Phase Shift not working on SSTC feedback

[Anders Mikkelsen]

I think this is a good excuse for coming up with something better, ideally using the same parts.

As I see it, there are two main goals. The first one is for the circuit to provide a predictable and adjustable amount of phase lead, ideally constant across frequency and input signal level. The second one is for the ciruit to have a wide dynamic range, triggering on weak signals, while handling the full 1+ A RMS of base current of a typical SSTC. Ideally the magnetizing inductance of the current transformer should not form the main phase shifting element, due to its loose tolerance, non-linearity and temperature coefficient, all due to the lack of a gap in the core.

Do you have any good ideas here Dave? I would lean towards either making the inverter self-oscillate close to the target operating frequency and letting the base current signal injection lock the oscillator, which should provide very clean startup behavior. Another option is to use resistive feedback around the 74HC14 to lower the input hysteresis, in order to provide good sensitivity without needing a non-linear terminating impedance for the CT. I won't have a chance to look deeper into this until later, but it's definitely an interesting topic. One of my main worries with any kind of base feedback is how it handles ground arcs.

[Simranjit]

Hi, I think at this point my problem is completely solved. My main concern was to minimize inductive spikes so I can push mosfets to their limits and improve performance of Tesla coil. I drawing over 9A@30V with my DC lab bench power supply. I added some more turn to long inductor to see what happens if more phase lead is added. I found sweet spot. I will attach pictures of waveforms of drain of mosfets.
I wound another fat inductor with multiple taps. I am also attaching pictures of fat inductor (200 turns) at different frequencies tested with signal generator.
Although I made the long inductor so that I could fit multiple small ferrite cores inside it. It worked purely by chance for me without ferrite rods. It would be really helpful if someone can let know if Q factor is also variable here and maybe give me some formulas to get a ball park idea of shape of inductor to be used.

I can post some more waveforms if it would help me figure out what's going on. At this point I have 3 inductors.
Air core Fat inductor: 20mm length * 60mm diameter
1) Multiple taps at 100, 200, 300, 400 turns
2) Exactly 2000 turns

Air core Long inductor: 90mm length * 10.5mm diameter
3) Over 2000 turns
 

[davekni]

Do you have any good ideas here Dave? I would lean towards either making the inverter self-oscillate close to the target operating frequency and letting the base current signal injection lock the oscillator, which should provide very clean startup behavior.

Definitely something I've thought about for the past few years, though more for a simple UD1.3 upgrade for DRSSTCs.  UD1.3 uses similar 74HC14 CT circuitry.  Certainly self-oscillation is my favorite solution.  I use that in all my TCs and in other H-bridge circuits feeding resonant transformer primaries.
74HC14s have loosely defined hysteresis parameters.  Given their very simple internal circuitry, stability of hysteresis is surprisingly reasonable.  Though I haven't explored wide temperature ranges.  Center of input range is always below Vdd/2.  To get 50% duty cycle self-oscillation, output needs a little biasing towards negative rail.  I've thought about posting some variation of the following circuit as its own topic, but didn't want to post until I'd actually built and tested it.  Reasonably confident it would work fine here.

 [ You are not allowed to view attachments ]

POT R2 adjusts duty cycle.  POT R4 adjusts frequency.  C1 sets frequency range.  POT R5 adjusts phase lead.  L4 sets range of phase lead adjustment.  R1 is there for two reasons.  One is it allows input node v1 to be constructed physically small to avoid excess noise such as capacitive coupling to secondary etc.  Second, it limits max frequency for given C1 value.  Keeps input impedance well above CT feedback burden impedance of R5+L4.  If frequency can't be set high enough, C1 value needs to decrease.

Hi, I think at this point my problem is completely solved.

Yes, looks good!

I wound another fat inductor with multiple taps.

Might be fine, but one caution for a multi-tap inductor:  Self-resonant frequency is determined by highest turn count even if using a tap with many fewer turns.  One way to reduce this effect some is to isolate taps as separate windings.  Connect in series only when more turns are needed.  Leave unused sections open at both ends of each winding.

good_phase_lead_long_inductor.jpg

Switching is right at peak secondary voltage, which is zero secondary current.  In SSTCs, primary current phase always lags secondary current.  Spikes are minimum here because bridge switching is slightly before primary current reaches zero.

