I have a bunch of these eBay-special flybacks I'd like to slap one to a
piece of FR4 and use for prototyping a thing at +/- 100, but the output
is kinda ugly and has about 3-500mV of ~70kHz noise I want to reduce
before it goes to the circuit.

<https://www.ebay.com/itm/293851632650>

Looking for a ballpark for a starting point for the design of the LC, I
have a number of ferrite toroid cores in stock so if it's feasible I'll
just wind it myself like a common-mode choke. Is there a preferable
material for this application? Is it better to have two Cs to ground, a
C across the rail (like a differential filter), or combination?

On Fri, 30 Jul 2021 09:36:07 -0400, bitrex <user@example.net> wrote:

>I have a bunch of these eBay-special flybacks I'd like to slap one to a
>piece of FR4 and use for prototyping a thing at +/- 100, but the output
>is kinda ugly and has about 3-500mV of ~70kHz noise I want to reduce
>before it goes to the circuit.
>
><https://www.ebay.com/itm/293851632650>
>
>Looking for a ballpark for a starting point for the design of the LC, I
>have a number of ferrite toroid cores in stock so if it's feasible I'll
>just wind it myself like a common-mode choke. Is there a preferable
>material for this application? Is it better to have two Cs to ground, a
>C across the rail (like a differential filter), or combination?

If you output +100 and -100, two Ls and 2 Cs would be better. Just L-C
twice, caps grounded.

--

John Larkin Highland Technology, Inc

The best designs are necessarily accidental.

On Saturday, July 31, 2021 at 12:28:09 AM UTC+10, jla...@highlandsniptechnology.com wrote:
> On Fri, 30 Jul 2021 09:36:07 -0400, bitrex <us...@example.net> wrote:
>
> >I have a bunch of these eBay-special flybacks I'd like to slap one to a
> >piece of FR4 and use for prototyping a thing at +/- 100, but the output
> >is kinda ugly and has about 3-500mV of ~70kHz noise I want to reduce
> >before it goes to the circuit.
> >
> ><https://www.ebay.com/itm/293851632650>
> >
> >Looking for a ballpark for a starting point for the design of the LC, I
> >have a number of ferrite toroid cores in stock so if it's feasible I'll
> >just wind it myself like a common-mode choke. Is there a preferable
> >material for this application? Is it better to have two Cs to ground, a
> >C across the rail (like a differential filter), or combination?
>
> If you output +100 and -100, two Ls and 2 Cs would be better. Just L-C
> twice, caps grounded.

But check for resonances. A small resistor in series with at least one of the inductors can sometimes be a very good idea.

--
Bill Sloman, Sydney

On 7/30/2021 10:28 AM, jlarkin@highlandsniptechnology.com wrote:
> On Fri, 30 Jul 2021 09:36:07 -0400, bitrex <user@example.net> wrote:
>
>> I have a bunch of these eBay-special flybacks I'd like to slap one to a
>> piece of FR4 and use for prototyping a thing at +/- 100, but the output
>> is kinda ugly and has about 3-500mV of ~70kHz noise I want to reduce
>> before it goes to the circuit.
>>
>> <https://www.ebay.com/itm/293851632650>
>>
>> Looking for a ballpark for a starting point for the design of the LC, I
>> have a number of ferrite toroid cores in stock so if it's feasible I'll
>> just wind it myself like a common-mode choke. Is there a preferable
>> material for this application? Is it better to have two Cs to ground, a
>> C across the rail (like a differential filter), or combination?
>
> If you output +100 and -100, two Ls and 2 Cs would be better. Just L-C
> twice, caps grounded.
>

The inductance values I'm coming up for a Butterworth that gives -40dB
of rejection at 70kHz, without coupling them end up requiring ~200 turns
on a FT50-43 toroid which seems physically impossible with magnet wire
of a gauge I can easily work with, even if I wanted to take the time to
do it. Maybe need different core material to go that way.

If I couple them that comes down by a factor of 4 which is doable on a
core that size, kinda.

