SMPS current transformer help

[Yzordderex]

Greetings,

I need to come up with a current transformer for a smps. Power supply
is a full bridge, phase shifted resonant zvs design. Will be used for
a modulator for a class e transmitter. Output 0-60v about 250-350w at
60v.

The switching frequency will be around 200 khz, and I expect waveform
to be bipolar pulses with a ramp on top. Peak currents with 115v
input design will be around 3 amps. Output of current transformer
going into a diode bridge using 4148 diodes. Output of bridge to be
about 2-3v at Ipeak.

I'm thinking of using a ferrite core with single turn on primary and
multiple turns on secondary with a load resistor at the output of the
4148 diode bridge.

I need to determine:
Core size
Core material
number secondary turns
Load resistance

My first guess would be to pick core that wouldn't saturate with max
N*I, but not enough experience with this type design. I'm hoping to
do away with the trial by error solution and start with something
close.

Design should (if possible) be something easy for the home
experimenter to wind.

regards,
Bob

 

[Harry Dellamano]

Greetings,

I need to come up with a current transformer for a smps. Power supply
is a full bridge, phase shifted resonant zvs design. Will be used for
a modulator for a class e transmitter. Output 0-60v about 250-350w at
60v.

The switching frequency will be around 200 khz, and I expect waveform
to be bipolar pulses with a ramp on top. Peak currents with 115v
input design will be around 3 amps. Output of current transformer
going into a diode bridge using 4148 diodes. Output of bridge to be
about 2-3v at Ipeak.

I'm thinking of using a ferrite core with single turn on primary and
multiple turns on secondary with a load resistor at the output of the
4148 diode bridge.

I need to determine:
Core size
Core material
number secondary turns
Load resistance

My first guess would be to pick core that wouldn't saturate with max
N*I, but not enough experience with this type design. I'm hoping to
do away with the trial by error solution and start with something
close.

Design should (if possible) be something easy for the home
experimenter to wind.

regards,
Bob

Hi Bob,
Instead of sensing current in the bridge load, transformer primary, with a
current transformer, why not put a 0.050 ohm 1W resistor in the return legs
to ground. A gain of 20 amp will give you 1V/A output. Much simpler,
cheaper, smaller, will not saturate, works if bridge gets stuck in DC mode.

harry

 

[Yzordderex]

Snip----
--unsnip

Hi Bob,
Instead of sensing current in the bridge load, transformer primary, with a
current transformer, why not put a 0.050 ohm 1W resistor in the return legs
to ground. A gain of 20 amp will give you 1V/A output. Much simpler,
cheaper, smaller, will not saturate, works if bridge gets stuck in DC mode.

harry

Harry,

Thanks for response.

Are you asking why not put resistors in the two lower mosfet sources,
or the center tap of the transformer secondary? Either might work,
but..

Control circuit is at earth ground. Secondary side of the bridge
driving the RF deck will not be at earth. Reason for the control ckt
being at ground is that microphone and audio processing circuitry
should be at earth ground for safety.

Resistors would probably work if my control circuit was at negative
bus. Could probably have used one and tie the lower mosfet sources
together. Only prob with resistor is that have to pay attention to
part inductance if high di/dt. Perhaps it would have taked an SMD
part, or a bunch of through hole parts in parallel to reduce
inductance.

I suppose I could have put control ckt at negative bus and used
optocoupler to couple audio to it. In any case pc control board has
been prototyped, stuffed, and tested. So just looking for a little
help to complete current sense transformer design.

thanks again
Bob
N9NEO

 

[JeffM]

Yup. Even with the differential amp you'll need, its going to be cheaper
--and WAY easier to engineer (not to mention getting/stocking parts).

 

[R.Legg]

Harry,

Thanks for response.

Are you asking why not put resistors in the two lower mosfet sources,
or the center tap of the transformer secondary? Either might work,
but..

Control circuit is at earth ground. Secondary side of the bridge
driving the RF deck will not be at earth. Reason for the control ckt
being at ground is that microphone and audio processing circuitry
should be at earth ground for safety.

