Big mo-fo ferrite HV transformer driving (another CCPS/SLR inverter) / Projects / Forums

Forums

4hv.org :: Forums :: Projects

« Previous topic | Next topic »

Big mo-fo ferrite HV transformer driving (another CCPS/SLR inverter)

previous 1 2 3 4

Move Thread

LAN_403

Steve Conner

Sun Nov 16 2008, 06:36PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

Those LC meters are pretty easily fooled.

uzzors2k

Sun Nov 16 2008, 08:49PM

Registered Member #95 Joined: Thu Feb 09 2006, 04:57PM
Location: Norway
Posts: 1308

That was my thought too, afterall I made the meter myself. So I used my Fres Finder and put the primary in series with a 220nF cap. The resonant frequency was 66kHz.... I'm not sure if I should ignore with this or not, seeing as an oil soaked transformer is not something I'll ever bother troubleshooting. Which would render too much valuable time and ferrite lost. Then again, when I put a a spare secondary which is nearly identical to the ones I'm using on the core, the primary inductance sunk once again, meaning it can't be a shorted turn.

I just wound a 4 turn winding on the core now with solid insulated wire, which reduced the primary inductance from 370ÂµH to 300ÂµH. The winding was left open circuit, like the others. This makes no sense to me what so ever. I'll try the Fres Finder method again. Hopefully it is my LC meter.

UPDATE: It seems the errors in measurement are due to self resonance. With nothing but the primary connected to my fres finder, there are parallel resonant peaks at frequencies in the range of 70 to 150kHz. Somehow the secondaries are contributing with some capacitance to the primary winding, making my transformer as touchy as a TC. At any rate, with just the primary and parasitic capacitances from itself and the HV secondaries, the resonant frequency is 73,5kHz. Since this is parallel resonant, I might have to consider a whole new method of driving the thing? Maybe it's what caused the problems I had with SLR before. With nothing on the transformer but the primary the resonant frequency is 150kHz.

Sulaiman

Mon Nov 24 2008, 09:06PM

Registered Member #162 Joined: Mon Feb 13 2006, 10:25AM
Location: United Kingdom
Posts: 3141

Power ferrites aren't irreversibly damaged by too much flux density so I doubt that's a problem.
From experiments with 1" x 1" x 4" ferrite blocks of my own as reference ;
(my ferrite is 3C85.. almost identical to N27)
A single block has Ui of about 118 uH/turn2 ... 8turns = 7.5 uH
Four blocks in a rectangle with 1.6mm total airgap c550 nH/turn2..8turns = 35 uH
Four blocks in a rectangle with minimal airgap ... c 3600 nH/turn2 .. 8turns = 230 uH
So the airgap is very important!
With a minimal airgap changing the clamping pressure significantly changes the inductance!

One way to significantly increase the secondary self-resonant frequency is the so-called 'split-diode'
winding technique as used in many TV flybacks.
Wind several smaller secondaries and put a diode between each,
the nett resonant frequency is very close to the resonant frequency of just one of the secondaries.
For full-wave rectification use four diodes per secondary winding,
a lot of diodes BUT
- each diode only 'sees' one secondary voltage MAX.
- each diode is therefore (much) cheaper.
- the secondary resonant frequency will be about the same as for one secondary winding
much much much higher than when all the inductance and capacitance is lumped together.

uzzors2k

Sat Dec 27 2008, 10:52PM

Registered Member #95 Joined: Thu Feb 09 2006, 04:57PM
Location: Norway
Posts: 1308

What a headache! I've made some progress, and should be able to get this thing working soon. I discovered that I don't need to feel constrained by the turns ratio inorder to get the voltages I want, as resonant rise and other unknown effects cause enough ringing to vastly increase output voltage. I reduced the turns ratio down to 1:27 and submerged the transformer under oil since last time. With the new turns ratio the self resonant frequency is 160kHz, better than it was at least and higher than I've planned on ever driving it. I've also upgraded the inverter to include it's own power supply and improved the gate driver section one more time. Just to get some arcs from it I ran it hard-switched at 25% duty and 145kHz. Here's the youtube link.

I daresay it looks promising!

Now for the problems. For SLR use I used a 15ÂµH external air core inductor and 22nF capacitor. All tests are done with the secondary shorted.
At 50V all is nice and I can tune for a nice buttocks waveform. Unlike the first time I did this, I've tired varying the frequency and duty cycle instead of relying on math and assumptions. At 320V however, things get worse. For some reason my OC protection trips at the correct duty cycle and frequency, even though the surge impedance is only 40 ohms (this figure checks out in both calculations and 50V tests). After some adjusting I was able to get it to run and the waveform looked ok. However my mosfets were getting a little warm, even though the current was only peaking at about 10A. And after just about 10s one half of the bridge died. The mosfets weren't all that warm yet either. I'm stumped and quite frankly tired of SLR, just as I imagine you guys are of this thread. Any take on mosfets dying while they were supposed to be soft-switching?

EDIT: I increased the primary turns to 15 for a 1:15 ratio, output voltage seems to be about 20kV now. The leakage inductance is 68ÂµH, so I don't need an external inductor anymore. Still with a 22nF primary resonant capacitor, I'm testing SLR at 60kHz +/- 10kHz. However when I tune for the correct waveform the mosfets start heating incredibly, much more than when hard-switching. I haven't bypassed the internal diodes, will they conduct if the frequency isn't just right or the duty cycle too long/short?

GeordieBoy

Tue Jan 06 2009, 01:38PM

Registered Member #1232 Joined: Wed Jan 16 2008, 10:53PM
Location: Doon tha Toon!
Posts: 881

Why did you choose MOSFETs instead of IGBTs? The heating of the MOSFETs may be due to forced reverse recovery of the body-drain diodes if you forgot to isolate them and bypass with external fast diodes?

The SLR converter with its zero-current-switching is really better suited to IGBTs that are co-packaged with fast diodes. The zero-current switching minimises tail-current losses in the IGBT and you get lower conduction losses and turn-on losses than with MOSFETs. The fast co-pack diode of modern IGBTs is also much better suited to SLR duty than the sloooooooow MOSFET body-drain diodes.

In the SLR converter the first half-cycle of current is carried by a switch, and the remaining half-cycle (of the opposite polarity) is carried by the free-wheel diode across that switch. Then the process repeats with the sinewave of current inverter, where the first half-cycle is carried by the other switch, and the second half-cycle is carried by it's tandem free-wheel diode. So with the output short-circuited, the average current is pretty much split evenly between the 2 diodes and 2 switches. (As the load is increased the Q of the resonant circuit falls and the balance of current sharing shifts in favour of the switches.)

-Richie,

uzzors2k

Tue Jan 06 2009, 03:59PM

Registered Member #95 Joined: Thu Feb 09 2006, 04:57PM
Location: Norway
Posts: 1308

Mosfets were used based on availability, but if a freewheel diode is needed to pass half of the cycle I'd better upgrade. It would definitely explain the increased dissipation as IRFP450s have over 1Âµs of recovery time.

I'll try IGBTs next then, with fast diodes.

previous 1 2 3 4

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob