"Thumper" / Projects / Forums

Forums

4hv.org :: Forums :: Projects

« Previous topic | Next topic »

"Thumper"

previous 1 2 3 4 next last

Move Thread

LAN_403

Bennem

Sun May 03 2009, 06:37PM

Registered Member #154 Joined: Sun Feb 12 2006, 04:28PM
Location: Westmidlands, UK
Posts: 260

Hi Finn,

Thanks for the info,
Do you think you could get the same results with a high Freq Res coil?
I'm going to try it with my small 300Khz coil as i have one of Steve's
older version circuits in that coil.

Mel

Finn Hammer

Sun May 03 2009, 07:43PM

Registered Member #205 Joined: Sat Feb 18 2006, 11:59AM
Location: SkÃ¸rping, Denmark
Posts: 741

Bennem wrote ...

Hi Finn,

Thanks for the info,
Do you think you could get the same results with a high Freq Res coil?
I'm going to try it with my small 300Khz coil as i have one of Steve's
older version circuits in that coil.

Mel

It certainly is not designed for that. But I can't see why not, for different reasons, though.
For the low frequency coils the reason is the long shut off delays of the bricks, mainly.
But in a high frequency coil, the delays are smaller, but so is the frequency, so the delay to period ratio might be comparable.

All I can say, give it a try. Worst case scenario: the thing stops to switch, no big deal, just rewire and make sparks all day!

I am pondering a new version.
First of all, there was the problem with oscillation due to equal input voltages. I can solve it by referring the one side of the burden a few millivolts away from - input, the other to + input. Better it by adding a bit of hysteresis.
This can not be done with an open collector output, so I have chosen a totem pole output comparator, AD790 which has some good features. Built in hysteresis, 500ÂµV, will accept up to +- 18V supply, fast, but not too fast, 40nS etc.

Bias the - input for adjustable offset. And maby looking a bit more nice in a way. Have to wait for the comparator, though.

EDIT:
Woops. Having read up on the datasheet of AD790: It has a 5Vlogic output totem pole. What a wonderfull comparator!, that means I can forget about the interfacing transistor. This is looking good indeed.

Cheers, Finn Hammer

Bennem

Mon May 04 2009, 08:04AM

Registered Member #154 Joined: Sun Feb 12 2006, 04:28PM
Location: Westmidlands, UK
Posts: 260

Cool.....I look forward to seeing your results!!

GeordieBoy

Tue May 05 2009, 11:36AM

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

Hi Finn,

Great work. You can actually vary the amount of phase-lead in the feedback signal from the CT just by making the inductive part of the burden impedance variable. i.e. Use a tiny adjustable pot-core or I.F. transformer type coil that has an adjustable ferrite slug. You should then be able to adjust the inductive part of the burden impedance to get the amount of phase-lead that you want.

Steve W is right about the dangers of too much phase-lead in the feedback path. Present DRSSTC's with no phase-lead compensation switch a load current that has leading power factor. To put this another way, the load current has already passed through zero and changed polarity before the message gets through the controller and tells the IGBTs to actually switch over. Funnily enough, this is the way it should be for IGBTs....

If you add too much phase-lead into the feedback path at the operating frequency you will end up with the IGBTs switching in advance of the load current passing through zero. This is generally bad for slow IGBTs as they have stored charge which will then be swept out in the form of current tailing. The further the IGBT switches ahead of the current zero, the higher the instantaneous current will be at turn-off and the greater the tail-current losses will be.

So in short it is always better to let the IGBT's collector current fall smoothly to zero, THEN turn off the IGBT when there can be no current tailing. Then, only when the first IGBT has been allowed sufficient time to turn-off, the opposing IGBT can be turned on.

The holy-grail of timing that you should be aiming for is this:

1. Turn off the current carrying IGBT as soon as possible after its collector current has fallen to zero and the load current has started to build in it's co-packaged free-wheel diode. (The device can be turned off with a very rapid slope because current has already commutated to the co-pack diode some time ago!)

2. Wait sufficient time for this IGBT to regain blocking ability before trying to turn on the opposing IGBT.

3. Turn on the opposing IGBT with a controlled slope so that the rising load current is now commutated from the other device's free-wheel diode to this IGBT in a smooth and controlled manner.

If all of these things are performed as quickly as possible after the current zero-crossing then the load current should not have had time to rise very much from its zero crossing. Therefore turn-on losses should not be very high, despite the controlled turn-on slope. Note that the controlled turn-on slope may actually reduce turn-on losses in the IGBTs due to forced reverse recovery in the opposing device's co-pack free-wheel diodes. Again, it is advisable to start the turn-on process as early as possible. The longer the load current is allowed to flow in the free-wheel diode the higher the current magnitude will rise, and the worse the reverse-recovery spike will be when the diode turns off.

The phase-lead network I told you about with the L+R burden impedance is great for compensating for inherent delays in the control and gate-drive paths. However, you should be careful not to introduce overall net phase-lead for the reasons stated above. It is also important to realise that the amount of time-lead in nanoseconds is dependent on the amplitude of the current signal being sensed. That is to say that the output of the comparator will switch further ahead of the current zeroes if the current being sensed is large, than if the current being sensed is small. The L part of the burden impedance basically adds a "differential" component to the otherwise proportional output from the CT. So the output is no longer proportional to the sinewave load current being sensed, but is instead proportional to the sinewave load current plus a certain amount of the differential of this sinewave load current. The more rapid the slew of the load current, the earlier the comparator will switch and the more "time-lead" is introduced.

