CD4046 PLL driver issues / Tesla Coils / Forums

Forums

4hv.org :: Forums :: Tesla Coils

« Previous topic | Next topic »

CD4046 PLL driver issues

1 2 next

Move Thread

LAN_403

Tonic

Wed Jul 23 2008, 08:23PM

Registered Member #528 Joined: Fri Feb 16 2007, 10:32PM
Location: Warsaw, Poland
Posts: 166

After messing with antenna+74HC14 driver, I've decided to try with Conner's PLL driver (click for scheme). The antenna driver somewhat worked, I've got even nice 20cm sparks with smoothed full-wave rectified power supply. But it had problems with starting oscilation, I've had to put power supply cord to autotransformer few times to get SSTC working. Also, antenna seems to have big problems with picking signals, when inverter works at low voltage - in my case 35V. I've killed few transistors, one was literally blown :) Sometimes diodes were stressed, some of them overheated, so I guessed it's caused by a primary voltage and current being out of phase.

Now, retutrning to the topic.

For testing, the feedback isn't plugged to CD4046 yet, and signal in pin (14) is connected to ground, preventing from picking noises. The issue is frequency which isn't high as expected. I've placed 470pF capacitor across 6 and 7 pins, 33 kohm resistor across 12 pin and ground, 47 kohm resistor across 11 pin and ground. Just like as Conner done and according to him, it should have give frequency setting range between 150kHz and 225 kHz. Well, the highest frequency I can set with 10k potentiometer, is merely ~60kHz. When I'm turning potentiometer to lowest settings, the frequency drops to about few hundreds Hz, then signal becomes a flat line on oscilloscope.

I've tried with 5 seriesed 220pF capacitors (44pF in total), but it didn't gave even a 200khz, which is a approximate resonance frequency of my secondary coil. I've tried with putting 10k potentiometer in place of 11 and 12 pin resistors, but didn't help. The supply voltage is 12V. I've tried with another samples. In below picture, you can see how I connected everything - I've checked it few dozen times, also I've used Uzzors2k's scheme for reference. The picture can be enlarged:

1).A bit off topic, is it normal for you to have not working feedback when TC works at low voltage (let's, say, 24-35V)?

2). How switching from oscilator mode to feedback mode looks for CD4046? When a signal is applied to CD4046, the oscillator shuts down and CD4046 passes only this signal to output?

]cd4046.jpg[/file]

Dylan

Sat Jul 26 2008, 12:54AM

Registered Member #1024 Joined: Sun Sept 23 2007, 10:56AM
Location: Northern NSW, Australia
Posts: 95

Mmm , I'd be triple checking all the basics, all component values, components themselves, supply voltage..........
Have you tried a new/different CD4046 chip?
If the circuit is right , and a known good one, then that really only leaves component failure/identificatin problems.
Also , there is a possability your scope may be throwing things off , if you have an alternative , try that too

Tonic

Sat Jul 26 2008, 12:08PM

Registered Member #528 Joined: Fri Feb 16 2007, 10:32PM
Location: Warsaw, Poland
Posts: 166

wrote ...

Mmm , I'd be triple checking all the basics, all component values, components themselves, supply voltage..........

Did it.

wrote ...

Have you tried a new/different CD4046 chip?

Did it :)

wrote ...

Also , there is a possability your scope may be throwing things off , if you have an alternative , try that too

Much less likely, the scope works pretty well for other circuits I have.

Anyway, replacing 100k resistor that goes to 10k potentiometer to a another 10k VR, helped with setting frequency. After messing with components, I could set oscillator to work from 2XX kHz to 1XX kHz. I have no any ideas why it was working weirdly. I guess I'm just bad at electronic..

I did use CT for feedback, instead of antena. 3cm OD good ferrite core, 1:90 ratio, clamped by 1N5819, coupled with 1uF capacitor, then "squared" by two 74HC14's Schmitt triggers.