It would be really helpful if someone can let know if Q factor is also variable here and maybe give me some formulas to get a ball park idea of shape of inductor to be used.

Q is not that critical for phase-lead inductor.  Resistor in series is intentionally reducing Q of the L+R combination.

[Simranjit]

Hi Dave,

I tested your circuit. It works beautifully when I supply sine wave with signal generator instead of CT. Even if signal generator is off and input is connected to positive clip of signal generator the output is stable at 90kHz. However, when I connect CT the frequency starts to decrease steadily. Also the amplitude of output drops to 1.4V I didn't connect power to my H bridge. Maybe I'm doing something wrong here. Any ideas ?

P.S. I'm using shottky diodes 1N5819. I also blew one of UCC drivers the one before capacitor. Not sure if the other one is ok. So I just removed that too. Currently testing it without UCC drivers.
I also tried with power applied to h bridge. However Tesla coil didn't oscillate. Possibly because output is not stable when connected to CT. Currently at work, I will try again when I get home with signal generator driving Tesla coil and then test the output of your circuit again.

[davekni]

I tested your circuit. It works beautifully when I supply sine wave with signal generator instead of CT. Even if signal generator is off and input is connected to positive clip of signal generator the output is stable at 90kHz. However, when I connect CT the frequency starts to decrease steadily.

CT output is nominally a current source.  It has high impedance.  Thus impedance of input is defined mostly by L1 and R1 in your older circuit, L4 and R5 in my schematic.  (BTW, I should have matched reference numbers to yours for clarity.  Didn't take time to do that.)  That is one possible (likely) reason for different behavior.  Would expect a bit of frequency increase, not progressively continuing decrease.  What values do you have for L4 and R5?

Also the amplitude of output drops to 1.4V I didn't connect power to my H bridge. Maybe I'm doing something wrong here. Any ideas ?

That sounds like a more serious issue.  If frequency was MUCH higher, could be dropping amplitude doe to scope's frequency response and/or 74HC14's frequency capability.  I'd measure 5V supply to make sure it isn't dropping.

P.S. I'm using shottky diodes 1N5819.

Thank you for that information.  Good choice, and allows eliminating R6 if you want.  Does add a little parallel capacitance, but not likely enough to be of concern at 90kHz.

I also blew one of UCC drivers the one before capacitor. Not sure if the other one is ok. So I just removed that too. Currently testing it without UCC drivers.

Most likely cause would be oscillating at MUCH higher frequency.  I'm wondering if it is actually oscillating at a high frequency and that signal is aliasing with scope sample rate, causing a scope trace that appears to be low frequency.
I suggest this for all TC testing:  Start at low duty cycle enable pulses (short pulses with long gap between pulses, such as 1 per second).  That way many issues that could cause part failure will not have time to heat parts to failing point.

Thinking about my circuit proposal more, I'm becoming more certain that issue is VERY high frequency oscillation.  Actually an issue with my circuit and probably many other variations that combine self-oscillation with current feedback and phase lead.  Impedance of phase lead inductor is high at high frequency.  When that impedance is higher than C1, frequency is determined by just 1N5819 capacitance and not by C1 value.

First thought for a fix is to add a 1k resistor across L4 or directly across CT terminals.  That will reduce phase lead some, but also keep impedance low enough to avoid high frequency oscillation.  I'll edit this reply if I think of any better fixes.  (My apology for suggesting a circuit that fried your driver chip.  I should have tested before posting.  But then would have been a year or two.  So also, thank you for testing it for me!)

[Simranjit]

(My apology for suggesting a circuit that fried your driver chip.  I should have tested before posting.  But then would have been a year or two.  So also, thank you for testing it for me!)

No need to apologize. I'm always happy to test any untested circuits. I spend most of time
"What if I do this, what if I do that, connect it here, connect it there"
I'm always frying my circuits anyways to learn. This way I remember it for next time and precautions I should take. I have been working on Tesla coil for over a year. So far I have blown over 30 mosfets and 7 UCC drivers 😆😆😆.