Resistors would probably work if my control circuit was at negative
bus. Could probably have used one and tie the lower mosfet sources
together. Only prob with resistor is that have to pay attention to
part inductance if high di/dt. Perhaps it would have taked an SMD
part, or a bunch of through hole parts in parallel to reduce
inductance.

I suppose I could have put control ckt at negative bus and used
optocoupler to couple audio to it. In any case pc control board has
been prototyped, stuffed, and tested. So just looking for a little
help to complete current sense transformer design.

thanks again
Bob
N9NEO

A simple current transformer won't handle DC motor current, or
unipolar switch duty cycles larger than about 90% (the remaining
period is required to reset the core).

Harry was giving you the straight goods.

There are a number of methods of getting DC value measurements using a
transformer, but none of them are simple enough to use in a
lower-power lower-cost sensing application.

If indication-only is required, the mosfet on-voltage could be
carefully sampled; series drops in board tracking can also serve as
crude sensors.

RL

 

[R.Legg]

Snip----


Resistors would probably work if my control circuit was at negative
bus. Could probably have used one and tie the lower mosfet sources
together. Only prob with resistor is that have to pay attention to
part inductance if high di/dt. Perhaps it would have taked an SMD
part, or a bunch of through hole parts in parallel to reduce
inductance.

I suppose I could have put control ckt at negative bus and used
optocoupler to couple audio to it. In any case pc control board has
been prototyped, stuffed, and tested. So just looking for a little
help to complete current sense transformer design.

Sorry - just read the posting a bit more carefully.

With a phase modulated primary, the output is a full-wave rectifier -
so you can put a current transformer there and full-wave rectify it's
isolated output to anywhere in the circuit you care to place it. Each
half wave reverses current flow, and the peak output is a reflection
of the DC output current in your choke.

For this there's basically one formula

Nmin= V x t / ( B x A)

V = peak current sensing output voltage plus two diode drops (fw
rectified sensor output) (volts)
t = 1 / 2 x conversion frequency (seconds)
B = 2xBsat of the core material in the current transformer (tesla)
Ferrite=0.6T
A = Cross sectional area of the current transformer core (meters^2)
N = minimum sense winding turns.

The more turns, the larger the current sensing resistor and it's
reporting voltage can be, and the lower the power loss.

More turns means more expense to fabricate. You might try some of the
double insulated through-hole commodity parts from Coilcraft,
Coiltronics, Pulse Engineering and the like.

Note that there could be a substantial contribution to output
rectifier winding leakage inductance created by the insertion of the
sensing transformer in this location. Take care in selecting effective
snubber positions.

Reverse recovery currents will likely also show up. A suitable RC
filter on the sensor or downstream should take care of the latter.

RL

 

[Fritz Schlunder]

Greetings,

I need to come up with a current transformer for a smps. Power supply
is a full bridge, phase shifted resonant zvs design. Will be used for
a modulator for a class e transmitter. Output 0-60v about 250-350w at
60v.

The switching frequency will be around 200 khz, and I expect waveform
to be bipolar pulses with a ramp on top. Peak currents with 115v
input design will be around 3 amps. Output of current transformer
going into a diode bridge using 4148 diodes. Output of bridge to be
about 2-3v at Ipeak.

I'm thinking of using a ferrite core with single turn on primary and
multiple turns on secondary with a load resistor at the output of the
4148 diode bridge.

I need to determine:
Core size
Core material
number secondary turns
Load resistance

My first guess would be to pick core that wouldn't saturate with max
N*I, but not enough experience with this type design. I'm hoping to
do away with the trial by error solution and start with something
close.

Design should (if possible) be something easy for the home
experimenter to wind.

regards,
Bob

I've never designed one of these either, I've always favored sense
resistors. Though I've been interested in them from time to time, I've
never read anything about their design simply because I've never seen any
information about them before (although I've read plenty of SMPS books/web
pages/app notes and electrical circuits books, but noone ever covers them).
Nevertheless after thinking about your post I think I understand their
operation.