On 7/30/2021 11:20 AM, bitrex wrote:
> On 7/30/2021 10:28 AM, jlarkin@highlandsniptechnology.com wrote:
>> On Fri, 30 Jul 2021 09:36:07 -0400, bitrex <user@example.net> wrote:
>>
>>> I have a bunch of these eBay-special flybacks I'd like to slap one to a
>>> piece of FR4 and use for prototyping a thing at +/- 100, but the output
>>> is kinda ugly and has about 3-500mV of ~70kHz noise I want to reduce
>>> before it goes to the circuit.
>>>
>>> <https://www.ebay.com/itm/293851632650>
>>>
>>> Looking for a ballpark for a starting point for the design of the LC, I
>>> have a number of ferrite toroid cores in stock so if it's feasible I'll
>>> just wind it myself like a common-mode choke. Is there a preferable
>>> material for this application? Is it better to have two Cs to ground, a
>>> C across the rail (like a differential filter), or combination?
>>
>> If you output +100 and -100, two Ls and 2 Cs would be better. Just L-C
>> twice, caps grounded.
>>
>
> The inductance values I'm coming up for a Butterworth that gives -40dB
> of rejection at 70kHz, without coupling them end up requiring ~200 turns
> on a FT50-43 toroid which seems physically impossible with magnet wire
> of a gauge I can easily work with, even if I wanted to take the time to
> do it. Maybe need different core material to go that way.
>
> If I couple them that comes down by a factor of 4 which is doable on a
> core that size, kinda.
>

The inductance comes down by a factor of 4, rather, so the windings gets
cut by half.

On Fri, 30 Jul 2021 11:20:20 -0400, bitrex <user@example.net> wrote:

>On 7/30/2021 10:28 AM, jlarkin@highlandsniptechnology.com wrote:
>> On Fri, 30 Jul 2021 09:36:07 -0400, bitrex <user@example.net> wrote:
>>
>>> I have a bunch of these eBay-special flybacks I'd like to slap one to a
>>> piece of FR4 and use for prototyping a thing at +/- 100, but the output
>>> is kinda ugly and has about 3-500mV of ~70kHz noise I want to reduce
>>> before it goes to the circuit.
>>>
>>> <https://www.ebay.com/itm/293851632650>
>>>
>>> Looking for a ballpark for a starting point for the design of the LC, I
>>> have a number of ferrite toroid cores in stock so if it's feasible I'll
>>> just wind it myself like a common-mode choke. Is there a preferable
>>> material for this application? Is it better to have two Cs to ground, a
>>> C across the rail (like a differential filter), or combination?
>>
>> If you output +100 and -100, two Ls and 2 Cs would be better. Just L-C
>> twice, caps grounded.
>>
>
>The inductance values I'm coming up for a Butterworth that gives -40dB
>of rejection at 70kHz, without coupling them end up requiring ~200 turns
>on a FT50-43 toroid which seems physically impossible with magnet wire
>of a gauge I can easily work with, even if I wanted to take the time to
>do it. Maybe need different core material to go that way.
>
>If I couple them that comes down by a factor of 4 which is doable on a
>core that size, kinda.

Coupled inductors, namely a common-mode choke, only filters
common-mode noise.

An LC Butterworth (or other classic form) filter assumes driving and
load impedances. One doesn't usually try to match those in a power
supply filter. A small L and a big electrolytic cap (and maybe a
ceramic too) will filter well and have a low Q.

maybe 10 uH and 100 uF, something like that. Do that in each leg.

--

John Larkin Highland Technology, Inc

The best designs are necessarily accidental.

