inductor sizing

[John Fields]

Gosh, you sure have a way with cliches.

 

[John Larkin]

On purpose, usually.

John

 

[Harry Dellamano]

Sounds like a deal as long as you promise not to drink it all in one night


Just tell me where to send and your favorite type!

cheers,
Jamie

Ok Jamie,
You sound like a nice person so here is a filter design that will meet your
needs with a bottle of tequila to be sent upon the design's approval. This
is the first LC section for one phase and will survive the 200KHz/440VDC
stress and present a nice 60Hz sine wave to the second LC section input
which you can design using high flux 60Hz material to meet you unspecified
output requirements. As stated earlier the THD depends on the control loop
and not the HF output filter.
For the 25uH/30Apk inductor use a Ferroxcube or similar EC-70 (3F3
material) core set and bobbin with 21 turns of #10 AWG LITZ wire wound in 3
layers of 7turns/layer. Insulate all for 3500VDC. The LITZ wire should be
made up with 256/#34 AWG wires. This will require a 0.21" gap across all
three legs of the core. Do not grind the center leg for a 0.42" gap because
this will cause increased Rac losses. The total dissipation will be <7.0W so
be aware.
For the shunt caps use 5X EPCOS 2.2uF/400VDC MetalPoly at DigiKey for
$1USD/100 PN B32594C6225J008. Irms in each cap is ~=1A at 200KHz which
MetalPoly types of this size can handle.

Enjoy,
Harry

 

[John Fields]

On purpose, usually.

 

[Tom Bruhns]

Interesting, because of the PWM and duty>90%, there are a ton of harmonics.
To do a good job attenuating the first harmonic the smaller LC would have to
start early so the fundamental trap may not be too effective. Do not quote
me on this!
Cheers,
Harry

One reason to stay with smaller inductor values is to minimize the
parasitics, so they perform well at high frequencies. Given the
200kHz fundamental, it doesn't take all that much inductance/
capacitance to kill the harmonics. There are "a ton of" harmonics at
50 percent duty cycle, too; it's more a matter of there being MORE
fundamental at 50 percent duty cycle. For killing harmonics, a
multiple-stage filter is far more efficient than a single L section.
You can add zeros to the filter if you want, to kill lower frequencies
(i.e. the fundamental) more effectively, and the tradeoff is lower
attenuation at higher frequencies (which are likely lower amplitude to
begin with, so the higher attenuation isn't really necessary there).

Although I probably wouldn't seriously do a power supply like this
with an air core coil, I thought it worthwhile to give as an example
because it sets an upper limit on the required coil size. I don't
really agree that an air coil would be "impossible" to shield; I'm
quite used to getting in excess of 120dB shielding on filters using
air core coils, and since it's a linear system, I'd expect to be able
to do that without too much trouble for such a power supply filter, if
I needed to. I made a crude estimate of the Q of an air core coil
like the one I suggested, and decided it would be fairly easy, using
Litz wire, to get it above 100 at 200kHz.

Cheers,
Tom

 

[Harry Dellamano]

One reason to stay with smaller inductor values is to minimize the
parasitics, so they perform well at high frequencies. Given the
200kHz fundamental, it doesn't take all that much inductance/
capacitance to kill the harmonics. There are "a ton of" harmonics at
50 percent duty cycle, too; it's more a matter of there being MORE
fundamental at 50 percent duty cycle. For killing harmonics, a
multiple-stage filter is far more efficient than a single L section.
You can add zeros to the filter if you want, to kill lower frequencies
(i.e. the fundamental) more effectively, and the tradeoff is lower
attenuation at higher frequencies (which are likely lower amplitude to
begin with, so the higher attenuation isn't really necessary there).

Although I probably wouldn't seriously do a power supply like this
with an air core coil, I thought it worthwhile to give as an example
because it sets an upper limit on the required coil size. I don't
really agree that an air coil would be "impossible" to shield; I'm
quite used to getting in excess of 120dB shielding on filters using
air core coils, and since it's a linear system, I'd expect to be able
to do that without too much trouble for such a power supply filter, if
I needed to. I made a crude estimate of the Q of an air core coil
like the one I suggested, and decided it would be fairly easy, using
Litz wire, to get it above 100 at 200kHz.

Cheers,
Tom

This filter is interesting because the first inductor is about protecting
the first cap's RMS ripple current at 200KHz. The design iterations are;
select a size and cost of first cap with as much Irms current as possible.
Calculate input inductor necessary to maintain the cap's ripple current. If
inductor size is too big go back and increase the cap's size and recalculate
the inductor. When you are happy with the size, cost and temp rise of the
first stage, then put on your 60Hz hat and design the second stage for
attenuation requirement.
The first stage air core sounds outrageous to me. The air core is a primary
winding with 220VAC, 25A at 200KHz. Every loop of wire within 6" of it would
be a secondary. A u-metal shield would be necessary. We have problems with
radiated magnetic fields when we gap a power transformer by >0.050", your
air core has a >2" gap.
Regards,
Harry

 

[Eeyore]

Also I need some toroid cores for 200kHz LC filters, 250uH and 50Amps
peak current, are these going to be huge cores and what core materials
would be best to use, ferrite or powder cores?