Unless I'm mistaken they opeate just like regular voltage conversion
transformers except they operate with current. In the case of voltage,
Vin*turns ratio=Vout is the formula to use. In the current transformer
Iin=(turns ratio)^2*Iout is the formula of interest. In your case Iin is
3Amps (your highest value) and you get to decide what Iout should be
depending upon your load resistor you select. Once you find the load
resistance (you somewhat arbitrarily select it) you can find the turns
ratio, and then with that you can find the secondary turns. With that
information you can use the formula provided by R. Legg to select a core
size that will not saturate.

The core material should probably be as high permeability ferrite as
feasible with low loss at your operating frequency. It should also have as
linear as possible a B versus magnetizing current curve. In other words
don't operate the core near saturation where things get non-linear for best
current measurement accurary.

So for example:

Abritrarily select the load resistor as 1k Ohms, to be placed after the
diode bridge. You want about a 3V signal at 3Amps primary current, so you
need a current through the 1K ohm resistor of 3mA. So Iout=3mA. Iin=3Amps.
n^2 therefore equals 3/0.003=1000. So n (the turns ratio)=32.

Since turns primary=1, turns secondary is 32 (although since flux coupling
with only one turn primary will likely produce high leakage, this may need
to be tweaked somewhat, but since I have no experience building these I
don't know by how much to expect). Now lets use R.Legg's formula. Lets
rewrite it like this for this example:

(cross sectional area)=(voltage)/[(2)(frequency)(turns)(delta B max)]

Everything in SI units.

Since we want to operate the core well away from saturation for linearity,
lets pick a delta B max of say 0.3Tesla instead of the 0.6T value mentioned
in R. Legg's post. The frequency is 200kHz as you gave it to us. The
secondary voltage produced needs to be 3V (3mA @ 1k load) plus two diode
drops when the primary current is 3A. This means the secondary voltage is
around 4.5V or so. However, since the primary current isn't a clean
squarewave of 3A constant magnitude with full effective duty cycle, but
often less than this for part of the cycle, if we use 4.5V as our voltage in
the forumla the cross sectional area will be larger than we really need.
Nevertheless since we don't precisely know what the current waveform will be
in worst case situations, we will just use that 4.5V value. Too large a
cross sectional area won't hurt, too small will.

So, plugging in:

Ae=(4.5V)/(2)(200000)(32)(0.3)
Ae=1.2*10^(-6) square meters.

By the way... Make sure the frequency figure you use is the frequency seen
by the transformer. In a full bridge topology the primary MOSFET switching
frequency is half the output rectifier/filter frequency. When some people
say they have a 200kHz full bridge, sometimes they mean if you hook an
oscilloscope up to any of the primary side MOSFETs you will see 200kHz,
while others mean if you look at the ripple current through the output
filter inductor you will see it as being at 200kHz (but only 100kHz if you
look at the MOSFETs on the primary side).

 

[Yzordderex]

Yup. Even with the differential amp you'll need, its going to be cheaper
--and WAY easier to engineer (not to mention getting/stocking parts).

Jeff,

Thanks for reply.

Once engineering for transformer is done, it's done. A small value
shunt resistor in SMD package is $0.26 in quantities of 100k. I
expect a through-hole part to be a bit more expensive, but somewhat
more readily available. Haven't checked Digi-key for either. Ferrite
core for application probably around $0.50 Cores are easy to get in
small quantities from Amidon.

Problem with SMD part is that it is smd. While the control circuits
with mosfet pulse drive transformers are on a pc board, the power
components are mounted on a chassis. Hard to mount smd part to
chassis. Problem with through-hole part is inductance. Any
inductance will disturb the waveform I'm trying to measure.

Since this project is geared towards the Ham/experimentor I'm not too
concerned if current sensing costs $0.26, or $0.96 Parts cost for
the existing power supply and modulator which most are running is (my
best guess) near $200.00 There is an expensive array of electrolytic
capacitors and a very expensive steel wound toroidal transformer.
These expensive parts take up lots of room too. I expect to get price
down to around $50 with this new design. I should be able to cut
volume down to 25%.