On 7/30/2021 12:05 PM, jlarkin@highlandsniptechnology.com wrote:
> On Fri, 30 Jul 2021 11:20:20 -0400, bitrex <user@example.net> wrote:
>
>> On 7/30/2021 10:28 AM, jlarkin@highlandsniptechnology.com wrote:
>>> On Fri, 30 Jul 2021 09:36:07 -0400, bitrex <user@example.net> wrote:
>>>
>>>> I have a bunch of these eBay-special flybacks I'd like to slap one to a
>>>> piece of FR4 and use for prototyping a thing at +/- 100, but the output
>>>> is kinda ugly and has about 3-500mV of ~70kHz noise I want to reduce
>>>> before it goes to the circuit.
>>>>
>>>> <https://www.ebay.com/itm/293851632650>
>>>>
>>>> Looking for a ballpark for a starting point for the design of the LC, I
>>>> have a number of ferrite toroid cores in stock so if it's feasible I'll
>>>> just wind it myself like a common-mode choke. Is there a preferable
>>>> material for this application? Is it better to have two Cs to ground, a
>>>> C across the rail (like a differential filter), or combination?
>>>
>>> If you output +100 and -100, two Ls and 2 Cs would be better. Just L-C
>>> twice, caps grounded.
>>>
>>
>> The inductance values I'm coming up for a Butterworth that gives -40dB
>> of rejection at 70kHz, without coupling them end up requiring ~200 turns
>> on a FT50-43 toroid which seems physically impossible with magnet wire
>> of a gauge I can easily work with, even if I wanted to take the time to
>> do it. Maybe need different core material to go that way.
>>
>> If I couple them that comes down by a factor of 4 which is doable on a
>> core that size, kinda.
>
> Coupled inductors, namely a common-mode choke, only filters
> common-mode noise.
>
> An LC Butterworth (or other classic form) filter assumes driving and
> load impedances. One doesn't usually try to match those in a power
> supply filter. A small L and a big electrolytic cap (and maybe a
> ceramic too) will filter well and have a low Q.
>
> maybe 10 uH and 100 uF, something like that. Do that in each leg.

The noise seems to be predominantly common-mode induced from the MOSFET
switching/parasitic capacitance into the ground common to both sides.
These bricks aren't isolated flybacks unfortunately they share a ground
between input and output.

But if 10u is all that's required don't really need to try to make the
inductance smaller by using the same core.

On Fri, 30 Jul 2021 12:35:22 -0400, bitrex <user@example.net> wrote:

>On 7/30/2021 12:05 PM, jlarkin@highlandsniptechnology.com wrote:
>> On Fri, 30 Jul 2021 11:20:20 -0400, bitrex <user@example.net> wrote:
>>
>>> On 7/30/2021 10:28 AM, jlarkin@highlandsniptechnology.com wrote:
>>>> On Fri, 30 Jul 2021 09:36:07 -0400, bitrex <user@example.net> wrote:
>>>>
>>>>> I have a bunch of these eBay-special flybacks I'd like to slap one to a
>>>>> piece of FR4 and use for prototyping a thing at +/- 100, but the output
>>>>> is kinda ugly and has about 3-500mV of ~70kHz noise I want to reduce
>>>>> before it goes to the circuit.
>>>>>
>>>>> <https://www.ebay.com/itm/293851632650>
>>>>>
>>>>> Looking for a ballpark for a starting point for the design of the LC, I
>>>>> have a number of ferrite toroid cores in stock so if it's feasible I'll
>>>>> just wind it myself like a common-mode choke. Is there a preferable
>>>>> material for this application? Is it better to have two Cs to ground, a
>>>>> C across the rail (like a differential filter), or combination?
>>>>
>>>> If you output +100 and -100, two Ls and 2 Cs would be better. Just L-C
>>>> twice, caps grounded.
>>>>
>>>
>>> The inductance values I'm coming up for a Butterworth that gives -40dB
>>> of rejection at 70kHz, without coupling them end up requiring ~200 turns
>>> on a FT50-43 toroid which seems physically impossible with magnet wire
>>> of a gauge I can easily work with, even if I wanted to take the time to
>>> do it. Maybe need different core material to go that way.
>>>
>>> If I couple them that comes down by a factor of 4 which is doable on a
>>> core that size, kinda.
>>
>> Coupled inductors, namely a common-mode choke, only filters
>> common-mode noise.
>>
>> An LC Butterworth (or other classic form) filter assumes driving and
>> load impedances. One doesn't usually try to match those in a power
>> supply filter. A small L and a big electrolytic cap (and maybe a
>> ceramic too) will filter well and have a low Q.
>>
>> maybe 10 uH and 100 uF, something like that. Do that in each leg.
>
>The noise seems to be predominantly common-mode induced from the MOSFET
>switching/parasitic capacitance into the ground common to both sides.
>These bricks aren't isolated flybacks unfortunately they share a ground
>between input and output.
>
>But if 10u is all that's required don't really need to try to make the
>inductance smaller by using the same core.
>

Those values were wild guesses, but not bad ones. They resonate at
about 5 KHz. With a typical electrolytic cap, Q will be low, and a
parallel ceramic cap won't change that much.

