Phase Shift not working on SSTC feedback

[Simranjit]

Hi everyone,

I am trying to implement phase shift on my SSTC. I am using CT for feedback and inductor for shifting phase. Phase shift is not working regardless of value of trimmer inductor. I will attach my circuit diagram and signal from oscilloscope. Yellow trace is probe placed near SSTC. Purple trace is input of GDT (or output of UCC drivers). My circuit is oscillating at 90kHz.

Thank you.

[AstRii]

With such high resistance of the 100k pot, the 70uH inductor won't have much of an effect. Even if you turn the pot to some lower resistance value and turn the inductor to maximum, the phase shift will be minimal.
Phase shift can be calculated as: arctan((omega*L)/R)

Try to decrease the loading resistor, something like 10Ohms would be more reasonable.

Btw. resistor R3 is not needed, it doesn't serve any purpose. The input impedance of 74HC14 is much much larger than 1kOhms.

I also recommend to bias the feedback voltage with a resistor divider like this:


This makes it easier for the feedback to kick in as the voltage is biased closer to the triggering threshold of the 74HC14. The input signal can then have lower amplitude to trigger the schmidt trigger. This is useful especially if you are using lower resistance shunt resistor.

[Simranjit]

This makes it easier for the feedback to kick in as the voltage is biased closer to the triggering threshold of the 74HC14. The input signal can then have lower amplitude to trigger the schmidt trigger. This is useful especially if you are using lower resistance shunt resistor.

I noticed when using using smaller value pot when I'm lowering the value of pot the signal starts to shift a little bit. If I decrease the resistance further the oscillations just die down. I will test with biasing resistors as you described to see if that helps.

[AstRii]

I noticed when using using smaller value pot when I'm lowering the value of pot the signal starts to shift a little bit. If I decrease the resistance further the oscillations just die down. I will test with biasing resistors as you described to see if that helps.

If that won't help, try reducing the winding count of the feedback transformer to allow for higher current through the shunt resistor and therefore higher voltage.

[davekni]

Btw. resistor R3 is not needed, it doesn't serve any purpose. The input impedance of 74HC14 is much much larger than 1kOhms.

R3 does have a purpose.  74HC14 input impedance is high only when input is between 0 and Vdd (+5V).  Beyond that range internal body diodes and/or substrate diodes conduct.  That conduction can cause 74HC14 to not operate correctly or worst-case latch up.  Using standard silicon 1N4148 diodes for D1 and D2, their Vf is similar to internal 74HC14 diodes.  Current is shared between internal and external diodes.  Adding R3 causes most of the current to flow through D1 and D2 rather than into 74HC14 input.  If not using R3, then diodes should be schottky so that their Vf is lower than internal 74HC14 diodes.

I also recommend to bias the feedback voltage with a resistor divider like this:

That helps.  My preferred version is to add a single resistor from 74HC14 output back to left side of R2.

Phase shift is not working regardless of value of trimmer inductor.

Lowering R or increasing L to get time constant within useful range is necessary.  However, to make it work like UD2.7 input, left side of C1 needs to connect to left side of L1 (J1-2) rather than right side of L1.  (As it is, has no effect at all until L1 is large compared to CT secondary inductance.)

[Simranjit]

However, to make it work like UD2.7 input, left side of C1 needs to connect to left side of L1 (J1-2) rather than right side of L1.  (As it is, has no effect at all until L1 is large compared to CT secondary inductance.)

That's what I was trying to replicate 😁😁😁

[Simranjit]

Hi again,
Now I'm even more confused. I couldn't make trimmer inductor work because I think the core of trimmer inductor was rated for much higher frequency 400khz and above.
I opted to use capacitor instead with a trimmer pot. I get maximum spark length (more current pulled from power supply) without any phase shift circuit. After adding phase shift circuit with a capacitor and a pot. The spark length decreases (low current pulled from power supply). Is that supposed to happen ? I thought I would get more spark length.

I'm getting more confused now. I will attach some pics of oscilloscope.

Also the drain signal doesn't look very good either when phase shift is not added. It looks better when phase shift is added.

Edit:
No phase shift: 5amps @ 22V
Phase shift: less than 1 amp @ 22V

P.S. I added 20k resistor from output to input of 74hc14. That definitely helped in sustaining oscillations.

[davekni]

I couldn't make trimmer inductor work because I think the core of trimmer inductor was rated for much higher frequency 400khz and above.