Now if I were to go with current sense resistor scheme and diff amp I
would still have to get signal from negative bus, or transformer
primary, to the control circuit at earth ground. I would have to
amplify the signal at the resistor. A single-ended amp might be ok to
gain up signal, but then would still have to get over to earth ground.
Would need a differential amp to do that. Remember that the signal I
am trying to get across needs a very high bandwith amplifier and will
have some noise on it. Can't really filter out noise as I will lose
my signal too.

I'm hoping I can wind a toroid with some number of turns, rectify and
stick a small value resistor on rectifier output to load the core.
While my first post said I would need to pay attention to N*I, I'm
thinking if the core secondary is loaded, probably don't have to pay
too much attention even to N*I. Just find a core geometry that will
fit turns and maybe a reasonable ferrite material. Probably don't
even have to pay too much attention to material as flux density *
frequency losses not so high.

So had I put all power components on a pc board, and referenced
controller to minus bus I would agree that resistor sensing would
probably have been best route.

With design as it is, I think (hope) transformer is best way.

thanks again,
Bob
N9NEO

 

[Fritz Schlunder]

I've never designed one of these either, I've always favored sense

resistors. Though I've been interested in them from time to time, I've
never read anything about their design simply because I've never seen any
information about them before (although I've read plenty of SMPS books/web
pages/app notes and electrical circuits books, but noone ever covers them).
Nevertheless after thinking about your post I think I understand their
operation.

Unless I'm mistaken they opeate just like regular voltage conversion
transformers except they operate with current. In the case of voltage,
Vin*turns ratio=Vout is the formula to use. In the current transformer
Iin=(turns ratio)^2*Iout is the formula of interest. In your case Iin is
3Amps (your highest value) and you get to decide what Iout should be
depending upon your load resistor you select. Once you find the load
resistance (you somewhat arbitrarily select it) you can find the turns
ratio, and then with that you can find the secondary turns. With that
information you can use the formula provided by R. Legg to select a core
size that will not saturate.

Whoops. I think I made a rather serious blunder (like I say I've never
built one of these before). The current between primary and secondary isn't
proportional to the turns ratio squared. I was thinking impedances for some
odd reason. The formula should have been Iin=(turns ratio)*Iout.

So perhaps we should change the example load resistance to something smaller
like 100 Ohms for instance.

3V/100 Ohms=30mA
Turns ratio=3A/30mA=100
Secondary turns=100
Ae=4.5/2*200000*0.3*100=3.8*10^(-7) square meters

 

[Paul Mathews]

Hi Bob,
Instead of sensing current in the bridge load, transformer primary, with a
current transformer, why not put a 0.050 ohm 1W resistor in the return legs
to ground. A gain of 20 amp will give you 1V/A output. Much simpler,
cheaper, smaller, will not saturate, works if bridge gets stuck in DC mode.

harry

Nice little 100:1 CTs are available from Pulse Engrg, Datatronics, and
many others at reasonable cost, but I agree with harry about the
sensing resistor having advantages. On the other hand, at this power
level, why bother with the complex, expensive, hard to troubleshoot
phase-shifted full bridge ZVS? You'll find forward converters much
easier to work with and nearly as efficient. With the availability of
higher performance MOSFETs and rectifiers, the full bridge ZVS is
seldom used anymore at less than 1kW or so.
Paul Mathews

 

[Paul Mathews]

Hi Bob,
Instead of sensing current in the bridge load, transformer primary, with a
current transformer, why not put a 0.050 ohm 1W resistor in the return legs
to ground. A gain of 20 amp will give you 1V/A output. Much simpler,
cheaper, smaller, will not saturate, works if bridge gets stuck in DC mode.