On Friday, July 30, 2021 at 9:35:31 AM UTC-7, bitrex wrote:

> The noise seems to be predominantly common-mode induced from the MOSFET
> switching/parasitic capacitance into the ground common to both sides.
> These bricks aren't isolated flybacks unfortunately they share a ground
> between input and output.

So, can you connect switch-side drive and ground through a CM choke? There's
a big gain if you WANT drive and return currents to sum to zero current;
passing power current through an output filter inductor, it's got DC bias...

30.07.21 19:35, bitrex wrote:
>On 7/30/2021 12:05 PM, jlarkin@highlandsniptechnology.com wrote:
>> On Fri, 30 Jul 2021 11:20:20 -0400, bitrex <user@example.net> wrote:
>>
>>> On 7/30/2021 10:28 AM, jlarkin@highlandsniptechnology.com wrote:
>>>> On Fri, 30 Jul 2021 09:36:07 -0400, bitrex <user@example.net> wrote:
>>>>
>>>>> I have a bunch of these eBay-special flybacks I'd like to slap one to a
>>>>> piece of FR4 and use for prototyping a thing at +/- 100, but the output
>>>>> is kinda ugly and has about 3-500mV of ~70kHz noise I want to reduce
>>>>> before it goes to the circuit.
>>>>>
>>>>> <https://www.ebay.com/itm/293851632650>
>>>>>
>>>>> Looking for a ballpark for a starting point for the design of the LC, I
>>>>> have a number of ferrite toroid cores in stock so if it's feasible I'll
>>>>> just wind it myself like a common-mode choke. Is there a preferable
>>>>> material for this application? Is it better to have two Cs to ground, a
>>>>> C across the rail (like a differential filter), or combination?
>>>>
>>>> If you output +100 and -100, two Ls and 2 Cs would be better. Just L-C
>>>> twice, caps grounded.
>>>>
>>>
>>> The inductance values I'm coming up for a Butterworth that gives -40dB
>>> of rejection at 70kHz, without coupling them end up requiring ~200 turns
>>> on a FT50-43 toroid which seems physically impossible with magnet wire
>>> of a gauge I can easily work with, even if I wanted to take the time to
>>> do it. Maybe need different core material to go that way.
>>>
>>> If I couple them that comes down by a factor of 4 which is doable on a
>>> core that size, kinda.
>>
>> Coupled inductors, namely a common-mode choke, only filters
>> common-mode noise.
>>
>> An LC Butterworth (or other classic form) filter assumes driving and
>> load impedances. One doesn't usually try to match those in a power
>> supply filter. A small L and a big electrolytic cap (and maybe a
>> ceramic too) will filter well and have a low Q.
>>
>> maybe 10 uH and 100 uF, something like that. Do that in each leg.
>
>The noise seems to be predominantly common-mode induced from the MOSFET
>switching/parasitic capacitance into the ground common to both sides.
>These bricks aren't isolated flybacks unfortunately they share a ground
>between input and output.
>
>But if 10u is all that's required don't really need to try to make the
>inductance smaller by using the same core.
>
>
Is it noise from the FET or from the transformer?

That size transformers look like the capacitance from the transformer would dominate

A solution could be to shield the core. Not a complete turn...

