Electronics
 
On Tuesday, June 20, 2023 at 9:28:39 AM UTC-7, John Larkin wrote:
On Tue, 20 Jun 2023 09:13:14 -0700 (PDT), whit3rd <whi...@gmail.com
wrote:

On 19/06/2023 11:25, John Larkin wrote:
If someone suggests a current mirror, it\'s probably a bad idea.

Yes. AIUI they\'re more something that gets built in to integrated
circuits since they don\'t work well when built from discrete components
because they\'re not exactly matched, and not thermally linked together.

Exact is a relative term; thermally linked is also a relative term, you can glue
two TO92\'s together and they ARE linked, thermally and physically: it
works better with C1841 NPN transistors than with PN2222\'s, because
the Japanese case has its metal tab oriented differently.


Most dual transistors are really two die, not thermally coupled.

That\'s a packaging issue.

Discrete transistor designs have the collector as the substrate, so
you can\'t easily electrically separate collectors of two transistors mounted
to the same metal plate.

A pure-silicon current mirror runs one transistor at 0.6ish volts and
low dissipation, and the other at whatever Vce and higher dissipation,
so tight thermal coupling is mandatory.

It takes four transistors to cascode that problem out of existence.
Really good monolithic pairs have many equivalent transistors
interleaved in a tricky pattern.

But, that\'s because they\'re on complex ICs that have heat sources
all around; and those IC transistors are TERRIBLE performers; usually
beta of about 40.

I guess you could bolt two TO-220s or TO-247s metal-to-metal with a
thin AlN insulator.

Better ways exist; some of the surface mount duals are in very thin packages,
less than half a mm thick, and you can glue a heat spreader on top of \'em.
Two TO220 tabs will keep the dice two or three mm apart, and you have
equal temperature AFTER heating up both tabs... heat doesn\'t just conduct,
it diffuses (long time delay for heating the tabs all the way through).
 
whit3rd <whit3rd@gmail.com> wrote:
On Tuesday, June 20, 2023 at 9:28:39 AM UTC-7, John Larkin wrote:
On Tue, 20 Jun 2023 09:13:14 -0700 (PDT), whit3rd <whi...@gmail.com
wrote:

On 19/06/2023 11:25, John Larkin wrote:
If someone suggests a current mirror, it\'s probably a bad idea.

Yes. AIUI they\'re more something that gets built in to integrated
circuits since they don\'t work well when built from discrete components
because they\'re not exactly matched, and not thermally linked together.

Exact is a relative term; thermally linked is also a relative term, you can glue
two TO92\'s together and they ARE linked, thermally and physically: it
works better with C1841 NPN transistors than with PN2222\'s, because
the Japanese case has its metal tab oriented differently.


Most dual transistors are really two die, not thermally coupled.

That\'s a packaging issue.

Discrete transistor designs have the collector as the substrate, so
you can\'t easily electrically separate collectors of two transistors mounted
to the same metal plate.

A pure-silicon current mirror runs one transistor at 0.6ish volts and
low dissipation, and the other at whatever Vce and higher dissipation,
so tight thermal coupling is mandatory.

It takes four transistors to cascode that problem out of existence.

Really good monolithic pairs have many equivalent transistors
interleaved in a tricky pattern.

But, that\'s because they\'re on complex ICs that have heat sources
all around; and those IC transistors are TERRIBLE performers; usually
beta of about 40.

Riiigghhht. All bipolar IC processes are the same, which is why the uA741
is still the best performing part.


--
Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC /
Hobbs ElectroOptics Optics, Electro-optics, Photonics, Analog Electronics
 
On Tue, 20 Jun 2023 18:40:25 -0000 (UTC), Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

John Larkin <jlarkin@highlandSNIPMEtechnology.com> wrote:
On Tue, 20 Jun 2023 09:13:14 -0700 (PDT), whit3rd <whit3rd@gmail.com
wrote:

On Tuesday, June 20, 2023 at 7:31:09?AM UTC-7, Phil Hobbs wrote:
On 2023-06-19 23:18, Gerhard Hoffmann wrote:
Am 19.06.23 um 23:30 schrieb Phil Hobbs:
John Larkin <jla...@highlandSNIPMEtechnology.com> wrote:
On Mon, 19 Jun 2023 17:08:58 +0100, Brian Gregory
void-invalid...@email.invalid> wrote:

On 19/06/2023 11:25, John Larkin wrote:
If someone suggests a current mirror, it\'s probably a bad idea.

Yes. AIUI they\'re more something that gets built in to integrated
circuits since they don\'t work well when built from discrete components
because they\'re not exactly matched, and not thermally linked together.

Exact is a relative term; thermally linked is also a relative term, you can glue
two TO92\'s together and they ARE linked, thermally and physically: it
works better with C1841 NPN transistors than with PN2222\'s, because
the Japanese case has its metal tab oriented differently.


Most dual transistors are really two die, not thermally coupled.

That\'s a packaging issue.