It seems that PLL works as it should be. When feedback is disabled and CD4046's signal in pin is grounded, it's square wave is solid, no matter what voltage is fed to half-bridge. At certain voltage (to be exact, at a very, very small range) I can get a pseudo sparks, thanks for CD4046's frequency being close to resonance frequency. When I apply feedback, fire at ~20V (that's a minimum voltage when SSTC starts to give sparks), then set 10k (it doesn't matter which one of two, remember I replaced 100k resistor) to moment I achieve biggest and stable arc. At this point, I could set variac voltage from 20V to 120V AC (I have only IRFP250's, unfortunately) and have sparks all time, so I guess that's how tuned PLL works. When I increase voltage, I increase sparks and it causes bigger load. Bigger load means lower resonance frequency. I could see on scope that CD4046 changes output's frequency, keeping SSTC sparking. The perfomance is very similiar, if not same, to 74HC14+antenna driver, so it would give a third argument for claiming a working PLL driver.

But.. a one thing bugs me. According to Conner's site, the half-bridge output voltage and primary current must be in phase, so zero voltage switching occurs. When I don't have sparks (at voltage lower than 20VAC), I could set in that way. But when SSTC starts to spark, it screws up, phase is shifted by 90*. Is it normal? I don't find MOSFETs and diodes to be stressed, but perhaps it's caused by a strong air-flow. Might it be a bad way I'm measuring? I was measuring secondary current, but when I was testing with current transformer placed on primary, both currents were in phase. For half-bridge output voltge, I've used CD4046's output, I thought since it's a chip that dictates, when gate drivers, MOSFETs and half-bridge switches. Or I should measure only primary and half-bridge output?

If yes, then another question:

1). I know rule to not plug probe to half-bridge connected to mains, because it causes short. With variac, this rule is still important? I know that I can use two probes technique, but with that I couldn't measure primary current and voltage simultaneously.

2). Can I plug probe to gate and source of MOSFET, when half-bridge is connected to mains/variac?

Steve Conner

Sat Jul 26 2008, 12:20PM

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

I'm beginning to wish I never published that PLL circuit! It's not perfect, but you guys don't seem to care whether a circuit works or not, as long as it's easy to build

First of all, you should be aware that, if you have an untuned primary SSTC, then a decent streamer load can make zero current switching impossible, because of magnetizing current. Under these conditions, the behaviour of a PLL running off primary current feedback is undefined. I knew this when I designed that circuit, so I used antenna feedback from the output voltage instead. But the phase shift between primary and secondary circuits depends on streamer loading too.

Second, the 4046 has is own self-biasing input amp which works better than any combination of 74HC14s.

Third, the circuit I used locks with a 90 degree phase shift. When used with an antenna, there's another phase shift that cancels this, but when used with a CT, you need to introduce one yourself. I placed a capacitor across the CT to add a 90' phase lag, and it seemed to work.

Marko

Sat Jul 26 2008, 03:00PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Hi steve, all,

1). I know rule to not plug probe to half-bridge connected to mains, because it causes short. With variac, this rule is still important? I know that I can use two probes technique, but with that I couldn't measure primary current and voltage simultaneously.

2). Can I plug probe to gate and source of MOSFET, when half-bridge is connected to mains/variac?

you can't, since variac is not an isolation transformer, you will create a short if you try to reference any point behind the bridge rectifier to ground. You need to either use isolation transformer or use two probes and subtract voltages if that is possible on your scope.

First of all, you should be aware that, if you have an untuned primary SSTC, then a decent streamer load can make zero current switching impossible, because of magnetizing current. Under these conditions, the behaviour of a PLL running off primary current feedback is undefined. I knew this when I designed that circuit, so I used antenna feedback from the output voltage instead. But the phase shift between primary and secondary circuits depends on streamer loading too.

Yes, primary current feedback is not possible because magnetizing current has linear rising slope and thus the overall sum would never be able to cross zero and there would be no feedback signal.
When we talk about CT feedback in SSTC's it's meant for secondary base current feedback.

The magnetizing current is always hard switched, but there is also additional hard switching of primary current due to phase shift between it and secondary base current. I think the phase shift will increase at first but after streamer is large enough I think it will be more or less constant since streamer is a constant current sink.

This phase shift may be tuned out by PLL, but not by direct feedback which is advantage of PLL. Magnetizing current is actually good and there would be no point removing it actually, since this would lose ZVS and induce high recovery losses in diodes.

Second, the 4046 has is own self-biasing input amp which works better than any combination of 74HC14s.