--
Klaus

On 8/4/2021 6:30 PM, Klaus Kragelund wrote:
> 30.07.21 19:35, bitrex  wrote:
>> On 7/30/2021 12:05 PM, jlarkin@highlandsniptechnology.com wrote:
>>> On Fri, 30 Jul 2021 11:20:20 -0400, bitrex <user@example.net> wrote:
>>>
>>>> On 7/30/2021 10:28 AM, jlarkin@highlandsniptechnology.com wrote:
>>>>> On Fri, 30 Jul 2021 09:36:07 -0400, bitrex <user@example.net> wrote:
>>>>>
>>>>>> I have a bunch of these eBay-special flybacks I'd like to slap one
>>>>>> to a
>>>>>> piece of FR4 and use for prototyping a thing at +/- 100, but the
>>>>>> output
>>>>>> is kinda ugly and has about 3-500mV of ~70kHz noise I want to reduce
>>>>>> before it goes to the circuit.
>>>>>>
>>>>>> <https://www.ebay.com/itm/293851632650>
>>>>>>
>>>>>> Looking for a ballpark for a starting point for the design of the
>>>>>> LC, I
>>>>>> have a number of ferrite toroid cores in stock so if it's feasible
>>>>>> I'll
>>>>>> just wind it myself like a common-mode choke. Is there a preferable
>>>>>> material for this application? Is it better to have two Cs to
>>>>>> ground, a
>>>>>> C across the rail (like a differential filter), or combination?
>>>>>
>>>>> If you output +100 and -100, two Ls and 2 Cs would be better. Just L-C
>>>>> twice, caps grounded.
>>>>>
>>>>
>>>> The inductance values I'm coming up for a Butterworth that gives -40dB
>>>> of rejection at 70kHz, without coupling them end up requiring ~200
>>>> turns
>>>> on a FT50-43 toroid which seems physically impossible with magnet wire
>>>> of a gauge I can easily work with, even if I wanted to take the time to
>>>> do it. Maybe need different core material to go that way.
>>>>
>>>> If I couple them that comes down by a factor of 4 which is doable on a
>>>> core that size, kinda.
>>>
>>> Coupled inductors, namely a common-mode choke, only filters
>>> common-mode noise.
>>>
>>> An LC Butterworth (or other classic form) filter assumes driving and
>>> load impedances. One doesn't usually try to match those in a power
>>> supply filter. A small L and a big electrolytic cap (and maybe a
>>> ceramic too) will filter well and have a low Q.
>>>
>>> maybe 10 uH and 100 uF, something like that. Do that in each leg.
>>
>> The noise seems to be predominantly common-mode induced from the
>> MOSFET switching/parasitic capacitance into the ground common to both
>> sides. These bricks aren't isolated flybacks unfortunately they share
>> a ground between input and output.
>>
>> But if 10u is all that's required don't really need to try to make the
>> inductance smaller by using the same core.
>>
>>
> Is it noise from the FET or from the transformer?

Beats me. How do I check?

> That size transformers look like the capacitance from the transformer
> would dominate
> A solution could be to shield the core. Not a complete turn...
>
> --
> Klaus

On Wednesday, August 4, 2021 at 3:30:55 PM UTC-7, Klaus Kragelund wrote:
> 30.07.21 19:35, bitrex wrote:
> >On 7/30/2021 12:05 PM, jla...@highlandsniptechnology.com wrote:

> >>>> On Fri, 30 Jul 2021 09:36:07 -0400, bitrex <us...@example.net> wrote:
> >>>>
> >>>>> I have a bunch of these eBay-special flybacks I'd like to slap one to a
> >>>>> piece of FR4 and use for prototyping a thing at +/- 100, but the output
> >>>>> is kinda ugly and has about 3-500mV of ~70kHz noise I want to reduce
> >>>>> before it goes to the circuit.

> >The noise seems to be predominantly common-mode induced from the MOSFET
> >switching/parasitic capacitance into the ground common to both sides.
> >These bricks aren't isolated flybacks unfortunately they share a ground
> >between input and output.
> >
> >But if 10u is all that's required don't really need to try to make the
> >inductance smaller by using the same core.

> Is it noise from the FET or from the transformer?
>
> That size transformers look like the capacitance from the transformer would dominate
>
> A solution could be to shield the core. Not a complete turn...

Also to consider, the inner winding's outer wrap should be ground (or unswitched power), and the outer winding's
inner wrap as well. That doesn't take hardware, just attention to which lead is which.

For radiated emission, a complete turn wrap IS appropriate, around not just the center leg
of the E's, but around the whole thing.