Core will work at lower frequency.  Issue would be winding enough turns around core to get desired inductance.

I opted to use capacitor instead with a trimmer pot. I get maximum spark length (more current pulled from power supply) without any phase shift circuit.

Without seeing a circuit, I don't know how you've connected a capacitor.  Scope plots suggest capacitor is delaying phase rather than making it earlier.

After adding phase shift circuit with a capacitor and a pot. The spark length decreases (low current pulled from power supply). Is that supposed to happen ? I thought I would get more spark length.

That's because your circuit is adding phase lag rather than phase lead.

Also the drain signal doesn't look very good either when phase shift is not added. It looks better when phase shift is added.

That's likely due to change in primary current rather than any direct consequence of added phase lag.

P.S. I added 20k resistor from output to input of 74hc14. That definitely helped in sustaining oscillations.

Great, glad that is working for you.

[Simranjit]

That's because your circuit is adding phase lag rather than phase lead.

I knew I was missing something. Thank you very much for clarification. I really appreciate it. I will do some more testing. I have quite a few medium sized N97 cores. I will try to make my own inductor for phase lead.

[davekni]

I have quite a few medium sized N97 cores. I will try to make my own inductor for phase lead.

N97 is a high permeability material, as are most ferrites.  To be a useful inductor, core needs to have a gap.  Gapped ferrite core inductors are common.  If using a normal toroid core, you need to cut a gap through core.

[Simranjit]

I have quite a few medium sized N97 cores. I will try to make my own inductor for phase lead.

N97 is a high permeability material, as are most ferrites.  To be a useful inductor, core needs to have a gap.  Gapped ferrite core inductors are common.  If using a normal toroid core, you need to cut a gap through core.

You just saved me a lot of time. Thank you again.

[AstRii]

To be a useful inductor, core needs to have a gap.

Could you elaborate on that? I thought gaps only serve a purpose to increase the saturation current of an inductor. Which is most likely not an issue here with these small currents in the feedback path.
 
Also it would be harder to create decent inductance with gapped core instead of non-gapped. What am I missing?

[davekni]

Could you elaborate on that? I thought gaps only serve a purpose to increase the saturation current of an inductor. Which is most likely not an issue here with these small currents in the feedback path.

Great question!  Saturation is usually the most important reason for gapping ferrite inductor cores.  I agree that saturation is unlikely to be an issue here.
The other reason for gapping ferrite inductor cores is to get stable inductance.  Ferrite material (including N97) permeability varies widely over temperature, DC bias level, and AC amplitude.
Temperature might not be a problem here if circuit remains at room temperature (core not adjacent power devices).  N97 permeability changes about 2:1 over 20C to 100C.
DC bias level is zero in this application.  But given material hysteresis, low amplitude AC use can depend some on history of past DC bias exposure.
AC amplitude is the key issue for this application.  Permeability change with amplitude is worst at low amplitudes where this core would be used.  There is little information for such low amplitudes in data sheets as cores aren't generally used at such low amplitudes.  Can easily be 2:1.
BTW, for any commercial designs, one more factor would be problematic, part-to-part variation of permeability.  N97 permeability is +-25%.

For this specific use with secondary current feedback, an ungapped N97 ferrite core might work well enough.  Secondary current changes less than primary current as power increases.  But I'd hate to have some finicky issue show up later when temperature is different or power increases or a ground strike changes secondary current.

Also it would be harder to create decent inductance with gapped core instead of non-gapped.

Yes, that is the down-side of either gapping a ferrite core or using a powdered iron core (which is iron with distributed gaps between particles).  It is a universal tradeoff with inductor design.  Larger gap requires more wire but results in more stable inductance.

[Simranjit]

Hi again,

I have some updates I want to share.

1) I couldn't cut slot in ferrite but my signal generator arrived today. The variable inductors I have do work at 90 kHz.

2) Before I was using 200:1 CT. So, I wound my own with 45:1 CT and 75:1 CT using N87 toroid. I tested them with my setup without any phase shift just to confirm that it is working. Tesla coil was oscillating at 350kHz. So I created a low pass filter with capacitor and pot to filter out high frequency. It started oscillating at 90kHz

3) I connected variable inductor with 47R resistor just as in ud2.7 I took core out of inductor to make sure phase shift was minimal. There is a high pitch sound coming from GDT with/without power applied to bridge. The result is same if core is inside it. As soon as I disconnect inductor the Tesla coil oscillates properly.

