Isolated, synchronous Cuk with dual primaries

[bitrex]

If I have an isolated Cuk topology with dual primaries where both input
voltages are floating, and I'd like to drive the single secondary from
only one primary supply at a time, with the other shut down not sinking
or sourcing any current.

I have a question about the series inductors on the primary and
secondary side which in the single primary, single secondary case are
often transformer-coupled to reduce output ripple. Is there a way to
time the switches S1, S2, S3 (assume I can synchronize them in an
isolated way via an optocoupler or something), and/or arrange the
secondary-side inductors L3, L4 in series/parallel such that they can be
coupled to their respective inductors on the two primaries, but only act
at any given time as would be expected in a Cuk with a single primary
winding, and don't back-feed the source on the other primary which is
supposed to be shut down.

Or do something else clever that hopefully doesn't require custom
magnetics. Or just don't couple them and have a single inductor on the
secondary that's a compromise value and deal with it.


<https://www.dropbox.com/s/qvrurtn67knulom/IMG_20200206_031516890.jpg?dl=0>

 

[bitrex]

On 2/6/20 5:03 AM, DecadentLinuxUserNumeroUno@decadence.org wrote:

bitrex <user@example.net> wrote in
news:FnR_F.13618$2Y3.10014@fx29.iad:

On 2/6/20 4:50 AM, bitrex wrote:
If I have an isolated Cuk topology with dual primaries where both
input voltages are floating, and I'd like to drive the single
secondary from only one primary supply at a time, with the other
shut down not sinking or sourcing any current.

I have a question about the series inductors on the primary and
secondary side which in the single primary, single secondary case
are often transformer-coupled to reduce output ripple. Is there a
way to time the switches S1, S2, S3 (assume I can synchronize
them in an isolated way via an optocoupler or something), and/or
arrange the secondary-side inductors L3, L4 in series/parallel
such that they can be coupled to their respective inductors on
the two primaries, but only act at any given time as would be
expected in a Cuk with a single primary winding, and don't
back-feed the source on the other primary which is supposed to be
shut down.

Or do something else clever that hopefully doesn't require custom
magnetics. Or just don't couple them and have a single inductor
on the secondary that's a compromise value and deal with it.


https://www.dropbox.com/s/qvrurtn67knulom/IMG_20200206_031516890.
jpg?dl=0

The 1:n coupling transformer might be a low-profile planar type
that's "wound" like the traces on a PCB with the core inserted
through the stack, rather than on a traditional core.


You got no PC with a spreadsheet app like LibreOffice? You could
hand trace(within the app) grid lines and lines in a spreadsheet and
print it out and post a pic of that. Way better than that was.
Hell, you could do a screenshot at that point.

I used to make signs for the lab in Excel all the time. Signs that
I would print, and laminate and even cut out and place on the floor
with the special floor tapes we were using for certain areas of the
lab. Schematics would be easy. Albeit not captured.

Hell you could use a capture app, you dont have to have a full
netlist.

What are you running, a 286?

I like pen and graph paper, I feel my "drafting abilities" rank
somewhere in-between Larkin and JanPan.

 

bitrex <user@example.net> wrote in
news:FnR_F.13618$2Y3.10014@fx29.iad:

On 2/6/20 4:50 AM, bitrex wrote:
If I have an isolated Cuk topology with dual primaries where both
input voltages are floating, and I'd like to drive the single
secondary from only one primary supply at a time, with the other
shut down not sinking or sourcing any current.

I have a question about the series inductors on the primary and
secondary side which in the single primary, single secondary case
are often transformer-coupled to reduce output ripple. Is there a
way to time the switches S1, S2, S3 (assume I can synchronize
them in an isolated way via an optocoupler or something), and/or
arrange the secondary-side inductors L3, L4 in series/parallel
such that they can be coupled to their respective inductors on
the two primaries, but only act at any given time as would be
expected in a Cuk with a single primary winding, and don't
back-feed the source on the other primary which is supposed to be
shut down.

Or do something else clever that hopefully doesn't require custom
magnetics. Or just don't couple them and have a single inductor
on the secondary that's a compromise value and deal with it.


https://www.dropbox.com/s/qvrurtn67knulom/IMG_20200206_031516890.
jpg?dl=0

The 1:n coupling transformer might be a low-profile planar type
that's "wound" like the traces on a PCB with the core inserted
through the stack, rather than on a traditional core.

You got no PC with a spreadsheet app like LibreOffice? You could
hand trace(within the app) grid lines and lines in a spreadsheet and
print it out and post a pic of that. Way better than that was.
Hell, you could do a screenshot at that point.