Third, the circuit I used locks with a 90 degree phase shift. When used with an antenna, there's another phase shift that cancels this, but when used with a CT, you need to introduce one yourself. I placed a capacitor across the CT to add a 90' phase lag, and it seemed to work.

Conner- now for a while I'm sure I was wrong with my initial idea of antenna phase shift, it is really no difference between base current CT and antenna feedback. Both are current sources and behave exactly the same, no weird phase shifts anywhere!

Antenna can be viewed as not a voltage, but a capacitive current divider - it is like a small capacitor paralleled with large capacitance of topload-ground; the current in it will always be in phase with secondary base current as long as it is shunted with relatively low resistnace to ground.

Considering input impedances of any silicon preamplifiers we use are always very low (which they must be, clamped by resistors, zeners, or diodes) in order to prevent huge voltages appearing on the input and frying it.

Significant phase shift can exist only if there is really, really high resistance between antenna base and ground, so high that it is significant fraction of topload-antenna reactance. In that case it would also drop significant fraction of topload-ground voltage which just doesn't happen in any of our circuits! If antenna was open circuit to ground, we would be measuring topload voltage and phase shift would indeed be 90 degrees.

I'm sorry for spreading misconceptions, but Richie led me to right way. The point is, I just don't see how would adding a parallel cap to CT make any difference to adding it to antenna providing the same current.

I hope Richie is somewhere around...

My belief why this circuit worked was because we could easily reduce the 90 degree phase shift by biasing the VCO input - everyone would just turn the potentiometer for max smoke as it can allow shifts between 0 and 180 degrees. It would be 90 degrees by default only if we had no bias on VCO in.

One phasing may have worked better than another simply because of phase shift caused by the primary.

But I may be wrong as well, I only began to understand these things.

For the end, there is a PLL circuit that apparently works very well, built by steve ward:

He used phase comparator 2 and thus got rid of any phase shifts, and PC output is already a sort of digital integrator. But, due to it's nature output of this PC can't be biased and thus he used a delay network between VCO out and comp in.

The bilateral switch is used to 'remember' the VCO frequency during off times, but it may as well be removed if interrupter is not used I think.

Ward also used 74HC14 gates at the input, apparently to de-noise the signal since phase comparator 2 is susceptible to noise.

In any case I'd prefer that circuit over Conner's, looking at it's sucess.

Marko

Steve Conner

Sat Jul 26 2008, 08:41PM

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

Marko, sorry for the confusion. When I said CT I meant a CT for primary current feedback. So the phase shift is the one between primary and secondary, not between antenna and CT. (I still think the phase shift of an antenna is more or less undefined, though, since the input impedances of the pickup circuits aren't pure resistive or capacitive.)

In other words I'm not exactly sure what's going on! As for Ward's circuit, I've never tried it. I have my own 2nd generation PLL driver that was complicated enough to scare everyone else off. My design goals were similar to Ward's but I went about it a bit differently. I don't know if the driver itself is much more complex, but I added lots of protection and metering.

I know Steve got great results driving a CW SSTC with his improved PLL circuit, though, so it might well be the one to go for.

uzzors2k

Sun Aug 03 2008, 12:46PM

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

I'm glad I stumbled over this thread. I was just about to start working on a primary current feedback PLL when I realized that I would have to factor in magnetizing current too! Bummer, I was looking for an excuse to do another SSTC project.

I've never found any information about Tesla coil behaviour when in resonance, more specifically phasings and such. I know the primary current and voltage are in phase at resonance, but what's the phase relation on the secondary side? I would assume there to be zero phase shift there too, but that's not the impression I've gotten over the years.

Marko

Sun Aug 03 2008, 03:55PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Uzzors wrote ...

I'm glad I stumbled over this thread. I was just about to start working on a primary current feedback PLL when I realized that I would have to factor in magnetizing current too! Bummer, I was looking for an excuse to do another SSTC project.

I've never found any information about Tesla coil behaviour when in resonance, more specifically phasings and such. I know the primary current and voltage are in phase at resonance, but what's the phase relation on the secondary side? I would assume there to be zero phase shift there too, but that's not the impression I've gotten over the years.