harry

Nice little 100:1 CTs are available from Pulse Engrg, Datatronics, and
many others at reasonable cost, but I agree with harry about the
sensing resistor having advantages. On the other hand, at this power
level, why bother with the complex, expensive, hard to troubleshoot
phase-shifted full bridge ZVS? You'll find forward converters much
easier to work with and nearly as efficient. With the availability of
higher performance MOSFETs and rectifiers, the full bridge ZVS is
seldom used anymore at less than 1kW or so.
Paul Mathews

 

[legg]

Abritrarily select the load resistor as 1k Ohms, to be placed after the
diode bridge. You want about a 3V signal at 3Amps primary current, so you
need a current through the 1K ohm resistor of 3mA. So Iout=3mA. Iin=3Amps.
n^2 therefore equals 3/0.003=1000. So n (the turns ratio)=32.

Is = Np/Ns x Ip

100:1 turns is 1:100 current

At 32 turns, 6A primary generates 190mA in the sense winding. A 3V
signal requires a 16R 1W resistor. This is fairly watty and may be
pushing 4148 sense-side full-wave rectifiers, though they only see 50%
duty.

Core cross-section for 5uSec (100KHz) is 1.3E-6m^2 or ~2mm^2.

The core is obviously going to be bigger than that (below a certain
size, transformer fab cost starts to go up again). The bigger the core
and the larger the turns, the farther away from saturation you'll run,
which is always a good idea if cycle skipping or other abnormalities
are expected.

Larger turns and a resulting larger magnetizing inductance also
reduces pulse droop in the sensor, due to magnetization current
subtracting from the output signal (remember that the ratio of Lm/R is
a time constant). It will ,however, result in higher leakage
inductance terms that will show up in the power rectifier circuit.

Nmin is a limit condition only and has more relevance in lower
frequency applications - but you always check it first.

With the sensed current reversing, or two reversed phase windings
carrying reversed currents, 2xsat flux swing is possible without loss
of volt-seconds.

If very fine wire or a large number of turns is used in the sense
winding, it will contribute to the volt-second burden.

RL

 

[Harry Dellamano]

[email protected] (Yzordderex) wrote in message

Sorry - just read the posting a bit more carefully.

With a phase modulated primary, the output is a full-wave rectifier -
so you can put a current transformer there and full-wave rectify it's
isolated output to anywhere in the circuit you care to place it. Each
half wave reverses current flow, and the peak output is a reflection
of the DC output current in your choke.

For this there's basically one formula

Nmin= V x t / ( B x A)

V = peak current sensing output voltage plus two diode drops (fw
rectified sensor output) (volts)
t = 1 / 2 x conversion frequency (seconds)
B = 2xBsat of the core material in the current transformer (tesla)
Ferrite=0.6T
A = Cross sectional area of the current transformer core (meters^2)
N = minimum sense winding turns.

The more turns, the larger the current sensing resistor and it's
reporting voltage can be, and the lower the power loss.

More turns means more expense to fabricate. You might try some of the
double insulated through-hole commodity parts from Coilcraft,
Coiltronics, Pulse Engineering and the like.

Note that there could be a substantial contribution to output
rectifier winding leakage inductance created by the insertion of the
sensing transformer in this location. Take care in selecting effective
snubber positions.

Reverse recovery currents will likely also show up. A suitable RC
filter on the sensor or downstream should take care of the latter.

RL

Hi RL,
I agree with above and Fritz S. second post on current transformers. How do
we fabricate said transformer to meet the 400KHZ frequency requirement? OP
states 200KHZ switcher or 400KHZ secondary. Current transformer needs 100:1
turns ratio to keep secondary power less than 1.0 watts, but how do we
control shunt capacity and leakage inductance to maintain the 600KHz
bandwidth necessary for decent waveform reproduction?
Most "of the shelf" current transformers are rated 60HZ and some 100KHz.
The OP should at least drop that primary to 100KHz. Do we have a 100KHz
limit in current transformers?

regards
harry

 

[Yzordderex]

Snip-----
---unsnip

Nice little 100:1 CTs are available from Pulse Engrg, Datatronics, and
many others at reasonable cost, but I agree with harry about the
sensing resistor having advantages. On the other hand, at this power
level, why bother with the complex, expensive, hard to troubleshoot
phase-shifted full bridge ZVS? You'll find forward converters much
easier to work with and nearly as efficient. With the availability of
higher performance MOSFETs and rectifiers, the full bridge ZVS is
seldom used anymore at less than 1kW or so.
Paul Mathews

Paul,

Thanks for reply. I've added more information about design as thread
has progressed. See my response to the resistors. Looks like you
wrote response awhile ago, but this google seems to want to post at
it's own convenience.