4) Using higher value of resistor with variable inductor will make it oscillate however then the phase lead will be minimal.

[davekni]

1) I couldn't cut slot in ferrite but my signal generator arrived today. The variable inductors I have do work at 90 kHz.

Yes, ferrite is a hard material.  I use a diamond wheel on my Dremel tool for cutting ferrite.

2) Before I was using 200:1 CT. So, I wound my own with 45:1 CT and 75:1 CT using N87 toroid. I tested them with my setup without any phase shift just to confirm that it is working. Tesla coil was oscillating at 350kHz. So I created a low pass filter with capacitor and pot to filter out high frequency. It started oscillating at 90kHz

Yes, SSTCs (coils using feedback from secondary) can oscillate at higher frequency modes of secondary.  Much like bugles can make multiple notes using the same resonant tube length.
Adding a low-pass filter adds phase lag, so not an ideal solution.  Hopefully other people with SSTC experience can offer suggestions.  My only SSTC is self-oscillating at close to operating frequency.  That starts operation at correct frequency.  One option for you is to make your self oscillation (20k resistor across 74HC14) run close to 90kHz.
Does your SSTC have a top load (ie. toroid)?  If so, how big is it compared to secondary?  I think higher mode oscillation is more of a problem for SSTCs with no or small top loads.

3) I connected variable inductor with 47R resistor just as in ud2.7 I took core out of inductor to make sure phase shift was minimal. There is a high pitch sound coming from GDT with/without power applied to bridge. The result is same if core is inside it. As soon as I disconnect inductor the Tesla coil oscillates properly.

74HC14 requires more input voltage than does comparitor of UD2.7.  Even with your 45 turn CT, current is likely not high enough to generate enough voltage across 47R.  Might be enough if C1 (1uF input coupling capacitor in your circuit) is reduced to get self-oscillation frequency close to 90kHz.  I'd guess a value around 1nF to 5nF.  Value of 20k resistor could then be tweaked to get frequency close enough.

4) Using higher value of resistor with variable inductor will make it oscillate however then the phase lead will be minimal.

Another option, at least for testing, is to try your original idea of winding a higher value inductor from ungapped N97 core.  As mentioned in my last response, inductance may vary, but perhaps be stable enough to get your circuit working.  Then you could make a better inductor sometime later.

Wishing you luck with your project and learning!

[Simranjit]

Hi everyone  ;

I finally got some good news. This feels like a milestone completed. Here is the story:

So I decided to wind my own inductor since I didn't had any Dremel rotary cutter. I ended up winding my own inductor over 1200 turns (I lost count) with 34AWG wire. I connected everything as shown in schematic but coil would not start oscillations. I changed potentiometer to 20k connected everything back together, put potentiometer to max resistance. Since the resistance is very high the phase lead would be minimal. When the coil started oscillating I slowly decreased resistance of potentiometer to zero Ohms Lo and behold I finally achieved phase lead  . I started drawing more current from power supply, spark length increased, inductive spikes became minimal. Oh it was beautiful. I know this is not maximum phase lead. I will wind another air core coil with bigger diameter, multiple taps and possibly more turns.

Although inserting ferrite cores in air core coil started introducing phase lag instead of phase lead. I think I need to figure that out too.

I was really excited and could not wait to share results with everyone. I really appreciate everyone's help who took time out of their personal lives to help people like me. Thank you very much to everyone from bottom of my heart.

P.S. Some nice guy from facebook marketplace also gave me 750W isolated AC transformer for real cheap. I can finally monitor my tesla coil (QCW) riding on AC mains ramp with oscilloscope. The above results are using my lab bench DC power supply. I'm still using 1:200 CT though LTE-99002.

[davekni]

I finally got some good news. This feels like a milestone completed. Here is the story:

Wonderful to hear!  Thank you for sharing good news.

Although inserting ferrite cores in air core coil started introducing phase lag instead of phase lead. I think I need to figure that out too.

That does sound puzzling.  I'm not aware of any ferrite material so electrically conductive that it would reduce inductance at 90kHz.  Perhaps it is coupling magnetic field from primary coil to inductor.

I'm still using 1:200 CT though LTE-99002.

Lower ratio CT would allow more phase lead with the existing inductor you constructed.