Hey Uzzors,

Yes, there is much misinformation spread about SSTC's, some of which is likely resulting from Richie Burnett's web pages being unfinished. Everyone read his writings about resonators directly base fed by voltage source and perfect ZCS at resonance which confused many, including me for quite a long time. Richie did actually mention most of these things on magnetizing current, but not about it's importance (probably trying to simplify things up). His site actually leaves impression that magnetizing current is something very bad which should be reduced as much as possible, which is not completely true.

Richie has posted up lots of new information but it's scattered on forum and didn't seem to get much attention for some reason.

I'll try to summarize the points in few sentences:

1. In ''normal'' SSTC, there is NO zero current switching, not even close, and it never was. The magnetizing current is always sawtooth slope; it is the minimum of current that can ever possibly be switched.

2. for an typical SSTC, the magnetizing curent is fairly high. Not uncommon for it to be over 50% of primary current value.

3. The most important that needs to be realized is that there is nothing wrong with it! Yes, ZCS is what we *don't want* in a SSTC.
Why? Because this allows us to have ZVS (a tradeoff, as usual.)

4. ZCS is really not important for a simple reason - mosfets are really *stupidly* fast. For most smaller mosfets, the magnetizing current of many amps can easily be switched into few Mhz with very little loss apart from conduction loss.

5. How can high magnetizing current induce ZVS? Actually, all that is important for power devices is to switch inductive load. Inductive load makes the current maintain direction as the first device turns off, deadtime passes, and the other one turns on.

this does two big things:

- the magnetizing current *discharges* the output capacitance of a device before it turns on, so it can turn on without wasting the energy stored in this capacitance;

- the freewheeling diodes end conduction by adjacent device turning on and changing the direction of current - diode only sees few volts in reverse and this greatly reduces recovery losses.

6. If any of these losses were allowed to occur, they would be enormous at tipycal SSTC frequencies and would kill mosfets quickly.

7. I verified this by trying to base drive a TC directly from a half-bridge of mosfets - losses were intense even at just few W of output power. I ended putting inductance in parallel with bridge output, practically as much as a typical primary coil!

8. After all factors are summed up, there's hardly anything to be improved in a SSTC. I couldn't think anything better than the air core transformer and secondary feedback circuit - just use those as they are. If you want longer, non-CW sparks, go for a DRSSTC.

Marko

uzzors2k

Sun Aug 03 2008, 05:31PM

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

It's starting to make sense. The primary in a standard SSTC is just an inductor so current will have to lag voltage. Primary current feedback should work in a DRSSTC where the primary is part of a resonant circuit tuned to the same fres as the secondary, meaning just R.

For reference, I found this site answered lots TC theory questions I had.

A feedback winding can provide a signal lagging the inverter voltage by 90 degrees, which sounds like a good control reference to me. Maybe I'll do a SSTC project yet.

Marko

Sun Aug 03 2008, 07:22PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Uzzors wrote ...

It's starting to make sense. The primary in a standard SSTC is just an inductor so current will have to lag voltage. Primary current feedback should work in a DRSSTC where the primary is part of a resonant circuit tuned to the same fres as the secondary, meaning just R.

For reference, I found this site answered lots TC theory questions I had.

A feedback winding can provide a signal lagging the inverter voltage by 90 degrees, which sounds like a good control reference to me. Maybe I'll do a SSTC project yet.

Not sure what you mean by this Uzzors - in any case, it must be understood that due to magnetizing current SSTC's can +never+ soft switch, and for that reason they can never use primary feedback.

Although primary current waveform might appear 'phase shifted', this may be misleading notion. Rather, primary current is best viewed as load current superimposed onto magnetizing current. Load current is sinusoidal, but magnetizing current is sawtooth and can never cross zero on it's own; in most typical SSTC's this will just prevent primary current from crossing zero at all.

This might work only if magnetizing current is far lower than load current, like if we used a ferrite transformer; this would though create problems if additional inductance (known as commutating inductance) isn't provided externally as surrogate.

Leakage inductance can't produce any phase shifts because it is canceled out by resonator capacitance, it is only the magnetizing inductance that actually distorts waveform, and sinusoidal currents are always in phase everywhere.

If you want to see what your magnetizing current looks like, simply feed your inverter with signal at desired frequency and connect the primary coil without a secondary to the output.

Finally, I absolutely don't see what is wrong with secondary base CT feedback. There may not be much already designed circuits for it, but Steve Ward's PLL should do great judging by his success.

Marko

1 2 next

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