I'll have to take a peek at CT offerings. I've seen some cute ones at
Digi-Key, but also expensive. This 250w unit is just my first one - a
breadboard so to speak. I hope to be able to eventually get to full
legal limit on 75 meter AM bands. That would be 1.5kW PEP (I think).
That is reason for full bridge topology. If I haven't already
mentioned it, this design is for a modulator for a Class E RF deck.
The input to this thing is not a reference voltage, but a microphone.
There are 250w RF deck designs on the web using a single-ended
resonant topology. That is the motivation for the initial 250W
output. I will probably do a push-pull RF deck for the full legal
limit.

Also thanks to others who have replied - Harry, R.Legg, Jeff, and
Fritz. I think I've got the exact info I asked for. I'll print out
info and wind some parts on Thursday . I'll let you all know what I
end up with.

regards,
Bob

 

[R.Legg]

"H

I agree with above and Fritz S. second post on current transformers. How do
we fabricate said transformer to meet the 400KHZ frequency requirement?

Shunt capacity will likely work to your advantage, in this case. I
think you'll find the waveform produced by 100:1 ratio on a 10 to
20mmOD ferrite toroid is sufficiently accurate for most applications.
I've seen them rival 100MHz instrument-grade clip-ons without
signifigant modification, with burdens below 60ohms. You will of
course need to check performance of the part (before committing
suicide on the production floor)in any event.

OP states 200KHZ switcher or 400KHZ secondary. Current transformer needs 100:1
turns ratio to keep secondary power less than 1.0 watts, but how do we
control shunt capacity and leakage inductance to maintain the 600KHz
bandwidth necessary for decent waveform reproduction?

Have you really already gotten to the point where you are having a
problem here, or are you just anticipating it?

Keep it small, keep the high-current loop through the part short,
cover the entire core with the output winding. Terminate with R close
to the core. Twist output leads to instrumentation sensing point,
before sizing any filtering components (Differential and/or
Common-mode). There's a lot of play to be had with such a low source
impedance and such a potentially high amplifier or comparator input
impedance - there's really no excuse for poor output definition or
excessive noise problems using an AC current transformer in this
frequency range.

Use the same methods currently anticipated to deal with power
transformer output leakage inductance to deal with any extra
introduced by the high-current loop.

Most "of the shelf" current transformers are rated 60HZ and some 100KHz.
The OP should at least drop that primary to 100KHz. Do we have a 100KHz
limit in current transformers?

The 'rating' is typically specified for an output voltage (with an
unstated percentage of droop) in a single-ended detection circuit;
they should really be rated in volt-seconds, with an internal and
external impedance specified, in order to remove this confusion from
the data sheet.

RL

 

[Yzordderex]

----Snip snip hea

Hi RL,
I agree with above and Fritz S. second post on current transformers. How do
we fabricate said transformer to meet the 400KHZ frequency requirement? OP
states 200KHZ switcher or 400KHZ secondary. Current transformer needs 100:1
turns ratio to keep secondary power less than 1.0 watts, but how do we
control shunt capacity and leakage inductance to maintain the 600KHz
bandwidth necessary for decent waveform reproduction?
Most "of the shelf" current transformers are rated 60HZ and some 100KHz.
The OP should at least drop that primary to 100KHz. Do we have a 100KHz
limit in current transformers?

regards
harry

Harry,

Current sense transformer will be on the primary side. I imagine I'll
run controller somewhere between 100kHZ-200kHz. The higher the better
for output filter design. The switching frequency will in fact show
up in the sidebands of the AM carrier, and FCC says they have to be so
many db down.

You bring up some good points. Leakage inductance and inter-turn
capacitance may be a problem. I don't think it will be a major prob
to design a current transformer that will work at a few hundred kHz.
Heheh, got to now that I've downplayed the resistor idea.

I have bought some units from Ion Physics that work very well for fast
pulses. They have one unit I think good for 1Hz-35mHz at 3db points.
They are very expensive - labratory applications only. The app note
for UCC3875 resonant controller has a typical application that uses
current sensor, so I figure I just gotta give it a go.

regards,
bob

 

[legg]

Current sense transformer will be on the primary side. I imagine I'll
run controller somewhere between 100kHZ-200kHz. The higher the better
for output filter design. The switching frequency will in fact show
up in the sidebands of the AM carrier, and FCC says they have to be so
many db down.

If you are going to put this thing on the phase modulated primary, the
simple circuit described previously will not function without added
complication, as the primary current is dc-coupled to any CT position.

The presence of a DC imbalance may require the use of gapped
structures or composite materials with lower permeability.

RL

 

[Paul Mathews]

Paul,

Thanks for reply. I've added more information about design as thread
has progressed. See my response to the resistors. Looks like you
wrote response awhile ago, but this google seems to want to post at
it's own convenience.

I'll have to take a peek at CT offerings. I've seen some cute ones at
Digi-Key, but also expensive. This 250w unit is just my first one - a
breadboard so to speak. I hope to be able to eventually get to full
legal limit on 75 meter AM bands. That would be 1.5kW PEP (I think).
That is reason for full bridge topology. If I haven't already
mentioned it, this design is for a modulator for a Class E RF deck.
The input to this thing is not a reference voltage, but a microphone.
There are 250w RF deck designs on the web using a single-ended
resonant topology. That is the motivation for the initial 250W
output. I will probably do a push-pull RF deck for the full legal
limit.

Also thanks to others who have replied - Harry, R.Legg, Jeff, and
Fritz. I think I've got the exact info I asked for. I'll print out
info and wind some parts on Thursday . I'll let you all know what I
end up with.

regards,
Bob

I've worked extensively with phase-shifted FWB designs up to 1.5kW,
and I must tell you that there are many ways to imbalance the primary
and cause 'flux walking'. If allowed to continue for any length of
time, this will typically cause some kind of catastrophic failure.
You will likely read that current mode control designs prevent flux
walking, and this is partly true: IF the feedback loop is within its
range of control, and IF the CM controller has sufficient bandwidth at
the operating point of the moment, the CM controller MAY maintain the
same peak current in each half cycle and prevent flux walking.
However, there are many circumstances in which the feedback loop has
gone 'clangbar'....is operating at one clamp limit or the other....and
it does nothing (other than max peak current limiting) to control
current pulse-by-pulse. IF your core and other circuits are
sufficiently robust, you can ride through these conditions (which can
occur on power up/down and during brownouts, for example). Some other
posters have alluded to this situation and suggested things like air
gaps in the core. (My own designs used toroid cores for their low
profile and superior cooling characteristics, and an air gap was not
an option. The toroid core also allowed me to easily control leakage
inductance rather than use an external resonating inductor.) At any
rate, this is the reason why many designers end up putting a capacitor
in series with the primary and using voltage mode control. There is
also another problem with ZVS versions of FWB: reverse voltage stress
on secondary rectifiers can be astonishingly high, as the resonant
ringing on the primary gets reflected through a relatively low turns
ratio. I was able to use active snubbers on the secondary to limit
reverse voltage and recover some energy, but rectifier voltage ratings
still had to be quite a bit higher than seen in hard switched version.
If you dig around in the recent literature, you'll also find several
papers questioning whether soft-switching is worthwhile at all. The
most useful pieces of test gear you can have are a good AD/DC current
probe and a HV differential probe for your scope. Have fun.
Paul Mathews