Fuxue Jin said:
Hi, everyone,
Looking for a solution to improve an air core transformer efficiency.
The transformer can not have a regular soft magnetic core, must be air.
Primary to secondary is about 1:2 to 1:3. Coil is wound on a bobbin or
tube with about .5 to 1" diameter.
In order to improve the efficiency, what's the key factor that needs to
be considered or adjusted? Turns ratio, coil size, operating frequency
or something else?
Thanks.
How much power do you need? What is your input voltage, and what is your
output voltage?
There was a thread here a couple of years back from a guy trying to make an
air core power transfer system (titled Energy transmission with coils). The
thread inspired me to play with the stuff myself (though reading back my
contributions to his thread show my own ignorance of the subject at the
time), and based upon my experience I would say:
You must use a full bridge or half bridge topology to drive the primary if
you want anywhere near decent power output and efficiency. Do not use a
single switch converter topology, it puts allot of stress on the switch and
will not yeild much power output at all. You should use as close to 50%
duty cycle each direction as feasible, but make sure to insert some dead
time in the gate drive signals to prevent cross conduction.
Generally speaking the higher the frequency the better, but with some
limitations. The problem with air core transformers is they have such
pitifully small magnetizing inductances. As a result, the magnitizing
currents ramps up to extremely large values very quickly. In order to keep
the magnetizing currents reasonable, you can either increase the number of
turns used (to increase the inductance) or increase the frequency. If you
increase the number of turns used, you inevetably must use a longer wire
which has more resistance (both DC and AC). So you get a reduced
magentizing current, but you end up with more winding resistance which
causes more dissipation for a given amount of current. The net result is
they, to a degree, cancel each other out. So you really need to use a quite
high frequency to get good results.
Why a half/full bridge only? The novelty of the half and full bridge
topologies (built using either MOSFETs alone or BJTs/IGBTs with antiparallel
diodes) is they can readily recapture the energy stored in the magnetizing
inductance and return it to the input DC bus capacitance. This is very
important since leakage inductances and magnetizing currents would cause
massive power dissipation.
In my experimentation I found the best efficiency (for my setup) was
obtainable with a switching frequency of around 300kHz. In my case I was
using a half bridge from a ~155V DC supply rail (simple mains rectified and
filtered), driving a transformer constructed by winding about 67uH (IIRC)
worth of plain 22AWG wire on a plastic wire bobbin (one from the Radio Shack
magnet wire 3 spool set). This served as the primary (which was to see
about 78V RMS AC since it was being driving by a half bridge with near full
duty cycle). I used an approximately 1:2 turns ratio, so the secondary had
approximately 250uH (again IIRC) worth of the same 22AWG plain 300V
insulated wire (also from Radio Shack). Together they just filled the
bobbin to the max, all very nice and neat. The half bridge switches were
IRF630N devices. With resistive loads of less than or equal to a plain 120V
40W lightbulb, the whole arrangement could probably easily have operated all
day with no extra cooling but plain small heatsinks. I would have estimated
the efficiency around 80% perhaps (combined MOSFETs and transformer). I
used a variac to turn down the DC bus voltage slightly (otherwise the lamp
voltage would increase to around 130V, which was quite close to the unloaded
secondary voltage) to keep the output at 120V. When using a 60W lightbulb
the heat dissipation in the transformer and MOSFETs with small heatsinks was
enough to likely need forced air cooling for continous operation. With a
100W lightbulb operation for a few minutes was reasonable before things
would get too hot and need to be shut down. Nevertheless the load
regulation was suprisingly good as it seemed the output voltage would drop
maybe 20V between no load and 100W lightbulb loaded. As such I could easily
drive the 100W lightbulb at its full 100W rating at 120V.
The transformer design is quite important. Not having to worry about
saturation and core loss is nice, but ultimately DC and AC winding
resistance become incredible problems. At low frequencies the magnetizing
current could easily become substantially larger than the load current for
any reasonably sized primary inductance. You can find out how much the
current will swing using the basic E=L*(dt/dt) formula. Very high
(relatively speaking) frequencies combined with small size litz wire will
likely yeild the best possible combination and highest efficiency. My
choice of 300kHz with 22AWG wire was not optimal. The skin depth at 300kHz
is much smaller than 22AWG wire is thick. If you aren't already familiar
with it, Texas Instruments magnetics design handbook will provide an
excellent introduction to skin depth effects.
http://focus.ti.com/docs/training/catalog/events/event.jhtml?sku=SEM401014
It probably would yeild good results to try to size the magnetizing current
to be around the same size as the load current itself. You do this by using
the E=L*(di/dt) forumla along with your switching frequency and the formula
for finding inductance of air core coils for given geometries (any good
physics textbook) and numbers of turns.
I'm not quite sure how your transformer would behave in close proximity to a
superconducting inductor. I might imagine some big forces might get
developed (else how do mag-lev trains work?). I'm afraid I lack experience
with superconductors (the local Radio Shack doesn't stock room temp.
superconducting wires yet, dang!). The toroid winding technique sounds like
it might have some appeal, though I suspect you would have most
disappointing results if you tried to wrap them in sections. You can do
that with a nice high permeability core, but I hightly doubt that would work
with an air core. Wrapping them right on top of each other aught to work
fine though I should think. It seems to provide quite good coupling (k
perhaps equal to or better than 0.9?) with solenoid types wrapped right on
top of each other.