I hope this helps explain the details of how an IGBT inverter operates with a series resonant load. Have fun experimenting.

-Richie,

Dr. Drone

Tue May 05 2009, 04:37PM

Registered Member #290 Joined: Mon Mar 06 2006, 08:24PM
Location:
Posts: 1673

Finn Hammer

Tue May 05 2009, 07:32PM

Registered Member #205 Joined: Sat Feb 18 2006, 11:59AM
Location: SkÃ¸rping, Denmark
Posts: 741

Richie,

I don't plan to shut any device off before current zero crossing (CZC)
But when the turn off delay is the longest of the switching parameters, up in the ÂµS region, ThatÂ´s why I thought it was prudent to initiate the turn off before CZC.
I read your advice about allowing the IGBT to recompose itself before opening the other one, and this is where the adjustable deadtime circuit comes in. Supported by generous gate resistance, to slow down the turn on. (I don't so easily forget the hardship this reverse recovery current caused us in the CCPS thread.

I had not thought about the effect of increasing current, creating a larger advance in timing, and this means that the design has to be optimised for the end of the burst. Maby the deadtime window can be long enough to accomodate the differences here.

Chris, I also read your objections, and you obviously have a point. Fatboy is hard to beat, and why mess with a proven design.

The 1968 Shelby Mustang GT500 KR was a very successfull design in it's time.

1241551243 205 FT68820 07 Ford Shelby Cobra Gt Oldnew 500

But it didn't stop further development effort, and the well deserved success of the Ward controller should not do so either.
Your many years in a craftmans environment must surely also have taught you another thing: Apprentices ask around with the masters that care to share insight, then they go and do what they want themselves.
But we take every bit of detail you said that can go into this new design, be it bypassing, tuning or basic circuit ideas.
I bet you did the same back in your rookie years.
And like I stated early in this thread, hell I can always go back to the basic controller as it is.

I got the AD790 comparator, and it works really well, the internal hysteresis, 500ÂµV did the trick, but biassing the negative input a couple millivolts is probably going to be needed.
The offset here has to be greater than any voltage induced in the burden coil, and this can only be determined in a running coil. I*l look into guarding the coil, and of course orienting it perpendicular to the main source of mutual induction: the primary coil.

I also got a bit of work done on the over current detector.
Here I am also going another route, but not much:
I always felt that it was not too much to ask, that it was a straightforward procedure to set the maximum current trip point. This is a place where I don't want to experiment, but rather would demand a well defined voltage point to set the trip current.

I do so by using a precision rectifier instead if a diode bridge,
I know, Chris, you are going to puke now, but really, it is just one op amp more instead of the diode bridge.
This rectifier rectifies a current transformers output, and the result is fed into a comparator, as already, only this way, the trip voltage is well defined. Like incorporating a measurement instrument in the controller.

I do read your warning about RF, the enemy. I will fight it by your own measures: Decoupling and layout.
No power pin on any device is going to be more than 3mm away from it's decoupling cap, there are going to be analogue ground planes separated from digital ground plane, decoupling and bypassing galore!

Allow me to scite a great piece of inspiration I read the other day: (it is prolly a classic already)

"The breadboard is both the designer's playground and proving ground. It is here that Reality resides, and paper (and computer designs) meet their ruler. More than anything else, breadboarding is an iterative procedure, an odd amalgam of experience guiding an innocent, ignorant, explorative spirit. A key is to be willing to try things out, sometimes for not very good reasons. Invent problems and solutions, guess carefully and wildly, throw rocks and see what comes loose. Invent and design experiments, and follow them wherever they lead".

Jim Williams wrote AN72 in 1998, and subtitled it:
Guidance to put civilised speed to work.

With that guidance, and that of yourself: honorouble list members and friends, I intend to persue this idea to success, or to it's pitifull failiure.

Cheers, Finn Hammer

Dr. Drone

Tue May 05 2009, 09:22PM

Registered Member #290 Joined: Mon Mar 06 2006, 08:24PM
Location:
Posts: 1673

Dr. Drone

Tue May 05 2009, 09:26PM

Registered Member #290 Joined: Mon Mar 06 2006, 08:24PM
Location:
Posts: 1673

GeordieBoy

Tue May 05 2009, 09:45PM

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

Hi Finn,

All of what you are doing sounds good to me.

Advancing the drive signal to compensate for delays in the control and gate-drive electronics sounds like a sensible approach. As for how much this will increase efficiency, I am not sure for a pulsed application like a DRSSTC. I can say with some certainty that it won't noticeably increase the spark length because efficiency will likely already be in the 90's of percent, but it may reduce the device heating considerably and also reduce reverse recovery problems.

Regarding the phase-lead's dependancy on current amplitude: I would optimise the circuit so that it switches with best performance when current is maximum. This is the normal situation when designing a resonant switched mode power supply. Maximise efficiency at full-power because this is when it matters most. Poor efficiency at low power doesn't imply as much dissipation so is generally not a problem.

Regarding the concerns about interference pickup of that little air-cored coil: I would replace it with a pot-core or IF transformer type with a semi-enclosed bobbin and adjustable core. This will be less susceptible to strong magnetic fields. It may also be wise to orientate it's axis orthogonally to the TC's primary and secondary axes to minimise any remaining coupling.

-Richie,

Steve Conner

Wed May 06 2009, 10:11AM

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

Hi all

Having an advance that changes with current may be a good thing, since the turn-off delays of IGBTs probably increase with current too.

My last word on the DRSSTC driver debate was the Mk2 PLL driver that I built.

previous 1 2 3 4 next last

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