User-agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:24.0) Gecko/20100101 Thunderbird/24.3.0
Em 12/02/2014 09:31, Udo Lenz escreveu:
Yes, but note that the energy mainly goes back into the driver, i.e. the
power supply caps, on the second beat. This is due to a phase change of
180 degrees between driving voltage and primary current between first
and second beat.
This is undesirable and can be avoided with primary feedback. Then
energy can
only go one way, i.e. from the driver to the primary.
About this energy return, I was using ultrafast free-wheeling diodes in
parallel with the mosfets, but even so the body diodes of the mosfets
were delaying the switching of the output voltage of the bridge by the
returning current. The fast diodes were not making any effect. I tried
today to insert schottky diodes in series with the drains of the
mosfets. This eliminated the delayed switching. I used those large
schottky diodes found in the output of PC power supplies (finally an use
for those things). A curiosity is that with the switching delayed by the
body diodes the observed current in the return beat is significantly
larger than the current when the schottky diodes are added. To verify if
something was adding losses, I measured the input power drained from the
AC line with a Kill-a-Watt meter, in no breakout conditions. The power
with the schottky diodes was smaller (~2/3), and so I think that the
larger current was caused by the phase of the bridge output voltage.
I don't think it is detuning. If you are running at the lower pole,
i.e. at
300kHz - 5% (due to coupling), the streamer capacitance will pull the
secondary
into tune with this frequency. That will lower your primary current.
This implies,
that you will stay way below the max current rating of 90A and get
weak streamers.
Lowering the driving frequency to keep the current up doesn't really
help either,
since your primary tank will get severely out of tune. You can put
lots of current
into a primary (series) tank only close to its resonance. You'd need
to reduce
primary fres to get the current up again.
After I added the schottky diodes I could observe something like this.
Exciting the system at the lower pole I see the current first increasing
and then decreasing as streamers develop.
If you run at the upper pole (300kHz + 5%), the streamer capacitance will
move the secondary fres even further from the driving frequency. That
will
allow the tank current to rise, but since you're away from your
secondary resonance,
little of the primary energy will go to the secondary.
I see the current ramping up, and some energy return after the end of
the burst. Streamers are similar to when the excitation is at the lower
pole. This mode is clearly inefficient.
With the low-impedance primary there is an important change. Tuning
between the resonances I get breakout easily. If I just extend the
burst length I see that the primary current increases as a kind of
ramp after a first beat, and streamers grow with burst length up to a
limit caused by detuning. If I tune at the resonances, I get less
impressive results and much larger primary current.
Yes, the lower primary impedance will cause a much higher primary
current.
But you will probably have to limit burst length in order to stay
within your
bridges current limit. The primary current will rise later in the
burst due to
detuning.
I've done some simulations and they indicate, that you need to limit
burst
length to 50us in order not to exceed the 90A limit.
I didn't connect again the low-impedance primary, that is quite too much
for my 20A mosfets. With the high-impedance primary and 30 cycles I see
50A with excitation at the lower pole and almost 100A with excitation at
the upper pole. Breakout even without a point in both cases.
I've also tried to do a simulation with streamer loads. The best I could
achieve was to use the 58uH primary with an increased MMC of 7nF
and driving it at about 250kHz. The simulation shows an increase of
arc length of more than 50% against the simulation with your settings.
Primary current would be limited to less than 90A. Burst length should
be 100us or more.
How are you simulating streamer load?
Antonio Carlos M. de Queiroz
_______________________________________________
Tesla mailing list
Tesla@xxxxxxxxxx
http://www.pupman.com/mailman/listinfo/tesla