I used to make signs for the lab in Excel all the time. Signs that
I would print, and laminate and even cut out and place on the floor
with the special floor tapes we were using for certain areas of the
lab. Schematics would be easy. Albeit not captured.

Hell you could use a capture app, you dont have to have a full
netlist.

What are you running, a 286?

 

[bitrex]

On 2/6/20 4:50 AM, bitrex wrote:

If I have an isolated Cuk topology with dual primaries where both input
voltages are floating, and I'd like to drive the single secondary from
only one primary supply at a time, with the other shut down not sinking
or sourcing any current.

I have a question about the series inductors on the primary and
secondary side which in the single primary, single secondary case are
often transformer-coupled to reduce output ripple. Is there a way to
time the switches S1, S2, S3 (assume I can synchronize them in an
isolated way via an optocoupler or something), and/or arrange the
secondary-side inductors L3, L4 in series/parallel such that they can be
coupled to their respective inductors on the two primaries, but only act
at any given time as would be expected in a Cuk with a single primary
winding, and don't back-feed the source on the other primary which is
supposed to be shut down.

Or do something else clever that hopefully doesn't require custom
magnetics. Or just don't couple them and have a single inductor on the
secondary that's a compromise value and deal with it.


https://www.dropbox.com/s/qvrurtn67knulom/IMG_20200206_031516890.jpg?dl=0

The 1:n coupling transformer might be a low-profile planar type that's
"wound" like the traces on a PCB with the core inserted through the
stack, rather than on a traditional core.

 

bitrex <user@example.net> wrote in
news:LKR_F.385162$K87.379597@fx46.iad:

On 2/6/20 5:03 AM, DecadentLinuxUserNumeroUno@decadence.org wrote:
bitrex <user@example.net> wrote in
news:FnR_F.13618$2Y3.10014@fx29.iad:

On 2/6/20 4:50 AM, bitrex wrote:
If I have an isolated Cuk topology with dual primaries where
both input voltages are floating, and I'd like to drive the
single secondary from only one primary supply at a time, with
the other shut down not sinking or sourcing any current.

I have a question about the series inductors on the primary and
secondary side which in the single primary, single secondary
case are often transformer-coupled to reduce output ripple. Is
there a way to time the switches S1, S2, S3 (assume I can
synchronize them in an isolated way via an optocoupler or
something), and/or arrange the secondary-side inductors L3, L4
in series/parallel such that they can be coupled to their
respective inductors on the two primaries, but only act at any
given time as would be expected in a Cuk with a single primary
winding, and don't back-feed the source on the other primary
which is supposed to be shut down.

Or do something else clever that hopefully doesn't require
custom magnetics. Or just don't couple them and have a single
inductor on the secondary that's a compromise value and deal
with it.


https://www.dropbox.com/s/qvrurtn67knulom/IMG_20200206_03151689
0. jpg?dl=0

The 1:n coupling transformer might be a low-profile planar type
that's "wound" like the traces on a PCB with the core inserted
through the stack, rather than on a traditional core.


You got no PC with a spreadsheet app like LibreOffice? You
could
hand trace(within the app) grid lines and lines in a spreadsheet
and print it out and post a pic of that. Way better than that
was. Hell, you could do a screenshot at that point.

I used to make signs for the lab in Excel all the time. Signs
that
I would print, and laminate and even cut out and place on the
floor with the special floor tapes we were using for certain
areas of the lab. Schematics would be easy. Albeit not
captured.

Hell you could use a capture app, you dont have to have a full
netlist.

What are you running, a 286?


I like pen and graph paper, I feel my "drafting abilities" rank
somewhere in-between Larkin and JanPan.

Yeah... where do I get off at, eh?

 

[Tim Williams]

Can't do it. With all windings coupled together somehow or another, they're
all (AC) powered. The switch will rectify passively -- assuming it's a
single MOSFET.

What's wrong with two independent supplies, outputs in parallel? Possibly
some control circuitry could be muxed, but even if not, you're just
duplicating a controller, not a huge deal.

You're duplicating all the power components at least, and not batting an eye
at that.

Or actually, if it's a secondary side control, with isolators to the primary
side, you can still use just one control, and an opto to detect which
primary is active to determine which secondary side to drive (if it should
be synchronous), and the unpowered primary side drivers just don't do
anything because, well, unpowered.

Chicken-and-egg of initial startup, but that's left as an exercise for the
student. :^)

(Back in the day, there were TL494 based, PP forward converters -- your
average ATX supply -- that did this, of a sort. Drive transformer for the
bridge, secondary side controller. They solved startup either by making the
bridge self-oscillating, or adding a small aux supply that also delivered 5V
standby for the motherboard.)

