PLL SSTC with Type 2 phase comparator problems (added some info) / Tesla Coils / Forums

Forums

4hv.org :: Forums :: Tesla Coils

« Previous topic | Next topic »

PLL SSTC with Type 2 phase comparator problems (added some info)

previous 1 2 3 4 next

Move Thread

LAN_403

Steve Conner

Tue Sept 10 2013, 02:42PM

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

Yes, that's it exactly!

If you want to use a second-order loop in an ordinary PLL application, you have to insert a zero in the loop filter to get any phase margin whatsoever. Otherwise you have two integrators in the feedback loop and a phase margin of 0.

I haven't done an analog PLL in a while, but recently I did a few second-order digital ones with IIR loop filters. I think of the loop filter as a PI controller, where the P term provides the zero for stability.

These arguments about poles and zeros only hold when the circuit is running in its linear region, but the process by which the PLL gets into lock in the first place is highly nonlinear. So the second-order loop has more finicky capture behaviour that can include conditional stability. Again not relevant to Tesla coil applications because the loop is never "out of lock" in the sense that an ordinary PLL can get out of lock. The two frequencies going into the phase detector are always the same, so it can always produce a non-zero output on average.

Dr. Dark Current

Tue Sept 10 2013, 05:38PM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

Guys, I have found the problem, but still not sure how to cure it.

The IC is not really faulty, it is just more complex than it seemed at first.
My problem is not the noise. It is the fact, that the Phase Comparator 2 has a "memory" which is 2 cycles, or 720 degrees "long". The 74HC4046 datasheet (which is not really the same IC, but the principle is the same) tells us, that the demodulated output voltage when using the PC2 depends on a phase difference of the signals from -360 to +360 degrees.

So what happens when there is a missing pulse on the input? The input signal phase has changed by 360 degrees. The input and VCO signals are still in phase, but the Phase Comparator sees this as a 360 degree shift and tries to correct it by actually delaying or advancing the VCO output by + or - 360 degrees (depends on where exactly the next pulse after the missing one comes). However, because the Tesla Coil system doesn't allow such a large phase shift within the VCO frequency range, the PLL will never lock the phase again unless the input is disrupted once more (for example by resetting the supply voltage). In an open-loop system such as an FM demodulation, this is not a problem, because the VCO output and PC input signals are not frequency-linked.

As for the different behavior of another IC manufacturer... Either that IC is somewhat damaged (as I have tried 3 more manufacturers and all behave exactly the same as the TI one), or it has different architecture which causes additional problems.

P.S. The missing pulse was just an example, during a ground spark there may be several missing, distorted, or wrong-frequency pulses. But if the signals (after the resonance is possible again) catch up with the wrong phase (+- 360 degrees), there the problem comes.

Steve Conner

Tue Sept 10 2013, 07:16PM

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

Sounds like another vote for PC1 with an external integrator!

Dr. Dark Current

Tue Sept 10 2013, 07:44PM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

Well, it seems I'l really have to leave this phase comparator alone, it doesn't look suitable to use in closed loop systems...

As for the PC1 with integrator: There are 2 possible problems - noise getting into the op amp and weak signal on the shunt capacitor. That's why I was trying to avoid this solution... The 4046 IC itself is pretty well immune to interference, I could run the coil with the cover of the driver off (for measurement purposes) and it ran just fine.

Steve Conner

Tue Sept 10 2013, 07:53PM

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

I've built about 8 or 9 drivers using PC1 with an integrator. It works fine for me!

I didn't even bother trying PC2 because connecting a sensitive edge-triggered state machine to a Tesla coil looked like a really bad idea.

In a previous design I used a shunt capacitor on the CT. It worked great until I used it to drive a series resonant induction heater with feedback from the inverter output current. The RF current was too much for the capacitor and it exploded.

Dr. Dark Current

Tue Sept 10 2013, 09:36PM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

Haha

Steve, do you use an inverting amplifier after the integrator in the PLL filter? Or does it work to swap the PC inputs?

Steve Conner

Wed Sept 11 2013, 08:02AM

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

This is the circuit I've been using since about 2005.

It uses an inverting amplifier in the loop.

I don't think you can swap the phase comparator inputs, because only one of the inputs has the amplifier and self-bias circuit. I use that to take the signal from the CT directly.

Dr. Dark Current

Wed Sept 11 2013, 08:53AM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

OK, thanks

RayeR

Thu Sept 12 2013, 01:53PM

Registered Member #333 Joined: Mon Mar 20 2006, 06:02PM
Location: Czech Republic
Posts: 45

Very interesting thread about PLL. About 10 years ago I was started to design my 1st PLL controller for DR-SSTC. That times was PLL usage very rare, most of coilers used just simple pickup antenna.
I had spent many hours playing with 74HC4046 and trying all 3 comparators. I found that PC1 worked best for me, Also maybe I failed with PC2 was due to noise because I started on universal board without shielding and then found how it is sensitive to RFI when power goes up...
Also I was messing with various phase shift issues. One thing I tried to add additional 90deg phase shift was connecting 2 PLLs in cascade with PC1 to get 90+90 but later I found I don't need this when using pickup antenna... I can see Steve is doing this by running 2f0 with divider. Probably better choice as it holds 90deg in whole range. I also found interesting the idea of integrator in the filter, maybe I'll try this in some new design but have much less time than before...
BTW did someone tried to implement fully digital PLL in some fast ARM MCU or FPGA logic?

Steve Conner

Thu Sept 12 2013, 03:58PM

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

Yes, I believe Steve Ward and Phillip Slawinski's latest driver is a digital PLL running on some sort of fast microcontroller, and a few people are working on FPGA-based drivers at the moment.

A MCU has too much latency to run anything but a PLL, but a FPGA can emulate a traditional feedback driver too, possibly using digital filters for phase lead.

I stayed with my old driver because it worked fine and I couldn't be bothered designing a new one.

I've nearly run out of PCBs though so maybe it is time to try something new.

previous 1 2 3 4 next

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