Discrete transistor designs have the collector as the substrate, so
you can\'t easily electrically separate collectors of two transistors mounted
to the same metal plate. To make a good-quality dual, epitaxial
silicon grown over oxide can be electrically isolated, OR you need
a seven-pin package, so you can bias the substrate (as an IC would
do) to create a depletion region. Things like LM13700 have the
specified bias on that most-negative pin in order to work,
as did the (hard-to-find) transistor multiples of yesteryear (CA3046, anyone?).

A pure-silicon current mirror runs one transistor at 0.6ish volts and
low dissipation, and the other at whatever Vce and higher dissipation,
so tight thermal coupling is mandatory.

Really good monolithic pairs have many equivalent transistors
interleaved in a tricky pattern.

I guess you could bolt two TO-220s or TO-247s metal-to-metal with a
thin AlN insulator.



Trouble is, thermal conduction is proportional to the temperature gradient,
and there’s no way for a spreader to make the heat flow preferentially into
the cooler die.

Heat always flows from warmer to cooler.

Two giant chips will hardly heat up, and the thermal conductivity
between them will be huge. There wouldn\'t be much gradient.

Not that I plan to build any discrete current mirrors.



They really need to be interleaved on a single chip for good matching
above, say, 50 uW dissipation.

That’s not at all hard to do, but virtually no one does it anymore.

There are various semicustom array offerings, but the devices available
aren’t exactly cutting edge, unfortunately.

Cheers

Phil Hobbs
 
On Wednesday, June 21, 2023 at 2:28:39 AM UTC+10, John Larkin wrote:
On Tue, 20 Jun 2023 09:13:14 -0700 (PDT), whit3rd <whi...@gmail.com
wrote:
On Tuesday, June 20, 2023 at 7:31:09?AM UTC-7, Phil Hobbs wrote:
On 2023-06-19 23:18, Gerhard Hoffmann wrote:
Am 19.06.23 um 23:30 schrieb Phil Hobbs:
John Larkin <jla...@highlandSNIPMEtechnology.com> wrote:
On Mon, 19 Jun 2023 17:08:58 +0100, Brian Gregory
void-invalid...@email.invalid> wrote:

On 19/06/2023 11:25, John Larkin wrote:
If someone suggests a current mirror, it\'s probably a bad idea.

Yes. AIUI they\'re more something that gets built in to integrated
circuits since they don\'t work well when built from discrete components
because they\'re not exactly matched, and not thermally linked together.

Exact is a relative term; thermally linked is also a relative term, you can glue
two TO92\'s together and they ARE linked, thermally and physically: it
works better with C1841 NPN transistors than with PN2222\'s, because
the Japanese case has its metal tab oriented differently.


Most dual transistors are really two die, not thermally coupled.

That\'s a packaging issue.

Discrete transistor designs have the collector as the substrate, so
you can\'t easily electrically separate collectors of two transistors mounted
to the same metal plate. To make a good-quality dual, epitaxial
silicon grown over oxide can be electrically isolated, OR you need
a seven-pin package, so you can bias the substrate (as an IC would
do) to create a depletion region. Things like LM13700 have the
specified bias on that most-negative pin in order to work,
as did the (hard-to-find) transistor multiples of yesteryear (CA3046, anyone?).

A pure-silicon current mirror runs one transistor at 0.6ish volts and
low dissipation, and the other at whatever Vce and higher dissipation,
so tight thermal coupling is mandatory.

A three transistor Wilson current mirror doesn\'t.,

Really good monolithic pairs have many equivalent transistors interleaved in a tricky pattern.

I guess you could bolt two TO-220s or TO-247s metal-to-metal with a thin AlN insulator.

The Wilson current mirror minimises the problem by the kind of intelligent design you don\'t seem to be able to follow.

--
Bill Sloman, Sydney
 
On Wednesday, June 21, 2023 at 11:03:49 AM UTC+10, John Larkin wrote:
On Tue, 20 Jun 2023 18:40:25 -0000 (UTC), Phil Hobbs <pcdhSpamM...@electrooptical.net> wrote:
John Larkin <jla...@highlandSNIPMEtechnology.com> wrote:
On Tue, 20 Jun 2023 09:13:14 -0700 (PDT), whit3rd <whi...@gmail.com> wrote:
On Tuesday, June 20, 2023 at 7:31:09?AM UTC-7, Phil Hobbs wrote:
On 2023-06-19 23:18, Gerhard Hoffmann wrote:
Am 19.06.23 um 23:30 schrieb Phil Hobbs:
John Larkin <jla...@highlandSNIPMEtechnology.com> wrote:
On Mon, 19 Jun 2023 17:08:58 +0100, Brian Gregory > <void-invalid....@email.invalid> wrote:
On 19/06/2023 11:25, John Larkin wrote:

<snip>

> Not that I plan to build any discrete current mirrors.

Very wise. If you can\'t understand why the Wilson current mirror solves most of the problems you are complaining about you should restrict yourself to approaches you can understand.

https://en.wikipedia.org/wiki/Wilson_current_mirror

--
Bill Sloman, Sydney
 

Welcome to EDABoard.com

Sponsor

Back
Top