I'm sorry for not giving further updates. I got carried away testing QCW on my tesla coils. Will try to test your circuit again soon😀😀😀.

[davekni]

Looked back at my circuit post:  It has even more errors.  I was definitely posting without sufficient thought.  D2 is backwards and C1 is on the wrong side of diodes.  Not a circuit to use even with those fixes.  High frequency oscillation remains an issue with that circuit.  And has another issue that phase lead changes with amplitude.

I've been trying circuit options in simulation (LTSpice).  The more I simulate, the more issues show up with every variation tried.  My conclusion is that it needs to look even more like UD2.7 input.  CT secondary needs to have much more current so that burden L+R impedance can be low.  Then a series resistor can isolate circuitry including clamp diodes from input without significantly changing input impedance.  In other words, CT burden impedance needs to be low followed by a series resistor at least several times higher value.  That way diode clamping (which is low impedance) doesn't drastically change CT burden impedance.

For secondary current feedback, CT will need to be 10:1 or lower to get good current.  You could wind 20 primary turns through your 200:1 CT if there is physical room.  Or use home-made CT.  Even for home made, might be good to have several (say 5) primary turns (monitoring TC secondary current) with 50 CT secondary turns for 10:1 ratio.  If you are interested in trying a 10:1 CT circuit, I'll make an 74HC14 self-oscillating option to work with that CT.  With simulation before posting to reduce chance of critical mistakes.

[Anders Mikkelsen]

I agree with Dave here, I experimented a bit with options as well and I came to the same conclusion that a proper comparator is the right way to go here, and the added cost is more than justified when looking at the big picture. I already had most of the following written down earlier today, so some of it is redundant in light of his last post.

Looking at the measurements using the 1000+ turn inductor, you can see that it indeed acts like a capacitor at 90 kHz. The phase shift is in the opposite direction compared to 20 kHz. The other waveforms with the squiggles on top of the square wave are less clear to me, what is being measured here?

The asymmetry of the 74HC14 is an interesting point, and one I had not considered. This, along with the way the series capacitor interacts with the feed impedance from an LC-terminated current transformer, is good motivation to spend the 50 cents it costs for a proper comparator. Since a comparator has two inputs, you can use one for the RC feedback to make it oscillate, and the second one for the feedback signal to avoid them interacting. If the budget is increased to 70 cents, you can get a double comparator and use the second half for primary overcurrent, which can be a good investment in circuit reliability.

I would argue it's a bit early to say that it's working well if you only have it running from 30 V DC. Most troubles with feedback show up once you have enough voltage for real damage to happen. If you've already blown 30 MOSFETs, it's a sign that it makes sense to make sure the drive scheme is robust before sacrificing more. As for steps to make this happen, I would either consider the standard RL phase lead into a proper comparator as done in the UD* drivers, or something self-oscillating built around a proper comparator. If you have some info on your current sense transformer, I can suggest component values to start with.

There seems to be a strong motivation for people to use the simplest possible circuit, even when it involves serious compromises on performance and reliability. I can see where this comes from, as I started out this way myself. Over time, I gravitated to more and more complex drive schemes, but found that people are reluctant to copy complex designs, and optimal designs often require careful tuning and a deep understanding of the circuit in order to do so. I think the whole community would benefit if a compromise is found, and some reference designs are made to be as simple as practical, but not any simpler than that. A good example is the Ward UD2x series of drivers, which have become the standard for DRSSTCs. I don't believe anything quite like it exists for ramped *SSTCs, as the Labcoatz design has some major deficiencies and people have had a lot of trouble with it evidenced by the numerous threads here discussing build experiences. I believe a self oscillating design based on injection locking has a lot of merits for this application, but we can discuss that in a separate threat do not derail this one too much.

[Simranjit]

Looked back at my circuit post:  It has even more errors.  I was definitely posting without sufficient thought.  D2 is backwards and C1 is on the wrong side of diodes.  Not a circuit to use even with those fixes.  High frequency oscillation remains an issue with that circuit.  And has another issue that phase lead changes with amplitude.