Tim

--
Seven Transistor Labs, LLC
Electrical Engineering Consultation and Design
Website: https://www.seventransistorlabs.com/

"bitrex" <user@example.net> wrote in message
news:VjR_F.192071$zI7.29237@fx37.iad...

If I have an isolated Cuk topology with dual primaries where both input
voltages are floating, and I'd like to drive the single secondary from
only one primary supply at a time, with the other shut down not sinking or
sourcing any current.

I have a question about the series inductors on the primary and secondary
side which in the single primary, single secondary case are often
transformer-coupled to reduce output ripple. Is there a way to time the
switches S1, S2, S3 (assume I can synchronize them in an isolated way via
an optocoupler or something), and/or arrange the secondary-side inductors
L3, L4 in series/parallel such that they can be coupled to their
respective inductors on the two primaries, but only act at any given time
as would be expected in a Cuk with a single primary winding, and don't
back-feed the source on the other primary which is supposed to be shut
down.

Or do something else clever that hopefully doesn't require custom
magnetics. Or just don't couple them and have a single inductor on the
secondary that's a compromise value and deal with it.


https://www.dropbox.com/s/qvrurtn67knulom/IMG_20200206_031516890.jpg?dl=0

 

[bitrex]

On 2/6/20 7:32 AM, Tim Williams wrote:

Can't do it.  With all windings coupled together somehow or another,
they're all (AC) powered.  The switch will rectify passively -- assuming
it's a single MOSFET.

That actually might be OK, if the primaries are each connected to a
battery in a series stack. Cuz the idea is a crude type of
balancing-charger where the bulk charger connects to the top of the
stack and charges until one battery in the stack is detected to hit a
threshold SoC first, then the bulk charger cuts out and the "winning
battery" is used as the energy source to push charge into the others.
when it drops below a threshold and the others haven't hit theirs yet
the bulk charger cuts back in. Wash, rinse, repeat.

So the "secondary" doesn't really need to have a switch I guess it can
just have a free-wheeling diode and connect to the top of the stack when
the bulk charger cuts out, then only the selected primary MOSFET runs
and the other MOSFET body didoes and the diode on the secondary free-wheel.

The whitepaper I saw accomplished this with a multi-winding flyback but
I'd prefer to use isolation scheme that doesn't have to store energy in
the interest of compactness.

What's wrong with two independent supplies, outputs in parallel?
Possibly some control circuitry could be muxed, but even if not, you're
just duplicating a controller, not a huge deal.

You're duplicating all the power components at least, and not batting an
eye at that.

Or actually, if it's a secondary side control, with isolators to the
primary side, you can still use just one control, and an opto to detect
which primary is active to determine which secondary side to drive (if
it should be synchronous), and the unpowered primary side drivers just
don't do anything because, well, unpowered.

Chicken-and-egg of initial startup, but that's left as an exercise for
the student. :^)

(Back in the day, there were TL494 based, PP forward converters -- your
average ATX supply -- that did this, of a sort.  Drive transformer for
the bridge, secondary side controller.  They solved startup either by
making the bridge self-oscillating, or adding a small aux supply that
also delivered 5V standby for the motherboard.)

Tim

 

[bitrex]

On 2/6/20 12:31 PM, bitrex wrote:

On 2/6/20 7:32 AM, Tim Williams wrote:
Can't do it.  With all windings coupled together somehow or another,
they're all (AC) powered.  The switch will rectify passively --
assuming it's a single MOSFET.

That actually might be OK, if the primaries are each connected to a
battery in a series stack. Cuz the idea is a crude type of
balancing-charger where the bulk charger connects to the top of the
stack and charges until one battery in the stack is detected to hit a
threshold SoC first, then the bulk charger cuts out and the "winning
battery" is used as the energy source to push charge into the others.
when it drops below a threshold and the others haven't hit theirs yet
the bulk charger cuts back in. Wash, rinse, repeat.

So the "secondary" doesn't really need to have a switch I guess it can
just have a free-wheeling diode and connect to the top of the stack when
  the bulk charger cuts out, then only the selected primary MOSFET runs
and the other MOSFET body didoes and the diode on the secondary free-wheel.

The whitepaper I saw accomplished this with a multi-winding flyback but
I'd prefer to use isolation scheme that doesn't have to store energy in
the interest of compactness.

In that variation I think a high-side PMOS is used to drive the flyback
on the primary sides which prevents the passive-rectification issue but
isn't a particularly elegant solution