Ouch ! Those will be BIG !

I just did some 25uH, 25 Amp toroids using powder cores and they're on
Micrometals 1.87 inch cores.

Graham

 

[Eeyore]

The inductor is for generating 120VAC at 60Hz:

I reckon you need a big laminated EI core for that.

Graham

 

[Tom Bruhns]

This filter is interesting because the first inductor is about protecting
the first cap's RMS ripple current at 200KHz. The design iterations are;
select a size and cost of first cap with as much Irms current as possible.
Calculate input inductor necessary to maintain the cap's ripple current. If
inductor size is too big go back and increase the cap's size and recalculate
the inductor. When you are happy with the size, cost and temp rise of the
first stage, then put on your 60Hz hat and design the second stage for
attenuation requirement.
The first stage air core sounds outrageous to me. The air core is a primary
winding with 220VAC, 25A at 200KHz. Every loop of wire within 6" of it would
be a secondary. A u-metal shield would be necessary. We have problems with
radiated magnetic fields when we gap a power transformer by >0.050", your
air core has a >2" gap.
Regards,
Harry

So, I agree with the first stage design. If I'm not mistaken, the
Wima MKP10 caps I mentioned in an earlier posting can handle the sort
of ripple current that even a very small inductance would yield, given
a small stack of parallel ones.

As for designing the following stage for 60Hz, I gotta disagree.
That's what control systems are for. But because of parasitics that
are bound to be in the layout and the parts, a second stage is very
useful for further reducing the 200kHz+harmonics, as I see it. I
specifically do NOT want a low-frequency filter there, or it will play
havoc with the control system design.

I'll repeat that it's unlikely I'd actually use an air core coil in
that application, but to me, it's far from outrageous. People have
been building resonant tank circuits using air-core coils that handle
a lot more power than we're talking about here (and at a lot higher
loaded Q, so there's much more energy stored in the coil's field), and
appropriate shielding hasn't been all that difficult to achieve.
Please correct me if I'm wrong, but as far as I know, linearity holds
at these still fairly modest power levels, so the shielding that gives
me 120dB attenuation for moderately large air core coils at low power
(I use them for high Q; air core for extremely low distortion) should
give me similar attenuation for considerably higher power.

Cheers,
Tom

 

[John Larkin]

Well, it's more fun than being intentionally obtuse. Pays better, too.

John

 

[Terry Given]

Jamie Morken wrote:




I reckon you need a big laminated EI core for that.

Graham

once it catches fire, go build Harrys inductor.

Cheers
Terry

 

[John Fields]

Well, it's more fun than being intentionally obtuse. Pays better, too.

 

[John Larkin]

What matters is electronic design. The "electronic" thing is the easy
part; most any college will teach you the mechanics of that in four
years, more theory than you're likely to need. The hard part is
"design", which involves branching out, exploring strange
possibilities, understanding a lot more than the theory, riffing on
silly ideas in all directions, alone or with others, hoping to evolve
or stumble on a gem.

Namely, design involves doing the things you consider to be
"intentionally devious" and "bobbing and weaving."

Now, what do you have against bucks?

John

 

[John Larkin]

once it catches fire, go build Harrys inductor.

Cheers
Terry

Eeyore's world ends at 20 KHz.

John

 

[Harry Dellamano]

So, I agree with the first stage design. If I'm not mistaken, the
Wima MKP10 caps I mentioned in an earlier posting can handle the sort
of ripple current that even a very small inductance would yield, given
a small stack of parallel ones.

As for designing the following stage for 60Hz, I gotta disagree.
That's what control systems are for. But because of parasitics that
are bound to be in the layout and the parts, a second stage is very
useful for further reducing the 200kHz+harmonics, as I see it. I
specifically do NOT want a low-frequency filter there, or it will play
havoc with the control system design.

I'll repeat that it's unlikely I'd actually use an air core coil in
that application, but to me, it's far from outrageous. People have
been building resonant tank circuits using air-core coils that handle
a lot more power than we're talking about here (and at a lot higher
loaded Q, so there's much more energy stored in the coil's field), and
appropriate shielding hasn't been all that difficult to achieve.
Please correct me if I'm wrong, but as far as I know, linearity holds
at these still fairly modest power levels, so the shielding that gives
me 120dB attenuation for moderately large air core coils at low power
(I use them for high Q; air core for extremely low distortion) should
give me similar attenuation for considerably higher power.