I've been trying circuit options in simulation (LTSpice).  The more I simulate, the more issues show up with every variation tried.  My conclusion is that it needs to look even more like UD2.7 input.  CT secondary needs to have much more current so that burden L+R impedance can be low.  Then a series resistor can isolate circuitry including clamp diodes from input without significantly changing input impedance.  In other words, CT burden impedance needs to be low followed by a series resistor at least several times higher value.  That way diode clamping (which is low impedance) doesn't drastically change CT burden impedance.

For secondary current feedback, CT will need to be 10:1 or lower to get good current.  You could wind 20 primary turns through your 200:1 CT if there is physical room.  Or use home-made CT.  Even for home made, might be good to have several (say 5) primary turns (monitoring TC secondary current) with 50 CT secondary turns for 10:1 ratio.  If you are interested in trying a 10:1 CT circuit, I'll make an 74HC14 self-oscillating option to work with that CT.  With simulation before posting to reduce chance of critical mistakes.

Hi Dave,
I'm also leaning more towards self-oscillating tesla coil because my tesla coil would randomly oscillate above 200kHz even when I'm using 2 stage low pass RC filter with secondary feedback which throws off my phase lead which in-turn leads to huge inductive spikes.
I made changes to your circuit according to your suggestions (see attached picture). I was able to successfully run my tesla coil at 90 kHz. I am able to successfully adjust phase shift using potentiometer attached to the inductor. Here are few things I noticed when my tesla coil is running at 90kHz:

1) Tesla coil is oscillating at 90kHz regardless of potentiometer value.
2) Frequency gets locked at 90kHz even when I'm using 100k-ohm for RV4 potentiometer no matter where potentiometer is set.
3) Duty cycle gets locked at 50%. No matter where RV2 potentiometer is set.
4) Signal on scope is bit shaky (like vibrating) and not clean.

I'm also attaching a suggestion circuit. Although this circuit oscillates quite nicely. I'm not sure if phase shifting will work with it. I have not tested it yet.

Edit: I tested the suggestion.jpg circuit with phase lead. It actually works. Waveform is quite stable and I have tested it over 40 times by turning off and turning on my tesla coil again and again.  It always oscillates at 90-91kHz.
P.S. I am using different primary in this test. So my waveform might look different from my previous posts.

[davekni]

1) Tesla coil is oscillating at 90kHz regardless of potentiometer value.
2) Frequency gets locked at 90kHz even when I'm using 100k-ohm for RV4 potentiometer no matter where potentiometer is set.

Self-oscillation with high-value feedback resistors (RV4 + R1) still works well to keep 74HC14 input biased between threshold voltages.  Self-oscillation frequency may be low with high-value feedback resistor.  Actual coil frequency takes over quickly when CT burden impedance is reasonably high (L2 + RV1).  Though high CT burden impedance may end up resulting in higher (reasonable) frequency self-oscillation.  Self-oscillation frequency is then determined by capacitance of D1 + D2 + 74HC14 input combined with feedback (RV4 + R1) resistance.  Perhaps that frequency happens to be close to 90kHz.

3) Duty cycle gets locked at 50%. No matter where RV2 potentiometer is set.

RV2 sets duty cycle for self-oscillation before CT signal takes over.  To see that duty cycle, run with no power to H-bridge.

4) Signal on scope is bit shaky (like vibrating) and not clean.

What signal are you triggering scope with?  Jitter may be just a scope triggering issue.

Edit: I tested the suggestion.jpg circuit with phase lead. It actually works. Waveform is quite stable and I have tested it over 40 times by turning off and turning on my tesla coil again and again.  It always oscillates at 90-91kHz.

Once CT signal takes over, frequency is determined by TC and not self-oscillation.  (Unless CT signal is weak and barely locking.)
Same for duty cycle.  Self-oscillation will be at non-50% duty cycle.  Once CT signal takes over, duty cycle approaches 50%.
For both of these circuits, phase lead is generated largely by 1nF input capacitor (C1 or C5) feeding low impedance of clamp diodes D1 and D2.  Phase lead depends on CT feedback amplitude.  Will likely work fine without inductor (L2 or L3).  Phase lead will be lower for first few cycles of oscillation while CT current builds up.