[Simranjit]

Hi,

Here are some further updates. Things are getting more interesting. So, I decided to make another air core inductor. This coil has diameter of 60mm and length 20mm. I decided to make length short because inductance is inversely proportional to length. I wound exactly 2000 turns with 34awg wire which would give inductance of 305mH roughly. Following same procedure as before I connected everything together. I slowly started turning down the potentiometer. After about 5 degree of phase lead it started to phase lag.
I think the same thing occurred with previous long inductor (under 1mH) when I started to insert ferrite core in it. It started showing phase lag instead of phase lead.


A few things come to mind (90kHz frequency):
1) Maybe I should have considered Q factor of inductor.
    Long inductor = under 1mH, Q = 18, Reactance
    Fat inductor  = 305mH, Q = 60
2) Inductive Reactance Xl (Low is preferred)

[Anders Mikkelsen]

It makes sense to revisit the whole circuit at this point, because there could be multiple problems interacting here.

The 1k series resistor before the diodes, although widely used in base current feedback SSTC circuits, doesn't make much sense to me, maybe I'm missing something? A resistor in series with a current source doesn't do anything except for causing power loss. When used with a real world current transformer, it will cause a phase shift by interacting with the magnetizing inductance of the transformer, and it might also cause saturation of the current transformer core. The voltage resulting voltage drop across the transformer will also lead to core losses, with corresponding phase shift. So there's a lot of unpredictable phase shift added, some of which depends on the signal level. And this is only while the clamping diodes are forward biased. For signal levels below where the clamping diodes act, the transformer is essentially terminated into the capacitance of the diodes, which is small if using 4148s, and significant if using 5819s for clamping. Phase shift is very hard to predict and all over the place.

If you have some data on the  core size, core material and number of turns for the base current CT, this can be coarsely estimated to get a ballpark idea. I would just terminate the current transformer into either a resistor of moderate value to GND, or a resistor and inductor in series as done for phase lead when using primary current feedback. With some data on the current transformer, I can suggest starting point values.

I know this resistor is commonly used, and I've seen it in more than a few base current feedback SSTC circuits, this doesn't make it correct however, and I suspect this is one of the reasons why a lot of people seem to struggle with base current feedback.

Your inductor of 1000+ turns is for sure not acting like and inductor at these frequencies, but its behavior will be dominated by its stray capacitance and the interaction of the stray capacitance with the inductance. It could be that this capacitance adds enough phase lag to compensate for the excessive phase lead of the badly terminated current transformer, at least enough to make it oscillate. If you have an impedance analyzer, or even a signal generator and oscilloscope, this can easily be investigated by making the inductor part of a voltage divider, and seeing how the amplitude and phase of the output signal depends on the input signal frequency.

[davekni]

This coil has diameter of 60mm and length 20mm. I decided to make length short because inductance is inversely proportional to length. I wound exactly 2000 turns with 34awg wire which would give inductance of 305mH roughly.

At 305mH, I'd guess its self-resonance is below 90kHz.  Inter-winding capacitance is likely more significant (lower impedance) than inductance at 90kHz.  Anders discusses this in his last paragraph too.

I think the same thing occurred with previous long inductor (under 1mH) when I started to insert ferrite core in it. It started showing phase lag instead of phase lead.

Have you tried scoping CT input as you adjust phase?  One possibility that comes to mind is that amplitude drops enough that 74HC14 input hysteresis is adding phase lag.

The 1k series resistor before the diodes, although widely used in base current feedback SSTC circuits, doesn't make much sense to me, maybe I'm missing something? A resistor in series with a current source doesn't do anything except for causing power loss. When used with a real world current transformer, it will cause a phase shift by interacting with the magnetizing inductance of the transformer, and it might also cause saturation of the current transformer core. The voltage resulting voltage drop across the transformer will also lead to core losses, with corresponding phase shift. So there's a lot of unpredictable phase shift added, some of which depends on the signal level. And this is only while the clamping diodes are forward biased. For signal levels below where the clamping diodes act, the transformer is essentially terminated into the capacitance of the diodes, which is small if using 4148s, and significant if using 5819s for clamping. Phase shift is very hard to predict and all over the place.

Yes, I generally agree.  I'm guessing that some builds happen to get useful phase lead due to CT parallel inductance.  Since it seems many more SSTC designs use antenna feedback rather than CT feedback, circuit is often lifted from antenna-feedback designs.  Even there, 1k resistor has little purpose.  Perhaps slows down edge enough to be helpful in cases of ESD or top load to antenna strikes.
However, with the added inductor and resistor, those become additional CT burden impedance.  1k is useful here.