Cheers,
Tom

I totally agree with your second stage design. When I stated "60 Hz hat", I
meant the core material can be 60Hz material but the corner frequency is
still up at maybe 30KHz with low parasitic capacity to get the required
200KHz attenuation. As we both stated 60Hz THD requirements must be meat
with loop design. Wima caps are a good alternative but cost will start to
soar.
As for air cores, I do not disagree with anything you have stated above.
but size matters to further this discussion. The current levels are very
high, wiring loops areas must be keep as small as possible so they no not
act as primaries nor secondary windings. In so doing the inductor is in very
close proximity with many other windings with no room for shielding.
I would be very interested in the size of the air core inductor compared to
the gapped ferrite which is about a three inch cube. The self resonate
frequency, which I will calculate for the ferrite, would be an interesting
data point.

Regards,
Harry

 

[John Fields]

What matters is electronic design. The "electronic" thing is the easy
part; most any college will teach you the mechanics of that in four
years, more theory than you're likely to need. The hard part is
"design", which involves branching out, exploring strange
possibilities, understanding a lot more than the theory, riffing on
silly ideas in all directions, alone or with others, hoping to evolve
or stumble on a gem.

Namely, design involves doing the things you consider to be
"intentionally devious" and "bobbing and weaving."

---
Not at all, and unless you're very naive, you know very well that's
not what I'm talking about.

What I consider to be intentionally devious is when you make a
statement like, "Namely, design involves doing the things you
consider to be "intentionally devious" and "bobbing and weaving." in
order to try to save face and steer the discussion in the way you
want even though you know there's no truth in what you say when you
try to make it seem like being intentionally devious and bobbing and
weaving are part of the design process.

It isn't, and that's not for me.

As far as I'm concerned, design is the search for truth and
deviousness and bobbing and weaving have no welcome place along that
path.

YMMV, since you seem to define deviousness and bobbing and weaving
in terms other than those generally accepted.

 

[John Larkin]

---
Not at all, and unless you're very naive, you know very well that's
not what I'm talking about.

What I consider to be intentionally devious is when you make a
statement like, "Namely, design involves doing the things you
consider to be "intentionally devious" and "bobbing and weaving." in
order to try to save face and steer the discussion in the way you
want even though you know there's no truth in what you say when you
try to make it seem like being intentionally devious and bobbing and
weaving are part of the design process.

It isn't, and that's not for me.

As far as I'm concerned, design is the search for truth and
deviousness and bobbing and weaving have no welcome place along that
path.

YMMV, since you seem to define deviousness and bobbing and weaving
in terms other than those generally accepted.

As I said, what I consider to be the design process is what you seem
to define as "intentionally devious", which you have now further
morphed into "ill-gotten." This gets better and better.

We design and build electronics. We have specs, manuals, and a price
list. If people want to buy our stuff, they can; if they don't like
the price, they can shop around for a better deal. If they want to try
a loner, to see if it works as claimed, we provide one. If anything
goes wrong, we fix it fast. If the customer isn't happy with one of
our products, we take it back and refund all their money. If they ask
a question, we tell them the truth.

The design part is a lot more "devious", but that's not the customers
concern. The customer pays for a product that performs a function.
Since we never sell designs, only products, so the customer has no
reason to see the process.

But I'm excessively proud of myself today. I learned how to mud
sheetrock, and it came out pretty good.

John

 

[John Larkin]

I totally agree with your second stage design. When I stated "60 Hz hat", I
meant the core material can be 60Hz material but the corner frequency is
still up at maybe 30KHz with low parasitic capacity to get the required
200KHz attenuation. As we both stated 60Hz THD requirements must be meat
with loop design. Wima caps are a good alternative but cost will start to
soar.
As for air cores, I do not disagree with anything you have stated above.
but size matters to further this discussion. The current levels are very
high, wiring loops areas must be keep as small as possible so they no not
act as primaries nor secondary windings. In so doing the inductor is in very
close proximity with many other windings with no room for shielding.
I would be very interested in the size of the air core inductor compared to
the gapped ferrite which is about a three inch cube. The self resonate
frequency, which I will calculate for the ferrite, would be an interesting
data point.

Regards,
Harry

Does an air-core toroid make any sense? I don't think I've ever seen
one. Micrometals does have a no-op core material, u=1, made out of
phenolic or something.

John

 

[Terry Given]

Eeyore's world ends at 20 KHz.

John

isnt that infra-dc for some of your "DC"-GHz MMICs?

Cheers
Terry

 

[John Larkin]

isnt that infra-dc for some of your "DC"-GHz MMICs?

Cheers
Terry

I'm not sure what "DC" actually means. The RF people commonly rate
their parts to DC, but all the test circuits include coupling caps,
and bias levels are seldom specified. We use one Hittite part that's
rated, in huge typefont, "DC to 3 GHz" but absolutely screws up below
100 MHz.

We're lately playing with compound amplifiers, where you split a
signal with lowpass/highpass filters, amplify the slow and fast parts
separately, and then combine the outputs. That can get a little
tricky.

John